Datasets:

Hendrik-a

/

arxiv

like 0

new necessary and sufficient conditions are given for the quantization of a class of periodic second order non - homogeneous ordinary differential equations in the complex plane in this paper . the problem is studied from the viewpoint of complex oscillation theory first developed by bank and laine ( 1982 , 1983 ) and gundersen and steinbart ( 1994 ) . we show that when a solution is complex non - oscillatory ( finite exponent of convergence of zeros ) then the solution , which can be written as special functions , must degenerate . this gives a necessary and sufficient condition when the lommel function has finitely many zeros in every branch and this is a type of quantization for the non - homogeneous differential equation . the degenerate solutions are of polynomial / rational - type functions which are of independent interest . in particular , this shows that complex non - oscillatory solutions of this class of differential equations are equivalent to the subnormal solutions considered in a previous paper of the authors ( to appear ) . in addition to the asymptotics of special functions , the other main idea we apply in our proof is a classical result of e. m. wright which gives precise asymptotic locations of large zeros of a functional equation .

1002.3853

let @xmath0 be a transcendental entire function , and that @xmath1 an entire function solution of the differential equation @xmath2 we use the @xmath3 and @xmath4 to denote the _ order _ and _ exponent of convergence of zeros _ , respectively , of an entire function @xmath1 . a solution is called _ complex oscillatory _ if @xmath5 and _ complex non - oscillatory _ if @xmath6 . reader can refer to @xcite or @xcite for the notation and background related to nevanlinna theory where this research originated . however , we shall not make use of the nevanlinna theory in the rest of this paper . for earlier treatments of various complex oscillation problems considered , we refer the reader to , for examples , @xcite . in this paper , we consider the complex oscillation problem of a class of non - homogeneous differential equations which includes the simple looking equation @xmath7 as a special case , where @xmath8 and @xmath9 . it is well - known that all solutions of the equation are entire , see @xcite . in ( * ? ? * theorem 1.2 ) , the authors gave several explicit solutions of ( [ e : chiang - yu - equation - speical - case ] ) in terms of the sum of the bessel functions of first and second kinds @xmath10 and @xmath11 , and the _ lommel function _ @xmath12 ( [ a : bessel ] ) . in fact , the general solution of ( [ e : chiang - yu - equation - speical - case ] ) can be written as @xmath13 where @xmath14 and @xmath12 is a _ particular integral _ of the non - homogeneous bessel differential equation @xmath15 the functions @xmath10 and @xmath11 are two _ linearly independent _ solutions of the corresponding homogeneous bessel differential equation of ( [ e : non - homogeneous - bessel - de ] ) . the lommel function @xmath16 , is a special function that plays important roles in numerous physical applications ( see e. g. @xcite ) , was first appeared to be studied by lommel in @xcite . we refer the reader to @xcite and the references therein for further discussion about the background and the applications of the lommel function . the authors previous work ( * ? ? ? * theorem 1.2 ) concerns the _ subnormality _ of the solutions of ( [ e : chiang - yu - equation - speical - case ] ) . we recall that an entire function @xmath1 is called _ subnormal _ if either @xmath17 holds , where @xmath18 denotes the usual maximum modulus of the entire function @xmath1 and @xmath19 is the nevanlinna characteristics of @xmath1 . we have shown that solutions of ( [ e : chiang - yu - equation - speical - case ] ) are subnormal , that is , if ( [ d : subnormal - definition ] ) holds , if and only if @xmath20 and either @xmath21 or @xmath22 holds for a non - negative integer @xmath23 in ( [ e : chiang - yu - equation - speical - case - solutions ] ) and the subnormal solutions have the form given by the formulae ( [ e : lommel - sol - special ] ) and ( [ e : ck - special ] ) . in other words , _ subnormal solutions and finite order solutions of ( [ e : chiang - yu - equation - speical - case ] ) are equivalent _ , see corollary [ r : equivalent - conditions ] below . this provides a new _ non - homogeneous function - theoretic quantization - type _ result for the equation ( [ e : chiang - yu - equation - speical - case ] ) whose explanation will be given in 8 . [ r : chiang - yu - results]the authors also generalized the above results to a much more general equation ( [ e : chiang - yu ] ) below ( ( * ? ? ? * theorem 1.4 ) ) , and a number of interesting corollaries . for examples , orders of growth of the entire functions @xmath24 and @xmath25 were determined and _ non - homogeneous function - theoretic quantization - type results _ were also obtained . see ( * ? ? ? * theorem 1.7 , 6 ) for details . for homogeneous differential equation ( [ e : homogeneous - de ] ) , it is known that a solution @xmath1 could have @xmath6 but its growth being _ not subnormal _ , that is , + @xmath26 . thus one may ask what is the relationship between subnormality of solutions and its exponent of convergence of zeros of ( [ e : chiang - yu - equation - speical - case ] ) . we show that the finiteness of the two measures are equivalent . here are our main results : [ t : chiang - yu - special ] let @xmath1 be a solution of ( [ e : chiang - yu - equation - speical - case ] ) . then @xmath6 if and only if @xmath27 in ( [ e : chiang - yu - equation - speical - case - solutions ] ) and either @xmath21 or @xmath22 for a non - negative integer @xmath23 and @xmath28,\end{aligned}\ ] ] where the coefficients @xmath29 , are defined by @xmath30.\end{aligned}\ ] ] the above result is a special case of the following theorem [ t : chiang - yu ] . we first introduce a set of more general coefficients . suppose that @xmath31 is a positive integer and @xmath32 are complex numbers such that @xmath33 are non - zero and at least one of @xmath34 , @xmath35 , being non - zero . [ t : chiang - yu ] let @xmath1 be an entire solution to the differential equation @xmath36f \notag \\ & = \sum_{j=1}^n \sigma_j l^{\mu_j+1}m^2{{\rm e}}^{[m(\mu_j+1)-n]z}. \end{aligned}\ ] ] then @xmath1 is given by @xmath37 . \end{aligned}\ ] ] moreover , suppose that all the @xmath38 are distinct . then we have @xmath6 if and only if @xmath27 and for each non - zero @xmath34 , we have either @xmath39 where @xmath40 is a non - negative integer and @xmath41,\ ] ] where @xmath42 are defined by @xmath43.\ ] ] as an immediate consequence of the theorems [ t : chiang - yu - special ] or [ t : chiang - yu ] , we get the following result which gives us information about the number of zeros of the lommel function @xmath16 in the sense of nevanlinna s value distribution theory : [ t : chiang - yu - lommel - zeros ] suppose that @xmath34 are complex constants such that at least one of @xmath34 is non - zero , where @xmath44 . suppose further that @xmath38 are distinct and @xmath45 are lommel functions of arbitrary branches given in lemma [ l : lommel - continuation - general-1 ] . then each branch of the function @xmath46 has finitely many zeros if and only if either @xmath47 or @xmath48 for non - negative integers @xmath40 . in particular , the special case @xmath49 implies that each branch of @xmath50 has finitely many zeros must satisfy either @xmath21 or @xmath22 for a non - negative integer @xmath23 . the lommel function @xmath50 , as in the cases of many classical special functions , has , in general , infinitely many branches , that is , its covering manifold has infinitely many sheets . the values of the function in different branches are given by so - called _ analytic continuation formulae_. such analytic continuation formulae in its full generality , first derived by the authors in @xcite , are given in lemma [ l : lommel - continuation - general-1 ] . [ r : equivalent - conditions ] suppose that @xmath1 is a solution of ( [ e : chiang - yu - equation - speical - case ] ) . then we have @xmath6 if and only if the solution @xmath1 is subnormal . [ r : bessel - lommel - independence - branches ] we note that for all values of @xmath51 and @xmath52 , @xmath53 and @xmath54 are entire functions in the complex @xmath55-plane . hence they are single - valued functions and so are _ independent of the branches _ of @xmath45 . the main idea of our argument in the proofs is based on the asymptotic expansions of special functions ( bessel and lommel functions ) , the analytic continuation formulae for @xmath12 ( those formulae were first discovered by the authors which play a very important role in @xcite and also in this paper ) , the asymptotic locations of the zero of the transcendental equation @xmath56 given by wright @xcite , @xcite and application of rouch s theorem on suitably chosen contours in the complex plane . this paper is organized as follows . we introduce the lommel transformation in 2 , which serves as a crucial step in our proof to transform the equation ( [ e : chiang - yu ] ) into the equation @xmath57 since we need to consider the _ different branches _ of the function ( [ e : general - soln - first - second ] ) in the proof of the theorem [ t : chiang - yu ] , so the analytic continuation formulae for @xmath12 come into play at this stage . we quote these formulae , which were derived in @xcite , in 3 for easy reference . besides , we need information about the zeros of the function @xmath58 , where @xmath59 and @xmath60 . it turns out that wright has already investigated the precise locations of zeros of the equation @xmath61 in @xcite , where @xmath62 . in fact , this problem is of considerable scientific interest , see for examples ( @xcite ) . since we need to modify wright s method in the preliminary construction of one of the contours used in the proof of the theorem [ t : chiang - yu ] , so we shall sketch wright s method in 4 . a detailed study of the zeros of the function @xmath58 will be given in 5 , followed by the proof of the theorem [ t : chiang - yu ] in 6 . a proof of the corollary [ t : chiang - yu - lommel - zeros ] is presented in 7 and a discussion about the non - homogeneous function - theoretic quantization - type result will be given in 8 . appendix a contains all the necessary knowledge about the bessel functions and the lommel functions that are used in this paper . lommel investigated transformations that involve bessel equations @xcite in 1868 . our standard references are @xcite , @xcite and @xcite . lommel considered the transformation @xmath63 and @xmath64 , where @xmath65 and @xmath66 are the new independent and dependent variables respectively , @xmath67 and @xmath68 . we apply this transformation to equation ( [ e : generalized - lommel ] ) to obtain a second order differential equation in @xmath66 whose general solution is @xmath69 . following the idea in @xcite , we apply a further change of variable by @xmath70 and @xmath71 to the said differential equation in @xmath66 and then replacing @xmath72 and @xmath73 by @xmath33 and @xmath74 respectively . this process yields ( [ e : chiang - yu ] ) . as we have noted in 1 that the general solution of ( [ e : non - homogeneous - bessel - de ] ) is given by a combination of the bessel functions of first and second kinds and the lommel function @xmath16 ( see @xcite ) , hence the general solution to ( [ e : generalized - lommel ] ) is @xmath75 where @xmath76 and @xmath77 . it is easily seen that @xmath27 if and only if @xmath78 . thus the general solution @xmath79 of ( [ e : chiang - yu ] ) assumes the form @xmath80 . \end{aligned}\ ] ] we first note that the lommel functions @xmath12 have a rather complicated definition with respect to different subscripts @xmath81 and @xmath52 ( in four different cases ) even in the principal branch . in this section , we shall not repeat the description of its definition , interested readers please refer to @xcite and the references therein . here we only record the analytic continuation formulae of @xmath16 and the proofs of them can be found in @xcite . the asymptotic expansion and the linear independence property of @xmath12 will be given in the appendix . let @xmath82 . we define the constants @xmath83\cos\big(\frac{\mu+\nu}{2}\pi\big),\\ k'_\pm & : = \pi 2^{\nu-2 } i{{\rm e}}^{-m\nu\pi i}\gamma(\nu)\big[u_{m-1}(\cos \nu\pi){{\rm e}}^{(m \pm 1)\nu\pi i}-m\big],\\ k''_\pm & : = -\frac{m\pi^2(m \pm 1)}{4 } , \end{split}\ ] ] where @xmath84 is the gamma function and @xmath85 is the _ chebyschev polynomials of the second kind_. when @xmath86 for any integer @xmath23 , then we have ( * ? ? ? * theorem 3.4)[l : lommel - continuation - general-1 ] let @xmath87 be an integer . 1 . we have + @xmath88,\notag\end{aligned}\ ] ] + where @xmath89 is a rational function of @xmath90 and @xmath91 given by + @xmath92[1+{{\rm e}}^{-(\mu-\nu)\pi i}]}.\end{aligned}\ ] ] 2 . furthermore , the coefficients @xmath89 and @xmath93 are not identically zero _ simultaneously _ for all @xmath94 and all non - zero integers @xmath87 . when either @xmath95 or @xmath96 is an odd negative integer @xmath97 , where @xmath23 is a non - negative integer , then we have another sets of analytic continuation formulae which are given by ( * ? ? ? * lemmae 3.6 , 3.8 , 3.10)[l : lommel - continuation - general-2 ] let @xmath87 be an integer . then we have 1 . if @xmath98 , then we have + @xmath99 . \end{aligned}\ ] ] 2 . if @xmath100 , then we have + @xmath101 . \end{aligned}\ ] ] 3 . we define @xmath102 and for every polynomial @xmath103 of degree @xmath31 , we define @xmath104 to be the polynomial containing the term of @xmath103 with odd powers in @xmath105 and @xmath106 . if @xmath107 is a positive integer @xmath31 , then we have + @xmath108\notag \\ & \quad+\overline{b}_n(\zeta)\big[k_+''h_0^{(1)}(\zeta)+k_-''h_0^{(2)}(\zeta)\big]\notag\\ & \qquad-\zeta\overline{c}_n(\zeta)\big[k_+''h_1^{(1)}(\zeta)+k_-''h_1^{(2)}(\zeta)\big]\big\},\notag \end{aligned}\ ] ] + where @xmath109 and @xmath110 are polynomials in @xmath105 of degree at most @xmath31 such that @xmath111 and when @xmath112 , that they satisfy the following recurrence relations : + @xmath113 suppose that @xmath114 is a non - zero complex number such that @xmath115 , where @xmath116 and @xmath117 . in 1959 , wright @xcite obtained precise asymptotic locations of the zeros of the equation @xmath118 in terms of rapidly convergent series by constructing the riemann surface of the inverse function of @xmath119 . this result of wright is of considerable scientific interest , particularly in the theory and various applications of difference - differential equations , see @xcite . for further applications of this equation , please refer to @xcite . in this section , we first describe wright s result in the lemma [ l : wright ] below . next we apply wrights result to obtain finer estimates of the real and imaginary parts of the solutions of ( [ e : wright - equation ] ) in the lemmae [ l : real - imaginary - bounds ] and [ l : real - imaginary - bounds - subseq ] below which we need to construct a certain contour needed in the proof of the main result in proposition [ l : g - number - zeros ] . suppose that @xmath31 is an integer . we let @xmath120 be solutions of equation ( [ e : wright - equation ] ) , where @xmath121 and @xmath122 are real , and are given in the following result of wright : [ l : wright ] let @xmath115 , where @xmath123 and @xmath117 . let @xmath124 be the sign of the non - zero integer @xmath31 . we define @xmath125 taking @xmath126 real . if @xmath127 is sufficiently large such that @xmath128 then the solutions of the equation ( [ e : wright - equation ] ) are given by @xmath129 where @xmath130 and @xmath131 is the sequence of polynomials defined by @xmath132 where @xmath87 is a positive integer . [ r : eta - properties ] we deduce from the definition ( [ e : wright - h_n - beta_n ] ) that @xmath133 for all @xmath31 sufficiently large and @xmath134 as @xmath135 . we also note that it follows from ( [ e : qm - definition ] ) that @xmath136 is a polynomial of degree at most @xmath87 . this and the series representation of @xmath137 show that @xmath138 , @xmath139 and @xmath140 as @xmath135 . as we have already mentioned in 1 that we would like to apply rouch s theorem to suitable contours . to construct one of these contours , it is necessary to derive accurate bounds for @xmath121 and @xmath122 from the following lemma [ l : real - imaginary - bounds ] of wright . we include the argument leading to the inequalities to familiarize our readers for later applications . @xcite[l : real - imaginary - bounds ] suppose that @xmath141 and @xmath142 are as defined in ( [ e : wright - h_n - beta_n ] ) . then the upper and lower bounds for the real and imaginary parts of the solutions to ( [ e : wright - equation ] ) are given , respectively by @xmath143 and @xmath144 it is easy to see that the inequalities ( [ e : imaginary - bounds ] ) follow easily from the definitions ( [ e : wright - h_n - beta_n ] ) , ( [ e : wright - solutions ] ) and the properties of @xmath137 in the remark [ r : eta - properties ] above . for the inequality ( [ e : real - bounds ] ) representing the real part of @xmath145 , we deduce from the power series of @xmath146 and equation ( [ e : wright - solutions ] ) that @xmath147 this implies that the inequalities @xmath148 hold when @xmath31 is sufficiently large . on the other hand , the lemma [ l : wright ] and remark [ r : eta - properties ] assert that @xmath149 when @xmath31 is sufficiently large . combining these two inequalities and the fact @xmath150 , we deduce @xmath151 and @xmath152 since @xmath153 . hence we deduce from the inequalities ( [ e : real - upper - bounds ] ) and ( [ e : real - lower - bounds ] ) the inequalities @xmath154 since @xmath155 , we must have @xmath156 for every non - zero integer @xmath31 . then the desired inequalities ( [ e : real - bounds ] ) follow from this fact , the inequalities ( [ l : real - imaginary - bounds - inequality-3 ] ) and the definition ( [ e : wright - h_n - beta_n ] ) . this completes the proof of the lemma . [ l : real - imaginary - bounds - subseq ] let @xmath87 be a fixed positive integer such that @xmath157 . we define @xmath158 for every real @xmath159 and let @xmath160 , where @xmath161 is a sufficiently large positive integer such that @xmath162 satisfies the inequalities ( [ e : condition - n ] ) . then we have @xmath163 since @xmath162 is large and negative , so the inequalities ( [ e : real - imaginary - bounds - subseq - inequalities ] ) for @xmath164 follows easily from the second set of inequalities in ( [ e : imaginary - bounds ] ) . we next observe that both inequalities @xmath165 hold for @xmath161 sufficiently large . hence it follows from the inequalities ( [ e : real - bounds ] ) that @xmath166 and @xmath167 completing the proof of the lemma . [ r : z - zero - rectangle]we remark from the inequalities ( [ e : real - imaginary - bounds - subseq - inequalities ] ) that the particular set of zeros @xmath168 of equation ( [ e : wright - equation ] ) must lie inside the rectangles whose vertices are given by the points @xmath169 , @xmath170 , @xmath171 and @xmath172 in the complex @xmath55-plane . in the proof of our main results , it will become clear in 6 that we need to know the locations of zeros of the auxiliary function @xmath173 where @xmath174 are non - zero complex constants such that @xmath175 , and @xmath176 takes the principal branch . will be discussed in the lemma [ l : contour - special ] . ] we apply the results from [ s : location - zeros ] to investigate the asymptotic locations of zeros for @xmath177 . to do so , we first transform the equation @xmath178 into the form of ( [ e : wright - equation ] ) , where @xmath179 let @xmath180 and @xmath181 , where @xmath31 is an integer , @xmath182 and @xmath183 . then it follows from ( [ e : z - a - definitions ] ) and @xmath120 that for sufficiently large positive or negative integers @xmath31 , @xmath184 in order to find precise asymptotic locations of zeros of the function ( [ e : g - definition ] ) , we first consider the particular case that @xmath185 in ( [ e : zeta - real - imaginary ] ) . this forces @xmath186 and @xmath187 where @xmath188 . therefore we obtain from lemma [ l : real - imaginary - bounds - subseq ] and remark [ r : z - zero - rectangle ] the following lemma . [ l : zeta - bounds ] let @xmath185 and @xmath162 be defined as in lemma [ l : real - imaginary - bounds - subseq ] . then for @xmath161 sufficiently large , we have @xmath189 in other words , the zeros @xmath190 must lie inside the rectangles @xmath191 whose vertices are given by the points @xmath192 , @xmath193 and @xmath194 in the @xmath105-plane . [ r : fourth - quadrant]in view of the lemma [ l : zeta - bounds ] , we easily see that , when @xmath185 , _ all _ such zeros @xmath190 lie in the fourth quadrant of the @xmath105-plane and the real part of each @xmath190 is increasing much faster than the imaginary part in such a way , so that the argument @xmath195 is always negative and @xmath196 as @xmath197 ( or as @xmath198 ) . now we are ready to define one of the contours that will be used in the proof of the theorem [ t : chiang - yu ] . for any given function @xmath199 of the form ( [ e : g - definition ] ) , the contour is formed by the curves @xmath200 and the line segments @xmath201 which are defined as follows : @xmath202 we join the line segments @xmath201 and the curves @xmath200 to form the contour @xmath203 for each integer @xmath161 . we then _ glue _ the @xmath203 together along each pair of @xmath201 and the resulting set is denoted by @xmath204 . then we define @xmath205 and @xmath206 as follows : @xmath207 thus we have the following result : [ l : g - number - zeros ] let @xmath161 be a large positive integer and @xmath60 . then the function @xmath177 as defined in ( [ e : g - definition ] ) has at least @xmath161 distinct zeros lying inside the contour @xmath205 and infinitely many zeros lie inside the set @xmath208 . it suffices to prove the first statement . we suppose first that @xmath185 so that the contour and the sets given by ( [ e : definition - contour - set ] ) are @xmath203 and @xmath209 respectively , and all the zeros @xmath190 lie in the fourth quadrant of the @xmath105-plane by remark [ r : fourth - quadrant ] . we note that for each @xmath210 , the vertices of the rectangle @xmath211 are given by @xmath212 , @xmath213 and @xmath214 . since we have @xmath215 where @xmath216 , it means _ geometrically _ that the upper ( resp . lower ) edge of @xmath211 is _ below _ _ above _ ) the curve @xmath217 ( resp . @xmath218 ) , see figure [ f : figure1 ] for an illustration . thus the contour @xmath203 contains all the @xmath161 rectangles @xmath211 , for @xmath210 . by the lemma [ l : zeta - bounds ] , each rectangle @xmath211 contains the zero @xmath219 , for @xmath210 . these zeros are _ distinct _ because we have @xmath220 whenever @xmath221 . hence the result follows in this particular case . next we suppose that @xmath222 . then it may happen that _ not _ all zeros @xmath190 lie in the fourth quadrant of the @xmath105-plane . in this general case , we rotate the @xmath105-plane through the angle @xmath223 to the @xmath224-plane , where @xmath225 . thus it follows from the relations ( [ e : zeta - real - imaginary ] ) that @xmath226 so that _ all _ the zeros @xmath227 of @xmath228 lie in the fourth quadrant of the @xmath224-plane by the lemma [ l : zeta - bounds ] and the remark [ r : fourth - quadrant ] . thus the argument in the first part applies to this general case with respect to the contour @xmath205 and the set @xmath208 , thus completing the proof of the proposition . suppose that @xmath1 is subnormal . then the remark [ r : equivalent - conditions ] asserts that we must have @xmath27 and one of the equations in ( [ e : conditions ] ) holds . thus according to ( [ e : lommel - sol ] ) , we have @xmath6 . this proves the sufficiency part of the theorem . in order to complete the proof of the theorem [ t : chiang - yu ] , that is , to find the values of @xmath229 under the assumption that @xmath6 , we consider the function @xmath1 in the form ( [ e : f - solution - hankel ] ) . then we _ first _ need the following result : [ t : exponent - of - convergence ] if @xmath230 or @xmath231 , then @xmath5 . we let @xmath232 in the form ( [ e : general - soln - hankel ] ) be the general solution of ( [ e : generalized - lommel ] ) . since the lemma [ l : lommel - independence ] asserts that the lommel functions @xmath45 , @xmath44 , are linearly independent over @xmath233 and that not all @xmath34 are zero , so the summand in ( [ e : general - soln - hankel ] ) is not identically zero . without loss of generality , we may assume that @xmath234 and the constants @xmath235 , in the theorem [ t : chiang - yu ] satisfy @xmath236 . in order to prove the theorem [ t : exponent - of - convergence ] , we show that the general solution ( [ e : general - soln - hankel ] ) has infinitely many zeros _ in the principal branch of @xmath237 and @xmath16 _ .. ] the idea of our proof is to apply asymptotic expansions of the corresponding special functions and rouch s theorem on suitably chosen contours . when @xmath238 , we substitute the asymptotic expansions ( [ e : hankel1-asy ] ) , ( [ e : hankel2-asy ] ) and ( [ e : lommel - asymptotic - expansion ] ) into the solution ( [ e : general - soln - hankel ] ) to yield and the lommel functions @xmath239 . ] @xmath240+\widehat{d}{{\rm e}}^{-i\zeta}\left[\sum_{k=0}^{p-1}\frac{(\nu,\,k)}{(2i\zeta)^k } + o(\zeta^{-p})\right ] \notag\\ & \qquad+\sum_{j=1}^n \widehat{\sigma}_j \zeta^{\mu_j-\frac{1}{2}}\left[\sum_{k=0}^{p-1}\frac{(-1)^k c_{k,\,j}}{\zeta^{2k } } + o(\zeta^{-2p+1})\right],\notag\end{aligned}\ ] ] where @xmath241 , @xmath242 , @xmath243 , @xmath244 , where @xmath44 . this gives , when @xmath245 , that @xmath246+\widehat{d}{{\rm e}}^{-i\zeta}\big[1+o(\zeta^{-1})\big ] + \sum_{j=1}^n\widehat{\sigma}_j\zeta^{\mu_j-\frac{1}{2}}\big[1+o(\zeta^{-1})\big].\end{aligned}\ ] ] we distinguish two main cases : * case i * : @xmath247 and * case ii * : @xmath248 , which will then be further split into different subcases . * case i * : we suppose that @xmath247 . without loss of generality , we may assume that @xmath249 so that @xmath59 . we choose the function ( [ e : g - definition ] ) to be @xmath250 where @xmath251 and @xmath252 . moreover , we assume that the chosen integer @xmath87 in the lemma [ l : real - imaginary - bounds - subseq ] also satisfies the inequality @xmath253 there are two subcases in * case i*. they are * subcase a * : @xmath185 and * subcase b * : @xmath254 . * * subcase a * : @xmath185 . then the definition shows that @xmath255 which implies that @xmath256 must be a real number such that @xmath257 and @xmath258 . we obtain from ( [ e : modify - soln ] ) and ( [ e : g - reduced ] ) that + @xmath259 \notag\\ & \qquad+o(\zeta^{\mu_n-\frac{3}{2}})+\sum_{j=1}^{n-1 } \widehat{\sigma}_j \zeta^{\mu_j-\frac{1}{2}}\left[1+o(\zeta^{-1})\right]\bigg| \notag\\ & \le \big|\widehat{d}{{\rm e}}^{-i\zeta}+\widehat{c}{{\rm e}}^{i\zeta}o(\zeta^{-1})+\widehat{d}{{\rm e}}^{-i\zeta}o(\zeta^{-1})\big| + o(|\zeta|^\kappa),\end{aligned}\ ] ] + where @xmath260 is defined by + @xmath261 + we show that the inequality + latexmath:[\[\label{e : main - inequality } + holds on the contour @xmath203 for all @xmath161 sufficiently large . in fact , it is always true that @xmath263 , @xmath264 and + @xmath265 and @xmath266 along the contour @xmath203 . if @xmath267 , then we have @xmath268 ( see ( [ e : case(i)-contour ] ) ) , where @xmath269 ; and if @xmath270 , then we have @xmath271 , where @xmath272 . we deduce that + @xmath273 + and for @xmath161 sufficiently large that + @xmath274 + on the one hand , for all sufficiently large @xmath161 , it follows from ( [ e : m - condition ] ) and ( [ e : estimates-2 ] ) that + @xmath275 + but the triangle inequality + @xmath276 ) and ( [ e : estimates-4 ] ) imply for @xmath277 for some positive integer @xmath278 that @xmath279where the lower estimate for the case @xmath267 is trivial when @xmath280 and the case when @xmath281 follows since the factor @xmath282 from ( [ e : estimates-4 ] ) is a constant . on the other hand , we obtain from the relations ( [ e : case-(i)-inequality ] ) , ( [ e : kappa - inequality ] ) , ( [ e : estimates-1 ] ) and ( [ e : estimates-2 ] ) that + @xmath283 + where @xmath284 , @xmath285 and @xmath286 are some fixed positive constants depending only on @xmath287 ( see lemma [ l : real - imaginary - bounds - subseq ] ) and @xmath278 . note that it is easy to check that @xmath288 and @xmath289 hold trivially , we deduce from the inequalities ( [ e : estimates-6 ] ) and ( [ e : estimates-3 ] ) that the inequality ( [ e : main - inequality ] ) holds on @xmath290 and @xmath217 , and then similarly it also holds on @xmath291 and @xmath218 . hence the desired inequality ( [ e : main - inequality ] ) holds on the contour @xmath203 . * * subcase b * : if @xmath254 , then it may happen as described in the proof of the proposition [ l : g - number - zeros ] that _ not _ all zeros @xmath190 lie in the fourth quadrant of the @xmath105-plane , where the integers @xmath162 are also defined in the lemma [ l : real - imaginary - bounds - subseq ] . however , one can rotate the @xmath105-plane through the angle ( @xmath292 ) as described in the proposition [ l : g - number - zeros ] ( see also its proof ) so that all such zeros can only lie in the fourth quadrant of the @xmath224-plane . + in this circumstance , we note that the inequalities ( [ e : case-(i)-inequality ] ) and ( [ e : triangle - inequality ] ) are now replaced by + @xmath293 + and + @xmath294 + respectively , where @xmath295 , @xmath228 and the constant @xmath260 is given by ( [ e : kappa - definition ] ) . moreover , the relations ( [ e : estimates-1 ] ) and the inequalities ( [ e : estimates-2 ] ) are replaced by + @xmath296 + and + @xmath297 + respectively , where @xmath298 , @xmath269 , @xmath272 , where @xmath290 and @xmath217 are defined in ( [ e : case(i)-contour ] ) . + now we further distinguish two cases between * ( 1 ) * @xmath299 and * ( 2 ) * @xmath300 . + * ( 1 ) * if @xmath299 , then @xmath301 and it follows from the inequalities ( [ e : m - condition ] ) and ( [ e : estimates-5 ] ) that the inequalities ( [ e : estimates-4 ] ) are replaced by + @xmath302 + for @xmath161 sufficiently large . thus the inequality ( [ e : triangle - inequality - modify ] ) together with ( [ e : estimates-7 ] ) yield , for @xmath303 for some sufficiently large positive integer @xmath304 , that + @xmath305 + * ( 2 ) * if @xmath300 , then we have @xmath306 and the inequalities ( [ e : estimates-4 ] ) are now replaced by + @xmath307 + for @xmath161 sufficiently large . therefore we deduce from these that + @xmath308 + on the one hand , these limits show that the inequality ( [ e : triangle - inequality - modify ] ) imply for @xmath309 for some sufficiently large positive integer @xmath310 that the inequalities ( [ e : estimates-8 ] ) hold in this case . on the other hand , it follows from the relations ( [ e : kappa - inequality ] ) , ( [ e : case-(i)-inequality - modify ] ) , ( [ e : estimates-9 ] ) and ( [ e : estimates-5 ] ) that the inequalities + @xmath311 + hold in this * subcase b * , where @xmath312 are some positive constants , @xmath313 and @xmath314 so that ( [ e : estimates-8 ] ) and ( [ e : estimates-10 ] ) imply the inequality + @xmath315 + holds on the contour @xmath205 for all @xmath161 sufficiently large . hence , our desired inequality ( [ e : main - inequality ] ) still holds in this general case after we transform the @xmath224-plane back to the @xmath105-plane . * case ii * : next , we suppose that @xmath248 . unfortunately , the contour @xmath203 and the auxiliary function defined in ( [ e : case(i)-contour ] ) and ( [ e : g - reduced ] ) respectively , do not seem to apply in this case . this is because the zeros of ( [ e : g - reduced ] ) distribute evenly on a straight line parallel to the real axis , and hence the region enclosed by the contour @xmath203 can only contain _ finitely many _ such zeros for every positive integer @xmath161 . we choose the alternative auxiliary function to be : @xmath316without loss of generality , we continue to assume that @xmath59 . it remains to construct a suitable contour that contains the zeros of ( [ e : g - hat - definition ] ) which are given by the following lemma . [ l : contour - special ] let @xmath161 be an integer . 1 . if @xmath317 , then the zeros of ( [ e : g - hat - definition ] ) are given by @xmath318 where @xmath319 are solutions of the quadratic equation @xmath320 , which are given by @xmath321 and @xmath322 are the principal arguments of @xmath319 . 2 . if @xmath323 , then the zeros of ( [ e : g - hat - definition ] ) are given by @xmath324 where @xmath325 and @xmath326 is the principal argument of @xmath327 . we omit its proof . [ r : contour - special - zeros ] we remark that none of the @xmath328 , @xmath329 or @xmath330 can be zero . otherwise , @xmath331 or @xmath332 would be zero which contradicts the assumption . moreover , let @xmath333 and @xmath334 be two horizontal straight lines on which the zeros of the equation ( [ e : zeros - g - hat - nonzero ] ) fall on , such that @xmath333 corresponds to the zeros of @xmath335 and @xmath334 corresponds to the zeros of @xmath336 in lemma [ l : contour - special](a ) . similarly , we let @xmath337 denote the straight line representing the zeros of the equation ( [ e : zeros - g - hat - zero ] ) in lemma [ l : contour - special](b ) . both @xmath338 and @xmath334 are parallel to the real axis in the @xmath105-plane . the construction of the contour is divided into different cases depending on whether the @xmath339 vanishes . they are , * subcase a * : @xmath317 and * subcase b * : @xmath323 . the * subcase a * is further divided into * ( 1 ) * @xmath340 and * ( 2 ) * @xmath341 . the * ( 2 ) * is divided into ( i ) @xmath342 and ( ii ) @xmath343 . now we can start the construction of the contour . * * subcase a * : suppose that @xmath317 . by the remark [ r : contour - special - zeros ] , we define a constant @xmath344 as follows : + @xmath345 + it is easy to see from ( [ e : d - definition ] ) that we must have @xmath346 . we distinguish two cases between * ( 1 ) * @xmath340 and * ( 2 ) * @xmath347 . + * ( 1 ) * if @xmath340 , then we define , for each integer @xmath277 for some suitably large positive integer @xmath278 , the line segments @xmath348 and @xmath349 as follows : + @xmath350 + then the line segments @xmath351 are concatenated to form the rectangular contour @xmath352 . we also form the set @xmath353 , see figure 2 . + + instead of inequality ( [ e : main - inequality ] ) , we shall show that the inequality + @xmath354 + holds on @xmath352 for all @xmath161 sufficiently large . + on the one hand , we note from the lemma [ l : contour - special](a ) , the remark [ r : contour - special - zeros ] and the definition ( [ e : case(ii)-contour ] ) that for every integer @xmath161 , exactly @xmath355 distinct zeros of @xmath356 lie _ inside _ the @xmath352 but all @xmath336 lie _ outside _ the @xmath352 , see figure [ f : figure2 ] for an illustration . therefore we must have the fact that @xmath356 _ does not _ pass through any zero along the @xmath352 for every positive integer @xmath161 . in other words , there exists a positive constant @xmath357 , depending only on @xmath161 , such that the inequality + @xmath358 + holds on @xmath352 for every positive integer @xmath161 . to obtain the desired inequality ( [ e : main - inequality - hat ] ) , we must show that the constant @xmath357 can be chosen _ independent _ of @xmath161 . to see this , we note that for each positive integer @xmath161 , we have + @xmath359 + and so + @xmath360 + hence this implies that @xmath357 can be chosen _ independent _ of @xmath161 . we denote this positive number to be @xmath361 , thus we have the inequality + @xmath362 + holds on @xmath352 for every positive integer @xmath161 . + on the other hand , it follows from ( [ e : modify - soln ] ) and the chosen function ( [ e : g - hat - definition ] ) that + @xmath363\big| \notag \\ & \le \frac{\delta_2|{{\rm e}}^{i\zeta}|}{|\zeta|}+\frac{\delta_3|{{\rm e}}^{-i\zeta}|}{|\zeta|}+o(|\zeta|^{-1 } ) + \sum_{j=1}^{n-1 } o(|\zeta|^{{{\rm re}\,}(\mu_j)-\frac{1}{2 } } ) , \end{aligned}\ ] ] + where @xmath364 and @xmath365 are some fixed positive constants . since @xmath248 , the definition ( [ e : kappa - definition ] ) implies that @xmath260 is negative and @xmath366 , @xmath367 for all @xmath368 . thus the relations ( [ e : relation-3 ] ) and ( [ e : relation-6 ] ) imply that + @xmath369 + holds on the contour @xmath352 , where @xmath370 and @xmath371 are two fixed positive constants independent of @xmath161 . hence we obtain from ( [ e : relation-5 ] ) and ( [ e : relation-7 ] ) that the desired inequality ( [ e : main - inequality - hat ] ) holds on @xmath352 for all sufficiently large @xmath161 . + * ( 2 ) * if @xmath347 , then the definition ( [ e : d - definition ] ) gives @xmath372 . in addition , all the @xmath373 and @xmath374 lie on the same straight line @xmath375 ( see remark [ r : contour - special - zeros ] ) so that for every integer @xmath161 , we have @xmath376 . thus , there are two possibilities : ( i ) @xmath342 and ( ii ) @xmath343 . * * if @xmath342 , then @xmath377 for every integer @xmath161 . hence the above contour @xmath352 and the argument leading to the inequality ( [ e : main - inequality - hat ] ) can be applied without any change . * * if @xmath343 , then it may happen that @xmath378 or @xmath379 passes through the zeros @xmath380 or @xmath381 , so we need to modify the contour @xmath352 defined in ( [ e : case(ii)-contour ] ) . in fact , we can replace @xmath382 and @xmath383 by @xmath384 and @xmath385 respectively in the definitions ( [ e : case(ii)-contour ] ) . we then denote the modified line segments by @xmath386 and @xmath387 respectively : + @xmath388 + then the contour and the infinite strip are defined similarly and denoted by @xmath389 and @xmath390 respectively . see figure 3 below . + + since @xmath372 in ( [ e : d - definition ] ) , thus the relations ( [ e : relation-3 ] ) and ( [ e : relation-4 ] ) are replaced by + @xmath391 + and + @xmath392 + respectively . thus , the inequalities ( [ e : relation-5 ] ) , ( [ e : relation-6 ] ) and ( [ e : relation-7 ] ) can be similarly deduced with a possibly different set of the positive constants @xmath393 . + it is trivial to check that there are totally @xmath394 distinct zeros inside the modified contour @xmath389 for every positive integer @xmath161 . * * subcase b * : suppose that @xmath323 . then it is easy to see that this can be regarded as the degenerated case in * subcase a(1)(i ) * with the constant @xmath395 and the straight line @xmath333 replaced by @xmath396 and @xmath337 respectively . we may now continue the proof of the theorem [ t : exponent - of - convergence ] . so rouch s theorem implies that , the functions @xmath397 and @xmath177 ( resp . @xmath356 ) have the same number of zeros inside @xmath203 ( resp . @xmath352 or @xmath389 ) . the proposition [ l : g - number - zeros ] ( resp . lemma [ l : contour - special ] ) asserts that @xmath177 ( resp . @xmath356 ) , and hence @xmath397 , has _ infinitely many distinct zeros _ inside @xmath209 ( resp . @xmath398 or @xmath390 ) . let @xmath399 denote the number of zeros of the function @xmath1 inside the set @xmath339 . then given any @xmath400 , there exists an infinite sequence @xmath401 of zeros of @xmath397 , and hence of @xmath232 , with @xmath402 inside @xmath209 ( resp . @xmath398 or @xmath390 ) such that @xmath403 for all sufficiently large @xmath31 . by the substitution @xmath404 , where @xmath405 and @xmath406 . then for choosing @xmath407 and @xmath408 such that @xmath409 as @xmath410 and @xmath411 for all positive integers @xmath31 , where @xmath412 is the principal argument of @xmath413 , we must have @xmath414 as @xmath410 and then @xmath415 as @xmath410 which implies that @xmath416 , thus completing the proof of the theorem . we can continue the proof of the theorem [ t : chiang - yu ] now . recall that @xmath417 , where @xmath232 is a solution to the equation ( [ e : generalized - lommel ] ) . so the requirement @xmath6 is _ independent of the branches _ of the function @xmath232 . it follows from the theorem [ t : exponent - of - convergence ] that we must have @xmath78 and hence so are @xmath27 . hence the solution ( [ e : f - solution ] ) is expressed in the form @xmath418 to complete the proof of the theorem [ t : chiang - yu ] , we need to prove that when @xmath34 is non - zero , then @xmath51 and @xmath52 must satisfy either @xmath419 where @xmath420 . following a similar idea as in @xcite , we have from the remark [ r : bessel - lommel - independence - branches ] that @xmath421 are entire functions in the @xmath55-plane and that each @xmath422 ( @xmath44 ) is _ independent of the branches _ of @xmath45 . we choose for a @xmath423 such that @xmath424 . so we can rewrite the solution ( [ e : lommel - sum ] ) as @xmath425 where the function @xmath45 belongs to the branch @xmath426 and the other lommel functions @xmath427 are in the principal branch @xmath428 and @xmath87 is an arbitrary but otherwise fixed non - zero integer . [ r:4.1]we note again that in the following discussion that we only consider the case @xmath429 . the other case @xmath430 can be dealt with similarly by applying the property that each @xmath45 is an even function of @xmath52 . suppose that @xmath429 for some non - negative integer @xmath431 . if @xmath432 or @xmath100 , then it follows from the lemma [ l : lommel - continuation - general-2](a ) and ( b ) that the solution ( [ e : lommel - sum - rewrite ] ) can be expressed in the form ( [ e : f - solution - hankel ] ) with @xmath433 and @xmath434 where @xmath435 and @xmath436 are the constants defined in ( [ e : ks - definition ] ) . in order to apply the theorem [ t : exponent - of - convergence ] , we may follow closely the argument used in ( * ? ? ? * proposition 4.4 ( i ) and ( ii ) ) , where if we have @xmath249 or @xmath437 for any integer @xmath87 , then we will obtain a contradiction to the free choice of the integer @xmath87 . hence @xmath5 and then either @xmath438 or @xmath339 must be zero . this implies that ( [ e : desired - equations ] ) holds , as required . if @xmath107 for a positive integer @xmath31 , then it follows from lemma [ l : lommel - continuation - general-2](c ) that the solution ( [ e : lommel - sum - rewrite ] ) ( with @xmath87 replaced by @xmath439 ) is given by @xmath440\bigg . \notag\\ & \qquad -(l{{\rm e}}^{mz})c_n(l{{\rm e}}^{mz } ) \big[k_+''h_1^{(1)}(l{{\rm e}}^{mz})+k_-''h_1^{(2)}(l{{\rm e}}^{mz})\big]\bigg\ } \notag\\ & \quad+\sum_{j=1}^n \sigma_j { { \rm e}}^{-nz}s_{\mu_j,\,\nu}(l{{\rm e}}^{mz } ) . \notag\end{aligned}\ ] ] it is obvious that the above expression is _ not _ in the form ( [ e : f - solution - hankel ] ) so that the theorem [ t : exponent - of - convergence ] does not apply in this case . in order to find an alternative approach to show @xmath5 , we show that the function @xmath441 defined by : @xmath442 \\ & \qquad-\zeta c_n(\zeta)\big[k_+''h_1^{(1)}(\zeta)+k_-''h_1^{(2)}(\zeta)\big]\big\}+\sum_{j=1}^n\sigma_j s_{\mu_j,\,\nu}(\zeta ) \notag\end{aligned}\ ] ] has _ infinitely many _ zeros in the principal branch of @xmath443 and @xmath45 . therefore we suppose that @xmath238 . then the asymptotic expansions ( [ e : hankel1-asy ] ) , ( [ e : hankel2-asy ] ) and ( [ e : lommel - asymptotic - expansion ] ) with putting @xmath245 yield @xmath444+\sum_{j=1}^n\widehat{\sigma}_j \zeta^{\mu_j-\frac{1}{2}}\big[1+o(\zeta^{-1})\big ] , \notag\end{aligned}\ ] ] where @xmath445 . to find the number of zeros of @xmath441 in @xmath238 , we need the following result : [ l : dn - degree ] suppose that @xmath31 is a positive integer . then at least one of @xmath446 or @xmath447 has degree @xmath31 . we let @xmath448 where @xmath449 are complex constants . we prove the lemma by induction on @xmath31 . when @xmath49 , we have @xmath450 , so the statement is true . assume that it is also true when @xmath451 for a positive integer @xmath161 . without loss of generality , we may assume that @xmath452 so that @xmath453 when @xmath454 , it follows from the recurrence relations ( [ e : recurrence - a_n - b_n - c_n ] ) for @xmath455 and @xmath110 that @xmath456 it is easy to check that the coefficients of the @xmath457 in @xmath458 are given by @xmath459 respectively . if @xmath460 , then we have @xmath461 so that both @xmath462 and @xmath463 are zero which certainly contradict to our inductive assumption ( [ e : inductive - assumption ] ) . hence we must have @xmath464 or @xmath465 , completing the proof of the lemma . we can complete the proof of the theorem now . we recall that we have assumed @xmath429 for some non - negative integer @xmath431 and @xmath107 for a positive integer @xmath31 , see the paragraphs following the remark [ r:4.1 ] . by the lemma [ l : dn - degree ] , we may suppose that @xmath466 and @xmath467 , where @xmath468 . then the expression ( [ e : h - hat ] ) induces @xmath469 + \widehat{c}_-k_-''{{\rm e}}^{-i\zeta}\big[1+o(\zeta^{-1})\big ] + \sum_{j=1}^n\widehat{\sigma}_j \zeta^{\mu_j-\frac{1}{2}}\big[1+o(\zeta^{-1})\big]\end{aligned}\ ] ] which is in the form ( [ e : modify - soln ] ) with @xmath332 and @xmath331 replaced by @xmath470 and @xmath471 respectively . therefore the proof of the theorem [ t : exponent - of - convergence ] can be applied without change to show that if at least one of @xmath472 or @xmath473 , then the function @xmath441 has infinitely many zeros in @xmath238 and thus @xmath5 , a contradiction . hence we conclude that @xmath474 can not be an odd negative integer . now we can apply the analytic continuation formula in the lemma [ l : lommel - continuation - general-1 ] with this fixed integer @xmath87 to get @xmath475 where @xmath476 is the polynomial as defined in the lemma [ l : lommel - continuation - general-1 ] . then the expressions ( [ e : lommel - sum - rewrite ] ) and ( [ e : lommel - continuation - generalpi ] ) give @xmath477 if either of the coefficients of @xmath478 and @xmath479 in the ( [ e : lommel - sum - modify - further ] ) is non - zero , then the theorem [ t : exponent - of - convergence ] again implies that @xmath5 which is impossible . thus we must have @xmath480 now we are ready to derive the equations ( [ e : desired - equations ] ) , we again recall that the value of @xmath481 must be independent of branches of the function @xmath482 which is equivalent to equations ( [ e : subnormal - zero - anypi ] ) hold for _ each _ integer @xmath87 . it is clear from the lemma [ l : lommel - continuation - general-1](b ) that @xmath483 and @xmath484 do not hold simultaneously for any integer @xmath87 . thus @xmath485 must hold , _ i.e. _ , when @xmath486 , @xmath487 where @xmath420 . hence we deduce the first and the second conditions in ( [ e : conditions ] ) from the first and the second equations in ( [ e : desired - equations ] ) respectively . a detailed deduction can be found in @xcite . this completes the proof of the theorem . if @xmath49 , then the assumption gives @xmath488 so that @xmath489 . if @xmath112 , then it follows from the lemma [ l : lommel - independence ] that @xmath490 . thus the function @xmath491 as defined in ( [ e : f - definition ] ) is non - trivial so that we may suppose that the function ( [ e : f - definition ] ) has finitely many zeros in every branch of @xmath105 . then the entire function @xmath492 is certainly a solution of the equation ( [ e : chiang - yu ] ) with @xmath493 and @xmath6 . hence the theorem [ t : chiang - yu ] implies that either @xmath47 or @xmath48 for non - negative integers @xmath40 , where @xmath44 . conversely , if either @xmath47 or @xmath48 for non - negative integers @xmath40 , where @xmath44 , then the remark [ r : lommel - terminate ] shows that each @xmath494 is a polynomial in @xmath495 so that @xmath45 has only finitely many zeros in every branch of @xmath105 , where @xmath44 . thus this implies that the function ( [ e : f - definition ] ) has finitely many zeros in every branch of @xmath105 . this completes the proof of the corollary [ t : chiang - yu - lommel - zeros ] . the explicit representation and the zeros distribution of an entire solution @xmath1 of either the equation @xmath496 or the equation @xmath497 were studied by bank , laine and langley @xcite , @xcite ( see also @xcite ) . later , ismail and one of the authors strengthened @xcite ( announced in @xcite ) their results . in fact , they discovered that the solutions of ( [ e : chiang - ismail-1 ] ) and ( [ e : chiang - ismail-2 ] ) can be solved in terms of bessel functions and coulomb wave functions respectively . besides , they identified that two classes of classical orthogonal polynomials ( bessel and generalized bessel polynomials respectively ) appeared in the explicit representation of solutions under the _ boundary condition _ that the exponent of convergence of the zeros of the solution @xmath1 is finite , _ i.e. _ , @xmath498 . this also results in a complete determination of the eigenvalues and eigenfunctions of the equations . we call such pheonomenon a _ function - theoretic quantization result _ for the differential equations ( [ e : chiang - ismail-1 ] ) and ( [ e : chiang - ismail-2 ] ) . it is also well - known that both equations have important physical applications . for examples , the eqn . ( [ e : chiang - ismail-1 ] ) is derived as a reduction of a non - linear schrdinger equation in a recent study of benjamin - feir instability phenomenon in deep water in @xcite , while the second eqn . ( [ e : chiang - ismail-2 ] ) is an exceptional case of a standard classical diatomic model in quantum mechanics introduced by p. m. morse in 1929 @xcite see @xcite for a historical background of the morse potential . and is a basic model in the recent @xmath499symmetric quantum mechanics research @xcite ( see also @xcite ) . in ( * ? ? ? * theorem 6.1 ) , the authors considered the following differential equation @xmath500 which is a special case of the equation ( [ e : chiang - yu ] ) when @xmath501 and @xmath49 in the theorem [ t : chiang - yu ] , where @xmath502 . they obtained the necessary and sufficient condition on @xmath503 so that the equation ( [ e : reduced - chiang - yu ] ) admits subnormal solutions which are related to classical polynomials and / or functions , _ i.e. _ , neumann s polynomials , gegenbauer s generalization of neumann s polynomials , schlfli s polynomials and struve s functions . this exhibits a kind of function - theoretic quantization phenomenon for non - homogeneous equations . now the following result holds trivially by our main theorem [ t : chiang - yu ] : [ t : quantizations ] with each choice of parameters as indicated in table [ mytable ] below , we have a necessary and sufficient condition on @xmath503 that depends on the non - negative integer @xmath23 so that the equation ( [ e : reduced - chiang - yu ] ) admits a solution with finite exponent of convergence of zeros . furthermore , the forms of such solutions are given explicitly in table [ mytable ] : .special cases of ( [ e : reduced - chiang - yu ] ) . [ cols="^,^,^,^ " , ] here @xmath504 and @xmath505 are the neumann polynomials of degrees @xmath506 and @xmath507 respectively ; @xmath508 is the schlfli polynomial and @xmath509 is the struve function , see @xcite . [ [ a : asymptotic - expansion ] ] an asymptotic expansion of @xmath12 and linear independence of lommel s functions ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ for large @xmath522 and @xmath523 , where @xmath23 is a positive integer . see also @xcite . as a result , we see that the asymptotic expansions ( [ e : hankel1-asy ] ) , ( [ e : hankel2-asy ] ) and ( [ e : lommel - asymptotic - expansion ] ) are valid _ simultaneously _ in the range @xmath238 . [ r : lommel - terminate]it is clear that ( [ e : lommel - asymptotic - expansion ] ) is a series in descending powers of @xmath105 starting from the term @xmath524 and ( [ e : lommel - asymptotic - expansion ] ) terminates if one of the numbers @xmath520 is an odd positive integer . in particular , if @xmath22 for some non - negative integer @xmath23 , then we have @xmath485 in the analytic continuation formula ( [ e : lommel - continuation - general ] ) and thus , in this degenerate case , the formula ( [ e : lommel - continuation - general ] ) becomes @xmath525 for every integer @xmath87 and @xmath523 . [ l : lommel - independence ] ( * ? ? ? * lemma 3.12 ) suppose @xmath112 , and @xmath51 and @xmath52 be complex numbers such that @xmath38 are all distinct for @xmath526 . then the lommel functions @xmath527 , @xmath528 are linearly independent . j. c. slater , morse s contribution to atmoic , molecular , and solid state physics " in _ in honors of philip m. morse _ , edited by h. feshbach and k. ingard , the m.i.t . press , camb . , mass . , and london , england , 1969 .

strong - coupling series expansions are calculated for the hamiltonian version of compact lattice electrodynamics in ( 2 + 1 ) dimensions , with 4-component fermions . series are calculated for the ground - state energy per site , the chiral condensate , and the masses of ` glueball ' and positronium states . comparisons are made with results obtained by other techniques .

hep-lat9710008

quantum electrodynamics in 2 + 1 dimensions ( qed@xmath0 ) has generated considerable interest over recent years . the model is super - renormalizable , but shares a number of important features with quantum chromodynamics ( qcd ) in 3 + 1 dimensions : it is believed to be confining at large distances ( in the quenched approximation , at least ) , while in the massless fermion limit it displays a chiral - like symmetry which is spontaneously broken@xcite . it is thus an ideal laboratory for testing non - perturbative methods of analysis . versions of the model may also be relevant to theories of the new high - t@xmath1 superconductors@xcite . the version with two - component massless fermions generates a dynamical mass for the photon through a chern - simons term@xcite . this complication can be avoided in the four - component version@xcite , where `` chiral '' symmetry is broken in the normal goldstone fashion , leading to a doublet of massless goldstone bosons analogous to the pion in qcd . for the four - component model with @xmath2 flavours of massless fermions , there has been a debate running for some time whether chiral symmetry is broken for all values @xmath2@xcite or whether there is a critical value @xmath3 above which no spontaneous symmetry - breaking takes place@xcite . we shall have nothing to say about this question . the four - component version has been studied by several different techniques , but we remain far from a complete understanding of the model . euclidean lattice monte carlo simulations have been performed by several groups@xcite . a number of authors@xcite have used schwinger - dyson techniques to study the chiral symmetry breaking , and allen and burden@xcite have also produced estimates of the bound - state meson spectrum at finite fermion masses . the hamiltonian lattice version has been studied by means of strong - coupling expansions@xcite and a loop expansion technique@xcite . a light - front approach has also been discussed@xcite ; and the non - relativistic limit has been analyzed in some detail@xcite . here we treat the hamiltonian lattice model by using linked - cluster techniques@xcite to generate further strong - coupling series , thus extending the previous results of burden and hamer@xcite ( hereafter referred to as i ) . it is well - known that euclidean monte carlo techniques are difficult and expensive to apply to models with dynamical fermions , and so it seems worthwhile to see if other techniques such as strong - coupling expansions can give useful information in such cases . a previous analysis of the schwinger model@xcite did indeed show that strong - coupling series approximants can converge well into the weak coupling region . while not quite as accurate as the exact finite - lattice technique , the series approach did give quantitative estimates of the lowest bound - state mass in the continuum limit , at about the 5 - 10% level of accuracy . the paper begins with an outline of the lattice formulation of the model in section ii , followed by a brief summary of the methods of calculation in section iii . our results are presented in section iv , discussing the ground - state energy , the chiral condensate , the `` glueball '' masses , and the spectrum of the bound - state mesons as a function of the bare fermion mass . our conclusions are summarized in section v. the continuum lagrangian density takes the standard form @xmath4 where @xmath5 and the lorentz indices @xmath6 , 1 or 2 . the electric coupling @xmath7 in ( 2 + 1 ) dimensions has the dimensions of @xmath8 . choosing the timelike axial gauge @xmath9 the hamiltonian is found to be @xmath10 where @xmath11 and @xmath12 note that the magnetic field @xmath13 has only one component in ( 2 + 1)d . here @xmath14 is taken as a single four - component dirac spinor@xcite , and @xmath15 , @xmath16 , @xmath17 are @xmath18 dirac matrices , for which we shall use the dirac representation where necessary : @xmath19 in the zero - mass limit , the hamiltonian ( [ eq4 ] ) possesses a global @xmath20 `` chiral '' symmetry@xcite , whose lie algebra is spanned by the matrices @xmath21 and @xmath22 . this symmetry is expected to be spontaneously broken@xcite , which should be manifested by a non - zero value of the chiral condensate @xmath23 . at large fermion masses , a non - relativistic analysis can be carried out@xcite . cornwall and tiktopoulos and sen@xcite showed that if the divergences were regulated by giving a mass @xmath24 to the photon , then at one - loop order the renormalized self - mass of the fermion is @xmath25 while the potential due to one - photon exchange between the electron and positron is : @xmath26 where @xmath27 is euler s constant . both quantities show logarithmic divergences , but these divergences cancel in the schrdinger equation for the positronium bound states @xmath28 \psi(\vec{r } ) = [ - { 1\over m } \vec{\nabla}_r^2 + { e^2\over 2 \pi } ( \gamma + \ln mr ) ] \psi ( \vec{r})\ ] ] numerical solutions of this equation@xcite give the ` binding energies ' of the lowest positronium states as @xmath29 for angular momentum @xmath30 , and @xmath31 for @xmath32 . at leading order the binding energies are independent of `` spin '' , so that each of these energy levels should be four - fold degenerate in the four - component fermion model . a ` staggered ' hamiltonian lattice formulation of this model has been discussed in reference i. the four components of the continuum fermion field fit naturally onto the four sites of a @xmath33 unit cell on the 2-dimensional spatial lattice , leading to a lattice hamiltonian as follows : @xmath34 where @xmath35 and @xmath36 \\ w_2 & & = - \sum_p ( u_p + u_p^{\dag } ) \end{aligned}\ ] ] here @xmath37 labels the sites , @xmath38 the links , @xmath39 the plaquettes and @xmath40 the directions on a square two - dimensional spatial lattice with spacing @xmath41 . the dimensionless coupling @xmath42 and mass parameter @xmath43 are defined in terms of their continuum counterparts @xmath7 and @xmath44 by @xmath45 while @xmath46 , and @xmath47 , @xmath48 . the term @xmath49 is the electric field term , @xmath50 is the fermion mass term , @xmath51 is the fermion kinetic energy , and @xmath52 is the magnetic field energy , involving the usual plaquette operator @xmath53 . the correspondences between the lattice fields and their continuum counterparts are for gauge fields @xmath54 where the link operator @xmath55\ ] ] while for the fermion field components @xmath56 \left [ \begin{array}{c } \xi_1 \\ \xi_2 \\ \xi_3 \\ \xi_4 \end{array}\right ] \to \left [ \begin{array}{c } \psi_1 \\ \psi_2 \\ \psi_3 \\ \psi_4 \end{array}\right ] \label{eq19}\ ] ] where@xcite @xmath57 and the components @xmath58 are assigned to sites of the @xmath33 unit cell as shown in figure 1 . the commutation relations between the lattice fields are @xmath59 = u_l \delta_{ll^{\prime}}\ ] ] @xmath60 = - u_l^{\dag } \delta_{ll^{\prime}}\ ] ] @xmath61 @xmath62 = [ e_l , \chi^{\dag } ( \vec{r } ) ] = [ u_l , \chi ( \vec{r})]= [ u_l , \chi^{\dag } ( \vec{r } ) ] = 0\ ] ] with these correspondences , it can be shown@xcite that the lattice hamiltonian ( [ eq11 ] ) reduces to the continuum hamiltonian ( [ eq4 ] ) in the naive continuum limit @xmath63 . the introduction of the lattice breaks the u(2 ) `` chiral '' symmetry down to a discrete symmetry generated by shifts of one lattice spacing@xcite . a unit shift in either the @xmath64 or @xmath65 direction leaves the kinetic term in the hamiltonian ( [ eq11 ] ) invariant , but alters the sign of the mass term . the corresponding continuum field transformations , from ( 19 ) , are @xmath66 or @xmath67 [ eq22 ] respectively . the hamiltonian ( [ eq11 ] ) acts an a fock space spanned by the usual strong - coupling basis@xcite . with each link is associated an integer electric flux @xmath68 such that @xmath69 . the operators @xmath70 and @xmath71 increase and decrease the flux on link @xmath38 by one unit , respectively . each site of the lattice can be in one of two fermionic states @xmath72 or @xmath73 obeying @xmath74 consider first the massless theory , @xmath75 . in the strong - coupling limit , the variable @xmath76 and the hamiltonian @xmath77 reduces to @xmath49 . the ground - state is then highly degenerate , having flux @xmath78 on each link , but with the fermionic state entirely arbitrary . this degeneracy is broken at the next order by the kinetic term @xmath51 , leaving only two degenerate states @xmath79 and @xmath80 whose fermionic content is @xmath81 and @xmath82 the chiral shifts of equation ( [ eq22 ] ) map these two states into each other . when the mass term @xmath50 is included , chiral symmetry is explicitly broken and state @xmath83 is favoured energetically . we thus take @xmath80 as the unperturbed strong - coupling ground state for both the massive and massless cases , and interpret this as the state with no fermion excitations present . an excitation on an odd or even site creates a positively or negatively charged fermion respectively , i.e. a positron or electron . the first - order perturbation @xmath51 creates or destroys an electron - positron pair on neighbouring sites , joined by a link of flux . the second - order perturbation term @xmath52 creates or destroys a plaquette of flux . gauge invariance ensures that for any state obtained from the unperturbed vacuum by application of the operators @xmath51 and @xmath52 , the net flux from any site is equal to the charge of the fermion at that site , i.e. , gauss law is obeyed . this theory is expected to display confinement@xcite , and the only fermionic states with finite energy are expected to be electrically neutral `` positronium '' bound states . in the strong - coupling limit , the lowest energy positronium states consist of an electron - positron pair on neighbouring sites , connected by a link of unit flux . there are eight translationally - invariant states of this type , corresponding to the eight links in the unit cell , and we need to identify the linear combinations of these states which correspond to eigenstates of the lattice symmetry operators . the corresponding procedure for meson states in four - dimensional euclidean lattice qcd has been discussed by golterman@xcite . the symmetry group of the lattice hamiltonian ( [ eq11 ] ) is composed of the following elements : @xmath84 this corresponds to spatial translational invariance in the continuum model . @xmath85 or @xmath86 these are the discrete lattice versions of `` chiral '' symmetry corresponding to equations ( [ eq22 ] ) . the massless hamiltonian is symmetric under these transformations , but not the massive one . a combination of a shift by one site in the @xmath64 direction and one site in the @xmath65 direction gives a diagonal shift @xmath88 this corresponds to a discrete @xmath89 rotation in the continuum fields @xmath90 this remains a symmetry of the massive hamiltonian also . let @xmath91 denote a lattice rotation by @xmath92 about a perpendicular axis , as shown by fig . 2 : @xmath93 where @xmath94 and @xmath95\ ] ] repeated rotations generate the rotational symmetry group of a square , with 4 elements . it corresponds to rotation in both space and `` spin '' in the continuum model . the `` axial parity '' inversion is discussed by burden and allen@xcite . in the continuum , it corresponds to the operations : @xmath97 and the vector field transforms as : @xmath98 where @xmath99 . a suitable representation for the fermion operator @xmath96 is the matrix @xmath100 . on the lattice , this is simply @xmath101 which is equivalent to @xmath102 , a rotation by @xmath103 in 2 + 1 dimensions . a reflection in the @xmath65 axis corresponds to @xmath104 where @xmath105 the charge conjugation operation is also discussed by burden and allen@xcite . in the continuum , it corresponds to the operators : @xmath107 and @xmath108 the fermion operator can be represented , up to an arbitrary phase , by the matrix @xmath17 . the translation to lattice variables is slightly involved , and is best handled in terms of the fields @xmath109 defined in equations ( [ eq19 ] ) , ( [ eq20 ] ) . we shall not go into further details here . the translationally invariant positronium eigenstates can be classified in terms of their eigenvalues under these symmetry operations . the group of square rotations is @xmath110 , the cyclic group of 4 elements . it has 4 irreducible representations , each with dimension 1 . the allowed eigenvalues of @xmath91 simply consist of the 4th . roots of unity ( i.e. powers of @xmath111 ) . the shift @xmath87 , reflection @xmath103 and axial parity @xmath96 each generate 2-element groups , with eigenvalues @xmath112 for the positronium states with translational symmetry . in the strong - coupling limit , the `` link excitations '' on the unit cell corresponding to the lowest energy positronium states consist of the eight states listed in table 1 , numbered according to figure 3 . these states transform into each other under the action of the symmetry operators . by taking linear combinations of these states , one can form eigenstates @xmath113 of the lattice symmetry operators , as listed in table 2 . note that the states with rotation eigenvalue @xmath114 or @xmath115 can not simultaneously be eigenstates of @xmath103 , because a reflection converts @xmath116 to @xmath117 , and vice versa . we have chosen to list states @xmath118 to @xmath119 which correspond to eigenstates of @xmath103 , and are thus symmetric or antisymmetric combinations of the states with @xmath120 . similarly , we have chosen to list states @xmath121 and @xmath122 which are eigenstates of @xmath106 , rather than @xmath91 . all eight states are degenerate in energy in the strong - coupling limit @xmath123 . the fact that the hamiltonian is symmetric under both rotations and reflections implies that the pair @xmath118 and @xmath124 will remain degenerate at all couplings , and likewise the pair @xmath125 and @xmath119 . the combination of rotation and charge conjugation symmetry implies that the pair @xmath126 and @xmath127 will also remain degenerate . in the ` naive continuum limit ' @xmath63 , when @xmath128 , one finds that the quartet of states @xmath129 to @xmath130 transcribe to simple combinations of quark and antiquark fields on the lattice , and correspond to ` vector ' states in the language of burden and allen@xcite , with @xmath131 , where @xmath132 is the ` total angular momentum ' . the states @xmath133 to @xmath134 , on the other hand , contain an admixture of gauge fields in the naive continuum limit , and have no direct counterparts in the catalogue of quark - antiquark states discussed by burden and allen . the state @xmath135 is a scalar state , having the same quantum numbers as the vacuum . some useful information on the ground - state properties , at least , can be gained by performing a `` weak - coupling '' expansion for the lattice system as @xmath136 . for the present model , this exercise was carried out in reference i. the ground - state energy per site has an asymptotic expansion in the weak - coupling limit : @xmath137^{1/2 } + o(1)\quad { \rm as } \quad y\to \infty \label{eq37}\ ] ] where the integral arises from a sum over the fermionic degrees of freedom . at very large @xmath65 one finds @xmath138 but for finite @xmath65 and @xmath43 it is more useful to evaluate the integral as it stands . the ground - state expectation value of the chiral condensate was also calculated@xcite as @xmath139^{-1/2 } + o(y^{-2 } ) \quad { \rm as}\quad y\to\infty \label{eq39}\ ] ] where @xmath140 in terms of the continuum chiral condensate . this quantity is given by the feynman hellmann theorem as @xmath141 where @xmath142 is the ground - state eigenvalue of @xmath77 . when the fermion mass @xmath44 becomes very large , it should be possible to study the model in a `` quenched '' approximation , where fermion loop diagrams are suppressed . in a staggered lattice formulation such as the present one , the suppression of all fermion loops will lead to the `` static '' limit , in which any fermion excitation is fixed at its initial lattice site , and apart from the mass term the remaining lattice hamiltonian is simply that of the pure gauge theory . to allow the fermions to move or migrate on the lattice , one has to go to the next order in powers of @xmath144 , the mass parameter , and allow second - order diagrams involving the excitation of an ( @xmath145 ) pair on neighbouring sites , as shown in figure 4 . figure 4a ) represents a loop diagram in the vacuum sector , and 4b ) a similar diagram in the one - fermion sector ; figure 4c ) illustrates a `` hopping '' diagram , in which an existing fermion hops two lattice sites . this allows fermion migration to take place . it is useful to define new fermion variables representing the excitations on the strong - coupling ground - state , namely : @xmath146 so that @xmath147 where @xmath148 is the strong - coupling ground state . the hamiltonian ( [ eq12 ] ) and link excitations ( table 1 ) are easily translated in terms of the new variables . in the quenched approximation outlined above , the effective hamiltonian in the vacuum sector is @xmath149 ( where @xmath150 is the number of sites on the lattice ) which is simply the pure gauge field hamiltonian plus a constant . the first constant term is the negative energy of the `` dirac sea '' on the lattice , while the second constant term is the contribution of the second - order diagram , figure 4a ) . more interesting is the effective hamiltonian in the @xmath151 sector , which takes the form : @xmath152 where @xmath153 + \phi^{\dag}(\vec{r}-\hat{2 } ) u_2^{\dag } ( \vec{r}- \hat{2 } ) [ u_2 ( \vec{r } ) \phi ( \vec{r}+ \hat{2 } ) + ( -1)^{r_2 + 1 } u_1^{\dag } ( \vec{r } ) \phi ( \vec{r}+ \hat{1 } ) ] \nonumber \\ & & + ( -1)^{r_2 + 1 } \phi^{\dag } ( \vec{r}+ \hat{2 } ) u_2 ( \vec{r } ) u_1 ( \vec{r } ) \phi ( \vec{r}+ \hat{1 } ) + { \rm h.c . } { \big \ } } \nonumber \\ & & + { 1\over 2 \mu } \sum_{\vec{r } { \rm ~even } } ( -1)^{\phi^{\dag } \phi ( \vec{r } ) } { \big \ { } \phi^{\dag } ( \vec{r}- \hat{1 } ) u_1 ( \vec{r}-\hat{1 } ) [ u_1 ( \vec{r } ) \phi ( \vec{r}+ \hat{1 } ) \nonumber \\ & & + ( -1)^{r_2 + 1 } ( u_2^{\dag } ( \vec{r}- \hat{2 } ) \phi ( \vec{r}- \hat{2 } ) + u_2 ( \vec{r } ) \phi ( \vec{r}+ \hat{2 } ) ] \nonumber \\ & & + \phi^{\dag}(\vec{r}-\hat{2 } ) u_2 ( \vec{r}-\hat{2 } ) [ u_2^{\dag } ( \vec{r } ) \phi ( \vec{r}+\hat{2 } ) + ( -1)^{r_2 + 1 } u_1 ( \vec{r } ) \phi ( \vec{r}+\hat{1 } ) ] \nonumber \\ & & + ( -1)^{r_2 + 1 } \phi^{\dag}(\vec{r}+\hat{2 } ) u_2^{\dag } ( \vec{r } ) u_1^{\dag } ( \vec{r } ) \phi ( \vec{r}+ \hat{1 } ) + { \rm h.c . } { \big \ } } \end{aligned}\ ] ] this clumsy expression is merely a constant , plus the pure gauge hamiltonian , plus `` hopping '' terms for the fermions in the six different paths allowed for a double hop on the staggered lattice . the associated phase factor @xmath154 accounts for the change of sign if the hopping fermion `` passes through '' an occupied site . these hopping terms correspond to the kinetic energy in the non - relativistic continuum hamiltonian . the effective hamiltonian is more complicated in form than the original lattice hamiltonian , but does not allow any further fermion excitations , and would therefore be quicker and easier to implement in numerical calculations . we have not attempted any such calculations as yet . to calculate the strong - coupling series for the model , we used nickel s cluster expansion method . the techniques necessary were reviewed in he et al @xcite , and will not be repeated here . in these calculations , the @xmath155 in eq . ( [ eq12 ] ) is taken as the unperturbed hamiltonian , diagonal in the basis of eigenvectors of @xmath156 , while the @xmath51 and @xmath52 in eq . ( [ eq12 ] ) then act as perturbations . to generate the series for the ground state energy , we need to generate a list of connected plaquette configurations , together with their lattice constants and embedding constants . since the first - order perturbation @xmath51 and second - order perturbation @xmath52 involve links and plaquettes , respectively , a cluster @xmath157 will contribute terms @xmath158 , where @xmath157 is given by @xmath159 where @xmath160 is the number of plaquettes in @xmath157 , and @xmath68 the number of links not contained in plaquettes . up to the order @xmath161 considered in the current paper , there is only one graph ( figure 5 ) which does not obey the above relation , it actually contributes to order @xmath161 ( due to the combination of @xmath52 on each plaquette and @xmath51 in outer links ) instead of @xmath162 according to eq.([eqord ] ) . there are a total of 5494 graphs which contribute up to order @xmath161 for the ground - state properties . the calculation of glueball masses involves a list of clusters , both connected and disconnected , with at least one plaquette in each graph . there are 457 graphs which contribute to order @xmath163 . the calculation of meson masses generally involves a list of both connected and disconnected clusters@xcite . the eight different links in the unit cell ( shown in figure 3 ) are not equivalent in those calculations , which means we can not identify clusters which are topologically equivalent , or even use rotation or reflection symmetry . thus the separate clusters proliferate enormously in this case : there are 164142 clusters contributing up to order @xmath163 . for the scalar meson mass @xmath164 , the eight different bond types are equivalent , so we only need actually one bond type , and there are only 569 clusters which contribute up to order @xmath165 . using the linked - cluster expansion method , series have been calculated for the ground - state energy per site up to order @xmath161 . the first few terms are : @xmath166 where @xmath142 is the ground - state eigenvalue of @xmath77 . the coefficients are listed for various fixed values of the dimensionless mass parameter @xmath167 in table iii . these coefficients agree with those of reference i up to @xmath168 . extrapolating these series into the weak - coupling region using integrated differential approximants@xcite , one obtains results as shown in figure 6 . for the large - mass case , @xmath169 , it can be seen that the strong - coupling approximants match on to the weak - coupling form ( [ eq37 ] ) very nicely at around @xmath170 . for the lower masses , the strong - coupling approximants do not converge well enough to establish a precise matching , but they are clearly quite consistent with the asymptotic form ( [ eq37 ] ) . it is noteworthy that the @xmath43-independent form ( [ eq38 ] ) is only attained for very weak couplings @xmath171 . the successful matching between the strong - coupling approximants and the weak - coupling asymptotic form gives some confidence that the series coefficients have been calculated correctly , and that the approximants converge well enough to provide useful information about the weak - coupling ( continuum ) behaviour . a more interesting quantity is the chiral condensate . the first few terms in the series are : @xmath172 further coefficients at fixed values of @xmath43 are listed in table iii . figure 7 shows the extrapolation of these series into the weak - coupling region , as compared with the weak - coupling form ( [ eq39 ] ) . once again , the large mass results @xmath173 match very well to the weak - coupling form , while the lower - mass ones have not yet attained it before convergence of the strong - coupling approximants is lost . the most interesting case is @xmath174 , the zero mass limit . a graph of @xmath175 against @xmath176 is shown in figure 8 . it can be seen that integrated differential approximants to the series fail to converge below about @xmath177 , although pad approximants behave in a more consistent fashion . if taken at face value , the pad approximants would indicate a very large value of the chiral condensate in the continuum limit @xmath178 , @xmath179 also shown in figure 8 , however , are some monte carlo estimates of the chiral condensate for the euclidean version of this model and hamiltonian coupling @xmath180 are equal , lacking information on a more precise connection . ] by burkitt and irving@xcite . their results are roughly compatible with ours at about @xmath181 , but show a dramatic decrease in magnitude beyond that point . since this occurs well below the region of convergence of the series approximants , the series provide little evidence either to confirm or deny this phenomenon . another monte carlo calculation of the chiral condensate in the non - compact version of the model has been carried out by dagotto , kogut and kocic@xcite . they found a quite different behaviour , in which @xmath182 ( where @xmath183 ) plunges rapidly toward zero at a low @xmath184 value around @xmath185 , and then levels out to a plateau at a very small value , around @xmath186 . although this value is only tracked to @xmath187 , they take this as evidence of chiral symmetry breaking in the continuum limit . burden and roberts@xcite have also obtained a schwinger - dyson estimate of the chiral order parameter , @xmath188 which is at least a similar order of magnitude to that of dagotto _ et al._@xcite . it differs by a full two orders of magnitude from the naive extrapolation above , equation ( [ padechi ] ) . it would be interesting to see a more extensive monte carlo simulation of the compact lattice model , to check whether the decrease seen by burkitt and irving@xcite is real , and whether the chiral condensate subsequently levels out at a small plateau value as in the noncompact model . there seems to be very little prospect that further series calculations could shed light on these questions . strong - coupling series for the ` photonball ' masses , @xmath189 and @xmath190 , corresponding in the strong - coupling limit to single plaquette excitations which are antisymmetric and symmetric under reflections , respectively , have been calculated to order @xmath163 . the leading terms for both series are : @xmath191 with the difference between the two series only emerging at order @xmath192 . further coefficients at fixed values of @xmath43 are listed in table iv . the coefficients up to @xmath193 were previous calculated in reference i. approximants to these strong - coupling series are graphed in figures 9 and 10 , and compared with the results for the pure gauge theory , which we have calculated previously@xcite to order @xmath194 . it can be seen that for large @xmath43 values the series behave very similarly to the pure gauge case , and appear to be decreasing exponentially towards zero as @xmath65 increases . our results are consistent with those of burkitt and irving@xcite , who found a systematic downward shift in @xmath189 for moderate quark masses , as seen for the case @xmath195 in figs . 9 and 10 . at small @xmath43 values , however , the behaviour appears to change somewhat . in the region @xmath196 , there is substantial overlap and level crossing between the glueball and meson states in the strong - coupling region , and the picture becomes more confused . the series approximants show an initial rise for the glueball masses in the strong - coupling region , followed by an apparent turnover , at least in the @xmath197 case , but the convergence is not sufficient to track the behaviour reliably into the weak - coupling region . strong - coupling series for the meson states discussed in section ii have been calculated to @xmath198 ( or to @xmath199 for @xmath164 ) . the first three terms for arbitrary @xmath43 are @xmath200 + o(y^6)\\ m_2 & = & m_4= 1 + 2 \mu + 10 y^2/(1 + 2 \mu ) - y^4 ( 283 + 42\mu + 12\mu^2 + 8\mu^3)/[3(1 + 2\mu)^3 ] + o(y^6)\\ m_3 & = & 1 + 2 \mu + 6 y^2/(1 + 2 \mu ) + y^4 ( -175 + 6\mu + 36\mu^2 - 8\mu^3)/[3(1 + 2\mu)^3 ] + o(y^6)\\ m_5 & = & m_6 = 1 + 2 \mu + 6 y^2/(1 + 2 \mu ) - y^4 ( 199 + 66\mu + 12\mu^2 + 8\mu^3)/[3(1 + 2\mu)^3 ] + o(y^6)\\ m_7 & = & m_8 = 1 + 2 \mu + 6 y^2/(1 + 2 \mu ) - y^4 ( 199 + 66\mu + 12\mu^2 + 8\mu^3)/[3(1 + 2\mu)^3]+o(y^6)\end{aligned}\ ] ] further terms at selected values of @xmath43 are given in table v. figure 11 graphs the various meson masses as functions of @xmath176 , at a fixed , large mass parameter @xmath201 . it can be seen that the series approximants for these masses do not converge below about @xmath202 . for the lowest vector meson states there is a suggestion that the mass reaches a peak at @xmath202 , and then turns downward . a crude linear extrapolation has been made to estimate the continuum limit , but the uncertainly in the estimate is large . the resulting values for the vector mass are graphed as a function of @xmath43 in figure 12 , along with the schwinger - dyson estimates of allen and burden@xcite , and the non - relativistic prediction@xcite . it can be seen that the lattice estimates of the positronium binding energy have large errors , especially at smaller @xmath43 , and lie about three times higher than the schwinger - dyson values . neither the schwinger - dyson nor the lattice estimates show any definite evidence of the logarithmic increase at large @xmath43 predicted by the non - relativistic theory . figure 13 shows the masses as functions of @xmath176 for the massless case @xmath75 . once again , convergence is lost at rather small @xmath65 values , around @xmath203 , and it is hardly possible to make useful estimates of the continuum limit . there is no sign of any of the masses dropping towards zero , or acting like a goldstone boson . this is because the expected goldstone boson states are the `` axiscalar '' and `` axipseudoscalar '' states@xcite , which are not among the single - link excitations in the strong - coupling limit which we have treated here ( see section iid ) . the goldstone bosons probably correspond to l - shaped double - link excitations , which will transform into each other under a single plaquette excitation . it would be interesting to study their behaviour , but we have not yet attempted such a study , owing to technical complications . new strong - coupling series have been presented for the ground - state energy and chiral condensate , along with the ` glueball ' and positronium masses , in 4-component hamiltonian lattice qed@xmath204 with full dynamical fermions . this represents the first attempt at a series calculation for the positronium states in this model . two major features are evident from these results . firstly , there is a very clear separation of scales between the ` glueballs ' and the positronium states as the continuum limit is approached . at large fermion mass @xmath43 , the positronium energies remain finite in the continuum limit , while the ` glueball ' masses scale exponentially towards zero , presumably corresponding in the limit to massless photon states . the same thing appears to happen at smaller @xmath43 values , although our evidence for the exponential decrease of the glueball mass is rather slim . this separation of scales is consistent with earlier discussions of the pure gauge model@xcite : in the nonrelativistic or static fermion limit , if one sticks to naive ` engineering ' dimension scales one will end up with a theory of free , massless photons as above , whereas if one renormalizes the scale as discussed by polyakov@xcite or gpfert and mack@xcite one will obtain a theory of free , _ massive _ bosons . the second feature is that the addition of dynamical fermions does not greatly affect the glueball masses at large @xmath43 : at @xmath195 , for instance , the only effect was a reduction of the glueball mass at fixed @xmath65 of order a percent or two . for @xmath205 the effect is more pronounced , however . in general , the results of these calculations were somewhat disappointing . the bulk ground - state energy per site converged well enough into the weak - coupling region to display a convincing match with analytic weak - coupling expansions@xcite , and the ` glueball ' masses converged well enough to justify the statements given above . other quantities , however , were not mapped out with such success . series approximants to the chiral condensate at @xmath75 only converged down to @xmath177 , and were unable to confirm or deny the rapid plunge in magnitude seen by burkitt and irving@xcite around @xmath170 . it will require more detailed monte carlo studies to confirm that this plunge really occurs , and to determine whether the chiral condensate then levels out at a small but finite value . series approximants to the positronium masses likewise only converged down to @xmath202 , even at large mass @xmath43 , making any extrapolations to the continuum limit very uncertain . our estimates of the continuum ` binding energy ' are thus of little more than qualitative accuracy . they lie about three times higher than the schwinger - dyson estimates of allen and burden@xcite , which are not expected to be very accurate at large mass @xmath43 in any case . neither set of results shows any definite sign of the logarithmic increase in energy with @xmath43 predicted by the non - relativistic analyses@xcite . once again , a detailed understanding of the positronium spectrum must await a more accurate study , by monte carlo or other methods . the question whether this model develops goldstone bosons in the massless limit @xmath75 was not explored . these ` axiscalar ' and ` axipseudoscalar ' states@xcite are not among the single - link excitations for which we have calculated strong - coupling expansions . they are most likely to be found among the ` double - link ' excitations . it is technically more difficult to calculate strong - coupling series for the double - link excitations ; but on the other hand , the series may well converge more quickly for these low - lying excitations . this might provide an interesting subject for further study . in general , however , the prospects for further series calculations look dim . in contrast to the schwinger model case@xcite , the strong - coupling series approximants do not converge far enough into the weak - coupling region to allow accurate extrapolations to the continuum limit . thus it seems there is still no real alternative to monte carlo methods , whatever their limitations , for the detailed investigation of lattice gauge theories in three and four dimensions . the strong - coupling series provide an accurate ` platform ' of results in the strong - coupling region , which can provide useful calibration points for other methods ; but their extrapolation into the weak - coupling region remains rather uncertain . e - mail address : [email protected] e - mail address : [email protected] e - mail address : [email protected] j.m . cornwall , phys . d * 22 * , 1452(1980 ) . pisarski , phys . rev . d*29 * , 2423(1984 ) . appelquist , m. bowick , d. karabali and l.c.r . wijewardhana , phys . lett . * 55 * , 1715(1985 ) ; phys . rev . d*33 * , 3704(1986 ) . g. baskaran and p.w . anderson , phys . rev . b*37 * , 580(1988 ) ; i. affleck and j. marston , phys . b*37 * , 3774(1988 ) ; e. dagotto , e. fradkin and a. moreo , phys . rev . b*38 * , 2926(1988 ) . s. deser , r. jackiw and s. templeton , ann . ( ny ) * 140 * , 372(1982 ) ; j. f. schonfeld , nucl . b * 185 * , 157(1981 ) ; a.j . niemi and g.w . semenoff , phys . * 51 * , 2077(1983 ) ; a.n . redlich , phys . d * 29 * , 2366(1984 ) . c.j . burden and a.n . burkitt , europhys . lett . * 3 * , 545(1987 ) . m.r . pennington and d. walsh , phys . b * 253 * , 246(1991 ) ; d.c . curtis , m.r . pennington and d. walsh , phys . b * 295 * , 313(1992 ) . e. dagotto , j.b . kogut and a. kocic , phys . lett . * 62 * , 1083(1989 ) ; nucl . b * 334 * , 279(1990 ) . t. appelquist , d. nash and l. wijewardhana , phys . lett . * 60 * , 2575(1988 ) ; d. nash , phys . lett . * 62 * , 3024(1989 ) . gusynin , a.h . hams and m. reenders , phys . d * 53 * , 2227(1996 ) . k. farakos and g. koutsoumbas , phys . b * 178 * , 260(1986 ) . a.n . burkitt and a.c . irving , nucl . b * 295 * , 525(1988 ) . burden , nucl . b * 387 * , 419(1992 ) ; p. maris , phys . d * 52 * , 6087(1995 ) . c.j . burden and c.d . roberts , phys . rev . d * 44 * , 540(1991 ) . allen and c.j . burden , phys . rev . d*53 * , 5842(1996 ) ; phys . d * 55 * , 4954(1997 ) . c.j . burden and c.j . hamer , phys . rev . d*37 * , 479(1988 ) . aroca and h. fort , phys . b * 317 * , 604(1993 ) . m. burkardt and a. langnau , phys . d * 44 * , 1187(1991 ) . a. tam , c.j . hamer and c.m . yung , j. phys . g * 21 * , 1463(1995 ) . cornwall and g. tiktopoulos , phys . d * 15 * , 2937(1977 ) ; d. sen , phys . d * 41 * , 1227(1990 ) . yung and c.j . hamer , phys . d * 44 * , 2595(1991 ) ; v.g . koures , j. comp . phys . * 128 * , 1(1996 ) . for a review , he , c.j . hamer and j. oitmaa , j. phys . a * 23 * , 1775(1990 ) . hamer , w.h . zheng and j. oitmaa , phys . * , 55(1997 ) . l. susskind , phys . d * 16 * , 3031(1977 ) . j. kogut and l. susskind , phys . rev . d*11 * , 395(1975 ) . golterman , nucl . b * 273 * , 663(1986 ) . for a review , guttmann , in _ phase transitions and critical phenomena _ , edited by c. domb and j.l . lebowitz ( academic , new york , 1989 ) , vol . 13 . hamer , w.h . zheng and j. oitmaa , phys . rev . d*45 * , 4652(1992 ) . l. gross , commun . * 92 * , 137(1983 ) ; c.j . hamer and w.h . zheng , phys . d * 48 * , 4435(1993 ) . a.m polyakov , phys . b * 59 * , 82(1975 ) . m. gpfert and g. mack , commun . phys . * 82 * , 545(1982 ) . ( 1200,700)(-200,-150 ) ( -50,0)(3,0)400 ( 0,-50)(0,3)400 ( -50,300)(3,0)400 ( 300,-50)(0,3)400 ( 50,50)1 ( 330,50)2 ( 50,330)4 ( 330,330)3 ( 400,150)(2,0)80 ( 550,0)(3,0)400 ( 600,-50)(0,3)400 ( 550,300)(3,0)400 ( 900,-50)(0,3)400 ( 550,-50)2 ( 930,-50)3 ( 550,330)1 ( 930,330)4 ( 0,0)(300,0)2 ( 0,300)(300,0)2 ( 0,0)(300,0)1 ( 600,0)(300,0)2 ( 600,300)(300,0)2 ( 600,300)(300,0)1 ( -100,-100)@xmath206 ( 1200,1000)(-200,-100 ) ( -50,0)(3,0)700 ( 0,-50)(0,3)700 ( -50,300)(3,0)700 ( 300,-50)(0,3)700 ( -50,600)(3,0)700 ( 600,-50)(0,3)700 ( -50,-50)1 ( 330,-50)2 ( 630,-50)5 ( -50,330)4 ( 330,330)3 ( 330,630)7 ( -80,630)8 ( 630,330)6 ( 130,-60)@xmath207 ( 430,-60)@xmath208 ( 130,240)@xmath209 ( 430,240)@xmath210 ( 20,150)@xmath211 ( 20,450)@xmath212 ( 320,150)@xmath213 ( 320,450)@xmath214 ( 0,0)(300,0)3 ( 0,300)(300,0)3 ( 0,600)(300,0)3 ( 1000,600)(-200,-100 ) ( 0,0)(200,400 ) ( 0,0)(0,2)200 ( 200,400)(0,-2)200 ( 90,-80)a ) ( 450,0)(0,3)400 ( 450,0)(0,3)200 ( 500,0)(200,400 ) ( 500,0)(0,2)200 ( 700,400)(0,-2)200 ( 580,-80)b ) ( 1000,0)(0,3)300 ( 1000,0)(0,3)100 ( 1000,300)(3,0)100 ( 1100,300)(0,-30)200 ( 1100,300)(0,-30)100 ( 1100,100)(30,0)100 ( 1200,100)(0,30)300 ( 1200,100)(0,30)150 ( 1100,-80)c ) rrrrrrrrr & & & & & & & & + + @xmath215 & 1 & 1 & 1 & 0 & 1 & 0 & 1 & 0 + @xmath216 & 1 & 1 & 1 & 0 & @xmath2171 & 0 & @xmath2171 & 0 + @xmath218 & @xmath2171 & 0 & 1 & 1 & 0 & 1 & 0 & 1 + @xmath219 & @xmath2171 & 0 & 1 & 1 & 0 & @xmath2171 & 0 & @xmath2171 + @xmath220 & @xmath2171 & 1 & @xmath2171 & 0 & 1 & 0 & @xmath2171 & 0 + @xmath221 & @xmath2171 & 1 & @xmath2171 & 0 & @xmath2171 & 0 & 1 & 0 + @xmath222 & @xmath2171 & 0 & 1 & @xmath2171 & 0 & @xmath2171 & 0 & 1 + @xmath223 & @xmath2171 & 0 & 1 & @xmath2171 & 0 & 1 & 0 & @xmath2171 + + @xmath91 & + 1 & @xmath217 & @xmath2171 & @xmath217 & @xmath217 & @xmath217 & @xmath217 & @xmath217 + @xmath87 & + 1 & @xmath2171 & + 1 & @xmath2171 & + 1 & + 1 & @xmath2171 & @xmath2171 + @xmath103&+1 & + 1 & + 1 & + 1 & @xmath2171 & + 1 & @xmath2171 & + 1 + @xmath106 & + 1 & + 1 & + 1 & @xmath2171 & @xmath2171 & @xmath2171 & @xmath2171 & @xmath2171 + @xmath96 & + 1 & + 1 & + 1 & + 1 & @xmath2171 & @xmath2171 & @xmath2171 & @xmath2171 + rrrrrr & & & & & + + 0 & 0.000000000000 & @xmath2172.500000000000@xmath224 & @xmath2175.000000000000@xmath224 & @xmath2171.000000000000 & @xmath2175.000000000000 + 1 & @xmath2172.000000000000 & @xmath2171.000000000000 & @xmath2176.666666666667@xmath224 & @xmath2174.000000000000@xmath224 & @xmath2179.523809523810@xmath225 + 2 & 1.350000000000@xmath226 & 1.250000000000 & 1.851851851852@xmath225 & @xmath2173.880000000000@xmath224 & @xmath2174.984882842026@xmath224 + 3 & @xmath2172.258095238095@xmath227 & @xmath2177.562500000000 & @xmath2171.088065843621 & @xmath2171.057163636364@xmath224 & @xmath2174.840190310415@xmath228 + 4 & 4.740493349632@xmath229 & 4.168190104167@xmath226 & 2.828630278758 & 1.440413481513@xmath224 & 3.805417269032@xmath225 + 5 & @xmath2171.145332120404@xmath230 & @xmath2172.641393774675@xmath227 & @xmath2178.251436202716 & @xmath2171.196372180167@xmath224 & 7.580050883728@xmath231 + 6 & 3.019112271993@xmath232 & 1.828909211780@xmath229 & 2.673208303367@xmath226 & 1.528725434873@xmath224 & @xmath2172.829817166242@xmath233 + 7 & @xmath2178.446045864104@xmath234 & @xmath2171.343961734536@xmath235 & @xmath2179.195676938217@xmath226 & @xmath2172.146264820882@xmath224 & @xmath2178.316158017118@xmath236 + 8 & 2.467210478469@xmath237 & 1.031183072419@xmath230 & 3.302984237354@xmath227 & 3.075754366619@xmath224 & @xmath2176.084851949294@xmath228 + 9 & @xmath2177.447992091066@xmath238 & @xmath2178.175829204890@xmath230 & @xmath2171.226060361259@xmath229 & @xmath2174.580923967202@xmath224 & @xmath2173.103254879714@xmath236 + 10 & 2.307292550322@xmath239 & 6.651573204917@xmath232 & 4.670066518532@xmath229 & 6.989367972569@xmath224 & 4.685274190981@xmath228 + 11 & @xmath2177.298475917954@xmath240 & @xmath2175.525241939727@xmath234 & @xmath2171.816217581154@xmath235 & @xmath2171.087547736352 & 2.784016199972@xmath236 + + 0 & 5.000000000000@xmath224 & 5.000000000000@xmath224 & 5.000000000000@xmath224 & 5.000000000000@xmath224 & 5.000000000000@xmath224 + 1 & @xmath2174.000000000000 & @xmath2171.000000000000 & @xmath2174.444444444444@xmath224 & @xmath2171.600000000000@xmath224 & @xmath2179.070294784581@xmath233 + 2 & 8.400000000000@xmath226 & 5.250000000000 & 1.037037037037 & 1.344000000000@xmath224 & 4.319187992657@xmath228 + 3 & @xmath2172.231056689342@xmath229 & @xmath2173.661718750000@xmath226 & @xmath2173.414650205761 & @xmath2171.856045336482@xmath224 & @xmath2171.490233956148@xmath228 + 4 & 6.534158638592@xmath235 & 2.809490017361@xmath227 & 1.222412306072@xmath226 & 2.632452778514@xmath224 & 1.645776229220@xmath231 + 5 & @xmath2172.025737847176@xmath232 & @xmath2172.284382049423@xmath229 & @xmath2174.639999014178@xmath226 & @xmath2173.921048572445@xmath224 & 2.005503745248@xmath231 + 6 & 6.518294767679@xmath234 & 1.929013101510@xmath235 & 1.831818873072@xmath227 & 6.165298249176@xmath224 & @xmath2172.127283455260@xmath236 + 7 & @xmath2172.153203741629@xmath237 & @xmath2171.672676257346@xmath230 & @xmath2177.429933326106@xmath227 & @xmath2179.978195968035@xmath224 & @xmath2172.070374769933@xmath236 + 8 & 7.252977297783@xmath238 & 1.479158204994@xmath232 & 3.074244834593@xmath229 & 1.648612396832 & 1.799802187362@xmath241 + 9 & @xmath2172.480301602121@xmath239 & @xmath2171.327976394916@xmath234 & @xmath2171.291627245890@xmath235 & @xmath2172.768670289809 & @xmath2179.300630428741@xmath241 + 10 & 8.584509655365@xmath240 & 1.206675388588@xmath242 & 5.492904223987@xmath235 & 4.708649820309 & 1.110110045268@xmath241 + 11 & @xmath2173.000444314331@xmath243 & @xmath2171.107246112597@xmath237 & @xmath2172.359089145008@xmath230 & @xmath2178.090428010991 & 7.857720150141@xmath241 + rrrrrr & & & & & + + 0 & 4.000000000000 & 4.000000000000 & 4.000000000000 & 4.000000000000 & 4.000000000000 + 1 & 5.333333333333 & 1.500000000000 & @xmath2171.066666666667 & @xmath2171.523809523810@xmath224 & @xmath2171.743489157677@xmath233 + 2 & @xmath2174.189814814815@xmath226 & @xmath2173.135416666667 & 4.769703703704 & @xmath2172.352469495735@xmath224 & @xmath2174.165631870342@xmath224 + 3 & 5.935307344636@xmath227 & @xmath2171.170910493827@xmath225 & @xmath2172.394715197442@xmath226 & @xmath2174.890552931893@xmath224 & @xmath2173.937708897154@xmath228 + 4 & @xmath2173.229319000985@xmath229 & 8.663828760713@xmath226 & 1.501264118830@xmath227 & 2.073466005647 & 4.945486284578@xmath224 + 5 & @xmath2176.430232339023@xmath230 & @xmath2171.617503042067@xmath229 & @xmath2178.784848865979@xmath227 & @xmath2175.575125327898 & 1.071338323552@xmath233 + + 0 & 4.000000000000 & 4.000000000000 & 4.000000000000 & 4.000000000000 & 4.000000000000 + 1 & 5.333333333333 & 1.500000000000 & @xmath2171.066666666667 & @xmath2171.523809523810@xmath224 & @xmath2171.743489157677@xmath233 + 2 & @xmath2174.306481481481@xmath226 & @xmath2174.302083333333 & 3.603037037037 & @xmath2171.401913616240 & @xmath2171.583229853701 + 3 & 6.055105324434@xmath227 & 1.838348765432@xmath225 & @xmath2172.776504371231@xmath226 & @xmath2173.473208512644@xmath224 & @xmath2171.358230163160@xmath233 + 4 & 9.053923337359@xmath229 & 5.354226523059@xmath226 & 1.574287286646@xmath227 & 1.695152914866 & 1.162208030611 + 5 & @xmath2171.377596830177@xmath232 & @xmath2175.970881550866@xmath227 & @xmath2179.103588055806@xmath227 & @xmath2177.260136671150@xmath224 & 2.449261377761@xmath233 + rrrrrr & & & & & + + 0 & 1.000000000000 & 1.000000000000 & 1.000000000000 & 1.000000000000 & 1.000000000000 + 1 & 1.400000000000@xmath226 & 1.000000000000@xmath226 & 6.000000000000 & 6.000000000000 & 6.000000000000 + 2 & @xmath2171.783333333333@xmath227 & @xmath2179.433333333333@xmath226 & @xmath2175.833333333333@xmath226 & @xmath2176.633333333333@xmath226 & @xmath2176.633333333333@xmath226 + 3 & 3.625151675485@xmath229 & 1.757293650794@xmath229 & 1.200151675485@xmath229 & 1.341849206349@xmath229 & 1.253207231041@xmath229 + 4 & @xmath2179.101099254509@xmath235 & @xmath2174.164808339711@xmath235 & @xmath2172.976672075609@xmath235 & @xmath2173.324256060799@xmath235 & @xmath2173.008509651089@xmath235 + 5 & 2.534993350264@xmath232 & 1.113036398112@xmath232 & 8.177458288471@xmath230 & 9.143840136437@xmath230 & 8.115398963424@xmath230 + 6 & @xmath2177.521807853058@xmath234 & & & & + + 0 & 2.000000000000 & 2.000000000000 & 2.000000000000 & 2.000000000000 & 2.000000000000 + 1 & 7.000000000000 & 5.000000000000 & 3.000000000000 & 3.000000000000 & 3.000000000000 + 2 & @xmath2172.683333333333@xmath226 & @xmath2171.283333333333@xmath226 & @xmath2176.833333333333 & @xmath2179.833333333333 & @xmath2179.833333333333 + 3 & 1.365322916667@xmath227 & 6.154340277778@xmath226 & 4.075451388889@xmath226 & 5.498784722222@xmath226 & 4.683784722222@xmath226 + 4 & @xmath2178.920838363922@xmath227 & @xmath2173.876159660218@xmath227 & @xmath2172.760344845403@xmath227 & @xmath2173.678279451885@xmath227 & @xmath2172.870075748181@xmath227 + 5 & 6.436060668280@xmath229 & 2.730071323411@xmath229 & 2.011538052494@xmath229 & 2.664691694041@xmath229 & 2.110416878518@xmath229 + 6 & @xmath2174.984333949987@xmath235 & & & & + + 0 & 3.000000000000 & 3.000000000000 & 3.000000000000 & 3.000000000000 & 3.000000000000 + 1 & 4.666666666667 & 3.333333333333 & 2.000000000000 & 2.000000000000 & 2.000000000000 + 2 & @xmath2179.740740740741 & @xmath2174.259259259259 & @xmath2171.740740740741 & @xmath2173.518518518519 & @xmath2173.518518518519 + 3 & 2.117987280210@xmath226 & 9.052263374486 & 5.541189674523 & 9.760905349794 & 6.595922184811 + 4 & @xmath2177.103051896801@xmath226 & @xmath2172.746737478771@xmath226 & @xmath2171.941553551960@xmath226 & @xmath2173.281454899895@xmath226 & @xmath2172.157173735275@xmath226 + 5 & 2.054506982964@xmath227 & 9.091443816105@xmath226 & 6.590006451270@xmath226 & 1.138794166995@xmath227 & 7.209915733050@xmath226 + 6 & @xmath2178.384073194275@xmath227 & & & & + + 0 & 5.000000000000 & 5.000000000000 & 5.000000000000 & 5.000000000000 & 5.000000000000 + 1 & 2.800000000000 & 2.000000000000 & 1.200000000000 & 1.200000000000 & 1.200000000000 + 2 & @xmath2173.229333333333 & @xmath2171.277333333333 & @xmath2172.213333333333@xmath224 & @xmath2171.181333333333 & @xmath2171.181333333333 + 3 & 3.119580472860 & 8.464108513709@xmath224 & 4.398750760351@xmath224 & 1.428440057720 & 9.870369807970@xmath224 + 4 & @xmath2174.249661491769 & @xmath2171.098495894429 & @xmath2178.941846899782@xmath224 & @xmath2172.414224882476 & @xmath2171.183255224582 + 5 & 9.281773558871 & 1.580697986782 & 1.236101545345 & 4.024451851449 & 2.130785627667 + 6 & @xmath2172.667077094545@xmath226 & & & & + + 0 & 2.100000000000@xmath226 & 2.100000000000@xmath226 & 2.100000000000@xmath226 & 2.100000000000@xmath226 & 2.100000000000@xmath226 + 1 & 6.666666666667@xmath224 & 4.761904761905@xmath224 & 2.857142857143@xmath224 & 2.857142857143@xmath224 & 2.857142857143@xmath224 + 2 & @xmath2175.901090594968@xmath224 & @xmath2173.564409890941@xmath224 & @xmath2171.625094482237@xmath224 & @xmath2173.620559334845@xmath224 & @xmath2173.620559334845@xmath224 + 3 & @xmath2172.582504019689@xmath233 & @xmath2171.053385459541@xmath225 & 1.090172643379@xmath225 & 3.819111465784@xmath225 & 1.199822627856@xmath225 + 4 & 2.457235107452@xmath224 & 1.646791697513@xmath224 & 7.346442898971@xmath225 & 1.373283015502@xmath224 & 1.747678732566@xmath224 + 5 & 6.900777782581@xmath225 & 1.213278624644@xmath225 & @xmath2177.752772094654@xmath233 & 5.188958501563@xmath233 & @xmath2178.047408934744@xmath233 + 6 & @xmath2171.752056769026@xmath224 & & & & +

the agn - host co - evolution issue is investigated here by focusing on the evolution of the [ ] emission - line profile . in order to simultaneously measure both [ ] line profile and circumnuclear stellar population in individual spectrum , a large sample of narrow emission - line galaxies is selected from the mpa / jhu sdss dr7 catalog . by requiring that 1 ) the [ ] line signal - to - noise ratio is larger than 30 , 2 ) the [ ] line width is larger than the instrumental resolution by a factor of 2 , our sample finally contains 2,333 seyfert galaxies / liners ( agns ) , 793 transition galaxies , and 190 starforming galaxies . in additional to the commonly used profile parameters ( i.e. , line centroid , relative velocity shift and velocity dispersion ) , two dimensionless shape parameters , skewness and kurtosis , are used to quantify the line shape deviation from a pure gaussian function . we show that the transition galaxies are systematically associated with narrower line widths and weaker [ ] broad wings than the agns , which implies that the kinematics of the emission - line gas is different in the two kinds of objects . by combining the measured host properties and line shape parameters , we find that the agns with stronger blue asymmetries tend to be associated with younger stellar populations . however , the similar trend is not identified in the transition galaxies . the failure is likely resulted from a selection effect in which the transition galaxies are systematically associated with younger stellar populations than the agns . the evolutionary significance revealed here suggests that both nlr kinematics and outflow feedback in agns co - evolve with their host galaxies .

1108.1252

the emission from narrow - line region ( nlr ) of active galactic nucleus ( agn ) is an important tool to study the relation between the activity of the central supermassive black hole ( smbh ) and the growth of its host galaxy in which the smbh resides , both because the nlr emission is mainly resulted from the illumination by the central agn and because the nlr kinematics is believed to be mainly dominated by the gravity of the bulge ( see review in wilson & heckman 1985 and references therein , whittle 1992a , b ; nelson & white 1996 ) . the gravity dominated kinematics motivates a number of previous studies to demonstrate that the line width of the agn s strong [ ] @xmath15007 emission line can be used as a proxy for the stellar velocity dispersion of the bugle ( e.g. , nelson & white 1996 ; nelson 2000 ; boroson 2003 ; komossa & xu 2007 ) . basing upon the tight @xmath2 relationship ( e.g. , tremaine et al . 2002 ; ferrarese & merritt 2000 ; magorrian et al . 1998 ; gebhardt et al . 2000 ; haring & rix 2004 ) , the proxy therefore allows one to easily estimate @xmath3 in a large sample of agns ( e.g. , grupe & mathur 2004 ; wang & lu 2001 ; komossa & xu 2007 ) . it is well known for a long time that the line profiles of the [ ] doublelets show a blue asymmetry with an extended blue wing and a sharp red falloff in a large fraction of agns ( e.g. , heckman et al . 1981 ; whittle 1985 ; wilson & heckman 1985 ; grupe et al 1999 ; tadhunter et al . 2001 ; veron - cetty et al . 2001 ; zamanov et al . 2002 ; komossa & xu 2007 ; xu & komossa 2009 ; greene & ho 2005 ; de roberties & osterbrock 1984 ; storchi - bergmann et al . 1992 ; arribas et al . 1996 ; christopoulou et al . 1997 ) . the blue asymmetry requires a narrow core gaussian profile ( @xmath4 ) with a blueshifted , broad gaussian component ( @xmath5 ) to reproduce the observed asymmetric profiles for both [ ] @xmath6 emission lines . the spectroscopic monitor revealed a variability time scale from one to ten years for the blue wings of the [ ] @xmath0 lines in two type i agns ( izw1 : wang et al . 2005 ; ngc5548 : sergeev et al . 1997 ) , which means that the blue wings are likely emitted from the intermediate - line region located between the traditional blr and nlr . in addition to the blue asymmetry , the redshifts of the [ ] doublelets are often found to be negative compared to the redshifts measured from both stellar absorption features and h@xmath7 emission line ( i.e. , [ ] blueshifts , e.g. , phillips 1976 ; zamanov et al .2002 ; marziani et al . 2003 ; aoki et al . 2005 ; boroson 2005 ; bian et al . 2005 ; komossa et al . although they are rare cases , the objects with strong [ ] blueshifts larger than 100@xmath8 are called `` blue outliers '' . the popular explanation of the observed [ ] emission - line profile is that the material outflow from central agn plays important role in reproducing the observed blue asymmetry and blueshift . with the advent of the high spatial resolution of hubble space telescope ( hst ) , spatially resolved spectroscopic observations of a few nearby seyfert 2 galaxies indicate that the nlrs show complicate kinematics , which could reproduce the observed [ ] line profiles by the radial outflow acceleration ( or deceleration ) and/or jet expansion ( e.g. , crenshaw et al . 2000 ; crenshaw & kraemer 2000 ; ruiz et al . 2001 ; nelson et al . 2000 ; hutchings et al . 1998 ; das et al . 2005 , 2006 , 2007 ; kaiser et al . 2000 ; crenshaw et al . 2010 , schlesinger et al . 2009 ; fischer et al . 2010 ; fischer et al . 2011 ) . recent systematical studies suggested that the blue asymmetry is related with the activity of the central smbh . veron - cetty et al . ( 2001 ) indicated that half of their sample of narrow - line seyfert 1 galaxies ( nls1s ) shows a broad and blueshifted [ ] @xmath15007 component in addition to the unshifted narrow core component . nelson et al . ( 2004 ) found a correlation between the blue asymmetry and eigenvector - i space by studying the [ ] @xmath15007 line profiles of the pg quasars . the quasars associated with larger blue asymmetries tend to be stronger emitters presumably having larger eddington ratios ( @xmath9 , where @xmath10 is the eddington luminosity , see also in xu et al . 2007 ; boroson 2005 ; greene & ho 2005 ; mathur & grupe 2005 ) . similar as the blue asymmetry , the [ ] blueshift is also found to be related with a number of agn properties . some authors claimed that the [ ] blueshift is directly correlated with @xmath9 ( e.g. , boroson 2005 ; bian et al . 2005 ) , although the correlation might not be the truth ( e.g. , aoki et al . marziani et al . ( 2003 ) pointed out that all the `` blue outliers '' have small h@xmath7 line widths ( @xmath11 ) and high @xmath9 ( see also in zamanov et al . 2002 ; komossa et al . 2008 ) . agns are now widely believed to co - evolve with their host galaxies , which is implied by the tight @xmath2 correlation ( see the citations in the first paragraph ) and by the global evolutionary history of the growth of the central smbh that traces the star formation history closely from present to @xmath12 ( e.g. , nandra et al . 2005 ; silverman et al . 2008 ; shankar et al . 2009 ; hasinger et al . 2005 ) . a number of studies recently provided direct evidence supporting the co - evolutionary scenario in which an agn evolves along the eigenvector - i space from a high @xmath9 state to a low @xmath9 state as the circumnuclear stellar population continually ages ( e.g. , wang et al . 2006 ; wang & wei 2008 , 2010 ; kewley et al . 2006 ; wild et al . 2007 ; davis et al . 2007 ) . the results of theoretical simulations indicate a possibility that a major merger between two gas - rich disk galaxies plays important role in the co - evolution of agns and their host galaxies ( e.g. , di matteo et al . 2007 ; hopkins et al . 2007 ; granato et al . detailed analysis even suggests a delay of @xmath13myr for the detectable agn activity after the onset of the star formation activity ( e.g. , wang & wei 2005 ; schawinski et al . 2009 ; hopkins et al . 2005 ; davis et al . 2007 ; wild et al . 2010 ) . the delay might be resulted from the feedback from either the central agn ( e.g. , hopkins et al . 2005 ) or the stellar winds ( e.g. , morman & scoville 1988 ) , and/or resulted from the purely dynamical origin ( hopkins 2011 ) . the outflow origin for both blue asymmetry and blueshift naturally gives us the hint that the [ ] emission - line profile ( and the inferred nlr kinematics ) co - evolves with the stellar population of the host galaxy . we here report a study that is an effort to examine the evolution of the [ ] @xmath15007 emission - line profile . in principle , both broad- and narrow - line agns are needed to be analyzed to give a complete study . although the [ ] line profiles can be easily measured in the spectra of typical type i agns , the host galaxy properties are hard to be determined because of the strong contamination from the central agn s continuum . since our aim is to study the relationship between the line profile and the host galaxy properties , the narrow emission - line galaxies from sloan digital sky survey ( sdss ) are adopted in our study because their spectra allow us to simultaneously measure both [ ] @xmath15007 line profile and stellar population properties in individual object . the high - order dimensionless line shape parameters are adopted by us to describe the line profile in details . the paper is organized as follows . the sample selection is presented in 2 . 3 describes the data reduction , including the stellar component removal , the line profile measurements , and the stellar population age measurements . our analysis and results are shown in 4 . the implications are discussed in the next section . a @xmath14 cold dark matter ( @xmath14cdm ) cosmology with parameters @xmath15 , @xmath16 , and @xmath17 ( spergle et al . 2003 ) is adopted throughout the paper . we select a sample of narrow emission - line galaxies from the value - added sdss data release 7 ( abazajian et al . 2009 ) mpa / jhu catalog ( heckman et al . 2004 ; kauffmann et al . 2003a , b , c ; tremonti et al . 2004 ; see heckman & kauffmann 2006 for a review ) , the catalog provides the measured spectral and physical properties of 927,555 individual narrow emission - line galaxies , including seyfert galaxies , liners , transition galaxies and star - forming galaxies . the integrated spectra of the narrow emission - line galaxies allow us to study the relationship between the [ ] @xmath15007 emission - line profiles and the circumnuclear stellar populations , because their continuum is dominated by the starlight components . at first , we select the galaxies whose spectra have median signal - to - noise ratio per pixel of the whole spectrum s / n@xmath1820 from the mpa / jhu sdss dr7 catalog to ensure that the stellar features could be properly separated from each observed spectrum . secondly , all the emission lines used in the traditional three baldwin - phillips - terlevich ( bpt ) diagnostic diagrams are required to be detected with a significant level at least 3@xmath19 . thirdly , the spectral profiles of the [ ] @xmath15007 emission lines are only measured for the galaxies whose [ ] lines have s / n larger than 30 . finally , the galaxies with redshifts within the range from 0.11 to 0.12 are removed in the subsequent spectral analysis . the removement is used to avoid the possible fake spectral profile caused by the poorly subtracted strong sky emission line [ ] @xmath15577 at the observer frame . after the selection based on the mpa / jhu catalog , we remove all the galaxies whose [ ] line widths are smaller than @xmath20 from our subsequent statistical comparisons according to our spectral profile measurements ( see section 3 for details ) . the mean intrinsic resolution used in the above criterion is adopted to be 65@xmath8 according to the sdss pipelines measurements ( e.g. , york et al . 2000 ) . there are two reasons for this instrumental resolution selection . at first , the cut on line width removes the low mass starforming galaxies without bugle . secondly , the cut can alleviate the impact on the measured line shape parameters caused by the instrumental resolution . for a pure gaussian profile , the intrinsic line width @xmath19 can be obtained by the equation @xmath21 , where @xmath22 and @xmath23 is the observed line width and the instrumental resolution , respectively . however , this relationship is only a first - order approximation for the line profiles that deviate from a pure gaussian profile . our analytic calculation indicates that the correction of the instrumental resolution depends on the deviation not only for the line width , but also for the other two high - order dimensionless line shape parameters ( see section 4 and appendix a for details ) . after removing the duplicates in the mpa / jhu catalog , there are in total 3,339 entries fulfilling the above selection criteria . the spectra with the highest s / n ratios are kept in our final sample in the duplicate removements . given the line fluxes reported in the mpa / jhu catalog , these galaxies are subsequently classified into three sub - samples according to the [ ] /h@xmath7 versus [ ] /h@xmath24 diagnostic diagram using the widely accepted classification schemes ( kauffmann et al . 2003a ; kewley et al . 2001 ) . out of the 3,339 galaxies , there are 2,333 seyfert galaxies / liners ( hereafter agns for abbreviation ) , 793 transition galaxies located between the empirical and the theoretical demarcation lines , and 190 star - forming galaxies . there are 5,856 star - forming galaxies with line width @xmath25 , whose line profiles are believed to be dominated by a symmetric gaussian function . these galaxies are used as a comparison sample to present the scatters of the measured high - order dimensionless line shape parameters for a symmetric gaussian line profile , because of the limited wavelength sampling and limited signal - to - noise ratio . these scatters are useful in determining how strong a line profile deviates from a symmetric gaussian profile . the three diagnostic bpt diagrams are shown in figure 1 for the final samples , in which the agns , transition galaxies and star - forming galaxies are symbolized by the red , green and blue solid squares , respectively . the contours represent the distributions of the star - forming galaxies in the comparison sample . the spectra of the selected objects are reduced and analyzed by our principal component analysis ( pca ) pipeline ( see wang & wei ( 2008 ) for details ) . at first , each spectrum is corrected for the galactic extinction using the extinction law with @xmath26 ( cardelli et al . 1989 ) , in which the color excess @xmath27 is taken from the schlegel , finkbeiner , and davis galactic reddening map ( schlegel et al . 1998 ) . secondly , each extinctions - corrected spectrum is then transformed to the rest frame , along with the flux correction due to the relativity effect , given the redshift provided by the sdss pipelines ( glazebrook et al . 1998 ; bromley et al . 1998 ) . the stellar absorption features are subsequently separated from each rest - frame spectrum by modeling the continuum and absorption features by the sum of the first seven eigenspectra . the eigenspectra are built from the standard single stellar population spectral library developed by bruzual & charlot ( 2003 ) . each of the spectra is fitted over the rest - frame wavelength range from 3700 to 7000 by a @xmath28 minimization , except for the regions with strong emission lines ( e.g. , balmer lines , [ ] @xmath294959 , 5007 , [ ] @xmath296548,6583 , [ ] @xmath296716 , 6731 , [ ] @xmath13727 , [ ] @xmath14363 and [ ] @xmath16300 ) . the rest - frame emission - line isolated spectra are used to parametrize the spectral profiles of their [ ] @xmath15007 emission lines , after the stellar components are removed from the observed spectra . the emission - line profile can be parametrized by many possible ways . the widely used ones include the fwhm and the second moment of the line that is defined as @xmath30 where @xmath19 is in units of @xmath31 , @xmath32 and @xmath33 is the line centroid ( i.e. , the first moment ) and the flux density of the continuum - subtracted line flux , respectively . for a pure gaussian profile , we have a relationship @xmath34 , which means that both fwhm and @xmath19 comparably describe the line broadening if the line profile is a gaussian function . however , as stated in greene & ho ( 2005 ) , @xmath19 contains more information on the line profile broadening if the profile deviates from a pure gaussian profile . heckman et al . ( 1981 ) defined an asymmetry index @xmath35 to measure the line asymmetry of [ ] emission lines , where @xmath36 and @xmath37 is the half width to the left and right of the line center defined as the 80% peak intensity level . basing upon the index @xmath38 , the authors found that the blue asymmetry is shown in about 80% of the 36 seyfert and radio galaxies . alternatively , the asymmetry index @xmath39/fwhm$ ] is widely used to measure the asymmetry of agn s broad emission lines , where @xmath40 are profile centroids measured at different levels ( see review in sulentic et al . in addition to the asymmetry , the emission - line shapes of agns vary from extremely peaked to very `` boxy '' . previous studies used the ratio of line widths at different levels to parametrize such line shapes ( e.g. , marziani et al . 1996 ) . to make the statistical study on [ ] line profile is feasible for a large sample , we adopt the high - order dimensionless line shape parameters to describe the line profile departures from a pure gaussian function ( see binney & merrifield ( 1998 ) for more details ) : @xmath41 where @xmath42 is the @xmath43-order moment defined as @xmath44 and @xmath19 is the second order moment defined in eq ( 1 ) , respectively . the first shape parameter @xmath45 is termed the `` skewness '' that measures the deviation from symmetry . a pure gaussian profile corresponds to @xmath46 . the emission - line shape with a positive value of @xmath45 shows a red asymmetry , and the shape with a negative value a blue asymmetry . the second parameter @xmath47 is termed the `` kurtosis '' that measures the symmetric deviation from a pure gaussian profile ( for a pure gaussian profile , we have @xmath48 ) . the emission line with peaked profile superposed on a broad base has a value of @xmath49 . a value of @xmath50 corresponds to a `` boxy '' line profile . we refer the readers to figure 11.5 in binney & merrifield ( 1998 ) for how the values of @xmath45 and @xmath47 vary with the line shapes . the bulk relative velocity shift of the [ ] emission line is measured relative to the h@xmath7 emission line in each emission - line isolated spectrum , because the h@xmath7 line can be easily detected and because the line shows very small velocity shift relative to the galaxy rest - frame ( e.g. , komossa et al . the velocity shift is calculated as @xmath51 , where @xmath52 and @xmath53 is the rest - frame wavelength of the [ ] @xmath0 emission line and the velocity of light , respectively . @xmath54 denotes the wavelength shift of the [ ] line with respect to the h@xmath7 line . the shift is determined from the measured line centroids ( @xmath55 ) and the rest - frame wavelengths in vacuum of both lines . with the definition of the velocity shift , a blueshift corresponds to a negative value of @xmath56 , and a redshift to a positive value . by following our previous studies again , we use the two lick indices , i.e. , the 4000 break ( @xmath57 ) and the equivalent width of h@xmath58 absorption feature of a - type stars ( h@xmath59 ) ( bruzual 1983 ; balogh et al . 1999 ) . the index h@xmath60 is defined by worthey & ottaviani ( 1997 ) as @xmath61 where @xmath62 is the flux within the @xmath63 feature bandpass , and @xmath64 the flux of the pseudo - continuum within two defined bandpasses : blue @xmath65 and red @xmath66 . ] , as the indicators of the ages of the stellar populations of the galaxies ( e.g. , heckman et al . 2004 ; kauffmann et al . 2003 ; kauffmann & heckman 2008 ; kewley et al . 2006 ; wang & wei 2008 , 2010 ; wild et al . both indices are measured in the removed stellar component for each spectrum , and are reliable age indicators until a few gyr after the onset of a burst ( e.g. , kauffmann et al . 2003 , bruzual & charlot 2003 ) . the sub - sample of partially obscured agns associated with broad h@xmath24 emission lines ( i.e. , seyfert 1.8/1.9 galaxies ) is selected from the parent agn and transition galaxy samples . although the existence of the broad h@xmath24 emission is direct indicative of the accretion activity of the central smbh , the underlying agn s continuum in these partially obscured agns could potentially decrease the measured two indices . these partially obscured agns are therefore removed from the subsequent analysis when the two indices are involved . following our previous studies ( wang & wei 2008 ) , the broad - line agns are selected by the means of the high - velocity wings of the broad h@xmath24 components on their blue side after the stellar features are removed from the spectra . the red wing is not adopted in the selection because of the superposition of the strong [ ] @xmath16583 emission line . the partially obscured agns are at first automatically selected by the criterion @xmath67 , where @xmath68 is the specific flux of the line wing averaged within the wavelength range from 6500 to 6350 in the rest - frame , and @xmath69 is the standard deviation of the continuum flux within the emission - line free region ranging from @xmath70 and @xmath71 . the automatically selected agns are then inspected one by one by eyes . in total , there are 174 broad - line seyfert galaxies / liners and 55 broad - line transition galaxies . the mpa / jhu catalog contains many duplicates because of the repeat observations . the duplicates allow us to roughly estimate the uncertainties for the line shape parameters and the stellar population age indicators . the duplicates are reduced and measured by the same method as described above . for the three main sub - samples ( i.e. , agns , transition galaxies and star - forming galaxies ) , the uncertainties are estimated to be @xmath72 and @xmath73 for the shape parameters @xmath45 and @xmath47 , respectively . the two lick indices have uncertainties : @xmath74 , and @xmath75h@xmath76 . for the comparison sample , the duplicates provide the uncertainties of @xmath77 and @xmath78 for the parameters @xmath45 and @xmath47 , respectively . after obtaining all the required parameters , the relation between the [ ] line profiles and the stellar population properties is examined in this section . the [ ] line profile is compared between the agns and the transition galaxies in this section . to avoid the possible systematics caused by the host galaxy properties , we compare the line profiles in a sample of matched galaxy pairs ( e.g. , kauffmann et al . the matched pairs are created according to the sdss dr4 mpa / jhu agn catalog . the catalog provides the measurements of galaxy properties for the agns that are classified according to the demarcation line given in kauffmann et al . we select the pairs that have @xmath79 , @xmath80 , @xmath81 , @xmath82 , and @xmath83 , where @xmath84 is the stellar mass in unit of @xmath85 , @xmath86 is the effective stellar mass - density ( @xmath87 , where @xmath88 is the half - light radius in the @xmath89-band ) in unit of @xmath90 , @xmath91 is the concentration index defined as @xmath92 the ratio of the radius enclosing 90% of the total flux to that enclosing 50% of the flux in the @xmath93-band , @xmath94 is the stellar velocity dispersion and @xmath89 is the redshift . because there are much more agns than the transition galaxies in our sample , a large fraction of the transition galaxies have a few of agn pairs . in these cases , one agn is assigned to a pair by a random selection , and the final distributions are built from 100 monte - carlo iterations . finally , there are totally 163 distinct galaxy pairs , in which no galaxy is assigned to a pair more than once . the left panel in figure 2 compares the cumulative distributions of the [ ] line widths in terms of the calculated velocity dispersions . the agns and transition galaxies are shown by the red - solid and green - dotted lines , respectively . the error - bars are resulted from our iterations . one can see clearly that the line profiles of the agns are systematically wider than that of the transition galaxies . a two - side kolmogorov smirnov test yields a maximum discrepancy of 0.282 with a corresponding probability that the two samples match of @xmath95 . in fact , the comparison of the parameter @xmath47 indicates that the difference in the distributions is likely due to the fact that the transition galaxies have systematically smaller values of @xmath47 than the agns ( see section 4.3 for more details ) . that means the transition galaxies are associated with weaker [ ] broad components compared to the agns , which implies that the feedback caused by the material outflow is less stronger in the transition galaxies than in the agns . the distributions of the [ ] @xmath0 relative velocity shifts defined in section 3.2.2 are compared in the right panel of figure 2 between the agns and transition galaxies . the symbols are the same as the left panel . a two - side kolmogorov smirnov test shows that the two samples are matched at a probability of 77% , although a careful examination suggests that the agns marginally tend to be associated with larger negative velocity shifts compared with the transition galaxies . the left - bottom panel in figure 3 shows the @xmath47 versus @xmath45 diagram . the agns , transition galaxies and star - forming galaxies are symbolized by the red , green and blue points , respectively . the over - plotted contours present the distribution of the star - forming galaxies in the comparison sample . the two crosses at the left - bottom corner show the typical uncertainties for both parameters as roughly estimated from the duplicates . the red - solid one indicates the uncertainties for the large line - width objects , and the black - dashed one for the comparison sample . the left - upper panel and the right - bottom panel shows the distributions of the parameters @xmath45 and @xmath47 for all the sub - samples ( including the comparison sample ) , respectively . the distributions are color - codded for the sub - samples as same as the left - bottom panel . as expected from an emission - line profile dominated by the instrumental resolution , the distribution of the comparison sample is strongly clustered around the point ( i.e. , @xmath46 and @xmath48 ) associated with a pure gaussian function . the last column in table 1 lists the average and median values of @xmath45 ( @xmath96 and @xmath97 ) for the comparison sample . the same values are listed in the last column in table 2 but for the parameter @xmath47 ( @xmath98 and @xmath99 ) . the dispersions is calculated to be 0.14 and 0.43 for the parameters @xmath45 and @xmath47 , respectively . the distribution therefore validates our spectral measurements because the distribution of the comparison sample is highly consistent with the prediction of a pure gaussian function . similar as the comparison sample , the statistics are tabulated in table 2 and 3 for both agns and transition galaxies as well . one can find from the tables that the line profiles of these galaxies systematically deviate from that of the control sample ( i.e. , deviate from a pure gaussian profile ) by not only smaller @xmath45 ( i.e. , a stronger blue asymmetry ) , but also larger @xmath47 ( i.e. , a stronger broad base ) . the main panel in figure 3 shows that both agns and transition galaxies form a sequence starting from the pure gaussian region to the upper - left corner . in fact , this phenomenon is qualitatively in agreement with the fact that two gaussian components , one representing the narrow line core and the other representing the blueshifted broad wing , are usually required to model the observed [ ] line profiles . a minor fraction of the objects deviate from the sequence by their positive values of @xmath45 . by inspecting the spectra of these objects one by one by eyes , we find that the deviations are mainly resulted from the contamination at the [ ] red wing caused by the weak low - ionization emission line @xmath15016 ( veron et al . 2002 ) . it is interesting to see that the star - forming galaxies with @xmath100 follow the sequence as well , although most of them have gaussian line profiles . in order to validate the deviations in these galaxies , the spectra with @xmath101 are then inspected one by one by eyes again . to understand the nature of these star - forming galaxies with blue asymmetric [ ] emission lines , we need to carefully model all the strong emission lines in individual spectrum . the issue is out of the topic of this paper , and will be examined in our subsequent studies . the results from the two - sides kolmogorov - smirnov tests are tabulated in table 3 and 4 for the parameters @xmath45 and @xmath47 , respectively . each entry contains the maximum absolute discrepancy between the used two sub - samples , and the corresponding probability that the two sub - samples match . although the agns and the transition galaxies show similar distributions for @xmath45 , the agns significantly differ from the transition galaxies in the distribution of parameter @xmath47 . the fraction of objects with strong broad [ ] wings is larger in the agns than in the transition galaxies previous studies frequently identified a strong correlation between the [ ] line widths and the [ ] blueshifts ( e.g. , bian et al . 2005 ; komossa et al . 2008 ; zamanov et al . the correlation is commonly explained by the material outflows . the similar correlations are also identified in iron coronal lines ( erkens et al . 1997 ) and optical complex ( hu et al . 2008 ) in typical type i agns . the left panel in figure 4 plots the velocity shift as a function of the line width for the agns ( red points ) and the transition galaxies ( green points ) . spearman rank - order tests show a marginal anti - correlation between the velocity shifts and the line widths for the agns ( with a spearman correlation coefficient @xmath102 at a significance level @xmath103 , where @xmath104 is the probability that there is no correlation between the two variables . ) , and , however , a stronger correlation for the transition galaxies ( with @xmath105 and @xmath103 ) . as an additional examination , the middle panel in the figure plots the velocity shift against the parameter @xmath45 . strong correlations between the two variables can be identified for both agns ( @xmath106 and @xmath103 ) and transition galaxies ( @xmath107 and @xmath103 ) , which means that larger the blueshift , stronger the blue asymmetry . one can see from the correlations that the objects with the largest velocity shifts deviate from the correlations . these objects instead show symmetrical line profiles . in order to further describe their line profiles , the right panel in the figure plots the velocity shift against the parameter @xmath47 . we find that these [ ] lines have `` boxy '' line profiles rather than peaked ones . in fact , a `` boxy '' line profile could be reproduced by the sum of two or more ( distinct ) peaks with comparable fluxes and line widths ( see examples of spectra in the figure 1 in greene & ho 2005 ) . the evolution effect on the [ ] line profile is examined in this section by using the two indices , @xmath57 and h@xmath108 , as the indicators of the ages of the circumnuclear stellar populations . as described in section 3.3 , the partially obscured agns are excluded from the analysis throughout this section . the two indices are plotted against the parameter @xmath45 ( the left panels ) and against the relative velocity shift ( the right panels ) for the agns in figure 5 . at the first glance of the four panels , we fail to directly identify any significant correlation between the line profile parameters and the stellar population ages . to examine the evolution effect in more details , we divide the agn sample into two groups according to their @xmath45 values : one group for the galaxies with @xmath109 and the another one for the galaxies with @xmath101 . the distributions of the @xmath57 values are compared between the two groups in the upper - left panel of figure 7 . the comparison indicates a significant difference between the two groups . ] . the agns with larger amount of blue asymmetry are systematically associated with younger stellar populations . a two - sides kolmogorov smirnov test indicates that the difference between the two distributions is at a significance level @xmath110 with a maximum absolute discrepancy of 0.22 . similar as the parameter @xmath45 , the agn sample is instead separated into two groups by the relative velocity shift at @xmath111 . we identify a marginal trend that larger the blue velocity shifts , younger the associated stellar populations ( see the bottom - left panel in figure 7 ) . the same statistical test yields a probability that the two groups match of @xmath112 ( with a maximum absolute discrepancy of 0.11 ) . figure 6 is the same as figure 5 but for the transition galaxies . the same methods as the agn sample are adopted to divide these galaxies into two groups . the right panels in figure 7 compare the distributions of @xmath57 between the two groups . the two - sides kolmogorov - smirnov tests suggest that the two distributions match at the probabilities @xmath113 for the @xmath45 separation and @xmath114 for the @xmath56 separation , which is much less significant than the agn sample . in fact , these results are easily understood because the transition galaxies are usually systematically younger than the agns ( see also in kewley et al . 2006 ; schawinski et al . 2007 ; wang & wei 2008 ) . in summary , our results indicate that the agns associated with young stellar populations show a wide range in their [ ] line profiles that vary from a blue asymmetrical shape to a gaussian function . in contrast , the agns associated with old stellar populations always show symmetrical line profiles . the evolution of [ ] @xmath15007 emission - line profile is studied by using a large sample of narrow emission - line galaxies selected from the mpa / jhu sdss dr7 catalog . the line - profile parameterizing allows us to reveal a trend that agns with more significant blue asymmetries tend to be associated with younger stellar populations . at the beginning of discussion , we argue that the trend is not driven only by the orientation effect ( antonucci 1993 ; elitzur 2007 ) . it is widely accepted that the commonly observed blue asymmetrical and blue - shifted [ ] line profiles are caused by the wind - driven nlr outflows . the orientation - driven scenario therefore results in an odd corollary that a fraction of the agns with young stellar populations have outflow directions closer to the line - of - sight of observers than the agns with old stellar populations , while the starlight components from the host galaxies are believed to be isotropic . the revealed trend can be alternatively driven by the activity of the central smbh . in fact , there is many observational evidence supporting that the [ ] line blue asymmetries and blueshifts are correlated with @xmath9 or eigenvector - i space ( e.g. , boroson & green 1992 ; boroson 2002 ) . ] in typical type i agns ( see the description and citations in section 1 ) . the relationship between the [ ] line profile and @xmath9 is examined in the current narrow - line agn sample to present a complete understanding . similar as the previous studies of narrow - line agns , the parameter @xmath115}}/\sigma_*^4 $ ] ( e.g. , heckman et al . 2004 ) is used as a proxy of @xmath116 , where @xmath115}}$ ] and @xmath94 is the [ ] @xmath0 line luminosity and the bulge velocity dispersion , respectively . the transition galaxies are excluded from the subsequent calculations because of the possible contamination caused by regions . the bolometric luminosity @xmath117 is transformed from @xmath115}}$ ] through the bolometric correction @xmath118}}\approx3500 $ ] ( see heckman et al . ( 2004 ) for details ) . by assuming the balmer decrement for the standard case b recombination and the galactic extinction curve with @xmath26 , the [ ] line luminosity is corrected for the local extinction that is inferred from the narrow - line ratio @xmath119 . the blackhole mass is estimated from the @xmath2 calibration : @xmath120 ( tremaine et al . 2002 ) , in which the velocity dispersion is measured for each spectrum through our pca fittings . the galaxies with @xmath121 are removed from the analysis because the sdss instrumental resolution is @xmath122 . the results are shown in figure 8 . the left panel presents an anti - correlation between @xmath115}}/\sigma_*^4 $ ] and @xmath57 . larger the eddington ratio , and younger the stellar population , which is consistent with the previous studies ( e.g. , kewley et al . 2006 ; kauffmann et al . the middle panel plots @xmath115}}/\sigma_*^4 $ ] as a function of the parameter @xmath45 . when compared with figure 5 , one can find a similar trend in the @xmath115}}/\sigma_*^4 $ ] vs. @xmath45 plot as that in the @xmath57 vs. @xmath45 plane . agns with large amount of blue asymmetry tend to have large @xmath9 . a much stronger trend can be identified in the right panel between @xmath115}}/\sigma_*^4 $ ] and the parameter @xmath47 . the trend indicates a fact that agns with larger eddington ratios tend to be associated with stronger [ ] broad components . however , we fail to find a significant trend between @xmath115}}/\sigma_*^4 $ ] and the bulk relative velocity shifts @xmath56 . spearman rank - order tests are performed to show either the eddington ratio or the stellar population is intrinsically related with the line asymmetry . the resulted correlation coefficient matrix is listed in table 5 . all the entries in the table have a probability of null correlation @xmath103 . although both @xmath123 and @xmath57 are correlated with @xmath45 at comparable significance levels , @xmath47 is more strongly correlated with @xmath123 than with @xmath57 . the stronger correlation therefore indicates that the trend between the line asymmetry and @xmath57 is likely driven physically by the evolution of the eddington ratio ( see below ) . the trend that agns with more significant blue asymmetry tend to be associated with younger stellar populations provides a piece of direct evidence supporting the co - evolution of the nlr kinematics and the growth of the host galaxy . agns with stronger outflows tend to be at their earlier evolutionary stage as inferred from the stellar population ages . the analysis in the above section further suggests that the trend is likely driven by @xmath9 . the important evolutionary role of @xmath9 has been frequently proposed in recent studies ( e.g. , wang & wei 2008 , 2010 ; kewley et al . 2006 ; heckman et al . 2004 ) . putting these pieces together yields an improved co - evolution scenario in which agn likely evolve from a high-@xmath124 state with strong outflow to a low-@xmath124 state with weak outflow as the circumnuclear stellar population continually ages . this evolutionary scenario is consistent with the current understandings of agns . lieghly et al . ( 1997 ) reported evidence of relativity outflows in three nls1s . in comparison with typical broad - line seyfert galaxies , nls1s are `` young '' agns ( mathur 2000 ) typical of smaller @xmath3 , higher @xmath124 , larger [ ] emission - line asymmetries , younger circumnuclear stellar populations possibly at post - starburst phase , and enhanced star formations as recently revealed by the observation of _ spitzer ( e.g. , boroson 2005 ; zamanov et al . 2002 ; boroson & green 1992 ; wang et al . 1996 ; boller et al . 1996 ; sulentic et al . 2000 ; boroson 2002 ; xu et al . 2007 ; zhou et al . 2005 ; wang & wei 2006 ) . in addition to the observational ground , the feedback from central agn is frequently involved in the modern numerical and sim - analytical galaxy evolution models to quench the star formation and to blow the gas or dust away through the powerful agn s wind ( e.g. , springel et al . 2005 ; di matteo et al . 2005 ; kauffmann & heckman 2008 ; fabian 1999 ; hopkins et al . 2005 , 2008a , b ; croton et al . 2006 ; somerville et al . 2008 ; khalatyan et al . 2008 ) . the numerical simulation done by hopkins et al . ( 2005 ) suggests a scenario that the powerful agn s wind is required to occur in the young agn phase in which the central smbh is heavily obscured by the starforming gas and dust . _ transition galaxies are the narrow emission - line galaxies located between the theoretical and empirical demarcation lines in the [ ] /h@xmath7 versus [ ] /h@xmath24 diagnostic diagram ( kauffmann et al . 2003 ; kewley et al . their emission - line properties are explained by the mixture of the contributions from both star formations and typical agns ( e.g. , kewley et al . 2006 ; ho et al . 1993 , see recent review in ho 2008 ) . our spectral profile analysis indicates that the transition galaxies differs from the seyfert galaxies in their [ ] emission - line profiles . the transition galaxies systematically show weaker [ ] broad wings , and narrower [ ] line widths than the seyfert galaxies , which implies that the two kinds of objects are different from each other in their mass outflow kinematics . the line profile comparison allows us to suspect that the transition galaxies `` bridge '' the agn - starburst co - evolution from early starburst - dominated phase to late agn - dominated phase ( e.g. , yuan et al . 2009 ; wang & wei 2006 ; schawinski et al . 2007 , 2009 ) . the evolutionary role of the transition galaxies has been reported in the recent studies by many authors . with the large sdss spectroscopic database , transition galaxies are found to be systematically associated with younger stellar populations than seyfert galaxies ( e.g. , kewley et al . 2006 ; wild et al . 2007 ; and also see figure 6 in this paper ) . westoby et al . ( 2007 ) found that the distribution of the h@xmath24 equivalent widths of the transition galaxies peaks at the valley between the red and blue sequences . by analyzing the mm - wavelength observation taken by the 30 m iram , schawinski et al . ( 2009 ) suggested that molecular gas reservoir is destructed by the agn feedback at early agn+starforming transition phase . we systematically examined the evolutionary issue of the [ ] @xmath0 emission - line profile by using a large sample of narrow emission - line galaxies selected from the mpa / jhu sdss dr7 value - added catalog . the sample is separated into three sub - samples ( i.e. , star - forming galaxies , transition galaxies and seyfert galaxies / liners ) basing upon the line ratios given in the catalog . two shape parameters , skewness ( @xmath45 ) and kurtosis ( @xmath47 ) , are additionally used to quantify the profile deviation from a pure gaussian . our analysis indicates that a ) the transition galaxies are systematically associated with narrower line widths and with weaker [ ] broad wings than the agns ; b ) the agns with stronger blue asymmetries tend to be associated with younger stellar populations . the evolutionary significance of the [ ] line profile suggests a co - evolution of the outflow feedback and the agn s host galaxy . we would like to thank the anonymous referees for his / her constructive comments in improving the paper . this study was supported by the national science foundation of china ( under grant 10803008 ) , and by the national basic research program of china ( grant 2009cb824800 ) . this study uses the sdss archive data that was created and distributed by the alfred p. sloan foundation . a toy model of emission - line profile is constructed here to illustrate the effect of the instrumental resolution on the measured line shape parameters . the model is constructed by assuming an observed [ ] emission line profile that is prescribed by the sum of two pure gaussian profiles : @xmath125 , where @xmath126 is a normalized gaussian distribution with the peak at zero and the variance of @xmath127 , @xmath128 is the same distribution but with the peak shifted at @xmath129 and the variance of @xmath130 , and @xmath24 is a parameter much less than unit . with this two gaussian components model , @xmath126 represents the narrow core of the emission , and @xmath131 the shifted , low - contrast broad wing ( we require @xmath132 for avoiding double peaked profile ) . the normalizations of the two distributions yield @xmath133 . the first moment ( i.e. , the line centroid ) is written as @xmath134 , which is not affected by the instrumental resolution . with the first moment , the high - order moments are written as @xmath135 the second moment of the total line profile is therefore inferred to be approximately the linear combination of the square of the second moments of the two distributions : @xmath136 . in the context of the toy model , the linear combination means that the instrumental resolution @xmath23 can be removed from the measured values according to the equation @xmath137 , which means that the removal depends on the relative strength of the low - contrast gaussian profile , i.e. , the parameter @xmath24 . the third moment of the total profile can be trivially obtained through the integration @xmath138 . by ignoring the terms with high - orders of @xmath24 , the third moment approximately equals to @xmath139 . the result suggests that the measured third moment depends not only on the relative shift of the low - contrast gaussian profile , but also on the relative strength denoted by the parameter @xmath24 . although the second term in the above result means that the third moment does not depend on the instrumental resolution in our toy model at the first order approximation : @xmath140 , the definition of the dimensionless shape parameter @xmath45 introduces a non - linear dependence on the instrumental resolution in the denominator ( see eq . 2 in the main text ) . by adopting the same approach , the fourth moment is calculated to be @xmath141 , where one sees that the second and third terms introduce again a non - linear dependence on the instrumental resolution . with this toy model , the galaxies whose [ ] @xmath15007 lines are narrower than 2@xmath23 are then dropped out from our statistical analysis in order to avoid the effect of the instrumental resolution on the measured line shape parameters . abazajian , k. n. , et al . 2009 , , 182 , 543 aoki , k. , kawaguchi , t. , & ohta , k. 2005 , , 618 , 601 antonucci , r. r. j. 1993 , , 31 , 473 arribas , s. , mediavilla , e. , & garcia - lorenzo , b. 1996 , , 463 , 509 baldwin , j. a. , phillips , m. m. , & terlevich , r. 1981 , pasp , 93 , 5 balogh , m. l. , et al . , 1999 , , 527 , 54 bian , w. h. , yuan , q. r. , & zhao , y. h. 2005 , , 364 , 187 binney , j. , & merrifield , m. 1998 , galactic astronomy , princeton , new jersey boller , t. , brandt , w. n. , & fink , h. 1996 , , 305 , 53 boroson , t. a. 2002 , , 565 , 78 boroson , t. a. 2003 , , 585 , 647 boroson , t. a. 2005 , , 130 , 381 boroson , t. a. , & green , r. f. 1992 , , 80 , 109 bromley , b. c. , et al . 1998 , , 505 , 25 bruzual , a. g. 1983 , , 273 , 105 bruzual , g. , & charlot , s. 2003 , , 344 , 1000 cardelli , j. a. , clayton , g. c. , & mathis , j. s. , 1989 , , 345 , 245 christopoulou , p. e. , holloway , a. j. , steffen , w. , mundell , c. g. , thean , a. h. c. , goudis , c. d. , meaburn , j. , & pedlar , a. 1997 , , 284 , 385 crenshaw , d. m. , & kraemer , s. 2000 , , 532 , 101 crenshaw , d. m. , et al . 2000 , , 120 , 1731 crenshaw , d. m. , schmitt , h. r. , kraemer , s. b. , mushotzky , r. f. , & dunn , j. p. 2010 , , 708 , 419 croton , d. j. , et al . 2006 , , 365 , 11 das , v. , et al . 2005 , , 130 , 945 das , v. , crenshaw , d. m. , kraemer , s. b. , & deo , r. p. 2006 , , 132 , 620 das , v. , crenshaw , d. m. , & kraemer , s. b. 2007 , , 656 , 699 davies , r. , mueller sanchez , f. , genzel , r. , tacconi , l. , hicks , e. , friedrich , s. , & sternberg , a. 2007 , , 671 , 1388 de robertis , m. m. , & osterbrock , d. e. 1984 , , 286 , 171 di matteo , t. , springel , v. , & hernquist , l. 2005 , nature , 433 , 604 elitzur , m. 2007 , in asp conf . 373 , the central engine of active galactic nuclei , ed . l. c. ho & j. -m . wang ( san francisco , ca : asp ) , 415 erkens , u. , appenzeller , i. , & wagner , s. 1997 , , 323 , 707 fabian , a. c. 1999 , , 308 , l39 ferrarese , l. , & merritt , d. 2000 , , 539 , 9 fischer , t. c. , crenshaw , d. m. , kraemer , s. b. , schmitt , h. r. , mushotsky , r. f. , & dunn , j. p. 2011 , , 727 , 71 fischer , t. c. , crenshaw , d. m. , kraemer , s. b. , schmitt , h. r. , & trippe , m. l. 2010 , , 140 , 577 gebhardt , k. et al . 2000 , , 539 , 13 glazebrook , k. , offer , a. , r. , & deeley , k. 1998 , , 492 , 98 granato , g. l. , de zotti , g. , silva , l. , bressan , a. , & danese , l. 2004 , , 600 , 580 greene , j. e. , & ho , l. c. 2005 , , 627 , 721 grupe , d. , beuermann , k. , mannheim , k. , & thomas , h , -c . 1999 , , 350 , 805 grupe , d. , & mathur , s. 2004 , , 606 , 41 haring , n. , & rix , h. w. 2004 , , 604 , 89 hasinger , g. , miyaji , t. , & schmidt , m. 2005 , , 441 , 417 heckman , t. m. 1981 , , 250 , 59 heckman , t. m. , & kauffmann , g. 2006 , newar , 50 , 677 heckman , t. m. , kauffmann , g. , brinchmann , j. , et al . 2004 . , 613 , 109 ho , l. c. 2008 , , 46 , 475 ho , l. c. , filippenko , a. v. , & sargent , w. l. w. 1993 , , 417 , 63 hopkins , p. f. 2011 , astro - ph / arxiv:1101.4230 hopkins , p. f. , hernquist , l. , martini , p. , cox , t. j. , robertson , b. , di matteo , t. , & springel , v. 2005 , , 625 , 71 hopkins , p. f. , bundy , k. , hernquist , l. , & ellis , r. s. 2007 , , 659 , 976 hopkins , p. f. , cox , t. j. , keres , d. , & hernquist , l. 2008a , , 175 , 390 hopkins , p. f. , hernquist , l. , cox , t. j. , & keres , d. 2008b , , 175 , 356 hu , c. , wang , j. m. , ho , l. c. , chen , y. m. , zhang , h. t. , bian , w. h. , & xue , s. j. 2008 , , 687 , 78 hutchings , j. b. , et al . 1998 , , 492 , 115 kaiser , m. e. , et al . 2000 , , 528 , 260 kauffmann , g. , & heckman , t. m. 2009 , , 2009 , 397 , 135 kauffmann , g. , heckman , t. m. , tremonti , c. , et al . 2003a , , 346 , 1055 kauffmann , g. , heckman , t. m. , white , s. d. m. , et al . 2003b , , 341 , 54 kauffmann , g. , heckman , t. m. , white , s. d. m. , et al . 2003c , , 341 , 33 kauffmann , g. , heckman , t. m. , de lucia , g. , brinchmann , j. charlot , s. , tremonti , c. , white , s , d. m. , & brinkmann , j. 2006 , , 367 , 1394 kewley , l. j. , dopita , m. a. , sutherland , r. s. , heisler , c. a. , & trevena , j. 2001 , , 556 , 121 kewley , l. j. , groves , b. , kauffmann , g. , & heckman , t. 2006 , , 372 , 961 khalatyan , a. , cattaneo , a. , schramm , m. , gottlber , s. , steinmetz , m. , & wisotzki , l. 2008 , , 387 , 13 komossa , s. , & xu , d. 2007 , , 677 , 33 komossa , s. , xu , d. , zhou , h. , storchi - bergmann , t. , & binette , l. 2008 , , 680 , 926 leighly , k. m. , mushotzky , r. f. , nandra , k. , & forster , k. 1997 , , 489 , 25 magorrian , j. , et al . 1998 , , 115 , 2285 mathur , s. 2000 , , 314 , l17 mathur , s. , & grupe , d. 2005 , , 633 , 688 marziani , p. , sulentic , j. w. , dultzin - hacyan , d. , calvani , m. , & moles , m. 1996 , , 104 , 37 marziani , p. , sulentic , j. w. , zamanov , r. , calvani , m. , dultzin - hacyan , d. , bachev , r. , & zwitter , t. 2003 , , 145 , 199 nandra , k. , laird , e. s. , & steidel , c. c. 2005 , , 360 , l39 nelson , c. h. 2000 , , 544 , 91 nelson , c. , plasek , a. , thompson , a. , gelderman , r. , & monroe , t. 2004 , aspc , 311 , 83 nelson , c. h. , weistrop , d. , hutchings , j. b. , crenshaw , d. m. , gull , t. r. , kaiser , m. e. , kraemer , s. b. , & lindler , d. 2000 , , 531 , 257 nelson , c. h. , & whittle , m. 1996 , , 465 , 96 osterbrock , d. e. astrophysics of gaseous nebulae and active galactic nuclei , ( mill valley : univ . science books ) phillips , m. m. 1976 , , 208 , 37 ruiz , j. r. , crenshaw , d. m. , kraemer , s. b. , bower , g. a. , gull , t. r. , hutchings , j. b. , kaiser , m. e. , & weistrop , d. 2001 , , 122 , 2961 schawinski , k. , thomas , d. , sarzi , m. , maraston , c. , kaviraj , s. , joo , s. , yi , s. k. , & silk , j. 2007 , , 382 , 1415 schawinski , k. , virani , s. , simmons , b. , urry , c. m. , treister , e. , kaviraj , s. , & kushkuley , b. 2009 , , 692 , 19 schlegel , d. , finkbeiner , d. p. , davis , m. 1998 , , 500 , 525 schlesinger , k. , pogge , r. w. , martini , p. , shields , j. c. , & fields , d. 2009 , , 699 , 857 sergeev , s. g. , pronik , v. i. , malkov , yu . f. , & chuvaev , k. k. 1997 , , 320 , 405 shankar , f. , bernardi , m. , & haiman , z. 2009 , , 694 , 867 silverman , j. d. , et al . 2008 , , 679 , 118 somerville , r. s. , hopkins , p. f. , cox , t. j. , robertson , b. e. , & hernquist , l. 2008 , , 391 , 481 storchi - bergmann , t. , wilson , a. s. , & baldwin , j. a. 1992 , , 396 , 45 spergel , d. n. , et al . 2003 , , 148 , 175 springel , v. , white , s. d. m. , jenkins , a. , et al . 2005 , nature , 435 , 629 sulentic , j. w. , marziani , p. , & dultzin - hacyan , d. 2000 , , 38 , 521 tadhunter , c. , wills , k. , morganti , r. , oosterloo , t. , & dickson , r. 2001 , , 327 , 227 tremaine , s. , gebhardt , k. , bender , r. , et al . 2002 , , 574 , 740 tremonti , c. a. , et al . 2004 , , 613 , 898 veilleux , s. , & osterbrock , d. e. 1987 , , 63 , 295 veilleux , s. , cecil , g. , & bland - hawthorn , j. 2005 , , 43 , 769 veron , p. , goncalves , a. c. , & veron - cetty , m. -p . 2002 , , 384 , 826 wang , j. , & wei , j. y. 2006 , , 648 , 158 wang , j. , & wei , j. y. 2008 , , 679 , 86 wang , j. , & wei , j. y. 2009 , , 696 , 741 wang , j. , & wei , j. y. 2010 , , 719 , 1157 wang , j. , wei , j. y. , & he , x. t. 2005 , newa , 10 , 353 wang , j. , wei , j. y. , & he , x. t. 2006 , , 638 , 106 wang , t. , brinkmann , w. , & bergeron , j. 1996 , , 309 , 81 wang , t. , & lu , y. 2001 , , 377 , 52 westoby , p. b. , mundell , c. g. , & baldry , i. k. 2007 , , 382 , 1514 whittle , m. 1985 , , 213 , 1 whittle , m. 1992a , , 387 , 109 whittle , m. 1992b , , 387 , 121 wilson , a. s. , & heckman , t. m. 1985 , in astrophysics of active galaxies and quasi - stellar objects , ed . j. s. miller ( mill valley : university science books ) wild , v. , et al . 2007 , , 381 , 543 wild , v. , heckman , t. m. , & charlot , s. 2010 , , 405 , 933 worthey , g. , & ottaviani , d. l. 1997 , , 111 , 377 xu , d. w. , & komossa , s. 2009 , , 705 , 20 xu , d. w , komossa , s. , zhou , h. y. , wang , t. g. , & wei , j. y. 2007 , , 670 , 60 york , d. g. , et al . 2000 , , 120 , 1579 yu , q. j. , & tremaine , s. 2002 , , 335 , 965 yuan , t. -t . , kewley , l. j. , & sanders , d. b. 2010 , , 709 , 884 zamanov , r. , marziani , p. , sulentic , j. w. , calvani , m. , dultzin - hacyan , d. , & bachev , r. 2002 , , 576 , 9 zhou , h. y. , wang , t. g. , dong , x. b. , wang , j. , & lu , h. 2005 , mem . soc . astron . italiana , 76 , 93 ccccc property & agns & transitions & sf galaxies/@xmath142 & sf galaxies/@xmath143 + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) + numbers & 2333 & 793 & 190 & 5843 + @xmath144 & -0.167 & -0.170 & -0.09 & 0.01 + @xmath145 & -0.165 & -0.167 & -0.08 & 0.01 + @xmath146 & 0.31 & 0.28 & 0.32 & 0.14 + ccccc property & agns & transitions & sf galaxies/@xmath142 & sf galaxies/@xmath143 + ( 1 ) & ( 2 ) & ( 3 ) & ( 4 ) & ( 5 ) + numbers & 2333 & 793 & 190 & 5843 + @xmath147 & 3.00 & 2.79 & 2.84 & 2.61 + @xmath148 & 2.83 & 2.64 & 2.56 & 2.58 + @xmath149 & 0.70 & 0.56 & 1.01 & 0.43 + cccc sub - sample & agns & transitions & sf galaxies/@xmath142 + & ( 1 ) & ( 2 ) & ( 3 ) + agns & - & 0.034(0.48 ) & 0.159(@xmath150 ) + transitions & - & - & 0.164(@xmath151 ) + sf galaxies/@xmath142 & - & - & - + cccc sub - sample & agns & transitions & sf galaxies/@xmath142 + & ( 1 ) & ( 2 ) & ( 3 ) + agns & - & 0.154(@xmath152 ) & 0.214(@xmath153 ) + transitions & - & - & 0.133(@xmath154 ) + sf galaxies/@xmath142 & - & - & - +

we derive new bounds on decaying dark matter from the gamma ray measurements of ( i ) the isotropic residual ( extragalactic ) background by and ( ii ) the fornax galaxy cluster by . we find that those from ( i ) are among the most stringent constraints currently available , for a large range of dm masses and a variety of decay modes , excluding half - lives up to to few seconds . in particular , they rule out the interpretation in terms of decaying dm of the spectral features in , and , unless very conservative choices are adopted . we also discuss future prospects for bounds from fornax which , contrary to the present constraints of ( ii ) , may allow for an interesting improvement and may become better than those from the current or future extragalactic data . * * * marco cirelli* , * emmanuel moulin* , * paolo panci* , + * pasquale d. serpico* , * aion viana* _http://ph - dep - th.web.cern.ch / ph - dep - th/[cern theory division ] , ch-1211 genve , switzerland _ + _http://ipht.cea.fr / en / index.php[institut de physique thorique ] , cnrs , ura 2306 & cea / saclay , f-91191 gif - sur - yvette , france _ + _http://irfu.cea.fr / en/[irfu / dsm ] , cea saclay , f-91191 gif - sur - yvette , france _ + _http://cp3-origins.dk/[cp3-origins ] & the danish institute for advanced study dias , + university of southern denmark , campusvej 55 , dk-5230 odense m , denmark _ + _http://lapth.in2p3.fr/[lapth ] , univ . de savoie , cnrs , b.p.110 , annecy - le - vieux f-74941 , france _

1205.5283

the possibility that dark matter ( dm ) , which constitutes most of the matter in the universe , consists of a particle that actually decays on a very long time scale has attracted much attention lately . this is because the decay time scale @xmath12 can be taken to be ` short ' enough that the decay products give signals in current high energy cosmic ray experiments . namely , if @xmath13 sec , decaying dm can be invoked to explain the excesses in the fluxes of positrons and electrons measured by , and @xcite . on the other hand , this value of @xmath12 is so much longer than the age of the universe that the slow decay does not make a dent in the overall cosmological dm abundance and does not spoil the agreement with a number of astrophysical and cosmological observations @xcite . irrespectively of this recent activity spurred by the charged cr anomalies , decaying dm is an interesting subject by itself . a long time studied case is the one of supersymmetric gravitino dm @xcite , which is unstable due to r - parity violation . more generically , in several particle physics models , high - scale suppressed operators may naturally mediate the decay of dm ( for a miscellaneous list of references see @xcite ) . from the phenomenological point of view , the main feature of decaying dm with respect to the ` more traditional ' annihilating dm is that it is less constrained by neutral messenger probes ( essentially gamma rays , but also neutrinos ) originating from dense dm concentrations such as the galactic center , the galactic halo or nearby galaxies . the reason is simple and well - known : while the signal originating from annihilating dm is proportional to the square of the dm density , for decaying dm the dependence is on the first power ; as a consequence , dense dm concentrations shine above the astrophysical backgrounds if annihilation is at play , but remain comparatively dim if dm is decaying . decaying dm ` wins ' instead , generally speaking , when large volumes are considered . this is why in the following we will focus on targets as large as galaxy clusters or , essentially , the whole universe . on the observational side , the and telescopes are making unprecedented progress in the field of gamma - ray astronomy , producing measurements of many different targets including those of interest for decaying dm . it is therefore a good time to assess the current status of the latter . to this aim , we will compute the predicted signal from decaying dm , for a variety of decay channels ( limiting ourselves to 2-body channels , in order to remain as model - independent as possible ) , and compare it to the gamma - ray measurements , deriving constraints on the decay half - life . in particular , in this paper , we make use of two distinct probes : a. [ dataexg ] the isotropic residual gamma ray flux recently measured by @xcite , which now extends from about 200 mev up to 580 gev . the high - energy portion of this measurement is labelled as ` preliminary ' by the collaboration , so it should be used with care . however , we note that the data are based on solid premises , since they are obtained with the same procedure already published in @xcite with an enlargement of the dataset . we will anyway show later the effect of considering or not the preliminary portion of the data . b. [ datafornax ] the recent observation in gamma rays of the fornax galaxy cluster by @xcite . we choose these probes for a number of reasons , besides the obvious fact that they are among the most recent ones . for what concerns the isotropic flux ( [ dataexg ] ) , it is known since a long time that it represents a powerful testbed for decaying dm ( see for instance @xcite ) . for what concerns ( [ datafornax ] ) , we are again motivated by the fact that large virialized objects such as galaxy clusters are a promising target for decaying dm ( as briefly mentioned above ) and , more in particular , by the fact that indeed stringent constraints have been recently derived using data @xcite : on the basis of this we explore the constraining power of a complemetary observatory such as . this is particularly interesting because upcoming erenkov telescopes offer very promising prospects of improvement in the near future ( as opposed to a space based gamma ray observatory such as , which will at most increase its statistic by a factor of order 2 ) . indeed , we will also study the sensitivity of , the upcoming large erenkov telescope array @xcite . incremental improvements to the data and the tools relevant for our analysis also comes from other aspects : a. [ dataepem ] new data on charged cosmic rays ( cr ) , which have been recently presented . @xcite has measured the @xmath14 fraction , confirming the notorious rise exposed by in 2008 @xcite and extending the measured spectrum to larger energies ( about 200 gev ) ; has also presented updated measurements of the total flux of @xmath15 @xcite as well as the measurement of the separated fluxes of @xmath16 and of @xmath17 @xcite ; has published @xcite the measurement of the total flux of @xmath15 too ; finally , has presented results on the flux of pure @xmath17 @xcite , of @xmath18 @xcite and of @xmath19 @xcite . b. [ ewcorrections ] electroweak ( ew ) corrections to the dm generated fluxes . as discussed in a number of recent works @xcite , the emission of the ew gauge bosons ( @xmath20 , @xmath21 ) from the sm particles emerging from the dm decay process can significantly modify the phenomenology . ( i ) the corrections are particularly relevant for large dm masses ( above a tev ) ; ( ii ) they can alter significantly the fluxes of charged cosmic rays and gamma rays , both in their spectral shape and in their amplitude , affecting especially the low energies portion @xcite . we include ew corrections in all our computations . c. [ improvedprop ] an improved propagation scheme for @xmath2 in the galaxy . in semi - analytic treatments of charged cr propagation in the galactic environment , usually employed in dm studies , the energy losses of @xmath16 and @xmath17 are generally considered as space - indepedent and their behaviour with energy @xmath22 is approximated as @xmath23 . however , a space dependence obviously arises from the different conditions ( of magnetic field and of ambient light ) in different points of the galaxy and an energy dependence different from @xmath23 occurs in the full klein - nishina regime of inverse compton scattering . we make use of the fluxes of @xmath2 as provided in @xcite , which include these proper space- and energy - dependences . this impacts both the analysis of the @xmath2 fits ( mildly ) and the fluxes of ics gamma rays from the galactic halo ( more significantly ) . some recent studies have performed work related to ours @xcite . the most closely related , as far as the isotropic gamma ray flux is concerned , is @xcite , in which a subset of us derived constraints on decaying dm with the data and the tools available at that time . we improve , with respect to that work , by including the updated data in [ dataexg ] and by the points ( [ dataepem ] ) , ( [ ewcorrections ] ) and ( [ improvedprop ] ) discussed above . there are also some other minor analysis differences on which we will comment in the text . among other works , ref . @xcite made a particularly detailed analysis : we consider the more recent data in [ dataexg ] and [ datafornax ] above and we include the inverse compton scattering contribution to the gamma ray flux in the galactic component of [ dataexg ] ; the points ( [ dataepem ] ) , ( [ ewcorrections ] ) and ( [ improvedprop ] ) are new in our paper ; we also consider a few more decay channels . for the cases which overlap with this and other studies , we will present a comparison in sec . [ comparison ] . the rest of this paper is organized as follows . in sec . [ charged fits ] we update charged cr fits by including ( [ dataepem ] ) , ( [ ewcorrections ] ) and ( [ improvedprop ] ) . in sec . [ isotropic ] we discuss the calculation of the constraints from the isotropic residual @xmath24-ray flux , while in sec . [ fornax ] we discuss those from the fornax cluster . in sec . [ results ] we present the combined results . in sec . [ conclusions ] we present our conclusions . as mentioned in sec . [ introduction ] , the anomalous , and data in @xmath16 and @xmath15 have been interpreted in terms of dm decay . we recall here briefly the main features of the experimental data and of their dm interpretations , without entering in the details of any specific particle physics model . we use the following data sets : * positron fraction @xcite , selecting only points with @xmath25 gev , in order both to avoid the low energy region affected by the uncertainty of solar modulation and to have a consistent overlap with the positron fraction data points ( the low energy data have very small error bars that would overconstrain the fit ) . * positron fraction @xcite . * @xmath15 total flux @xcite , provided in the low energy ( le ) and high energy ( he ) samples . * @xmath15 total flux @xcite , also provided in a lower energy portion and a higher energy one . * @xmath15 total flux @xcite , which however consists of only 6 data points , with error bars larger than those of and at the same energy and therefore has no effect on the global fit . * @xmath19 flux @xcite . we perform the fit to these data using the dm generated fluxes as provided in @xcite , which include the features discussed in points ( [ ewcorrections ] ) and ( [ improvedprop ] ) above . in looking for the best fitting regions , we scan over the propagation parameters of charged cosmic rays and over the uncertainties on the slope and normalization of the astrophysical electron , positron and antiproton background . note that a simple power law is not expected to necessarily provide a good background model , especially given that the high energy part of the electron spectrum should be dominated by the contribution from local sources . on the other hand , due to the limited number of points available at high - energy and their relatively large statistical error , the background power - law flux is significantly constrained by the lower energy data and the assumption is sufficient in providing a good fit , when combined with the bumpy spectrum from dark matter . we have assumed a nfw profile for the galactic dm halo ( the specific parameters of which are discussed in sec . [ isotropic ] ) , but other choices would have given almost indistinguishable results . we refer to @xcite and @xcite for more details . [ fig : example ] presents the spectra results for one specific case : the best fit point for the decay mode dm@xmath26 . the allowed regions on the plane @xmath27@xmath12 , for the same channel , are instead shown in fig . [ fig : exclusionexample ] . we show the 95.45 % and 99.999 % c.l . regions for the fit to positron and antiproton data only ( green and yellow bands ) and for the whole datasets , i.e. including ( @xmath28 ) ( red and orange blobs ) . in fig . [ fig : exclusion ] , which we will discuss in more detail later , we show the global allowed regions for all the channels that we consider . an example of the signals in different channels from a decaying dm candidate that fits the charged cr anomalies : we show the positron fraction ( upper left ) , all leptons ( upper right ) , antiproton ( lower left ) and isotropic gamma ray ( lower right ) spectra for the best fit candidate dm @xmath29 , namely @xmath30 gev , @xmath31 sec ( marked by a white cross in fig.s [ fig : exclusionexample ] and [ fig : exclusion ] ) . the ` 2008 ' datapoints are reported for completeness , but we use only the more recent ` 2010 ' in the fits ( see text for details ) . in each panel , the dm contributions are dashed and the astrophysical background is shaded gray . in the lower right panel , the dotted ( dashed ) lines refer to the flux neglecting ( including ) extragalactic absorption , see sec . [ isotropic ] . _ , title="fig:",scaledwidth=50.0% ] an example of the signals in different channels from a decaying dm candidate that fits the charged cr anomalies : we show the positron fraction ( upper left ) , all leptons ( upper right ) , antiproton ( lower left ) and isotropic gamma ray ( lower right ) spectra for the best fit candidate dm @xmath29 , namely @xmath30 gev , @xmath31 sec ( marked by a white cross in fig.s [ fig : exclusionexample ] and [ fig : exclusion ] ) . the ` 2008 ' datapoints are reported for completeness , but we use only the more recent ` 2010 ' in the fits ( see text for details ) . in each panel , the dm contributions are dashed and the astrophysical background is shaded gray . in the lower right panel , the dotted ( dashed ) lines refer to the flux neglecting ( including ) extragalactic absorption , see sec . [ isotropic ] . _ , title="fig:",scaledwidth=50.0% ] + an example of the signals in different channels from a decaying dm candidate that fits the charged cr anomalies : we show the positron fraction ( upper left ) , all leptons ( upper right ) , antiproton ( lower left ) and isotropic gamma ray ( lower right ) spectra for the best fit candidate dm @xmath29 , namely @xmath30 gev , @xmath31 sec ( marked by a white cross in fig.s [ fig : exclusionexample ] and [ fig : exclusion ] ) . the ` 2008 ' datapoints are reported for completeness , but we use only the more recent ` 2010 ' in the fits ( see text for details ) . in each panel , the dm contributions are dashed and the astrophysical background is shaded gray . in the lower right panel , the dotted ( dashed ) lines refer to the flux neglecting ( including ) extragalactic absorption , see sec . [ isotropic ] . _ , title="fig:",scaledwidth=50.0% ] an example of the signals in different channels from a decaying dm candidate that fits the charged cr anomalies : we show the positron fraction ( upper left ) , all leptons ( upper right ) , antiproton ( lower left ) and isotropic gamma ray ( lower right ) spectra for the best fit candidate dm @xmath29 , namely @xmath30 gev , @xmath31 sec ( marked by a white cross in fig.s [ fig : exclusionexample ] and [ fig : exclusion ] ) . the ` 2008 ' datapoints are reported for completeness , but we use only the more recent ` 2010 ' in the fits ( see text for details ) . in each panel , the dm contributions are dashed and the astrophysical background is shaded gray . in the lower right panel , the dotted ( dashed ) lines refer to the flux neglecting ( including ) extragalactic absorption , see sec . [ isotropic ] . _ , title="fig:",scaledwidth=50.0% ] + illustrative example ( for the channel dm@xmath29 ) of the impact on the fit and constrained regions following from different assumptions and choices , as discussed in the text . in fig . [ fig : exclusion ] , on the other hand , we will show only the final regions and our fiducial constraints . _ , title="fig:",scaledwidth=50.0% ] illustrative example ( for the channel dm@xmath29 ) of the impact on the fit and constrained regions following from different assumptions and choices , as discussed in the text . in fig . [ fig : exclusion ] , on the other hand , we will show only the final regions and our fiducial constraints . _ , title="fig:",scaledwidth=50.0% ] the typical decay time scales that are required for the global fit are of the order of @xmath32 to @xmath33 seconds . moreover , as well known @xcite , only ` leptophilic ' channels allow a global fit : for the quark and gauge boson channels , the few tev decaying dm needed by ( @xmath28 ) is in conflict with @xmath19 data . the impact of the new data and of the improved analysis tools on the identification of the best fit dm properties is overall not big . for instance , in fig . [ fig : exclusionexample ] we show the dotted contours of the formerly preferred regions ( taken from @xcite ) , which are not very different . however , some features can readily be identified : * the fit region for the positron and antiproton signals ( yellow and green , see fig . [ fig : exclusionexample ] ) starts now at larger masses , due to the fact that @xmath16 data reach higher energies than the former ones from . * an allowed fit region now appears also for the channel dm@xmath34 ( see fig . [ fig : exclusion ] ) ; it was not present in previous analyses such as @xcite . this is due to a number of concurring reasons , among which the fact that the inclusion of the ew corrections ( [ ewcorrections ] ) and the refined propagation ( [ improvedprop ] ) smoothens out the @xmath28 spectrum and allow a decent fit to the @xmath35 1 tev hump , the fact that we now remove from the fit the low energy @xmath16 datapoints and , also , that we are now employing for data the full error bars ( slightly larger than before ) . * ew corrections imply the presence of a non - zero @xmath19 flux even for leptonic decay channels ( see e.g. the third panel of fig . [ fig : example ] ) . however , the yield is not enough to appreciably affect the fit regions . a comment on the selection of the datasets and their compatibility is also in order . we use the full set of data listed at the beginning of this section , but we also checked that dropping the data by on the positron fraction , i.e. using only data from and ( on the positron fraction and the pure @xmath16 and @xmath17 fluxes ) , does not significantly change the fit regions . similarly , the regions are not significantly modified if we adopt for the positron astrophysical background the recent determination in @xcite ( instead of the one conventionally used , from @xcite ) . this is expected , since the predictions in @xcite and in @xcite are very similar in the range of energies in which we are interested . the measurements in ( [ dataexg ] ) by the satellite correspond to the ( maximal ) residual , isotropic gamma - ray flux present in their data . its origin can be in a variety of different phenomena , both in the form of unresolved sources and in the form of truly diffuse processes ( see @xcite and reference therein ) . dm decays can also contribute to this isotropic flux , with two terms : 1 ) an extragalactic cosmological flux , due to the decays at all past redshifts ; 2 ) the residual emission from the dm halo of our galaxy . the former is of course truly isotropic , at least as long as one neglects possible nearby dm overdensities . the latter is not , but its minimum constitutes an irreducible contribution to the isotropic flux . in formul , the predicted differential dm flux that we compare with isotropic diffuse @xmath24-ray data is therefore given by @xmath36 for typical dm decay channels and for any dm halo profile , we find that the two contributions are of comparable amplitude , as it can be seen e.g. in the lower right panel of fig . [ fig : example ] . we now move to discuss the calculations of the two contributions separately , reproducing and updating the discussion of @xcite . the extragalactic flux is given , in terms of the earth - measured photon energy @xmath37 , by @xmath38 where @xmath39 is the decay rate . here the hubble function @xmath40 , where @xmath41 is the present hubble expansion rate . @xmath42 , @xmath43 and @xmath44 are respectively the dark matter , matter and cosmological constant energy density in units of the critical density , @xmath45 . the gamma ray spectrum @xmath46 , at any redshift @xmath47 , is the sum of two components : ( i ) the high energy contribution due to the prompt @xmath24-ray emission from dm decays and ( ii ) the lower energy contribution due to inverse compton scatterings ( ics ) on cmb photons of the @xmath16 and @xmath17 from those same decays . using eq . ( [ smoothedmap ] ) , the extragalactic flux can therefore be computed in terms of known quantities for any specified dm mass @xmath27 and decay channel . in particular , since the flux essentially consists of the integral over @xmath48 , at the first power , the result is not affected by the formation of dm halos , the history and the properties of which are highly uncertain . this is in contrast to the case of annihilating dm , where instead these issues affect the predictions by orders of magnitude ( see e.g. @xcite ) . we take the resulting fluxes from @xcite , but we improve the treatment of the effect of the finite optical depth of the universe , as we now move to discuss . the factor @xmath49 in eq . ( [ smoothedmap ] ) accounts for the absorption of high energy gamma rays due to scattering with the extragalactic uv background light : essentially , a high energy photon from dm hits a lower energy uv photon and produces an energetic electron ( or positron ) which , in turn , makes inverse compton scattering on the cmb and therefore reinjects a continuum of lower energy @xmath24-rays . the process amount to a subtraction of @xmath24-ray flux from the higher energies and a redistribution towards the lower part of the spectrum . as an element of novelty in this analysis , we take this process fully into account , using the energy and redshift dependent optical depth of the universe ( denoted @xmath50 ) of @xcite . the effect is sizable at energies @xmath51gev : it can reduce the high energy flux by about one order of magnitude . indeed , in fig . [ fig : example ] , lower right panel , we show with dotted lines the fluxes computed neglecting absorption and with dashed lines the effect of including it . summing over the galactic component ( which is of course not affected by absorption ) reduces the impact of the effect , which can however remain sizable . this will affect the constraints derived below , especially for channels which feature a large prompt contribution . the flux from the galactic halo , coming from a generic direction @xmath52 , is given by the well known expression @xmath53 \ , \frac{d n}{d e_\gamma}\ , , \label{fluxdec}\ ] ] i.e. as the integral of the decaying dm density piling up along the line of sight individuated by the direction @xmath52 . here @xmath54 is the dm distribution in the milky way , for which we will always take a standard navarro - frenk - white @xcite profile @xmath55 with parameters @xmath56 kpc and @xmath57 gev/@xmath58 @xcite . in principle one could consider other choices of profiles and parameters , but , for the case of decaying dm and considering that we will not be interested in the regions of the galactic center ( gc ) where profiles differ most ( quite the opposite : we will focus on the anti - gc ) , these choices make a negligible difference . the coordinate @xmath59 , centered on the gc , reads @xmath60 , where @xmath61 kpc is the most likely distance of the sun to the gc and @xmath62 is the angle between the direction of observation in the sky and the gc . the spectrum @xmath46 consists again of two components : the prompt one and the ics one . we compute both using the tools in @xcite . in particular , for the ics flux we use the generalized halo functions for the ic radiative process provided there , which take into account the full energy losses discussed in ( [ improvedprop ] ) . this is another refinement with respect to @xcite , which had computed the @xmath2 energy losses towards the anti - gc including cmb only : it was thus missing synchrotron losses ( which can account for up to roughly a third of the total losses , in the range 8.33 kpc @xmath63 15 kpc ) and ics losses on local ambient light other than the cmb ( which are instead less important ) . the treatment that we now adopt represents the most refined one we can afford , at the current state - of - the - art of these semi - analytical computations . as indicated in eq . ( [ eq : main ] ) and in the discussion above it , we need to determine the minimum of the flux in eq . ( [ fluxdec ] ) . for the prompt contribution , the minimum is obviously located where the line - of - sight density of dm is the lowest , just because the @xmath24-ray emission density traces the dm density by definition . this coincides with the direction of the anti galactic center ( anti - gc ) . for what concerns the ics contribution , on the other hand , the situation is more complicated . in this case the source of @xmath24-rays is the population of dm - originated @xmath2 , and therefore the minimum depends on what is assumed for the propagation and final distribution of the latter ones in the galactic halo . for instance , assuming a thin diffusive halo with @xmath64 kpc ( see e.g. @xcite for a discussion and an overview of standard values ) implies that the layer of ics emitting @xmath2 is particularly shallow in the zenith ( and nadir ) direction and therefore the ics @xmath24-ray flux is minimal along lines of sight pointing directly above ( and below ) the solar system . on the other hand , for a thick @xmath65 kpc diffusive halo , the ics emission from the directions orthogonal to the galactic plane is greatly enhanced and the minimum is located again at the anti - gc . in view of this complicated morphology , we choose to locate the minimum always at the anti - gc : @xmath66 this is supported , along the lines of the discussion above , by a number of considerations , among which : ( i ) thin diffusive halos are anyhow disfavored by other arguments ( see e.g. @xcite ) ; ( ii ) a number of other uncertain astrophysical variables would enter in a detailed determination of a true minimum : for instance if the _ radial _ size of the diffusive halo is smaller than usually assumed ( @xmath3520 kpc ) , then the ics flux at the anti - gc is reduced ; ( iii ) this discussion only affects the galactic ics contribution ( not the galactic prompt and the extragalactic contributions ) , thus limiting the impact on the derived total flux and having an importance ( for the bounds we will derive below ) only for those configurations in which the ics saturates the constraints . we thus compute @xmath67 as discussed above and we compare it with the data of ( [ dataexg ] ) . we show an example of such a comparison in fig . [ fig : example ] , lower right panel . it is pretty clear ( even by inspection , in the case of this example ) that the dm signal does not agree in shape with the data , which are instead well fit by a simple power law @xcite . for this reason , we are driven to derive constraints on the maximum dm signal , and therefore the minimum @xmath12 , admitted by the data . there are however several possible ways to compute such constraints . we discuss a few of them in the following and we illustrate them in the specific example of fig . [ fig : exclusionexample ] . * _ dm signal only . _ the most conservative option consists in demanding that the dm signal alone ( eq . ( [ eq : main ] ) ) does not exceed any single one of the data points by more than a chosen significance , which we set at 3@xmath68 . this is overly conservative for a number of reasons . first of all , since it supposes a vanishingly small astrophysical background . a simple inspection of the dm prediction proves that this assumption is physically untenable ; the bumpy shape of the signal is so different from the featurless observations that the ( bulk of the ) latter must be explained by astrophysical processes , leaving only a subleading role for signals from dm . second , given the smooth nature of the dm signal and of the data , it is clear that an excess of 3@xmath68 in one point is often accompanyed by similar excesses in neighboring points , so that the ` global significance ' of the exclusion is actually higher . for instance , the best - fit model example in fig . [ fig : example ] is only barely excluded by such a procedure , despite the clear tension with several data points . * _ dm signal + power - law background . _ a more realistic option consists in assuming that the astrophysical background consists of a power - law and demand that the sum of astrophysical background and dm signal does not exceed a chosen global significance . note that one expects several potential sources to contribute to this flux , like unresolved blazars , star - forming galaxies or electromagnetic cascades from ultra - high energy cosmic ray losses . in general , a combination of them is not expected to produce ( a priori ) an exact power - law , which may be inaccurate also for some of the different contributions taken alone . nonetheless , within the current precision we find that the data of @xmath69 alone are very well described by a power law with index -0.41 ( very similar to the one found in resolved blazars ) and normalization 1.02 @xmath70 gev/(@xmath71 sec sr ) . next , for a given @xmath27 we add the dm signal , whose normalization is controlled by @xmath72 . we let the normalization of the power law background vary within a factor of 2 from the central value specified above and its index within 0 and -1 ( this choice is astrophysically plausible , although varying the parameters in a broader range would not change the results shown ) . we then compute the @xmath73 to the data , marginalizing over these parameters . we compute in this way 95% c.l . limits on @xmath72 . * _ dm signal + unconstrained background . _ a variation of the procedure described in the previous point consists in avoiding to commit on the functional form of the background and assume that it consists of an arbitrary function which describes well the data ( namely : it is consistent with each data point within a small significance , e.g. 1@xmath68 ) . we then demand that , adding the signal , the @xmath74 does not exceed a given value , which in this case we fix very conservatively at 25 . for the practical case at hand , however , this procedure and the previous one yield similar results , as one can see on the example in fig . [ fig : exclusionexample ] . this is not surprising , since the isotropic residual datapoints are indeed very well described by a power law . the procedure _ ` dm signal + power - law background ' _ is the one that we will adopt to obtain the fiducial constraints shown in our results ( see fig . [ fig : exclusion ] ) . this procedure is ` fiducial ' also in the sense that it matches the analysis we do for charged cr anomalies ( see sec . [ charged fits ] ) and therefore fit regions and constraints are essentially consistent with each other in fig . [ fig : exclusion ] . before moving on , we discuss how the constraints would be modified if we removed from the analysis the datapoints above 100 gev , that are labelled as ` preliminary ' by the collaboration , i.e. if we limited the data to those published during 2010 in @xcite only . in fig . [ fig : exclusionexample ] the dotted gray line corresponds to how the fiducial constraints would be modified . we see that , as expected , the limits become looser by a factor of a few . galaxy clusters are the largest gravitationally bound structures in the universe , 80% of their total mass being in the form of dark matter . although they are located at much larger distances than other popular targets such as dwarf satellites of the milky way , they turn out to be attractive environments to search for dm due to their predicted high dm luminosity . in practice , which cluster is the most promising depends on a number of factors , and most notably the astrophysical background . indeed , standard astrophysical gamma - ray emission is expected , both from high energy cosmic rays interacting in the intra cluster medium ( the dominant contribution ) or from electrons in the ambient radiation field ( a subdominant contribution , see @xcite for a review ) . despite these predictions , no gamma - ray from these processes has been detected so far @xcite . radio galaxies lying at the gravity center of clusters such as m87 in virgo and ngc1275 in perseus are also @xmath24-ray emitters : these have actually been directly observed . a few galaxy clusters have been observed by : the most attractive of them for dm searches are virgo , coma and fornax . although virgo is closeby , the high energy gamma - ray emission from m87 prevents searches in the inner region . coma is among the most massive galaxy clusters but it has been shown that it is not a privileged environment because of a relatively high cr - induced gamma - ray emission @xcite . fornax is located at 19 mpc @xcite and shows favorably - low expected astrophysical background ( see , for instance , @xcite ) . given its location near the tropic of capricorn , the instrument is best - suited to observe fornax with respect to other currently - operating imaging atmospheric erenkov telescopes ( iacts ) . the predicted dm @xmath24-ray flux from fornax can be easily obtained by integrating eq.([fluxdec ] ) over the observational solid angle @xmath75 , and of course replacing @xmath54 with @xmath76 , the dm distribution in the cluster . there are a few different methods to determine @xmath76 ( and in general the dm content of galaxy clusters ) . a widely used approach is based on x - ray measurements on the gravitationally bound hot intracluster gas . assuming a nfw profile @xcite , the latter is entirely defined from the virial mass extracted form the hiflugcs catalog catalog @xcite and the virial mass and the concentration relationship found in ref . @xcite , in @xmath77cdm cosmology . other choices for the profile are possible , e.g. the cored burkert profile @xcite , which may be particularly motivated by the fact that baryon physics in the inner part of clusters may significantly alter the predictions of @xmath77cdm simulations : several processes invoking baryons in galaxy formation such as dynamical friction , agn feedback or gas outflows , may flatten the dm cusps in dm cores @xcite . for the case of fornax , various dark matter halo models have been considered in @xcite ( and references therein ) . however , as opposed to the annihilation case , the gamma - ray flux from decaying dm is less dependent on the dm distribution inside the target due to the simple linear dependence of the astrophysical factor with dm density . varying over the profiles discussed in @xcite , the difference in the flux factor is less than a factor of 3 for a given opening integration angle . another approach to determine the dm content of clusters consists in using dynamical tracers ( see again @xcite and references therein ) . in the following , we take as ` fiducial ' the x - ray based determination of the dm profile , hereafter referred to as the rb02 profile @xcite , which also well agrees with tracer dynamics at large distances . in this case , the dm distribution is inferred from a x - ray concentration - virial mass relationship is usually defined as the enclosed mass within a radius where the density reaches @xmath78 , with @xmath79 the critical density of the universe and @xmath80 taking values between 100 and 500 . although @xmath81 can be used to derive specific galaxy cluster properties , the x - ray concentration - mass relation is particularly well verified for @xmath80 = 200 @xcite . ] and the uncertainty on the mass determination is about 10% @xcite . as an alternative choice , we also consider the dw01 profile of @xcite , which is fully based on the dynamical tracers method . this generates a predicted dm @xmath24-ray flux @xmath353 times smaller than rb02 . next , we discuss the choice of @xmath82 . in contrast to annihilating dark matter for which most often the smallest opening angle provide the most sensitive searches , at least for cuspy profiles , decaying dark matter searches require optimization of the opening angle to guarantee the highest signal - to - noise ratio . as it is straightforward from eq.([fluxdec ] ) , the luminosity scales with the size of the solid integration angle . on the other hand , background is increasing as well . we find that the optimization of the signal - to - noise ratio versus the opening integration angle for the dark matter halo profile rb02 implies the best integration region to be 0.5@xmath83 , which corresponds to a solid angle of @xmath84 sr . we also mention that an additional component to the decaying dark matter spectrum needs in principle to be considered , particularly for leptonic final states : that of inverse compton emission ( indeed , as opposed to dwarf galaxies , in clusters the electrons lose energy primarily through ics on the ambient radiation field and produce additional gamma rays in the final state ) . we include it in the computations for the dm @xmath29 channel , but , due to the energy working range of , this component becomes important only for very large dm masses ( above @xmath8530 tev ) and its effect is therefore not visible in our results . in order to extract exclusion limits on the dark matter lifetime from gamma - ray astronomical observation with iacts , the background needs to be determined to constrain the decaying dm luminosity . the background is calculated in a region referred to as the off region , and the signal region as the on region . both on and off regions depend on the observation mode and are specific to the iact instrument . as for the background level , it is taken simultaneously , _ i.e. _ in the same data - taking observing conditions , to the signal events in order to allow for the most accurate estimate . in the ` standard ' on - off method , one has to take into account the background flux in the astrophysical factor and subsequently in the upper limit calculation . for the on region of 0.5@xmath83 and considering a off region in an annulus around the on region with an inner and outer radii of 0.5@xmath83 and 1.0@xmath83 respectively , this reduces the decaying dm flux in the on region of about 40% . an ` improved ' background estimate procedure called the _ template _ method allows to avoid this reduction . in such method the background events are determined in the on region but from selecting hadron - like events ( see @xcite for details on the analysis procedure ) . we indeed make use of it in this work . has observed fornax for a total of 14.5 hours at low zenith angle to allow for best sensitivity to low dark matter masses . figure [ fig : exclusion ] presents the exclusion plots and therefore summarizes our main results . one can see that the constraints from the isotropic @xmath24-ray data exclude decaying dm with a lifetime shorter than @xmath32 to few @xmath86 seconds , depending on its mass and the precise channel . therefore , in particular , they rule out the charged cr fit regions , for all the channels . as illustrated in the example in fig . [ fig : exclusionexample ] , and inspecting fig . [ fig : example ] , adopting the more conservative constraint procedure may marginally reallow a portion of the fit regions , for the dm @xmath29 , but leaving a clear tension . on the other hand , removing the ` preliminary ' data of @xcite and keeping only those published in 2010 in @xcite still allows to exclude the cr fit regions ( as illustrated in fig . [ fig : exclusionexample ] for the @xmath87 ; the other channels are less critical and remain safely excluded ) . the constraints from rise gently as a function of the mass , essentially as a consequence of the fact that the measured flux rapidly decreases with energy . they also depend ( mildly , a factor of a few at most ) on the decay channel , as a consequence of the different @xmath24-ray yield and the different shape of the dm signal in each channel . the constraints from fornax remain subdominant , roughly one order of magnitude below the ones . however , for the case of the dm @xmath88 channel , the bound also reaches the cr fit region and essentially confirms the exclusion . the constraints are cut at low masses by the dying out of the sensitivity in to low energy photons . they do not look competitive with respect to even for larger masses . the regions on the parameter space @xmath27@xmath12 that are excluded by the and constraints and that can be explored by , together with the regions of the global fit to the charged cr data , for different decay channels._,title="fig:",scaledwidth=45.0% ] the regions on the parameter space @xmath27@xmath12 that are excluded by the and constraints and that can be explored by , together with the regions of the global fit to the charged cr data , for different decay channels._,title="fig:",scaledwidth=45.0% ] + the regions on the parameter space @xmath27@xmath12 that are excluded by the and constraints and that can be explored by , together with the regions of the global fit to the charged cr data , for different decay channels._,title="fig:",scaledwidth=45.0% ] the regions on the parameter space @xmath27@xmath12 that are excluded by the and constraints and that can be explored by , together with the regions of the global fit to the charged cr data , for different decay channels._,title="fig:",scaledwidth=45.0% ] + the regions on the parameter space @xmath27@xmath12 that are excluded by the and constraints and that can be explored by , together with the regions of the global fit to the charged cr data , for different decay channels._,title="fig:",scaledwidth=45.0% ] the regions on the parameter space @xmath27@xmath12 that are excluded by the and constraints and that can be explored by , together with the regions of the global fit to the charged cr data , for different decay channels._,title="fig:",scaledwidth=45.0% ] in this study we have focussed on 2-body particle - antiparticle decay modes ( dm @xmath89 , where @xmath90 is a charged lepton ) , typical of scalar dm . we do not address 3-body decays such as dm @xmath91 , since those are model dependent . for fermionic dm , decay channels such as dm @xmath92 are possible : these ( in first approximation ) can be analysed in our framework with a trivial combination of the dm @xmath93 and dm @xmath94 channels . another possibility which has been recently considered is mixed ( ` democratic ' ) leptonic channels , such as dm @xmath95 33% @xmath96 + 33% @xmath87 + 33% @xmath97 : these can of course be approximately derived on the basis of our individual bounds . we here comment on the relative strength of the constraints in fig . [ fig : exclusion ] with respect to bounds from other analyses or other targets . with respect to the isotropic @xmath24-ray constraints of @xcite , the bounds derived here are stronger by a factor of 2 to 3 . the reasons of this are discussed at length in the text and essentially amount to : updated datasets , more refined dm analyses and the adoption of a more realistic constraint procedure ( some of these effects pull towards a weakening while others for a strengthening ) . with respect to the work in @xcite , our constraints from the isotropic background are somewhat stronger than their corresponding ones ( up to about a factor of 5 for the dm @xmath26 channel ) , likely due to the several differences in the analysis discussed in sec . [ introduction ] . in addition , @xcite presents bounds from the observation of the the fornax cluster by : these are less stringent than our isotropic background constraints but more powerful than our -based constraints at moderate masses . at the largest masses , our -based constraints pick up and match theirs ( for the @xmath98 channel ) , as expected from the different operational energy range of versus . the bounds from clusters other than fornax are less powerful , according to the analysis in @xcite . ref . @xcite has derived constraints from several clusters using data too , but their procedure is questioned in @xcite , and in any case it is now superseded . on the other hand , the preliminary constraints shown ( for the @xmath99 channel only ) in @xcite , obtained with a combination of several clusters in , exceed our bounds by a factor of 2 . bounds from probes other than the isotropic flux and clusters do not generally achieve the same constraining power . e.g. recently @xcite finds less stringent limits for the leptophilic channels , focussing on the milky way halo . within the context of observations performed by iact , we note that the decay lifetime constraints obtained with galaxy clusters are stronger than those from dwarf galaxies . even for the ultra - faint dwark galaxy segue 1 , which is believed to be the most promising dwarf in the northern hemisphere , the constraints are 2 orders of magnitude higher for a dark matter particle mass of 1 tev @xcite . we also estimate that the constraints that we derive are stronger than those that can come from neutrino observation of the galactic center ( see for instance @xcite ) . precisely computing those ones for all channels and in the same analysis framework of this study , however , is beyond the scope of this work , and would also probably benefit from a better knowledge of the data ( especially the reconstructed neutrino energy ) , which is not currently available . in summary : with the possible exception of the preliminary bounds from a combination of clusters by for the @xmath98 channel , the constraints that we derive from the isotropic @xmath24-ray flux are the most stringent to date . as we have seen , currently the constraints on decaying dm from the satellite are dominant with respect to those from . however , while the former may increase its statistics by at most a factor of a few , for the latter there are prospects of developments in the mid - term future . the next - generation iact will be a large array composed of a few tens to a hundred telescopes located on two sites , one in each hemisphere @xcite . the goal is to improve the overall performances of the present generation : one order - of - magnitude increase in sensitivity and enlarge the accessible energy range both towards the lower and higher energies allowing for an energy threshold down to a few tens of gev . from the actual design study of , the effective area will be increased by a factor 10 and a factor 2 better in the hadron rejection is expected . the calculation of sensitivity is performed with a background - only hypothesis , where the non - detection of a signal by would imply that the signal in the on region consists of only misidentified hadron showers . the background level is estimated by integrating eq . ( 4 ) of @xcite after multiplying it by the effective area . the upper limit on the number of gamma - ray events is calculated at 95% c.l . according to the method from @xcite assuming five off regions . this calculation follows the methodology described in @xcite , from which the effective area is extracted . the 95% c.l . sensitivity on the decay lifetime is then given by @xmath100 } \times \frac{m_{\rm dm } \ , n_{\gamma}^{\rm 95\% c.l.}}{t_{\rm obs}\ , \int_{0}^{m_{\rm dm}/2 } a_{\rm cta}(e_{\gamma } ) \ , \frac{dn}{d e_{\gamma}}(e_\gamma ) \ , d e_{\gamma } } \ , , \ ] ] where @xmath101 is the limit on the number of gamma - ray events , @xmath102 is the cta effective area and @xmath103 the observation time . figure [ fig : exclusion ] shows the 95% c.l . sensitivity of on the decay lifetime for the rb02 halo profile for 50h observation time and @xmath104 sr . to conclude , we also mention that a technique which could allow for significant improvements in the exploration of the parameter space of decaying dm using clusters is the one of stacking the observation of a large number of different clusters , as recently discussed in @xcite . the authors find that improvements of up to 100 can be theoretically achieved , albeit this factor is @xmath355 for more realistic background - limited instruments . decaying dark matter has come to the front stage recently as an explanation , alternative to annihilating dm , for the anomalies in cr cosmic rays in , and . but , more generally , decaying dm is a viable possibility that is or can naturally be embedded in many dm models . it is therefore interesting to explore its parameter space in the light of the recent observational results . we discussed the constraints which originate from the measurement of the isotropic @xmath24-ray background by and of the fornax cluster by , for a number of decaying channels and over a range of dm masses from 100 gev to 30 tev . we improved the analysis over previous work by using more recent data and updated computational tools . we found ( see fig . [ fig : exclusion ] ) that the constraints by rule out decaying half - lives of the order of @xmath32 to few @xmath33 seconds . these therefore exclude the decaying dm interpretation of the charged cr anomalies , ( at least ) for all 2-body channels , at least adopting our fiducial constraint procedure . the constraints by are generally subdominant . for the dm @xmath105 channel , they can however also probe the cr fit regions and essentially confirm the exclusion . we thank daniela borla tridon for communications on data , carlos de los heros for communications on data and alejandro ibarra for useful discussions . this work is supported by the french national research agency anr under contract anr 2010 blanc 041301 and by the eu itn network unilhc . ps thanks cern s theory division for hospitality during the development of this work . chen and f. takahashi , jcap * 0902 * ( 2009 ) 004 [ arxiv : [ hep - ph ] ] . yin , q. yuan , j. liu , j. zhang , x .- j . bi and s .- h . zhu , phys . rev . d * 79 * ( 2009 ) 023512 [ arxiv : [ hep - ph ] ] . k. ishiwata , s. matsumoto and t. moroi , phys . b * 675 * ( 2009 ) 446 [ arxiv : [ hep - ph ] ] . a. ibarra , d. tran , jcap * 0902 * ( 2009 ) 021 . [ arxiv : [ hep - ph ] ] . f. ferrer , c. nipoti and s. ettori , phys . d * 80 * ( 2009 ) 061303 [ arxiv : [ astro-ph.co ] ] . w. buchmuller , l. covi , k. hamaguchi , a. ibarra and t. yanagida , jhep * 0703 * ( 2007 ) 037 [ [ hep - ph ] ] . d. eichler , phys . * 63 * ( 1989 ) 2440 . a. arvanitaki , s. dimopoulos , s. dubovsky , p. w. graham , r. harnik , s. rajendran , phys . * d79 * ( 2009 ) 105022 [ arxiv : [ hep - ph ] ] . c. arina , t. hambye , a. ibarra and c. weniger , jcap * 1003 * ( 2010 ) 024 [ arxiv : [ hep - ph ] ] . talk by m. ackermann at the http://agenda.albanova.se/conferencedisplay.py?confid=2600[tevpa 2011 conference ] ( 1 - 5 aug 2011 , stockholm , sweden ) . a. a. abdo _ et al . _ [ fermi - lat collaboration ] , phys . rev . lett . * 104 * ( 2010 ) 101101 [ arxiv : [ astro-ph.he ] ] . a. abramowski _ et al . _ [ hess collaboration ] , astrophys . j. * 750 * ( 2012 ) 123 [ arxiv : [ astro-ph.he ] ] . f. takayama and m. yamaguchi , phys . b * 485 * ( 2000 ) 388 [ ] . j. m. overduin and p. s. wesson , phys . * 402 * ( 2004 ) 267 [ ] . g. bertone , w. buchmuller , l. covi and a. ibarra , jcap * 0711 * ( 2007 ) 003 [ arxiv : [ astro - ph ] ] . by s. zimmer and j. conrad at the . s. zimmer , j. conrad , for the fermi - lat collaboration and a. pinzke , arxiv : [ astro-ph.he ] . x. huang , g. vertongen and c. weniger , jcap * 1201 * ( 2012 ) 042 [ arxiv : [ hep - ph ] ] . see http://www.cta-observatory.org[www.cta-observatory.org ] . m. ackermann _ et al . _ [ the fermi lat collaboration ] , phys . rev . * 108 * ( 2012 ) 011103 [ arxiv : [ astro-ph.he ] ] . o. adriani _ et al . _ [ pamela collaboration ] , nature 458 , 607 - 609 , 2009 , arxiv : . see also o. adriani , g. c. barbarino , g. a. bazilevskaya , r. bellotti , m. boezio , e. a. bogomolov , l. bonechi and m. bongi _ et al . _ , astropart . * 34 * ( 2010 ) 1 [ arxiv : [ astro-ph.he ] ] . m. ackermann _ et al . _ [ fermi lat collaboration ] , phys . d * 82 * ( 2010 ) 092004 [ arxiv : [ astro-ph.he ] ] . d. b. tridon , p. colin , l. cossio , m. doro , v. scalzotto and f. t. m. collaboration , arxiv : [ astro-ph.he ] . o. adriani _ et al . _ [ pamela collaboration ] , phys . * 106 * ( 2011 ) 201101 [ arxiv : [ astro-ph.he ] ] . o. adriani _ et al . _ [ pamela collaboration ] , science * 332 * ( 2011 ) 69 [ arxiv : [ astro-ph.he ] ] . o. adriani _ et al . _ [ pamela collaboration ] , phys . * 105 * ( 2010 ) 121101 [ arxiv : [ astro-ph.he ] ] . see also : n. f. bell , j. b. dent , t. d. jacques and t. j. weiler , phys . d * 78 * ( 2008 ) 083540 [ arxiv : [ hep - ph ] ] . more recent analyses and a more complete appraisal of the phenomenological importance of the effect have been performed in : m. kachelriess , p. d. serpico and m. a. solberg , phys . d * 80 * ( 2009 ) 123533 [ arxiv : [ hep - ph ] ] ; p. ciafaloni and a. urbano , phys . d * 82 * ( 2010 ) 043512 [ arxiv : [ hep - ph ] ] . the antiproton constraints ( especially for leptophilic channels ) have been considered in m. garny , a. ibarra and s. vogl , jcap * 1107 * ( 2011 ) 028 [ arxiv : [ hep - ph ] ] and m. garny , a. ibarra and s. vogl , arxiv : [ hep - ph ] . the related phenomenon of gluon emissions is also addressed in m. asano , t. bringmann and c. weniger , arxiv : [ hep - ph ] . p.ciafaloni , d.comelli , a.riotto , f.sala , a.strumia , a.urbano , jcap * 1103 * ( 2011 ) 019 , arxiv : . m. cirelli , g. corcella , a. hektor , g. hutsi , m. kadastik , p. panci , m. raidal and f. sala _ et al . _ , jcap * 1103 * ( 2011 ) 051 [ arxiv : [ hep - ph ] ] ; the numerical results are available from http://www.marcocirelli.net/pppc4dmid.html[this website ] . c. -r . chen , s. k. mandal and f. takahashi , jcap * 1001 * ( 2010 ) 023 [ arxiv : [ hep - ph ] ] . a. ibarra , d. tran and c. weniger , phys . d * 81 * ( 2010 ) 023529 [ arxiv : [ hep - ph ] ] . m. cirelli , p. panci , p. d. serpico , nucl . . * b840 * ( 2010 ) 284 - 303 . [ arxiv : [ astro-ph.co ] ] . j. ke , m. luo , l. wang and g. zhu , phys . b * 698 * ( 2011 ) 44 [ arxiv : [ hep - ph ] ] . m. luo , l. wang and g. zhu , arxiv : [ hep - ph ] . f. aharonian _ et al . _ [ h.e.s.s . collaboration ] , phys . lett . 101 ( 2008 ) 261104 [ arxiv : ] . f. aharonian et al . [ h.e.s.s . collaboration ] , arxiv : [ astro-ph.he ] . m. cirelli , m. kadastik , m. raidal and a. strumia , nucl . b * 813 * ( 2009 ) 1 [ arxiv : [ hep - ph ] ] . j. lavalle , mon . not . * 414 * ( 2011 ) 985l [ arxiv : [ astro-ph.he ] ] . i. v. moskalenko and a. w. strong , astrophys . j. * 493 * ( 1998 ) 694 [ ] . e. a. baltz and j. edsjo , phys . d * 59 * ( 1998 ) 023511 [ ] . a. a. abdo _ et al . _ [ fermi - lat collaboration ] , jcap * 1004 * ( 2010 ) 014 [ arxiv : [ astro-ph.co ] ] . p. d. serpico , e. sefusatti , m. gustafsson and g. zaharijas , mon . not . soc . * 421 * , 87l ( 2012 ) arxiv : [ astro-ph.co ] . j. f. navarro , c. s. frenk and s. d. m. white , astrophys . j. * 462 * ( 1996 ) 563 [ arxiv : ] . t. bringmann , f. donato and r. a. lineros , jcap * 1201 * ( 2012 ) 049 [ arxiv : [ astro-ph.ga ] ] . p. blasi , s. gabici and g. brunetti , int . j. mod . a * 22 * ( 2007 ) 681 [ ] . f. aharonian [ hess collaboration ] , a&a , * 495 * , 27 , arxiv : [ astro - ph ] . f. a. aharonian [ the hess collaboration ] , a&a , * 502 * , 437 arxiv : [ astro-ph.co ] . j. aleksic _ et al . _ [ the magic collaboration ] , astrophys . j. * 710 * ( 2010 ) 634 [ arxiv : [ astro-ph.he ] ] . m. ackermann , m. ajello , a. allafort , l. baldini , j. ballet , g. barbiellini , d. bastieri and k. bechtol _ et al . _ , jcap * 1005 * ( 2010 ) 025 [ arxiv : [ astro-ph.co ] ] . m. ackermann _ et al . _ [ fermi - lat collaboration ] , astrophys . j. , * 717 * , l71 arxiv : [ astro-ph.he ] . t. h. reiprich and h. boehringer , astrophys . j. * 567 * ( 2002 ) 716 [ ] . t. e. jeltema , j. kehayias and s. profumo , phys . d * 80 * ( 2009 ) 023005 [ [ astro - ph ] ] . a. pinzke , c. pfrommer and l. bergstrom , phys . * 103 * ( 2009 ) 181302 [ arxiv : [ astro-ph.he ] ] . a. pinzke , c. pfrommer and l. bergstrom , phys . d * 84 * ( 2011 ) 123509 [ arxiv : [ astro-ph.he ] ] . d. a. buote , f. gastaldello , p. j. humphrey , l. zappacosta , j. s. bullock , f. brighenti and w. g. mathews , astrophys . j. * 664 * ( 2007 ) 123 [ ] . a. burkert , iau symp . * 171 * ( 1996 ) 175 [ astrophys . j. * 447 * ( 1995 ) l25 ] [ ] . see also : p. salucci and a. burkert , arxiv : . g. gentile , p. salucci , u. klein , d. vergani and p. kalberla , mon . not . soc . 351 ( 2004 ) 903 [ arxiv : ] and p. salucci , a. lapi , c. tonini , g. gentile , i. yegorova and u. klein , mon . not . 378 ( 2007 ) 41 [ arxiv : ] . a. el - zant , i. shlosman and y. hoffman , astrophys . j. , * 560 * , 636 , . d. martizzi , r. teyssier , b. moore and t. wentz , arxiv : [ astro-ph.co ] . g. m. voit , rev . * 77 * ( 2005 ) 207 [ ] . l. dugger , t. e. jeltema and s. profumo , jcap * 1012 * ( 2010 ) 015 [ arxiv : [ astro-ph.he ] ] . r. abbasi _ et al . _ [ icecube collaboration ] , phys . d * 84 * ( 2011 ) 022004 [ arxiv : [ astro-ph.he ] ] . the cta consortium ( m. actis et al . ) , exper . astron . * 32 * ( 2011 ) 193 [ arxiv : [ astro-ph.im ] ] . a. viana , c. m. medina , j. penarrubia , p. brun , j. f. glicenstein , k. kosack , e. moulin and m. naumann - godo _ et al . _ , astrophys . j. * 746 * ( 2012 ) 77 [ arxiv : [ astro-ph.he ] ] . w. a. rolke , a. m. lopez and j. conrad , nucl . instrum . a * 551 * ( 2005 ) 493 [ physics/0403059 ] . c. combet , d. maurin , e. nezri , e. pointecouteau , j. a. hinton and r. white , phys . d * 85 * , 063517 ( 2012 ) [ arxiv : [ astro-ph.co ] ] .

gravitational wave detectors are typically described as responding to gravitational wave metric perturbations , which are gauge - dependent and correspondingly unphysical quantities . this is particularly true for ground - based interferometric detectors , like ligo , space - based detectors , like lisa and its derivatives , spacecraft doppler tracking detectors , and pulsar timing arrays detectors . the description of gravitational waves , and a gravitational wave detector s response , to the unphysical metric perturbation has lead to a proliferation of false analogies and descriptions regarding how these detectors function , and true misunderstandings of the physical character of gravitational waves . here we provide a fully physical and gauge invariant description of the response of a wide class of gravitational wave detectors in terms of the riemann curvature , the physical quantity that describes gravitational phenomena in general relativity . in the limit of high frequency gravitational waves , the riemann curvature separates into two independent gauge invariant quantities : a `` background '' curvature contribution and a `` wave '' curvature contribution . in this limit the gravitational wave contribution to the detector response reduces to an integral of the gravitational wave contribution of the curvature along the unperturbed photon path between components of the detector . the description presented here provides an unambiguous physical description of what a gravitational wave detector measures and how it operates , a simple means of computing corrections to a detectors response owing to general detector motion , a straightforward way of connecting the results of numerical relativity simulations to gravitational wave detection , and a basis for a general and fully relativistic pulsar timing formula .

1310.2871

it has long been understood that , in general relativity gravitational phenomena are the physical manifestation of spacetime curvature . nevertheless , gravitational wave detectors , which make physical measurements , are typically described as responding to spacetime metric perturbations , which are coordinate gauge dependent and correspondingly unphysical quantities . just as early attempts to understand gravitational waves in terms of metric perturbations led to confusion regarding whether such waves existed or how they might be generated , so attempts to describe how gravitational wave detectors respond to metric perturbations lead to wooly statements and , sometimes , outright misconceptions @xcite . by way of contrast , gravitational waves described as spacetime curvature perturbations are , in a well - defined sense @xcite , physically unambiguous quantities ; correspondingly , describing the gravitational wave detector response in terms of the detector s interaction with spacetime curvature may be expected to be , if nothing else , conceptually more satisfying and physically more revealing . here we derive and describe the response of a wide class of gravitational wave antennas including pulsar timing arrays , spacecraft doppler tracking , and both ground- and space - based laser interferomteric detectors in a way that relies solely upon physical measurements and the physical properties of spacetime as described by the riemann curvature tensor . the resulting expression of the detector response , and each term that comprises it , is separately gauge invariant and has a clear physical interpretation . we show that when the gravitational waves can be described as a gauge independent curvature perturbation of a background spacetime the wave contribution to the response is wholly embodied in an integration of a projection of riemann curvature tensor perturbation along specific null geodesics of the unperturbed spacetime . our principle result is directly applicable to gravitational wave detection via pulsar timing or spacecraft doppler tracking ; is the building - block upon which a physically and pedagogically satisfying description of the response of interferometeric antennas may be based ; can form the basis for a simplified and general derivation of a fully relativistic pulsar timing formula ; and may be used to simplify the use of numerial relativity simulations to aid in the analysis or interpretation of gravitational wave detector observations . @xcite described the operation of the first practical gravitational wave detector in terms of riemann curvature induced excitations of the oscillation modes of a high quality - factor metal bar . throughout the 1960 s a variety of mechanical gravitational wave antenna configurations were proposed ( a qualitative summary of many of these configurations is given in ( * ? ? ? 1013 ) ) and their responses were generally described in an approximate way through a coupling to the riemann curvature as well . by the early 1970s , however , a description of gravitational waves in terms of metric perturbations had taken hold . this was the case from the start for interferometric gravitational wave antennas @xcite and their kin : spacecraft doppler tracking @xcite and pulsar timing @xcite . this trend of describing gravitational wave antenna response to metric perturbations has generally persisted throughout the literature to this day @xcite . as is well known , however , metric perturbations depend on coordinate gauge choices . while the freedom to choose a gauge may be exploited to simplify some computations , it is perilous to give physical meaning to partial or intermediate results of these calculations . nevertheless , throughout the literature research and pedagogical one finds numerous ( and sometimes conflicting ) descriptions of how a gravitational wave physically produces a signal in a detector based on an interpretation of the terms in calculations that depend on a choice of coordinate gague . for example , it is routinely claimed ( though we give only single , recent examples from the literature ) that gravitational waves physically move freely - falling test masses in an ideal interferometric detector ( * ? ? ? 21 ) ; gravitational waves `` push '' the mirrors of an ideal interferometric detector apart and together ( * ? ? ? * lecture 1 page 1 ) ; spacetime curvature influences light differently than it does mirror separations in an interferometer ( * ? ? ? * lecture 1 page 2 ) ; gravitational waves alternately redshift and blueshift the light in an interferometer detector @xcite ; it is the direct affect of gravitational waves on the earth and a distant pulsar that we measure when we detect when we observe gravitational waves via pulsar timing @xcite ; and it is the affect of gravitational waves on the trajectory of the light passing between a pulsar and earth that leads to their detection @xcite . in fact , as is apparent by appropriate application of the equivalence principle , each of these statements is incorrect ; that they are made at all is the result of ascribing physical significance to gauge - dependent quantities . other authors have explored alternative derivations or expressions of the gravitational wave detector response in terms of the riemann curvature @xcite ; however , these discussions either start from the metric perturbation in a preferred gauge or apply , in an approximate way , the geodesic deviation equation to describe the deviation vector between non - geodesics . additionally , these descriptions ( as well as most formulations coupling to the metric perturbation ) generally make the crucial assumption that the background spacetime is minkowski or that the different detector components ( beamsplitter and end mirrors for interferometers or the earth and a pulsar for pulsar timing ) are at some coordinate rest in a preferred gauge . the description of the response we present here makes no assumptions regarding the geometry of the background spacetime and involves , from beginning to end , only physical measurements and gauge - independent quantities : i.e. , it is valid in an arbitrary background spacetime and never requires or invokes any special gauge or gauge - dependent quantities . in section [ sec : doppler ] we provide a general , geometrically motivated derivation for the observed phase evolution of a remote clock . in section [ sec : gwaves ] we review how and when spacetime curvature may be physically and unambiguously separated into background and gravitational wave perturbation contributions , each of which is separately gauge independent , and show that , when such a distrinction is possible , the gravitational wave contribution to the observed clock phase is entirely due to the gravitational wave contribution to the curvature . we discuss the meaning and application of our results in section [ sec : discussion ] . we end in section [ sec : conclusions ] with some brief conclusions and directions for future study . gravitational wave detection via pulsar timing @xcite , spacecraft doppler tracking @xcite , and ground- or space - based laser interferometry @xcite all involve measuring the phase of a remote clock via an electromagnetic signal that propagates through spacetime along ( piecewise ) null geodesics . the location of the clock , the receiver / observer , and the path taken by the electromagnetic record of the phase as it is transferred from the clock to the receiver / observer differ ; however , the measured record of the clock phase following its propagation along the null trajectory is the defining characteristic of these detectors . in this section we develop a fully relativistic expression for evolution of that record of the clock s phase in terms of the clock s intrinsic phase evolution , the clock and receiver / observer spacetime trajectories , the null geodesic trajectories connecting the clock and observer , and the spacetime curvature along those geodesics . in section [ sec : gwaves ] we separate - out from this expression just that part of the response due to gravitational waves . referring to figure [ fig : geodev ] , identify the world lines of the clock and the receiver / observer . denote the clock s 4-velocity @xmath0 and the observer s 4-velocity @xmath1 . focus attention on the future - directed null geodesic congruence that connects the clock s world line ( in the past ) and the receiver / observer s world line ( in the future ) . identify each null geodesic in the congruence by the clock s ( monatonically increasing ) phase @xmath2 on the clock s world line where it intersects that null geodesic . let @xmath3 be an affine parameter distance along each null - geodesic measured from the clock s world line . in this way @xmath2 and @xmath3 are coordinate functions on the world sheet described by the null geodesic congruence . next note that , on the receiver / observer s world line , @xmath4 where @xmath5 and @xmath6 are the coordinate vectors corresponding to the coordinate functions @xmath3 and @xmath2 . the _ observed _ rate of phase is thus @xmath7 evaluating the right - hand side of equation [ eq : dphi : dtau ] requires knowing how @xmath6 varies along the geodesics described by @xmath5 . noting that the coordinate vector field @xmath8 is a null geodesic congruence whose deviation vectors are @xmath6 we can solve the equation of geodesic deviation to find @xmath6 along @xmath5 . for our purposes it is convenient to write the geodesic deviation equation as a set of first - order equations [ eq : k ] @xmath9 to complete the specification of the problem requires boundary conditions on @xmath6 and @xmath10 . we set the boundary conditions on @xmath6 on the clock s world line , where @xmath11 @xmath12 is the clock s 4-velocity , and @xmath13 is the clock s instantaneous frequency . the boundary condition on @xmath10 can be set in terms of the receiver / observer s trajectory and the observed direction to the clock . to do so , note that , since @xmath2 and @xmath3 are coordinate functions , @xmath14 next note that , on the receiver / observer s world line , @xmath15 where @xmath16 is the apparent direction to the clock as measured in the receiver / observer s proper reference frame . combining equations [ eq : kflip ] , [ eq : u ] and [ eq : plambda ] , find @xmath17\,.\end{aligned}\ ] ] lastly , take advantage of the freedom to choose the affine parameter scale to set @xmath18 thereby obtaining the boundary condition @xmath19 where @xmath20 is the receiver / observer s 4-acceleration and @xmath21 is the angular velocity ( proper motion ) of the clock as measured by the receiver / observer . to recapitulate , the clock phase record as observed at the receiver evolves as @xmath22 where @xmath23 is the receiver s 4-velocity , @xmath5 is the tangent to the null geodesic along which the clock phase record is propagated and @xmath6 is the deviation associated with the null geodesic field @xmath5 . the deviation vectors are computed via the equation of geodesic deviation @xmath24 with boundary conditions and null - geodesic affine parameter normalization @xmath25 a more physically revealing description of the observed clock phase may be found by focusing on the clock phase s observed doppler , or @xmath26 . our objective in this subsection is to find the observed doppler and show that it can be expressed in terms of the receiver / observer s acceleration , the clock s intrinsic frequency and frequency derivative ( @xmath13 and @xmath27 ) , the clock s apparent transverse velocity and most importantly for the gravitational wave signal the curvature along the null geodesics connecting the clock and receiver / observer . begin by adopting the affine parameter normalization in equation [ eq : lambdascale ] and writing the observed clock doppler as [ eq : doppler ] @xmath28\cdot\nabla\left({{\partial_{\lambda}}}\cdot{{\partial_{\phi}}}\right)\\ & = \left(\frac{d\phi}{d\tau}\right)^3 \left[{\nabla_{\bm{{{\partial_{\phi}}}}}}\left({{\partial_{\lambda}}}\cdot{{\partial_{\phi}}}\right)\right]_{\lambda=\lambda_r } + \left[\left(\frac{d\phi}{d\tau}\right)^2 \left(\frac{d\lambda}{d\tau}\right ) \left({{\partial_{\lambda}}}\cdot{\bm{k}}\right)\right]_{\lambda=\lambda_r}\\ & = \left(\frac{d\phi}{d\tau}\right)^3 \left [ { \nabla_{\bm{{{\partial_{\phi } } } } } } \left({{\partial_{\lambda}}}\cdot{{\partial_{\phi}}}\right ) \right]_{\lambda=\lambda_r}\ , . \label{eq : dphi}\end{aligned}\ ] ] in arriving at equation [ eq : dphi ] we have taken advantage of the affine parameter normalization ( eq . [ eq : lambdascale ] ) , the resolution of @xmath23 in terms of the ( @xmath6 , @xmath5 ) basis ( eq . [ eq : u ] ) , and the orthogonality of @xmath5 and @xmath29 : @xmath30 next note that , since @xmath29 and @xmath5 are orthogonal ( eq . [ eq : korthogonality ] ) , @xmath31 also @xmath32 i.e. , both @xmath33 and @xmath34 are constant along the null geodesics of the congruence . thus [ eq : retarded ] @xmath35 with the understanding that the right- and left - hand sides of equation [ eq : retarded ] are evaluated on the same null geodesic . taking advantage of equation [ eq : retarded ] expand the bracketed term in equation [ eq : dphi ] : @xmath36_{\lambda=0}\\ \label{eq : bracketed : b } & = \left({{\partial_{\phi}}}\cdot{\bm{k}}\right)_{\lambda=0 } + \frac { { { \partial_{\lambda}}}\cdot{\bm{a}}_c } { \left(d\phi / dt\right)^2 } + \frac { { { \partial_{\lambda}}}\cdot{{\partial_{\phi } } } } { d\phi / dt } { \nabla_{\bm{v}}}\left(\frac{d\phi}{dt}\right)^{-1}\\ \label{eq : bracketed : c } & = \frac{{{\partial_{\lambda}}}\cdot{\bm{a}}_c}{\left(d\phi / dt\right)^2 } + \left(\frac{d\phi}{d\tau}\right)^{-1}\left(\frac{d\phi}{dt}\right)^{-2 } \frac{d^2\phi}{dt^2 } \nonumber\\ & \qquad\qquad { } + \left.\left({{\partial_{\phi}}}\cdot{\bm{k}}\right)\right|_{\lambda=\lambda_r } - \int_0^{\lambda_{r}}\left[{\bm{k}}^2-{\bm{r}}({{\partial_{\phi}}},{{\partial_{\lambda}}},{{\partial_{\phi}}},{{\partial_{\lambda}}})\right]d\lambda\,.\end{aligned}\ ] ] in proceeding from equation [ eq : bracketed : a ] to [ eq : bracketed : b ] use the boundary condition on @xmath6 ( eq . [ eq : bcphi ] ) ; in passing on to equation [ eq : bracketed : c ] use @xmath37d\lambda\ , , \label{eq : int}\end{aligned}\ ] ] where the integration path is along a constant @xmath2 null geodesic . the boundary term @xmath38 may be re - expressed in terms of the receiver / observer s 4-acceleration : [ eq : bterm ] @xmath39 in passing from equation [ eq : bterm : a ] to [ eq : bterm : c ] take advantage of the expression for @xmath23 resolved on the @xmath40 coordinate basis ( eq . [ eq : u ] ) and the choice of affine parameter normalization ( eq . [ eq : lambdascale ] ) . bringing together equations [ eq : dphi ] , [ eq : bracketed : c ] and [ eq : bterm : c ] , find @xmath41d\lambda } \right . \nonumber\\ & \qquad\qquad \left . \vphantom { \left(\frac{d\phi / d\tau}{d\phi / dt}\right)^2 \left.\left({{\partial_{\lambda}}}\cdot{\bm{a}}_c\right)\right|_{\lambda=0 } - \left.\left({{\partial_{\lambda}}}\cdot{\bm{a}}_r\right)\right|_{\lambda=\lambda_r } -\left(\frac{d\phi}{/d\tau}\right)^2 \int_0^{\lambda_{r } } \left[{\bm{k}}^2-{\bm{r}}({{\partial_{\phi}}},{{\partial_{\lambda}}},{{\partial_{\phi}}},{{\partial_{\lambda}}})\right]d\lambda } -\left(\frac{d\phi}{d\tau}\right)^2 \int_0^{\lambda_{r } } \left[{\bm{k}}^2-{\bm{r}}({{\partial_{\phi}}},{{\partial_{\lambda}}},{{\partial_{\phi}}},{{\partial_{\lambda}}})\right]d\lambda \right\}\,\end{aligned}\ ] ] where we have relaxed the assumption that @xmath42 . in equation [ eq : riemannresponse ] all quantities are understood to be evaluated on the same @xmath2-constant geodesic of the congruence . equation [ eq : riemannresponse ] relates the doppler in the clock s phase record to the clock s intrinsic properties ( @xmath13 , @xmath27 ) and trajectory ( @xmath43 ) , the trajectory of the sensing apparatus ( @xmath44 and @xmath20 ) , the relative motion between the clock and receiver / observer expressed in the changing null geodesic traveled by the record as it propagates from the clock to the measuring apparatus [ @xmath45 and the spacetime properties along that null geodesic [ @xmath46 . in section [ sec : gwaves ] we identify when the integral of @xmath47 appearing in equation [ eq : riemannresponse ] can be separated , in a gauge invariant fashion , into a contribution that is associated with the gravitational waves and another that is associated with the background spacetime curvature . equation [ eq : riemannresponse ] describes how the clock s observed doppler , determined from the record propagated from clock to measuring apparatus , depends on the clock s intrinsic frequency and frequency evolution , the 4-velocity and 4-acceleration of the clock and measuring apparatus , and the riemann curvature along the null geodesics connecting the clock and measuring apparatus . included are gravitational wave effects . in this section we show that the gravitational wave effects are , when the radiation wavelength is much shorter than the scale of the background curvature , wholly due to the perturbation to the riemann integrated along the _ unperturbed _ null geodesic trajectory linking the clock and receiver / observer . the first difficulty to overcome is the identification of gravitational radiation as a distinct spacetime phenomena . in general relativity all truly gravitational phenomena - including gravitational radiation - are described by spacetime s riemann curvature tensor . owing to the non - linear nature of the einstein equations a complete distinction between stationary and dynamical gravitational phenomena is , in general , impossible . nevertheless , isaacson @xcite showed that gravitational waves emerge naturally as a distinct physical phenomena when a multi - scale perturbation expansion of the riemann curvature is possible . to this end , in subsection [ sec : mpt ] we briefly review the necessary elements of perturbation theory : i.e. , how the choice of a bijective map ( gauge ) allows us to give meaning to background and perturbation , and how the perturbation changes when we change the gauge . in subsection [ sec : gwpert ] we show that a short wavelength perturbation in the riemann curvature is a tensor ( i.e. , it is gague invariant and , correspondingly , physically unambiguous ) that satisfies the wave equation on the background spacetime : i.e. , it _ is _ a gravitational wave . having identified gravitational waves as a gauge - invariant , physical phenomena , we next ask how the introduction of a gravitational wave perturbation affects @xmath26 as given in equation [ eq : riemannresponse ] . in addition to its contribution to the riemann curvature , the gravitational wave perturbation also affects the null - geodesics along which the record of the clock s phase propagates : i.e. , the gravitational wave perturbation affects , in principle , every term that appears on the righthand side of equation [ eq : riemannresponse ] . in subsection [ sec : gwresp ] we find our principal result : the dominant contribution to the observed evolution of the clock phase @xmath26 owing to a gravitational wave perturbation is entirely due to the gravitational wave perturbation to the riemann curvature integrated along the unperturbed null geodesic trajectory linking the clock and the receiver / observer . in general relativity , spacetime is a manifold with a metric . when we perturb spacetime we perturb the manifold and metric . to compare tensors on the perturbed and unperturbed manifolds i.e. , to identify and study the perturbation requires a one - one and onto map between the two manifolds . following @xcite consider a one - dimensional family of spacetime manifolds and metrics , @xmath48 , smoothly depending on the dimensionless parameter @xmath49 . refer to the @xmath50 manifold and metric as the background and write them without the tilde : i.e. , write @xmath51 for @xmath52 and @xmath53 for @xmath54 . to compare tensor fields on @xmath55 with their background counterparts introduce a pull - back @xmath56 . the pull - back of any tensor @xmath57 can be written as a taylor expansion in @xmath49 : @xmath58\end{aligned}\ ] ] where @xmath59 & = \left.\frac{d^k\varphi^*_{\epsilon}\widetilde{\bm{t}}}{d\epsilon^k}\right|_{\epsilon=0}.\end{aligned}\ ] ] we identify @xmath60 $ ] as the order @xmath61 perturbation of @xmath62 . the pull - back of @xmath63 on @xmath55 is a field on @xmath51 . _ in particular , covariant derivatives of @xmath64 , or of @xmath62 and its perturbations @xmath65 $ ] , are always taken with respect to background metric @xmath66 . _ for our application we are particularly interested in the perturbation expansion of the metric and the riemann tensor ( and its contractions ) , which we write @xmath67 where we have introduced the additional shorthand @xmath68\\ { \bm{r}}^{(1 ) } & = \delta^1_{\varphi}[{\bm{r}}].\end{aligned}\ ] ] physical quantities do nt depend on coordinate choices ; correspondingly , @xmath62 is physical only if it has a measurable effect that is independent of the choice of coordinates used to measure it . consider a one - dimensional family of coordinate transformation @xmath69 generated by the vector field @xmath70 : i.e. , for coordinate function @xmath71 @xmath72 with @xmath73 @xcite showed that the transformation of @xmath74 under @xmath75 is described by the expansion @xmath76 where @xmath77 denotes the lie derivative with respect to @xmath78 : @xmath79 correspondingly , we conclude that @xmath62 is physical only if @xmath80 vanishes for all @xmath81 . following isaacson @xcite , consider a background spacetime with curvature radius @xmath82 a small ( order @xmath49 ) amplitude gravitational wave perturbation with curvature radius @xmath83 : i.e. , in abstract index notation @xcite , [ eq : expansion ] @xmath84 where @xmath85 the perturbation of the metric connection @xmath86 $ ] uses its definition in terms of the metric tensor : @xmath87 & = \gamma\indices{^{a}_{be } } + \epsilon\gamma\indices{^{(1)a}_{be } } + \mathcal{o}(\epsilon^2)\end{aligned}\ ] ] which scale with @xmath88 and @xmath89 as @xmath90 corresponding to our desire to interpret the perturbation as a gravitational wave we suppose that the perturbation wavelength is very much less than the scale on which the background varies : i.e. , @xmath91 . the ratio of the magnitude of @xmath92 to @xmath93 is then @xmath94 . now consider an order @xmath95 coordinate gauge transformation generated by the vector field @xmath96 . the difference between the perturbed riemann curvature in the old and the new gauge is given by the lie derivative of @xmath97 with respect to @xmath98 . the order @xmath49 gauge terms are @xmath99 where an overbar represents a quantity in the new gauge and covariant derivatives with respect to the background metric are denoted by a vertical bar ( @xmath100 ) . the generator of the gauge transformation is independent of @xmath49 and @xmath101 ; so @xmath102 correspondingly , the gauge terms are @xmath103 , which is @xmath104 relative to @xmath105 . we conclude , with isaacson @xcite , that as long as @xmath106 the first - order perturbation to the riemann @xmath105 ( and , similarly , the ricci @xmath107 and ricci scalar @xmath108 ) is physical : i.e. , gauge independent . the riemann tensor satisfies a non - linear wave equation that , in vacuum , may be written @xmath109 where @xmath110\nonumber\\ & \qquad{}- r\indices{^a_{bc}^f}r\indices{_{afde } } - \left [ r\indices{^a_{bd}^f}r\indices{_{acfe } } - r\indices{^a_{be}^f}r\indices{_{acfd } } \right]\end{aligned}\ ] ] is quadratic in the riemann ( * ? ? ? * eq . 6.9 - 10 ) . focusing on the @xmath111 terms in this equation we have @xmath112 thus , for @xmath113 the perturbation satisfies the vacuum wave equation on the background . we conclude that as long as @xmath113 i.e. , the curvature radius of the perturbation is small compared to the curvature scale of the background the first - order perturbation to the riemann is a gauge independent wave on the background described by the unperturbed spacetime . having established that short wave perturbations to the riemann are gauge independent and satisfy the wave equation on the background spacetime we ask how , for short - wave perturbations , the observed clock phase is affected by the presence of gravitational waves . as found in section [ sec : diffeq ] the observed clock phase is given by ( see eq . [ eq : riemannresponse ] ) @xmath114d\lambda } \right . \nonumber\\ & \qquad\qquad \left . \vphantom { \left(\frac{d\phi / d\tau}{d\phi / dt}\right)^2 \left.\left({{\partial_{\lambda}}}\cdot{\bm{a}}_c\right)\right|_{\lambda=0 } - \left.\left({{\partial_{\lambda}}}\cdot{\bm{a}}_r\right)\right|_{\lambda=\lambda_r } -\left(\frac{d\phi}{/d\tau}\right)^2 \int_0^{\lambda_{r } } \left[{\bm{k}}^2-{\bm{r}}({{\partial_{\phi}}},{{\partial_{\lambda}}},{{\partial_{\phi}}},{{\partial_{\lambda}}})\right]d\lambda } -\left(\frac{d\phi}{d\tau}\right)^2 \int_0^{\lambda_{r } } \left[{\bm{k}}^2-{\bm{r}}({{\partial_{\phi}}},{{\partial_{\lambda}}},{{\partial_{\phi}}},{{\partial_{\lambda}}})\right]d\lambda \right\}\,\end{aligned}\ ] ] where @xmath115 and @xmath116 are the 4-accelerations of the receiver / observer and the clock , @xmath5 is the null - geodesic field connecting the clock and receiver / observer , and @xmath117 and @xmath118 are the clock and receiver / observer proper times as measured at either end of the connecting null geodesics . of the terms on the right - hand side of equation ( [ eq : riemannresponserecap ] ) the gravitational wave perturbation does not affect @xmath13 , @xmath27 , which are intrinsic properties of the clock measured in the clock s proper reference frame . the remaining terms @xmath119 , @xmath115 , @xmath116 , @xmath5 , @xmath10 , @xmath120 , and @xmath97 are all perturbed and contribute to the perturbation of @xmath26 . in appendix [ sec : scaling ] we evaluate the scaling of all the contributions to the perturbation of equation [ eq : riemannresponserecap ] with respect to the gravitational wave s amplitude ( described by @xmath95 ) , its angular frequency ( @xmath89 ) and the ratio of the radiation wavelength to the background curvature scale ( @xmath121 ) . at first order in @xmath49 the terms in equation [ eq : riemannresponserecap ] scale as [ eq : scalings ] @xmath122 } & \sim \mathcal{o}(1)\\ { \delta^1_\varphi\left[\frac{d\phi / dt}{{{\partial_{\lambda}}}\cdot{\bm{u } } } \left(\frac{d\phi / d\tau}{d\phi / dt}\right)^2 \left.\left({{\partial_{\lambda}}}\cdot{\bm{a}}_c\right)\right|_{\lambda=0}\right ] } & \sim\mathcal{o}(1)\\ { \delta^1_\varphi\left[\frac{d\phi / dt}{{{\partial_{\lambda}}}\cdot{\bm{u}}}\left.\left({{\partial_{\lambda}}}\cdot{\bm{a}}_r\right)\right|_{\lambda=\lambda_r}\right ] } & \sim \mathcal{o}(1)\\ { \delta^1_\varphi\left[\frac{d\phi / dt}{{{\partial_{\lambda}}}\cdot{\bm{u}}}\left(\frac{d\phi}{d\tau}\right)^2 \int_0^{\lambda_{r } } { \bm{k}}^2 d\lambda\right ] } & \sim \mathcal{o}(\omega^2\delta)\\ { \delta^1_\varphi\left[\frac{d\phi / dt}{{{\partial_{\lambda}}}\cdot{\bm{u}}}\left(\frac{d\phi}{d\tau}\right)^2 \int_0^{\lambda_{r } } { \bm{r}}({{\partial_{\phi}}},{{\partial_{\lambda}}},{{\partial_{\phi}}},{{\partial_{\lambda}}})d\lambda\right ] } & \sim\mathcal{o}(\omega^2 ) . \end{aligned}\ ] ] thus we find our principal result : in the limit of high frequency gravitational waves ( i.e. , @xmath123 : the radiation wavelength is much smaller than the background curvature scale ) the contribution to the observed evolution of the clock phase @xmath26 owing to the gravitational waves is given by the gravitational wave contribution to the riemann curvature integrated along the unperturbed null geodesic trajectory linking the clock and receiver / observer , or @xmath124 where @xmath108 is the gravitational wave contribution to the riemann curvature and all other terms correspond to their unperturbed quantities . our expression for the observed doppler , @xmath125d\lambda } \right . \nonumber\\ & \qquad\qquad \left . \vphantom { \left(\frac{d\phi / d\tau}{d\phi / dt}\right)^2 \left.\left({{\partial_{\lambda}}}\cdot{\bm{a}}_c\right)\right|_{\lambda=0 } - \left.\left({{\partial_{\lambda}}}\cdot{\bm{a}}_r\right)\right|_{\lambda=\lambda_r } -\left(\frac{d\phi}{/d\tau}\right)^2 \int_0^{\lambda_{r } } \left[{\bm{k}}^2-{\bm{r}}({{\partial_{\phi}}},{{\partial_{\lambda}}},{{\partial_{\phi}}},{{\partial_{\lambda}}})\right]d\lambda } -\left(\frac{d\phi}{d\tau}\right)^2 \int_0^{\lambda_{r } } \left[{\bm{k}}^2-{\bm{r}}({{\partial_{\phi}}},{{\partial_{\lambda}}},{{\partial_{\phi}}},{{\partial_{\lambda}}})\right]d\lambda \right\}\,,\end{aligned}\ ] ] ( see eq . [ eq : riemannresponse ] ) is manifestly gauge - invariant . each term on the right hand side of equation [ eq : dopplerdisc ] is either an intrinsic property of the clock ( @xmath13 , @xmath27 ) or a spacetime scalar . equation [ eq : dopplerdisc ] was derived entirely in terms of the physical measurements made at the receiver / observer or at the clock . at no point was it necessary to identify special coordinates , fix a gauge , or introduce privileged observers . furthermore , each term is composed from quantities that are physical observables or , equivalently , well - defined spacetime properties : * @xmath43 and @xmath20 are the clock and receiver / observer 4-acceleration , just as would be measured by an inertial accelerometer co - located with the clock or the receiver ; * @xmath5 is the tangent to the null geodesic that connects clock and receiver ; * @xmath6 is defined everywhere on the clock world line in terms of the clock 4-velocity and intrinsic frequency ( see eq . [ eq : bcphi ] ) , and along each of the null geodesic connecting the clock and the receiver / observer via the equation of geodesic deviation ; * @xmath29 is defined everywhere along the receiver / observer worldline in terms of the receiver s 4-acceleration and the clock s observed proper motion ( see eq . [ eq : bclambda ] ) and then everywhere along the null geodesics via the equation of geodesic deviation ; and * @xmath126 is the spacetime riemann curvature . finally , when the gravitational radiation wavelength is short compared to the background curvature lengthscale , the gravitational wave contribution to the riemann curvature is a physical , gauge invariant property of spacetime . under these conditions equation [ eq : responseperturbation ] gives the leading contribution to the observed doppler , @xmath127 where @xmath128 is the gravitational wave contribution to the riemann curvature and the integration is along the unperturbed null geodesic trajectories of the background spacetime . each of the terms on the right hand side of equation [ eq : dopplerdisc ] also has a clean , unambiguous physical interpretation . the first term , involving @xmath27 , accounts for the intrinsic evolution of the clock s frequency . the second and third terms are the contributions owing to the relativistic generalization of the classical doppler effect associated with the receiver / observer acceleration along the path from clock to receiver / observer . the remaining two terms , involving integrals along the null geodesic from the clock to the receiver , describe the light - time contributions to the doppler . recalling ( eq . [ eq : k ] ) @xmath129 the term in @xmath130 accounts for what we term _ kinematic _ light - time contributions : kinematic because they are associated with the spacetime null path evolution without regard to the cause behind the path evolution . finally , the term involving the integral of @xmath47 is the light - time contribution owing to the effects of general relativistic gravity along the path . since the contribution from the integral of @xmath47 along the null geodesics is the only contribution that vanishes when `` gravity '' vanishes ( i.e. , spacetime is flat ) and , as was shown in section [ sec : scaling ] , dominates the gravitational wave contribution when waves are present , it may be fairly said to encapsulate the unique contribution of relativistic gravity to the observed doppler . it is instructive to compare our expression for the observed clock doppler to the timing model used today for pulsar timing . in that model @xcite the time - varying contributions to the expected pulse arrival times are identified individually and , for relativistic contributions , computed perturbatively in a combination of specific coordinate systems that are related to each other through appropriate coordinate transformations @xcite . all effects relativistic and non - relativisitc are grouped together in a small set of high - level `` delays '' ( sometimes `` corrections '' ) of common conceptual or perturbative origin . the delays associated with the vacuum propagation of the electromagnetic pulse are referred to as the rmer , einstein and shapiro delays @xcite . the rmer delay terms are strictly `` geometric '' e.g. , the light travel time across the solar system ( more precisely , between the pulse arrival at the solar system barycentre and the observatory ) without taking into account any relativistic effects . the rmer delay terms also includes light - time corrections owing to pulsar proper motion ( i.e. , the shklovskii effect @xcite ) , corrections owing to parallax , and corrections owing to aberration ( associated with earth s transverse velocity ) . the einstein delay accounts for the special - relativistic time - dilation and gravitational redshift effects as earth moves in its orbit and through the gravitational potential of sol and luna . finally , the shapiro delay describes additional delays owing to the shapiro effect @xcite associated with sol and other solar system bodies . ( if the pulsar is part of a binary system there are an additional three terms corresponding to these same effects in the pulsar s binary system . ) when equation [ eq : dopplerdisc ] is applied to pulsar timing the world lines describing the pulsar and earth s spacetime trajectories are geodesic ; so , ignoring the displacement of the observatory from earth s worldline , the acceleration terms in equation [ eq : dopplerdisc ] vanish . of the three remaining terms , one is due to the pulsar s period evolution evaluated at the retarded time and adjusted by @xmath131 , which includes some of what are presently computed as the einstein delay corrections . recalling equation [ eq : kdisc ] for @xmath29 , the integral of @xmath130 along the pulsar - earth null geodesic accounts for contributions owing to the change in the null geodesic path as a function of receiver / observer time : e.g. , the rmer delay and the shklovskii effect @xcite . writing @xmath119 as @xmath132 we see that the einstein delay corrections to these terms are already included . finally , the shapiro delay ( and , of course , gravitational wave ) corrections are part of the integral of @xmath47 over the null geodesic . a fuller discussion of how equation [ eq : dopplerdisc ] leads to a pulsar timing formula is part of a work in progress . pulsar timing for gravitational wave detection is generally aimed at detecting gravitational waves with periods on order years @xcite . in earth s neighborhood the background curvature scale owing to sol is on order two light - months ; in the neighborhood of the pulsar the curvature scale is much smaller . for pulsar timing , then , the condition @xmath134 , which is required to distinguish gravitational waves and background , does not hold near earth or near the pulsar . nevertheless , in the case of pulsar timing the failure of the condition @xmath123 near earth ( or the pulsar ) is not significant . the key to understanding how this is so is to recognize that the gravitational wave response is contributed by the _ integral of _ @xmath47 _ along the null geodesic from the pulsar to the earth - bound observatory . _ over most of that path @xmath123 : i.e. , it is over only a very small fraction of that path near , near its endpoints , that there is any ambiguity over what is wave and what is background . since the mass of sol and the mass of the pulsar are of the same order the spacetime curvature at the same distance from each is also of the same order . without loss of generality , focus attention on the importance of the spacetime curvature owing to sol in earth s neighborhood . the background curvature scale a distance @xmath135 from sol is @xmath136 ; so , @xmath137 if we require @xmath138 then it is only at distances @xmath139 where @xmath140 is the angle between the wave propagation direction and the earth - pulsar line - of - sight , that we need be concerned . ( for waves propagating with @xmath141 this is large ; however , @xmath142 ; so , when @xmath135 might be large the gravitational wave contribution vanishes . ) this corresponds to a fraction @xmath143 of a radiation wavelength and a similarly small fraction of the contribution to the integral of @xmath47 . we conclude that the failure to satisfy @xmath123 near either earth or the pulsar does not significantly affect the use of pulsar timing for gravitational wave detection with wavelengths greater than the spacetime curvature in the solar system . do similar problems arise when we treat gravitational waves as ( tt gauge ) metric perturbation ? they do , but in those treatments the problems have been overlooked . in conventional treatments the gravitational wave contribution to pulsar arrival times is proportional to the difference of the anti - derivative of what is supposed to be the tt gauge metric perturbation at earth and the retarded - time anti - derivative at the pulsar . we say `` supposed '' because the metric perturbations at earth and at the pulsar must be computed as if spacetime at earth the pulsar is flat : i.e. , that the pulsar , earth , sol and the solar system have no effect whatsoever on the metric perturbation representation of the gravitational wave perturbation . this is far from the case , especially near the pulsar . reviewing equation [ eq : dopplerdisc ] , however , we see that the conventional result comes about when @xmath123 over the null geodesic connecting the pulsar and earth except for a small ( relative to the gravitational radiation wavelength ) portion of the path near its endpoints . equation [ eq : dopplerdisc ] may also be applied to find a gauge invariant description of the response of laser interferometric gravitational wave detector . referring to figure [ fig : ifo ] , which shows in schematic a single - pass delay line interferometer , the interferometer s response is the phase difference @xmath144 of the clock s image reflected from the @xmath145-arm beamsplitter ( @xmath146 ) and the @xmath147-arm beamsplitter ( @xmath148 ) . equation [ eq : dopplerdisc ] describes how the phases @xmath146 and @xmath148 are observed at the output port . it also describes how @xmath146 and @xmath148 are determined from the light phase at the interferometer input port . the response of an interferometric detector thus follows from the iterative use of equation [ eq : dopplerdisc ] . a detailed discusson of interferometric detector response is part of work in progress . -arm and @xmath147-arms end - mirrors are shown as vertical , solid lines . the receiver / observer at the interferometer output port records the phase difference in the light arriving along the null paths , represented by the dotted lines . each of the end - mirrors may be considered its own clock , slaved to the laser illuminating the input port of the interferometer . to describe this interferometric detector use equation [ eq : dopplerdisc ] to determine @xmath149 and @xmath150 on the respective end - mirror worldlines , and then use equation [ eq : dopplerdisc ] again to determine @xmath151 how these are observed back at the beamsplitter.,scaledwidth=50.0% ] in an arbitrary gauge the @xmath111 perturbation to the riemann curvature can be expressed in terms of the metric perturbation @xmath152 as @xmath153 where , as before , latin indices @xmath154 , @xmath155 , @xmath156 , @xmath157 and @xmath158 denote abstract indices . in minkowski space the background curvature @xmath159 vanishes . if , as is customary , we assume that in the minkowski background global lorentz frames the receiver / observer is undergoing force - free motion , then we may introduce tt coordinates in the lorentz frame in which the receiver / observer is at rest . in those coordinates @xmath160 and the coordinate components of the riemann relevant for determining the gravitational wave response are @xmath161 where the indices in equation [ eq : ttgauge ] denote coordinate components . thus , when tt gauge exists we recover the usual representation of gravitational waves in terms of the tt gauge metric perturbation . in the view taken here , however , this association validates the tt - gauge description of gravitational waves and not the other way around . an advantage of our gauge - independent treatment is that it applies even in environments where special gauges e.g. , tt gauge do not exist . numerical relativity exploits the coordinate gauge freedom to place the einstein equations in forms that allow for stable evolution and the use of standard solution techniques ( e.g. , `` black - box '' elliptic , parabolic of first - order hyperbolic system solvers ) @xcite . to convert the results of these simulations to a tt - gauge metric perturbation is a non - trivial excercise that introduces additional , non - trivial numerical errors ; however , a direct product of numerical relativity simulations is the connection and the riemann tensor , from which the response of laser interferometric , spacecraft doppler tracking or pulsar timing gravitational wave detector response can be determined without the need for solving the gauge equations necessary to transform the simulation results into tt gauge . expressing the detector response in terms of the gauge independent riemann , as opposed to the metric perturbation in a specific gauge , eliminates unecessary steps and associated numerical errors when using numerical relativity simulations to aid in the analysis or interpretation of gravitational wave observations . with equations [ eq : dopplerdisc ] , [ eq : dopplerpertdisc ] and the intermediate results summarized in appendix [ sec : scaling ] we are in a position to address the mistatements and misconceptions , highlighed in section [ sec : intro ] , that have appeared in the pedagogical and research literature regarding how gravitational wave detectors function and what they respond to . as is common in these discussions we specialize to a minkowski background with the clock and receiver / observer at relative rest in the background . in this case equation [ eq : dopplerdisc ] and [ eq : dopplerpertdisc ] together reduce to @xmath162 where the integration path is along the appropriate ( piecewise ) null geodesics of the background spacetime . an important property of this result is its agreement with the einstein equivalence principle @xcite , which asserts that true gravitational effects are discernible only in experiments of sufficiently large scale . this should be no less the case for gravitational waves than for all other gravitational phenomena . in equation [ eq : dopplermink ] we see that the magnitude of the gravitational wave effect depends directly on the length of the path traversed by the electromagnetic wave . viewed from the perspective of gravitational wave detection via pulsar timing ( or spacecraft doppler tracking ) , the gravitational wave contribution to the observed pulse arrival time at earth depends on the riemann curvature along the path traversed by the electromagnetic signal from the distant pulsar : i.e. , contrary to the shibboleth ( * ? ? ? * e.g. , ) that pulsar timing measures the effect of gravitational waves near earth and the pulsar , _ the spacetime properties in the immediate neighborhood of the earth and the pulsar are irrelevant for the detection of gravitational waves via pulsar timing . _ for a graphical illustration of this latter point , refer to figure [ fig : curvature ] , which shows the worldlines of a clock and a receiver , propagating freely in a spacetime that is flat in their respective neighborhoods . a `` lump '' of curvature arises between them , persists for some time , and then evaporates . as shown , the curvature leads to variations in the pulse arrival times at the receiver even though spacetime in the neighborhood of the clock and receiver is always flat and described by the minkowskii metric : i.e. , there is no `` metric perturbation '' on either worldline . for discussion.,scaledwidth=50.0% ] we conclude that it is gravitational wave induced difference in light travel time as it propagates along the unperturbed null trajectories connecting the clock and observer that leads to the detector response . there is no meaningful or observable sense in which the gravitational waves accelerate or move the detector components ( cf . @xcite ) ; no significant contribution to the resonse that arises from the deflection of the electromagnetic waves by the gravitational waves ( cf . @xcite ) ; the riemann curvature tensor does not influence or affect light in any special way ( cf . finally , red / blue - shifting of the light in a detector depends only on the choice of observer and can , by appropriate choice of observer , be made to vanish ( cf . @xcite ) . the equation of geodesic deviation arises in our analysis because the vector @xmath6 is the deviation vector associated with the null geodesic congruence @xmath5 along which the phase record of the clock is propagated to the observer ( see sec . [ sec : doppler ] ) . other discussions of detector response also make use of the geodesic deviation equation ( * ? ? ? * ; * ? ? ? * e.g. , ) . what makes the discussion and result here unique ? the starting point for the discussion here is the physical measurement made at the receiver observer . the analysis that follows is fully relativistic , invokes no special coordinate systems , gauges , or privileged observers , and ends with a fully general result that involves only the intrinsic properties of the clock , the receiver / observer , their respective worldlines , and the spacetime geometry . the result is valid independent of the detector size relative to the radiation wavelength and even when a distinction between the gravitational radiation and the background curvature is ambiguous . in the analysis presented here the equation of geodesic deviation arises as a matter of course and not as a fundamental precept . in contrast , @xcite begin by asserting that the geodesic deviation equation describes the instantaneous spatial separation ( in some ill - defined detector `` rest - frame '' ) between the coordinate stationary world lines of the clock ( laser / beamsplitter ) and the receiver / observer ( end mirror ) , which may not be geodesics . the spatial separation divided by the light - speed @xmath156 , is then supposed to give the light travel time from clock to observer . finally , the observed phase is calculated under the assumption that the gravitational radiation wavelength is much longer than the detector size or light - storage time . in these descriptions geodesic deviation is applied heuristically to characterize a coordinate dependent non - deviation vector between two non - geodesics . previous analyses that depend upon or involve in some way the geodesic deviation equation , while doubtless having some heuristic value , offer , at best , results valid only in very special circumstances : a minkowski background , detectors small compared to a radiation wavelength and with light storage times short compared to the radiation period . they are not generalizable to backgrounds that are not minkowski , detectors that approach or exceed a radiation wavelength ( e.g. , lisa , spacecraft doppler tracking or pulsar timing ) , or detectors whose light storage time approaches or exceeds a radiation period ( e.g. , ligo and khz frequency gravitational waves associated with neutron star binary coalescence ) . finally , these derivations , through their focus on special coordinate systems where clock and receiver position and spatial separation can be said to play an important role , suggest that , in some way , gravitational radiation violates the equivalence principle and can be measured at a point . in this way the heuristic picture they offer is not that different than one focused upon metric perturbations . most modern gravitational wave detectors including laser interferometers , pulsar timing arrays , spacecraft doppler tracking rely upon the waves effect on light travel time to detect the wave s presence . the response of these detectors is conventionally expressed in terms of a metric perturbation description of the gravitational waves . metric perturbations , however , are coordinate gauge dependent quantities . their use obscures the physical nature of the gravitational wave phenomena and the principles that governs the operation of these detectors @xcite and lead to `` explanations '' of detector function that are at odds with the equivalence principle , which is the most basic assertion of general relativity . the response of modern gravitational wave detectors may also be expressed in terms of their interaction with spacetime curvature , which is the physically unambiguous property of spacetime that , in general relativity , underlies all gravitational phenomena . such an expression of the detector response , and each term that comprises it , is separately gauge invariant and has a clear physical interpretation . the expression for the response is manifestly consistent with the equivalence principle : i.e. , the role of all gravitational phenonmena in the response wave or otherwise clearly involves measurements made over finite spacetime intervals . the gravitational wave contribution to the response is separately identifiable and the way in which it leads to light - time fluctuations in the detector avoids all the suspect analogies ( moving mirrors , redshifting and blueshifting of light , etc . ) that plague descriptions of detector functioning based upon metric perturbations . contributions from other physical phenomena also appear naturally : e.g. , in the case of pulsar timing , all geometrical and kinematical effects ( rmer delay , parallax , aberration , shklovskii effect @xcite ) , special relativistic and general relativistic effects ( special and general relativistic einstein delay , shapiro time delay and gravitational wave contributions ) all appear naturally . the `` curvature - based '' response function describes all light - time based detectors in all regimes : i.e. , terrestrial laser interferometric detectors , whose physical dimensions and light storage times are shorter than the typical radiation wavelength @xcite ; proposed space - based detectors based upon the lisa heritage , where the detector size and light storage time is on - order the radiation wavelength @xcite ; and spacecraft doppler and pulsar timing , where the radiation wavelength is very much shorter than the detector size or light storage time @xcite . it meshes well with the use of numerical relativity calculation to determine detector response : i.e. , it does not require extracting from these calculations gauge dependent quantities that are not directly calculated in order to compute the director response . in companions papers we shall explore in greater depth how expressing the gravitational wave response in terms of the riemann curvature can be applied to delay - line and fabrey - perot interferometric detectors ; the role of non - uniform motion ( e.g. , rotation ) in lisa - heritage detectors ; the effects of micro - lensing in pulsar timing gravitational wave detection @xcite ; and the use of the riemann curvature response function to provide an alternative derivation of a general - relativistic pulsar timing formula . sometime in the next decade gravitational wave observations will begin to constrain our understanding of astronomical phenomena : when that moment arrives , gravitational wave astronomy will be born . the acceptance of gravitational wave observations as a tool of astronomical discovery can only be hastened by an intuitive , simple and physically correct understanding of gravitational waves and their detection : what they are , how they are generated , how they propagate , and how they interact with a detector . expressing the gravitational wave detector response in terms of the riemann curvature is but one step in that direction . lsf thanks the aspen center for physics for their hospitality and thanks david garfinkle and john baker for discussions . this work was supported by national science foundation grant numbers 09 - 40924 and 09 - 69857 awarded to the pennsylvania state university and supported by the natural science and research council of canada . in section [ sec : gwresp ] we concluded that the gravitational wave contribution to @xmath26 depends only on the integral of the gauge independent perturbation to the riemann curvature along the unperturbed null geodesic connecting clock and receiver / observer . our conclusion rests upon the scaling of the different contributions to the righthand side of equation [ eq : riemannresponserecap ] with respect to the radiation s amplitude ( described by the small parameter @xmath49 ) , it s angular frequency ( @xmath89 ) , and the ratio of the background curvature scale ( @xmath88 ) to the radiation wavelength ( @xmath163 ) . in this appendix we establish those scalings , which we summarized in equations [ eq : scalings ] . recall @xmath166 the clock is presumed to be much smaller than a radiation wavelength . since @xmath13 is an intrinsic property of the clock it is unaffected by the perturbation . taking advantage of equations [ eq : orderpuv ] we conclude @xmath167 & = \mathcal{o}(1).\end{aligned}\ ] ] the accelerations @xmath115 and @xmath43 are the result of applied forces : i.e. , they are unaffected by the gravitational wave perturbation . taking advantage of equations [ eq : orderpuv ] , @xmath169 & \sim \mathcal{o}(1 ) . \end{aligned}\ ] ] the perturbation of an integral of the form @xmath173 involves the perturbation of the integration @xmath174 and @xmath120 , the integration path @xmath175 , and the scalar function @xmath176 . in our case the integration endpoints are defined to be where @xmath178 is the point on the receiver / observer worldline along the null geodesic path @xmath175 leaving the clock worldline at @xmath179 . by definition @xmath180 at @xmath179 . denoting the perturbed path and the perturbed integration endpoint @xmath120 write the perturbed integral as @xmath182\ , d\lambda\nonumber\\ & \qquad{}+\int_0^{\epsilon\delta\lambda_r^{(1 ) } } f^{(0)}(x^{(0)}(\lambda_r^{(0)}+\lambda))d\lambda . \end{aligned}\ ] ] the last term on the right hand side of equation [ eq : pertint ] may be further expressed @xmath183 since the perturbation @xmath184 is @xmath185 the path perturbation @xmath186 and the integration endpoint perturbation @xmath187 are also @xmath185 . correspondingly , @xmath188 & = \int_0^{\lambda^{(0)}_r}\left [ f^{(1)}(x^{(0)\mu}(\lambda ) ) + \delta x^{(1)\mu}\frac{\partial f^{(0)}}{\partial x^\mu}(x^{(0)\mu}(\lambda ) ) \right]\ , d\lambda + \delta\lambda_r^{(1)}f^{(0)}(\lambda_r^{(0)})\end{aligned}\ ] ] 57ifxundefined [ 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 10.1007/bf02723171 [ * * , ( ) ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1007/s10714 - 007 - 0415 - 5 [ * * , ( ) ] link:\doibase 10.1103/physrevd.79.022002 [ * * ( ) , 10.1103/physrevd.79.022002 ] @noop _ _ ( , ) link:\doibase 10.1103/physrev.166.1263 [ * * , ( ) ] link:\doibase 10.1103/physrev.117.306 [ * * , ( ) ] @noop _ _ ( , , ) @noop * * , ( ) http://www.ligo.caltech.edu/ligo_web/rescomm/rescomm_home.html/docs/p/p720002-01-r.pdf[__ ] , ( , , ) @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1086/157593 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.98.111102 [ * * , ( ) ] link:\doibase 10.1103/physrevd.78.042003 [ * * , ( ) ] link:\doibase 10.1103/physrevd.78.122002 [ * * , ( ) ] link:\doibase 10.1103/physrevlett.110.171102 [ * * , ( ) ] @noop _ _ , ( , , ) http://elmer.tapir.caltech.edu/ph237/ [ `` , '' ] ( ) , link:\doibase 10.1119/1.18578 [ * * , ( ) ] link:\doibase 10.1088/0004 - 637x/752/1/54 [ * * , ( ) ] link:\doibase 10.1103/physrevd.79.084030 [ * * , ( ) ] , @noop _ _ ( , ) link:\doibase 10.1088/0004 - 637x/693/2/1113 [ * * , ( ) ] , link:\doibase 10.1146/annurev.aa.24.090186.002541 [ * * , ( ) ] @noop * * , ( ) http://www.livingreviews.org/lrr-2006-1 ( cited on 28 sep 2008 ) [ * * ( ) ] link:\doibase 10.12942/lrr-2010 - 1 [ * * ( ) , 10.12942/lrr-2010 - 1 ] link:\doibase 10.12942/lrr-2011 - 5 [ * * ( ) , 10.12942/lrr-2011 - 5 ] @noop * * , ( ) @noop * * , ( ) link:\doibase 10.1103/physrev.166.1272 [ * * , ( ) ] link:\doibase 10.1007/s10714 - 009 - 0773 - 2 [ * * , ( ) ] @noop * * , ( ) , @noop _ _ ( , , ) @noop _ _ ( , , ) @noop * * , ( ) http://archive.numdam.org/archive/aihpa/aihpa_1986__44_3/aihpa_1986__44_3_263_0/aihpa_1986__44_3_263_0.pdf [ * * , ( ) ] link:\doibase 10.12942/lrr-2003 - 5 [ * * ( ) , 10.12942/lrr-2003 - 5 ] link:\doibase 10.1111/j.1365 - 2966.2006.10302.x [ * * , ( ) ] link:\doibase 10.1111/j.1365 - 2966.2006.10870.x [ * * , ( ) ] http://archive.numdam.org/archive/aihpa/aihpa_1986__44_3/aihpa_1986__44_3_263_0/aihpa_1986__44_3_263_0.pdf [ * * , ( ) ] @noop * * , ( ) link:\doibase 10.1103/physrevlett.13.789 [ * * , ( ) ] @noop `` , '' ( ) , link:\doibase 10.1088/0264 - 9381/18/17/202 [ * * ( ) , 10.1088/0264 - 9381/18/17/202 ] @noop `` , '' ( ) link:\doibase { 10.1088/0264 - 9381/27/8/084006 } [ * * ( ) , ] , link:\doibase 10.1088/1742 - 6596/203/1/012074 [ * * , ( ) ] http://stacks.iop.org/0264-9381/25/184017 [ * * , ( ) ] link:\doibase 10.1088/0264 - 9381/26/9/094014 [ * * , ( ) ] http://sci.esa.int/science-e/www/object/index.cfm?fobjectid=48364[__ ] , ( ) @noop * * , ( ) link:\doibase { 10.1088/0264 - 9381/27/8/084013 } [ * * ( ) , ] , @noop * * , ( ) @noop * * , ( )

we relate the notion of matroid pathwidth to the minimum trellis state - complexity ( which we term trellis - width ) of a linear code , and to the pathwidth of a graph . by reducing from the problem of computing the pathwidth of a graph , we show that the problem of determining the pathwidth of a representable matroid is np - hard . consequently , the problem of computing the trellis - width of a linear code is also np - hard . for a finite field , we also consider the class of-representable matroids of pathwidth at most , and correspondingly , the family of linear codes over with trellis - width at most . these are easily seen to be minor - closed . since these matroids ( and codes ) have branchwidth at most , a result of geelen and whittle shows that such matroids ( and the corresponding codes ) are characterized by finitely many excluded minors . we provide the complete list of excluded minors for , and give a partial list for . matroids , pathwidth , linear codes , trellis complexity , np - hard . 05b35 , 94b05

0705.1384

the notion of pathwidth of a matroid has received some recent attention in the matroid theory literature @xcite , @xcite . this notion has long been studied in the coding theory literature , where it is used as a measure of trellis complexity of a linear code @xcite , @xcite , @xcite . however , there appears to be no standard coding - theoretic nomenclature for this notion . it has been called the state complexity of a code in @xcite , but the use of this term there conflicts slightly with its use in @xcite . so to avoid ambiguity , we will give it a new name here _ trellis - width _ which acknowledges its roots in trellis complexity . the relationship between matroid pathwidth and code trellis - width can be made precise as follows . to an arbitrary linear code @xmath4 over a finite field @xmath0 , we associate a matroid , @xmath5 , which is simply the vector matroid , over @xmath0 , of any generator matrix of the code . recall that in coding theory , a matrix @xmath6 is called a generator matrix of a code @xmath4 , if @xmath4 is the rowspace of @xmath6 . consequently , the matroid @xmath5 does not depend on the actual choice of the generator matrix , and so is a characteristic of the code @xmath4 . the code @xmath4 may in fact be viewed as a representation over @xmath0 of the matroid @xmath5 . the trellis - width of @xmath4 is simply the pathwidth of @xmath5 ; we will give the precise definition of matroid pathwidth in section [ pw_section ] . it has repeatedly been conjectured in the coding theory literature that computing the trellis - width of a linear code over a fixed finite field @xmath0 is np - hard @xcite , @xcite , ( * ? ? ? * section 5 ) . this would imply that the corresponding decision problem ( over a fixed finite field @xmath0 ) given a generator matrix for a code @xmath4 over @xmath0 , and a positive integer @xmath1 , deciding whether or not the trellis - width of @xmath4 is at most @xmath1 is np - complete . this decision problem has been given various names `` maximum partition rank permutation '' @xcite , `` maximum width '' @xcite and `` trellis state - complexity '' @xcite . an equivalent statement of the trellis - width conjecture above is the following : given a matrix @xmath7 over @xmath0 , the problem of computing the pathwidth the vector matroid @xmath8 $ ] is np - hard . in this paper , we prove the above statement for any fixed field @xmath0 , not necessarily finite . our proof is by reduction from the problem of computing the pathwidth of a graph , which is known to be np - hard @xcite , @xcite . thus , in particular , computing the trellis - width of a linear code over @xmath0 is np - hard , which settles the aforementioned coding - theoretic conjecture . the situation is rather different if we weaken the trellis - width decision problem above by _ not _ considering the integer @xmath1 to be a part of the input to the problem . in other words , for a fixed finite field @xmath0 , and a _ fixed _ integer @xmath9 , consider the following problem : + _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ given a length-@xmath10 linear code @xmath4 over @xmath0 , decide whether or not @xmath4 has trellis - width at most @xmath1 . + _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ the equivalent decision problem for matroid pathwidth would be to decide ( for a fixed finite field @xmath0 and integer @xmath9 ) whether or not a given @xmath0-representable matroid has pathwidth at most @xmath1 . based on results from the structure theory of matroids @xcite , we strongly believe that these problems are solvable in polynomial time . in the process of studying matroids of bounded pathwidth , we observe that for any finite field @xmath11 and integer @xmath9 , the class , @xmath12 , of @xmath13-representable matroids having pathwidth at most @xmath1 , is minor - closed and has finitely many excluded minors . as a relatively easy exercise , we show that the list of excluded minors for @xmath14 consists of with ground set @xmath15 and rank function @xmath16 to be @xmath17 for @xmath18 . therefore , what we consider to be matroids of pathwidth one would be matroids of pathwidth two in @xcite , @xcite . ] @xmath19 , @xmath20 , @xmath21 and @xmath22 . unfortunately , the problem of finding excluded - minor characterizations of @xmath12 for @xmath23 becomes difficult very quickly . we give a list of excluded minors for @xmath24 , which is probably not complete . the rest of the paper is organized as follows . in section [ prelim_section ] , we lay down the definitions and notation used in the paper . in section [ np_section ] , we prove that , for any fixed field @xmath0 , the problem of computing the pathwidth of an @xmath0-representable matroid is np - hard , and therefore , so is the problem of computing the trellis - width of a linear code over @xmath0 . finally , in section [ bounded_pw_section ] , we consider the class of matroids @xmath12 . we give the complete lists of excluded minors for @xmath14 and the corresponding family of linear codes over @xmath13 having trellis - width at most one . we also give a partial list of excluded minors for @xmath24 . we assume familiarity with the basic definitions and notation of matroid theory , as expounded by oxley @xcite . the main results and proofs in this paper will be given in the language of matroid theory , rather than that of coding theory , as it is easier to do so . however , as our results may be of some interest to coding theorists , we make an effort in this section to provide the vocabulary necessary to translate the language of matroid theory into that of coding theory . definitions of coding - theoretic terms not explicitly defined here can be found in any text on coding theory ( _ e.g. _ , @xcite ) . let @xmath4 be a linear code of length @xmath10 over the finite field @xmath25 . the dimension of @xmath4 is denoted by @xmath26 , and the coordinates of @xmath4 are indexed by the integers from the set @xmath27 = \{1,2,\ldots , n\}$ ] as usual . we will also associate with the coordinates of @xmath4 a set , @xmath28 , of _ coordinate labels _ , so that there is a bijection @xmath29 \rightarrow e(\cc)$ ] label sequence _ of @xmath4 is defined to be the @xmath10-tuple @xmath30 , where @xmath31 . for notational convenience , we will simply let @xmath32 denote the label sequence of @xmath4 . unless specified otherwise ( as in the case of code minors and duals below ) , we will , by default , set @xmath28 to be @xmath27 $ ] , and @xmath32 to be the @xmath10-tuple @xmath33 . in such a case , the label of each coordinate is the same as its index . given a code @xmath4 over @xmath13 , specified by a generator matrix @xmath6 , we define its _ associated matroid _ @xmath5 to be the vector matroid , @xmath34 $ ] , of @xmath6 . we identify the ground set of @xmath5 with @xmath28 . note that if @xmath6 and @xmath35 are distinct generator matrices of the code @xmath4 , then @xmath34 = m[g']$ ] , and hence , @xmath5 is independent of the choice of generator matrix . thus , any generator matrix of @xmath4 is an @xmath13-representation of @xmath5 . conversely , if @xmath36 is an @xmath13-representable matroid , and @xmath6 is an @xmath13-representation of @xmath36 , then @xmath37 for the code @xmath4 generated by @xmath6 . thus , each @xmath13-representable matroid is associated with some code @xmath4 over @xmath13 . for any code @xmath4 , the dual code , @xmath38 , is specified to have the same label sequence as @xmath4 , _ i.e. _ , @xmath39 . it is a particularly nice fact ( * ? ? ? * theorem 2.2.8 ) that the matroids associated with @xmath4 and @xmath38 are dual to each other , _ i.e. _ , @xmath40 . given a @xmath41 , we will denote by @xmath42 ( resp . @xmath43 ) the code obtained from @xmath4 by puncturing ( resp . shortening at ) those coordinates having labels in @xmath44 . thus , @xmath45 . a _ minor _ of @xmath4 is a code of the form @xmath46 for disjoint subsets @xmath47 . a minor of @xmath4 that is not @xmath4 itself is called a _ proper minor _ of @xmath4 . the coordinates of a minor of @xmath4 retain their labels from @xmath28 . more precisely , we set @xmath48 , and take the label sequence of @xmath46 to be the @xmath49-tuple obtained from @xmath50 by simply removing those entries that are in @xmath51 . the operations of puncturing and shortening correspond to the matroid - theoretic operations of deletion and contraction , respectively : for @xmath41 , @xmath52 we will find it convenient to use @xmath53 to denote the restriction of @xmath4 to the coordinates with labels in @xmath44 , _ i.e. _ , @xmath54 , where @xmath55 denotes the set difference @xmath56 . this allows us to express the rank function , @xmath57 , of the matroid @xmath5 as follows : for @xmath41 , @xmath58 . two length-@xmath10 linear codes @xmath4 and @xmath59 over @xmath13 are defined to be _ equivalent _ if there is an @xmath60 permutation matrix @xmath61 and an invertible @xmath60 diagonal matrix @xmath62 , such that @xmath59 is the image of @xmath4 under the vector space isomorphism @xmath63 defined by @xmath64 . informally , @xmath59 is equivalent to @xmath4 if @xmath59 can be obtained by first multiplying the coordinates of @xmath4 by some nonzero elements of @xmath13 , and then applying a coordinate permutation . in such a case , we write @xmath65 . the equivalence class of codes equivalent to @xmath4 will be denoted by @xmath66 $ ] . it is clear that if codes @xmath4 and @xmath59 are equivalent , then their associated matroids are isomorphic . we remark that code equivalence has been defined above according to the coding - theoretic convention . note that , under this definition , if @xmath59 is obtained by applying an automorphism of the field @xmath13 to @xmath4 , then @xmath4 and @xmath59 would in general be considered to be inequivalent . a family , @xmath67 , of codes over @xmath13 is said to be _ minor - closed _ if , for each @xmath68 , any code equivalent to a minor of @xmath4 is also in @xmath67 . a code , @xmath69 , over @xmath13 is said to be an _ excluded minor _ for a minor - closed family @xmath67 , if @xmath70 , but every proper minor of @xmath69 is in @xmath67 . it is easily verified that if @xmath67 is a minor - closed family , then a code @xmath4 is in @xmath67 iff no minor of @xmath4 is an excluded minor for @xmath67 . given a collection , @xmath71 , of @xmath13-representable matroids , define the code family @xmath72 evidently , if @xmath71 is a minor - closed class of @xmath13-representable matroids , then @xmath73 is also minor - closed . in this case , if @xmath74 is the set of all excluded minors for @xmath71 , then @xmath75 is the set of all excluded minors for @xmath73 . the definitions in this section rely on the notion of the connectivity function of a matroid . let @xmath36 be a matroid with ground set @xmath76 and rank function @xmath77 . its _ connectivity function _ , @xmath78 , is defined by @xmath79 for @xmath80 . note that @xmath81 , and @xmath82 . it should be pointed out that in the matroid theory literature , the prevalent definition of the connectivity function adds a ` + 1 ' to the expression we have given . we have chosen not to follow suit in order that we can give a minimum - fuss definition of trellis - width below . the connectivity function is non - negative , _ i.e. _ , @xmath83 for all @xmath80 , and submodular _ i.e. _ , @xmath84 for all @xmath85 . it is monotone under the action of taking minors if @xmath86 is a minor of @xmath36 , then for all @xmath87 , @xmath88 . finally , the connectivity function of a matroid is identical to that of its dual , _ i.e. _ , @xmath89 for all @xmath80 . given an ordering @xmath90 of the elements of @xmath36 , define the _ width _ of the ordering to be @xmath91 } \l_m(e_1,e_2,\ldots , e_i)$ ] . ( for simplicity of notation , we use @xmath92 instead of @xmath93 . ) the _ pathwidth _ of @xmath36 is defined as @xmath94 , the minimum being taken over all orderings @xmath90 of @xmath76 . an ordering @xmath90 of @xmath76 such that @xmath95 is called an _ optimal _ ordering . since @xmath96 , it is clear that @xmath97 . another useful and easily verifiable property of pathwidth is that , for matroids @xmath98 and @xmath99 , the pathwidth of their direct sum , @xmath100 , equals @xmath101 . the property of pathwidth most important for our purposes is stated in the following lemma . + if @xmath86 is a minor of @xmath36 , then @xmath102 . [ minor_pathwidth_lemma ] let @xmath103 be an optimal ordering of @xmath76 . it is enough to show the result in the case when @xmath104 or @xmath105 for some @xmath106 $ ] . in such a case , consider the ordering @xmath107 of @xmath108 . for @xmath109 , we have @xmath110 . for @xmath111 , we have @xmath112 it follows that @xmath113 , and hence , @xmath102 . + the _ trellis - width _ of a linear code @xmath4 over @xmath13 is defined to be @xmath114 . for a discussion of the motivation and practical implications of this definition , we refer the reader to ( * ? ? ? * section 5 ) . the pathwidth of a matroid is an upper bound on its branchwidth , a more well known measure of matroid complexity . the branchwidth of a matroid is defined via cubic trees . cubic tree _ is a tree in which the degree of any vertex is either one or three . the vertices of degree one are called _ leaves_. a _ branch - decomposition _ of a matroid @xmath36 is a cubic tree , @xmath115 , with @xmath116 leaves , labelled in a one - to - one fashion by the elements of @xmath36 . each edge @xmath117 of such a branch - decomposition @xmath115 connects two subtrees of @xmath115 , so @xmath118 has two components . we say that edge @xmath117 _ displays _ a subset @xmath80 if @xmath119 is the set of labels of leaves of one of the components of @xmath118 . the _ width _ of an edge @xmath117 of @xmath115 is defined to be @xmath120 , where @xmath119 is one of the label sets displayed by @xmath117 . the _ width _ of @xmath115 is the maximum among the widths of its edges . the _ branchwidth _ of @xmath36 is the minimum among the widths of all its branch - decompositions . note that if @xmath115 is the branch - decomposition of @xmath36 shown in figure [ branch_decomp_fig ] , then the width of @xmath115 is precisely @xmath121 . indeed , the width of any edge of @xmath115 is either @xmath122 or @xmath123 for some @xmath106 $ ] . now , for any @xmath106 $ ] , @xmath124 the inequality above arising from the submodularity of @xmath78 . since @xmath125 , either @xmath126 or @xmath123 is at least as large as @xmath122 . therefore , the width of @xmath115 is given by @xmath127 } \l_m(e_1,\ldots , e_i ) = w_m(e_1,\ldots , e_n)$ ] . it follows that the branchwidth of @xmath36 is upper - bounded by @xmath128 . in this section , we prove that for any fixed field @xmath0 , the problem of computing the pathwidth of an @xmath0-representable matroid @xmath36 , given a representation of @xmath36 over @xmath0 , is np - hard . we accomplish this by reduction from the known np - hard problem of computing the pathwidth of a graph @xcite , @xcite . the notion of graph pathwidth was introduced by robertson and seymour in @xcite . let @xmath129 be a graph with vertex set @xmath130 . an ordered collection @xmath131 , @xmath132 , of subsets of @xmath130 is called a _ path - decomposition _ of @xmath129 , if * @xmath133 ; * for each pair of adjacent vertices @xmath134 , we have @xmath135 for some @xmath136 $ ] ; and * for @xmath137 , @xmath138 . the _ width _ of such a path - decomposition @xmath139 is defined to be @xmath140 } |v_i| - 1 $ ] . the _ pathwidth _ of @xmath129 , denoted by @xmath141 , is the minimum among the widths of all its path - decompositions . a path - decomposition @xmath139 such that @xmath142 is called an _ optimal _ path - decomposition of @xmath129 . let @xmath0 be an arbitrary field . given a graph @xmath129 with vertex set @xmath130 , our aim is to produce , in time polynomial in @xmath143 , a matrix @xmath7 over @xmath0 such that @xmath141 can be directly computed from @xmath144)$ ] . the np - hardness of computing graph pathwidth then implies the np - hardness of computing the pathwidth of an @xmath0-representable matroid . the obvious idea of taking @xmath7 to be a representation of the cycle matroid of @xmath129 does not work . as observed by robertson and seymour @xcite , trees can have arbitrarily large pathwidth ; however , the cycle matroid of any tree is @xmath145 for some @xmath10 , and @xmath146 . what actually turns out to work is to take @xmath7 to be a representation of the cycle matroid of a certain graph constructible from @xmath129 in polynomial time , as we describe next . let @xmath147 be a graph defined on the same vertex set , @xmath130 , as @xmath129 , having the following properties ( see figure [ gprime ] ) : * @xmath147 is loopless ; * a pair of distinct vertices is adjacent in @xmath147 iff it is adjacent in @xmath129 ; and * in @xmath147 , there are exactly two edges between each pair of adjacent vertices . it is evident from the definition that @xmath148 is a path - decomposition of @xmath129 iff it is a path - decomposition of @xmath147 . therefore , @xmath149 . define @xmath150 to be the graph obtained by adding an extra vertex , henceforth denoted by @xmath151 , to @xmath147 , along with a pair of parallel edges from @xmath151 to each @xmath152 ( see figure [ gbar ] ) . clearly , @xmath150 is constructible directly from @xmath129 in @xmath153 time . but more importantly , the pathwidth of the cycle matroid , @xmath154 , of @xmath150 relates very simply to the pathwidth of @xmath129 , as made precise by the following proposition . + @xmath155 . + [ pw_prop ] before proving the result , we present some of its implications . for any field @xmath0 , @xmath154 is @xmath0-representable . indeed , if @xmath156 is any directed graph obtained by arbitrarily assigning orientations to the edges of @xmath150 , then the vertex - arc incidence matrix of @xmath156 is an @xmath0-representation of @xmath154 ( * ? ? ? * proposition 5.1.2 ) . it is easily verified that such an @xmath0-representation of @xmath154 can be constructed directly from @xmath129 in @xmath157 time . now , suppose that there were a polynomial - time algorithm for computing the pathwidth of an arbitrary @xmath0-representable matroid , given an @xmath0-representation for it . then , given any graph @xmath129 , we can construct an @xmath0-representation , @xmath7 , of @xmath154 , and then compute the pathwidth of @xmath8 = m(\ocg)$ ] , all in polynomial time . therefore , by proposition [ pw_prop ] , we have a polynomial - time algorithm to compute the pathwidth of @xmath129 . however , the graph pathwidth problem is np - hard . so , if there exists a polynomial - time algorithm for it , then we must have @xmath158 . this implies the following result . + let @xmath0 be a fixed field . the problem of computing the pathwidth of @xmath8 $ ] , for an arbitrary matrix @xmath7 over @xmath0 , is np - hard . + [ pw_np_thm ] as a corollary , we have that computing the trellis - width of a code is np - hard . + let @xmath0 be a fixed finite field . the problem of computing the trellis - width of an arbitrary linear code over @xmath0 , specified by any of its generator matrices , is np - hard . + [ tw_np_cor ] the remainder of this section is devoted to the proof of proposition [ pw_prop ] . since @xmath149 , for the purpose of our proof , we may assume that @xmath159 . thus , from now until the end of this section , we take @xmath129 to be a loopless graph satisfying property ( p3 ) above . note that @xmath150 also satisfies ( p3 ) . for each pair of adjacent vertices @xmath160 in @xmath129 or @xmath150 , we denote by @xmath161 and @xmath162 the two edges between @xmath163 and @xmath164 . let @xmath130 and @xmath15 denote the sets of vertices and edges of @xmath129 , and let @xmath165 and @xmath166 denote the corresponding sets of @xmath150 . we thus have @xmath167 , and @xmath168 . set @xmath169 , so that @xmath170 . note that since @xmath150 is connected ( each @xmath152 is adjacent to @xmath151 ) , we have @xmath171 . we will first prove that @xmath172 . let @xmath131 be a path - decomposition of @xmath129 . we need the following fact about @xmath139 : for each @xmath173 $ ] , @xmath174 the above equality follows from the fact that a path - decomposition , by definition , has the property that for @xmath137 , @xmath138 . for @xmath173 $ ] , let @xmath175 be the set of edges of @xmath129 that have both their end - points in @xmath176 . by condition ( ii ) in the definition of path - decomposition , @xmath177 . now , let @xmath178 , so that @xmath179 . + an ordering @xmath103 of the elements of a matroid @xmath36 is said to _ induce _ an ordered partition @xmath180 of @xmath76 if for each @xmath173 $ ] , @xmath181 @xmath182 , where @xmath183 ( and @xmath184 ) . + let @xmath185 be any ordering of @xmath166 that induces the ordered partition @xmath186 , where for each @xmath173 $ ] , @xmath187 . we claim that the width of @xmath188 is at most one more than the width of the path - decomposition @xmath139 . + @xmath189 . + [ width_lemma ] observe first that @xmath190 } \max_{1 \leq k \leq n_j - n_{j-1 } } \l_m\left(\bigcup_{i < j } e_i \cup \{e_{n_{j-1}+1 } , \ldots , e_{n_{j-1}+k}\}\right ) \nonumber \\ & \leq & \max_{j \in [ t ] } \max_{e ' \subset e_j } \l_m \left(\bigcup_{i < j } e_i \cup e'\right ) . \label{wm_eq}\end{aligned}\ ] ] let @xmath191 for some @xmath173 $ ] and @xmath192 , and consider @xmath193 . since @xmath150 is a connected graph , @xmath194 . if @xmath164 is a vertex of @xmath150 incident with an edge in @xmath119 , then @xmath195 so , the subgraph of @xmath150 induced by @xmath119 has its vertices contained in @xmath196 . therefore , @xmath197 . next , consider @xmath198 . reasoning as above , the subgraph of @xmath129 induced by @xmath199 has its vertices contained in @xmath200 . hence , @xmath201 . therefore , we have @xmath202 the last equality arising from ( [ vj_eq ] ) . hence , carrying on from ( [ wm_eq ] ) , @xmath203 } |v_j| = w_{\cg}(\cv ) + 1,\ ] ] as desired . + the fact that @xmath204 easily follows from the above lemma . indeed , we may choose @xmath139 to be an optimal path - decomposition of @xmath129 . then , by lemma [ width_lemma ] , there exists an ordering @xmath103 of @xmath76 such that @xmath205 . hence , @xmath206 . + we prove the reverse inequality in two steps , first showing that @xmath207 , and then showing that @xmath208 . + @xmath209 . + [ pw_lemma ] clearly , if @xmath131 is a path - decomposition of @xmath129 , then @xmath210 is a path - decomposition of @xmath150 . hence , choosing @xmath139 to be an optimal path - decomposition of @xmath129 , we have that @xmath211 . for the inequality in the other direction , we will show that there exists an optimal path - decomposition , @xmath212 , of @xmath150 such that @xmath213 for all @xmath214 $ ] . we then have @xmath215 being a path - decomposition of @xmath129 , and hence , @xmath216 . let @xmath217 be an optimal path - decomposition of @xmath150 , and let @xmath218 and @xmath219 . since @xmath220 for @xmath221 , we must have @xmath222 for each @xmath223 $ ] . we claim that @xmath224 is a path - decomposition of @xmath150 . we only have to show that @xmath225 , and that for each pair of adjacent vertices @xmath226 , @xmath227 for some @xmath223 $ ] . to see why the first assertion is true , consider any @xmath228 , @xmath229 . since @xmath151 is adjacent to @xmath164 , and @xmath230 is a path - decomposition of @xmath150 , @xmath231 for some @xmath136 $ ] . however , @xmath222 iff @xmath223 $ ] , and so , @xmath231 for some @xmath223 $ ] . in particular , @xmath232 for some @xmath223 $ ] . for the second assertion , suppose that @xmath160 is a pair of vertices adjacent in @xmath150 . obviously , @xmath233 for some @xmath173 $ ] . suppose that @xmath234 $ ] . we consider the case when @xmath235 ; the case when @xmath236 is similar . as @xmath225 , there exist @xmath237 $ ] such that @xmath238 and @xmath239 . without loss of generality ( wlog ) , @xmath240 . if @xmath241 , then there exists @xmath223 $ ] such that @xmath227 . if @xmath242 , we have @xmath243 and @xmath244 . hence , @xmath245 as well , and so once again , we have an @xmath223 $ ] such that @xmath246 . thus , @xmath224 is a path - decomposition of @xmath150 , with the property that @xmath222 for all @xmath223 $ ] . it must be an optimal path - decomposition , since it is a subsequence of the optimal path - decomposition @xmath230 . + to complete the proof of proposition [ pw_prop ] , it remains to show that @xmath208 . we introduce some notation at this point . recall that the two edges between a pair of adjacent vertices @xmath163 and @xmath164 in @xmath150 ( or @xmath129 ) are denoted by @xmath161 and @xmath162 . we define @xmath247 @xmath248 and @xmath249 , where @xmath151 is the distinguished vertex in @xmath250 . thus , @xmath251 and @xmath252 . note that , by construction of @xmath150 , @xmath253 , where @xmath254 denotes the closure operator of @xmath36 . we will need the fact that there exists an optimal ordering @xmath103 of @xmath166 that induces a certain ordered partition of @xmath166 of the form @xmath255 where for each @xmath173 $ ] , @xmath256 , @xmath257 , @xmath258 , and @xmath259 . this will follow from a re - ordering argument given further below . but first , we make some simple observations about orderings of @xmath166 . given an ordering of @xmath166 , we may assume , wlog , that for each pair of adjacent vertices @xmath226 , @xmath161 appears before @xmath162 in the ordering ; we denote this by @xmath260 . we call such an ordering of @xmath166 a _ normal _ ordering . + let @xmath103 be a normal ordering of @xmath166 . then , for @xmath261 , we have * @xmath262 iff @xmath263 ; and * @xmath264 iff @xmath265 . + [ conn_fn_lemma ] we only prove ( a ) , as the proof of ( b ) is similar . it is easy to deduce from the definition of the connectivity function that @xmath262 iff @xmath263 and @xmath266 . now , if @xmath263 , then @xmath267 for some @xmath160 . ( if not , _ i.e. _ , if @xmath268 , then since @xmath260 , we must have @xmath269 , and so , @xmath270 , a contradiction . ) therefore , @xmath271 contains @xmath162 , and hence , @xmath272 . we have thus shown that if @xmath263 , then @xmath266 . part ( a ) of the lemma now follows . + we now describe a procedure that takes as input a normal ordering of @xmath166 , and produces as output a re - ordering of @xmath166 with certain desirable properties . + + _ input _ : a normal ordering @xmath103 of @xmath166 . + _ initialization _ : @xmath273 . + [ cols= " < , < " , ] + denote by @xmath274 the final output generated by the above algorithm . set @xmath275 , and for @xmath276 $ ] , @xmath277 . stepping through the algorithm , one may easily check that @xmath274 has the following property : for @xmath278 , if @xmath279 , then @xmath280 , and if @xmath281 , then @xmath282 the following claim can be readily deduced from this property , and we leave the details to the reader . + ( a ) the ordering @xmath274 induces an ordered partition of @xmath166 of the form @xmath255 where for each @xmath173 $ ] , @xmath256 , @xmath257 , @xmath258 and @xmath259 . moreover , for each @xmath226 , @xmath283 iff @xmath284 . + ( b ) for the ordered partition in ( a ) , we have for each @xmath173 $ ] , @xmath285 + [ re - ordering_claim ] the crucial property of @xmath274 is the following . + if @xmath274 is the output of the re - ordering algorithm in response to the input @xmath103 , then @xmath286 . + [ re - ordering_lemma ] steps 01 of the algorithm go through @xmath10 iterations , indexed by @xmath287 . in the @xmath288th iteration , step 1 is given a normal ordering @xmath103 , in response to which it produces an ordering @xmath289 , which is also normal . to prove the lemma , it is enough to show that @xmath290 . so , suppose that the algorithm is in its @xmath288th iteration ( @xmath278 ) . we first dispose of the case when @xmath291 . then , @xmath292 for some @xmath293 such that @xmath294 . observe that if @xmath295 or if @xmath296 , then @xmath297 is just a re - ordering of @xmath298 , and hence @xmath299 . so , consider @xmath300 . in this case , since @xmath302 , we have @xmath303 . on the other hand , @xmath304 hence , @xmath305 . therefore , for any @xmath306 $ ] , we have shown that there exists a @xmath307 $ ] such that @xmath308 . it follows that @xmath290 . + we must now deal with the case when @xmath309 , _ i.e. _ , @xmath310 . note that if @xmath311 , then since @xmath312 , we have @xmath313 . therefore , @xmath314 , which means that @xmath315 , a contradiction . therefore , there must exist some @xmath293 such that @xmath316 . let @xmath317 be the least integer @xmath318 such that there exists @xmath319 for some @xmath293 . by choice of @xmath317 , we have @xmath320 . for this @xmath321 , we again have @xmath322 as before , if @xmath295 or if @xmath296 , then @xmath299 . for @xmath323 , we have @xmath324 as @xmath325 . and as in ( [ rm_eq1 ] ) , @xmath326 . hence , @xmath305 . we are left with @xmath327 . note that by choice of @xmath317 , @xmath328 . therefore , @xmath329 since ( [ rm_eq1 ] ) again applies , we have that @xmath330 observe that , since @xmath263 , by lemma [ conn_fn_lemma](a ) , @xmath331 furthermore , by choice of @xmath317 , @xmath332 , but @xmath333 , which together imply that @xmath334 . hence , again by lemma [ conn_fn_lemma](a ) , @xmath335 therefore , from ( [ l_eq1])([l_eq3 ] ) , we find that for @xmath336 or @xmath337 , we have @xmath338 . we claim that for @xmath339 , we have @xmath340 , so that by induction , @xmath341 . this would then imply , via ( [ l_eq1 ] ) and ( [ l_eq3 ] ) , that @xmath342 . thus , for any @xmath306 $ ] , we have a @xmath307 $ ] such that @xmath343 . therefore , @xmath290 , which would complete the proof of the lemma . to prove our claim , it is enough to show that when @xmath339 , we have @xmath344 . indeed , it then follows from lemma [ conn_fn_lemma](b ) that @xmath345 . so , suppose that @xmath346 for some @xmath339 . then , @xmath347 for some @xmath226 . ( otherwise , if @xmath348 , then since @xmath349 , we would have @xmath344 . ) note that @xmath350 ; otherwise , the fact that @xmath351 is a flat of @xmath36 would imply that @xmath352 . so , @xmath353 . suppose that @xmath354 for some @xmath228 . then , @xmath355 , which contradicts the choice of @xmath317 . therefore , @xmath356 , meaning that @xmath347 for some @xmath134 . now , if @xmath357 , then @xmath344 , as @xmath358 is a triangle in @xmath150 . so , wlog , @xmath359 . by choice of @xmath317 , @xmath360 therefore , @xmath361 . but now , @xmath362 , as @xmath363 is a triangle in @xmath150 . however , @xmath364 ; otherwise , we would have @xmath365 , which , since @xmath351 is a flat and @xmath363 is a triangle , would imply that @xmath269 . thus , @xmath366 for some @xmath367 . as already noted , @xmath362 , and so once again , our choice of @xmath317 is contradicted . therefore , our assumption that @xmath346 always leads to a contradiction , from which we conclude that the assumption is false . this completes the proof of the lemma . + we can now furnish the last remaining piece of the proof of proposition [ pw_prop ] . + @xmath208 . + [ pw_lemma2 ] let @xmath103 be an optimal ordering of @xmath166 . wlog , @xmath103 may be assumed to be normal . let @xmath274 be the output of the re - ordering algorithm in the response to the input @xmath103 . then , @xmath274 has the properties listed in claim [ re - ordering_claim ] , and , by lemma [ re - ordering_lemma ] , is also an optimal ordering of @xmath166 . now , @xmath274 induces an ordered partition @xmath368 of @xmath166 , as in claim [ re - ordering_claim](a ) . for @xmath173 $ ] , define @xmath369 , and @xmath370 . letting @xmath371 $ ] and @xmath372 $ ] denote the subgraphs of @xmath150 induced by @xmath373 and @xmath374 , respectively , set @xmath375 ) \cap v(\ocg[y_j'])$ ] . in other words , @xmath176 is the set of vertices common to both @xmath371 $ ] and @xmath372 $ ] . it is easily checked that @xmath131 is a path - decomposition of @xmath150 . note that @xmath376)| + |v(\ocg[y_j'])| - |\ov|.\ ] ] we next observe that @xmath371 $ ] and @xmath372 $ ] are connected graphs . from claim [ re - ordering_claim](b ) , we have that @xmath377 . therefore , for any edge @xmath161 ( or @xmath162 ) in @xmath378 , both @xmath379 and @xmath380 must be in some @xmath381 , @xmath382 . thus , in @xmath371 $ ] , each vertex @xmath229 is adjacent to @xmath151 , which shows that @xmath371 $ ] is connected . consider any vertex @xmath229 in @xmath372 $ ] , such that @xmath383 . then , @xmath384 for some @xmath385 . so , @xmath386 for some @xmath387 . by claim [ re - ordering_claim](b ) , this implies that either @xmath389 or @xmath390 . hence , either @xmath391 or @xmath392 . however , @xmath393 can not be in @xmath394 , since @xmath383 , and so , @xmath391 . thus , @xmath395 forms a path in @xmath372 $ ] from @xmath151 to @xmath164 . it follows that @xmath372 $ ] is connected . therefore , @xmath396)| - 1 ) + ( |v(\ocg[y_j'])| - 1 ) - ( \ov-1 ) = |v_j| - 1.\end{aligned}\ ] ] hence , @xmath397 } \l_m(y_j ) = \max_{j \in [ t ] } |v_j| - 1 = w_\ocg(\cv ) \geq \pw(\ocg),\ ] ] which proves the lemma . + the proof of proposition [ pw_prop ] is now complete . theorem [ pw_np_thm ] shows that the following decision problem is np - complete . + rl * problem : * & matroid pathwidth + + * instance : * & an @xmath398 matrix @xmath7 over @xmath0 , and an integer @xmath9 . + * question : * & is there an ordering @xmath103 of the elements of @xmath399 $ ] , + & such that @xmath400 ? + similarly , corollary [ tw_np_cor ] shows that the corresponding decision problem for code trellis - width ( over a fixed finite field @xmath0 ) is np - complete . in this section , we consider the situation when the parameter @xmath1 above is a fixed constant , and therefore , not considered to be part of the problem instance . in contrast to the np - completeness of matroid pathwidth , we believe that the following decision problem and its coding - theoretic counterpart are solvable in polynomial time . + rl * problem : * & weak matroid pathwidth + + * instance : * & an @xmath398 matrix @xmath7 over @xmath13 . + * question : * & is there an ordering @xmath103 of the elements of @xmath399 $ ] , + & such that @xmath400 ? + our optimism above stems from the fact that the property of having pathwidth bounded by @xmath1 is preserved by the minors of a matroid . to be precise , let @xmath12 be the class of matroids representable over the finite field @xmath11 , that have pathwidth at most @xmath1 . by lemma [ minor_pathwidth_lemma ] , @xmath12 is minor - closed . since pathwidth is an upper bound on the branchwidth of a matroid , all matroids in @xmath12 have branchwidth at most @xmath1 . now , geelen and gerards have shown that if @xmath71 is any minor - closed class of @xmath13-representable matroids having bounded branchwidth , then @xmath71 has finitely many excluded minors ( * ? ? ? * theorem 1.4 ) . as a result , we have the following theorem . + for any integer @xmath9 and finite field @xmath13 , @xmath12 has finitely many excluded minors . consequently , the code family @xmath401 also has finitely many excluded minors . + [ pwq_thm ] theorem [ pwq_thm ] shows that deciding whether or not a given @xmath13-representable matroid @xmath36 belongs to @xmath12 can be accomplished by testing whether or not @xmath36 contains as a minor one of the finitely many excluded minors of @xmath12 . the minor - recognition conjecture of geelen , gerards and whittle ( * ? ? ? * conjecture 1.3 ) states that , for any fixed @xmath13-representable matroid @xmath86 , testing a given @xmath13-representable matroid for the presence of an @xmath86-minor can be done in polynomial time . so , if this conjecture is true and there is evidence to support its validity @xcite then membership of an @xmath13-representable matroid in the class @xmath12 can be decided in polynomial time . hence , assuming the validity of the minor - recognition conjecture , weak matroid pathwidth is solvable in polynomial time . while the finiteness of the list of excluded minors for @xmath12 implies , modulo the minor - recognition conjecture , the existence of a polynomial - time algorithm for weak matroid pathwidth , an actual implementation of such an algorithm would require the explicit determination of the excluded minors . as a relatively easy exercise , we prove the following theorem . + a matroid is in @xmath14 iff it contains no minor isomorphic to any of the matroids @xmath19 , @xmath20 , @xmath21 and @xmath22 . + [ p1q_thm ] we first verify the easy `` only if '' part of the above theorem . + @xmath19 , @xmath20 , @xmath21 and @xmath22 are not in @xmath14 . + [ tc1_prop1 ] if @xmath402 is any ordering of the elements of @xmath403 , then @xmath404 . it follows that @xmath405 . now consider @xmath406 . for any ordering @xmath407 of @xmath408 , we have @xmath409 , with equality iff @xmath410 is a triangle , in which case @xmath411 is a triad . it follows that @xmath412 . hence , @xmath413 . the proof for @xmath414 is very similar . for any @xmath415 with @xmath416 , @xmath417 , since @xmath418 has no circuits of size less than 4 . therefore , for any ordering @xmath407 of @xmath419 , @xmath420 . thus , @xmath421 , and by duality , @xmath422 as well . + we now prove the `` if '' part of theorem [ p1q_thm ] . for the duration of the proof , we take @xmath36 to be a matroid that contains no minor isomorphic to the matroids listed in the statement of the theorem . since @xmath20 is a minor of each of the matroids @xmath423 , @xmath424 , @xmath425 , @xmath426 , @xmath427 and @xmath428 , @xmath36 contains none of these as minors . therefore , @xmath429 for some planar graph @xmath129 ( cf . * theorem 13.3.1 and proposition 5.2.6 ) ) . evidently , we may take @xmath129 to be connected as a graph . since @xmath14 is closed under direct sums , we may assume that @xmath36 is 2-connected . therefore , @xmath129 is either a graph consisting of a single vertex with a self - loop incident with it , or @xmath129 is a loopless graph . in the former case , @xmath430 , which is in @xmath14 . so , we may assume that @xmath129 is loopless . if @xmath129 has exactly two vertices , then @xmath431 for some @xmath10 , which is also in @xmath14 . hence , we may assume that @xmath432 , in which case , @xmath129 is 2-connected as a graph ( * ? ? ? * corollary 8.2.2 ) . moreover , if @xmath433 is any geometric dual of @xmath129 , then , since @xmath434 is 2-connected , by the same argument as above , we may assume that @xmath433 is also 2-connected as a graph . at this point , we need the following definition . we call a graph an _ umbrella _ if it is of the form shown in figure [ umbrella ] . formally , an umbrella is a graph @xmath435 that consists of a circuit on @xmath436 vertices @xmath437 , and in addition , for each @xmath438 $ ] , zero or more parallel edges between @xmath439 and @xmath440 . note that @xmath441 is a simple path , where @xmath441 denotes the graph obtained from @xmath435 by deleting the vertex @xmath439 and all edges incident with it . returning to our proof , we have @xmath429 for a loopless , 2-connected , planar graph @xmath129 , such that any geometric dual of @xmath129 is also 2-connected . + @xmath129 has a geometric dual @xmath433 that is isomorphic to an umbrella . + [ umb_lemma1 ] we prove the lemma using the concept of an outerplanar graph . a planar graph is said to be _ outerplanar _ if it has a planar embedding in which every vertex lies on the exterior ( unbounded ) face . we will refer to such a planar embedding of the graph as an _ outerplanar embedding_. outerplanar graphs were characterized by chartrand and harary @xcite as graphs that do not contain @xmath442 or @xmath418 as a minor . + _ proof of lemma [ umb_lemma1 ] _ : since @xmath443 contains no @xmath20- or @xmath21-minor , @xmath129 can not contain @xmath442 or @xmath418 as a minor . therefore , by the chartrand - harary result mentioned above , @xmath129 is outerplanar . let @xmath433 be the geometric dual of an outerplanar embedding of @xmath129 . let @xmath151 be the vertex of @xmath433 corresponding to the exterior face of the outerplanar embedding of @xmath129 . by a result of fleischner _ et al . _ ( * ? ? ? * theorem 1 ) , @xmath444 is a forest . in fact , since @xmath433 is 2-connected , @xmath444 is a tree . we claim that no vertex of @xmath444 has degree greater than two , and hence , @xmath444 is a simple path . indeed , suppose that @xmath444 has a vertex @xmath163 adjacent to three other vertices @xmath445 . since @xmath446 is 2-connected , there are paths @xmath447 , @xmath448 and @xmath449 in @xmath433 from @xmath450 , @xmath451 and @xmath452 , respectively , to @xmath151 that do not pass through @xmath163 . also , since @xmath444 is a tree , these paths must be internally disjoint in @xmath433 . the graph @xmath433 thus has a subgraph as depicted in figure [ deg3_obs ] . but this subgraph is obviously contractible to @xmath418 , and hence @xmath433 has @xmath418 as a minor . however , this is impossible , as @xmath434 does not have @xmath21 as a minor . thus , @xmath444 is a simple path . the two degree - one vertices ( end - points ) of this path must be adjacent to @xmath151 in @xmath433 ; otherwise , @xmath433 is not 2-connected . it follows that @xmath433 is isomorphic to an umbrella . + to complete the proof of theorem [ p1q_thm ] , we show that @xmath453 , so that by duality , @xmath454 . this is done by the following lemma . if @xmath435 is an umbrella , then @xmath455 . [ umb_lemma2 ] let @xmath435 be an umbrella on @xmath436 vertices @xmath456 , where @xmath439 is the vertex such that @xmath457 is a simple path . for @xmath438 $ ] , let @xmath458 denote the set of edges between @xmath439 and @xmath440 . also , for @xmath459 $ ] , let @xmath460 denote the edge between @xmath461 and @xmath462 . consider any ordering of the edges of @xmath435 that induces the ordered partition @xmath463 let @xmath464 , with @xmath465 ( @xmath119 may be empty ) . note that the subgraph , @xmath466 $ ] , of @xmath435 induced by the edges in @xmath44 is incident only with vertices in @xmath467 . therefore , setting @xmath468 , @xmath469)|-1 \leq j$ ] . similarly , the subgraph of @xmath435 induced by the edges in @xmath470 is incident only with vertices in @xmath471 , and so , @xmath472 . thus , @xmath473 , and it follows that @xmath474 . being graphic , @xmath36 is @xmath13-representable , and hence , @xmath475 . + this completes the proof of theorem [ p1q_thm ] . + as a corollary to the theorem , we give a coding - theoretic characterization of the code family @xmath476 . in coding theory , an @xmath13-representation of a uniform matroid is called a _ maximum - distance separable ( mds ) code_. for any field @xmath13 , the matrices @xmath477 , @xmath478 and @xmath479 below are @xmath13-representations of @xmath20 , @xmath21 and @xmath22 , respectively . @xmath480;\ ] ] @xmath481 ; \ \ g^*_{2,3 } = \left [ \begin{array}{cccccc } 1 & -1 & 0 & -1 & 1 & 0 \\ 1 & 0 & -1 & -1 & 0 & 1 \\ \end{array } \right].\ ] ] the matroids @xmath20 , @xmath21 and @xmath22 , being binary , are uniquely representable over @xmath13 , in the matroid - theoretic sense ( * ? ? ? * section 6.3 and theorem 10.1.3 ) . we let @xmath482 , @xmath483 and @xmath484 denote the codes over @xmath13 generated by the matrices @xmath477 , @xmath478 and @xmath479 , respectively . + let @xmath13 be an arbitrary finite field . a linear code @xmath4 over @xmath13 has trellis - width at most one iff it contains no minor equivalent to any of the following : * a @xmath485 $ ] mds code ; * a code obtainable by applying an automorphism of @xmath13 to one of the codes @xmath482 , @xmath483 and @xmath484 . + [ cp1q_cor ] the problem of finding the complete set of excluded minors for @xmath12 quickly becomes difficult for @xmath23 . the main obstacle is that we may only assume the basic property of 2-connectedness for such excluded minors . the class @xmath12 is not even closed under 2-sums , so excluded minors for the class need not be 3-connected . an illustration of this is given by the following result , which provides a partial list of excluded minors for @xmath24 . + for any finite field @xmath13 , the matroids @xmath423 , @xmath424 , @xmath425 , @xmath426 , @xmath427 , @xmath428 , and @xmath443 , where @xmath129 is any of the planar graphs in figure [ p2q_ex_minors ] , are excluded minors for @xmath24 . if @xmath486 , then @xmath487 is also an excluded minor for @xmath24 . + [ p2q_prop ] we omit the proof , as it is only a matter of verifying that for each matroid @xmath36 listed in the proposition , @xmath488 , but @xmath489 for any @xmath490 . we point out that the cycle matroids of all but the three leftmost graphs in figure [ p2q_ex_minors ] are not 3-connected . the author would like to thank jim geelen for contributing some of his ideas to this paper , and alexander vardy for pointers to the prior literature on trellis complexity . 99 s. arnborg , d.g . corneil and a. proskurowski , `` complexity of finding embeddings in a @xmath491-tree , '' _ siam j. alg . disc . _ , vol . 8 , pp . 277284 , 1987 . bodlaender , `` a tourist guide through treewidth , '' _ acta cybernetica _ , vol . 11 , pp . 123 , 1993 . g. chartrand and f. harary , `` planar permutation graphs , '' _ ann . henri poincar sec . b _ , vol . iii , no . 4 , pp . 433438 , 1967 . fleischner , d.p . geller and f. harary , `` outerplanar graphs and weak duals , '' _ j. indian math . 38 , pp . 215219 , 1974 . forney jr . , `` dimension / length profiles and trellis complexity of linear block codes , '' _ ieee trans . theory _ , vol . 40 , no . 6 , pp . 17411752 , nov . j. geelen , b. gerards and g. whittle , `` towards a matroid - minor structure theory '' , to appear in _ combinatorics , complexity and chance . a tribute to dominic welsh _ , g. grimmett and c. mcdiarmid , eds . , oxford university press , 2007 . available online at ` http://homepages.cwi.nl/\simbgerards/personal/papers/towards_welsh.pdf ` . j. geelen , b. gerards and g. whittle , `` on rota s conjecture and excluded minors containing large projective geometries , '' _ j. combin . theory , ser . b _ , vol . 96 , pp . 405425 , 2006 . j. geelen and g. whittle , `` branch - width and rota s conjecture , '' _ j. combin . theory , ser . b _ , vol . 86 , no . 2 , pp . 315330 , nov . r. hall , j. oxley and c. semple , `` the structure of 3-connected matroids of path width three '' , _ europ . j. combin . _ , vol . 28 , pp . 964989 , 2007 . g.b . horn and f.r . kschischang , `` on the intractability of permuting a block code to minimize trellis complexity , '' _ ieee trans . inform . theory _ , vol . 42 , no . 6 , pp . 20422048 , nov . 1996 . k. jain , i. mndoiu and v.v . vazirani , `` the ` art of trellis decoding ' is computationally hard for large fields , '' _ ieee . trans . inform . theory _ , vol . 44 , no . 3 , pp . 12111214 , may 1998 . n. kashyap , `` a decomposition theory for binary linear codes , '' submitted to _ ieee trans . inform . theory_. arxiv e - print cs.dm/0611028 . f.j . macwilliams and n.j.a . sloane , _ the theory of error - correcting codes _ , north - holland , amsterdam , 1977 . muder , `` minimal trellises for block codes , '' _ ieee . trans . inform . theory _ , vol . 34 , no . 5 , pp . 10491053 , sept . oxley , _ matroid theory _ , oxford university press , oxford , uk , 1992 . n. robertson and p.d . seymour , `` graph minors . i. excluding a forest , '' _ j. combin . theory , ser . b _ , vol . 35 , pp . 3961 , 1983 . a. vardy , `` trellis structure of codes , '' in _ handbook of coding theory _ , r. brualdi , c. huffman and v. pless , eds . , amsterdam , the netherlands : elsevier , 1998 .

the canonical front form hamiltonian for non - abelian su(n ) gauge theory in 3 + 1 dimensions is mapped non - perturbatively on an effective hamiltonian which acts only in the fock space of a quark and an antiquark . emphasis is put on dealing with the many - body aspects of gauge field theory , and it is shown explicitly how the higher fock - space amplitudes can be retrieved self - consistently from solutions in the-space . the approach is based on the novel method of iterated resolvents and on discretized light - cone quantization , driven to the continuum limit . it is free of the usual perturbative tamm - dancoff truncations in particle number and coupling constant , and respects all symmetries of the lagrangian including covariance and gauge invariance . it is emphasized that vertex regularization as opposed to fock - space regularization allows an explicit renormalization , which yields linear confinement . the string constant must be determined from experiment . = 23.5 cm = 16.5 cm = -2.0 cm = -0.34 cm = -0.34 cm preprint mpih - v26 - 1997 + submitted to physical review d + pacs - index : 11.10ef , 11.15tk , 12.38lg , 12.40yx +

hep-ph9707361

the hamiltonian bound - state problem in a field theory is notoriously difficult . procedures like those of schwinger and dyson or of bethe and salpeter are not easy to cope with in practice as reviewed recently in @xcite . in the fifties , the difficulties were considered so enormous that the hamiltonian approach was given up altogether in favor of feynman s action oriented approach . its modern successors like the lattice gauge theories @xcite have arrived at a stage of making quantitative predictions @xcite . in contrast , phenomenological models @xcite are closer to experiment and have the different objective to classify the bulk of empirical data . they leave little doubt that a heavy meson contains primarily a pair of constituent quarks and _ not _ an infinity of sea particles as suggested by the perturbative treatment of quantum field theory . how can one reconcile such models with the need to understand hadronic structure from a covariant theory ? particularly , how can one reconcile the constituent quark model ( cqm ) with the quantum field theory of chromodynamics ( qcd ) ? there are several reasons why the front form of hamiltonian dynamics @xcite is one of the very few candidates ( see @xcite ) . particularly the simple vacuum and the simple boost properties @xcite are in contrast with the complicated vacuum and the complicated boosts in the conventional hamiltonian theory ( see also @xcite and @xcite ) . nowadays it is known that even the front form vacuum is not simple @xcite , but it is still simpler than in the instant form : the problem can at least be formulated mathematically @xcite . the success of discretized light - cone quantization @xcite particularly in 1 + 1 dimensions has stirred hope that its apparent simplicities carry over to 3 + 1 dimensions @xcite . the two major obstacles for such an endeavor are the following : ( 1 ) the hamiltonian matrix increases exponentially fast with the particle number and the transverse momenta in the fock states . truncating the fock space to 2 particles like tamm @xcite and dancoff @xcite invokes perturbation theory , violates gauge and lorentz invariance , and generates non - integrable singularities . _ ad hoc _ procedures must be introduced to make things work , both in the original @xcite and the light - cone work @xcite . ( 2 ) the unrestricted integration over the transverse momenta creates divergences , which are regularized by some cut - off . the unphysical dependence on the regularization parameters has to be removed by renormalization . this is difficult enough in a perturbative context , but serious challenges appear when dealing with the non - perturbative solutions of a hamiltonian ( see @xcite ) . in the present work , these problems are addressed and partially solved by adopting systematically the point of view that the bound states of qcd are eigenstates of the canonical field - theoretical hamiltonian in the light - cone gauge @xmath1 . the zero modes emphasized in previous work @xcite are omitted here . in section [ sec:2 ] a method for diagonalizing numerically a very large matrix , say with dimension @xmath2 , by mapping the problem equivalently to the diagonalization of a reasonably small matrix , say with dimension @xmath3 , is proposed , using the theory of effective interactions @xcite . the reduction to comparatively small matrices is more than a technical issue and requires one to cope with the many - body aspects of a field theory . in the present work , the many - body aspects are kept as resolvents which _ are not expanded as perturbative series_. they are treated in section [ sec:3 ] by the method of iterated resolvents . this novel method requires the development of considerable formal apparatus which was given earlier @xcite . in the continuum limit the matrix equations become integral equations , which are similar to those solved numerically already @xcite , but here they are derived without assuming that the coupling constant is small and without truncating the fock space . the important issues of vertex regularization and renormalization are discussed in sections [ sec:4 ] and [ sec:5 ] lead ultimately to ( linear ) confinement . the approach is summarized and discussed in section [ sec:6 ] . in canonical field theory the four commuting components of the energy - momentum four - vector @xmath4 are constants of the motion . in the front form of hamiltonian dynamics @xcite they are denoted by @xmath5 , see for example @xcite . its spatial components @xmath6 and @xmath7 do not depend on the interaction and are diagonal operators in momentum representation . their eigenvalues are the sums of the single particle momentum components , _ i.e. _ @xmath8 each particle has four - momentum @xmath9 , sits on its mass - shell @xmath10 and therefore has @xmath11 . each particle state `` @xmath12 '' is then characterized by six quantum numbers @xmath13 . the first three specify the particle s spatial momentum and @xmath14 its helicity @xmath15 or @xmath16 . for qed that s all there is , but for qcd a quark is specified further by its color and flavor , @xmath17 and @xmath18 , respectively . correspondingly , gluons are specified by the five quantum numbers @xmath19 , with @xmath20 being the glue index . the temporal component @xmath21 depends on the interaction and is a very complicated non - diagonal operator . it propagates the system in the light - cone time @xmath22 , _ i.e. _ @xmath23 and is the proper front - form hamiltonian @xcite . the contraction of @xmath24 is the operator of invariant mass - squared , _ i.e. _ @xmath25 it is a lorentz scalar and referred to somewhat improperly but conveniently as the ` light - cone hamiltonian @xmath26 ' @xcite , or shortly @xmath27 . its matrix elements ( and eigenvalues ) somewhat unusually carry the dimension of an invariant - mass - squared . in this work one aims at a representation in which @xmath28 is diagonal , @xmath29 since the hamiltonian does not conserve particle number , the eigenfunctions @xmath30 are complicated superpositions of particle number projections with contributions like @xmath31 or @xmath32 . the eigenvalue equation ( [ eq:2.2 ] ) stands thus for an infinite set of coupled integro - differential equations which are even difficult to write down explicitly @xcite . in order to analyze the structure of the eigenvalue equation , we apply a technical trick , referred to as discretized light cone quantization ( dlcq ) @xcite : one imposes periodic boundary conditions on the lagrangian fields . in consequence one maps the infinite set of coupled integral equations onto a _ finite set _ of coupled matrix equations . the kernel of the integral equation ( [ eq:2.2 ] ) becomes a finite matrix , which in principle can be diagonalized by numerical methods . v & 4 & 5 & 6 & 4 & 5 & 6 & 7 & 8 + k & n@xmath33 & sector & n & 1 & 2 & 3 & 4 & 5 & 6 & 7 & 8 & 9 & 10 & 11 & 12 & 13 + 1 & 2 & @xmath34 & 1 & & & v & & & & & & & & & & + 2 & 2 & @xmath35 & 2 & & & v&&v & & & & & & & & + 2 & 3 & @xmath36 & 3 & v&v&&v&&v & & & & & & & + 2 & 4 & @xmath37 & 4 & & & v&&&&v & & & & & & + 3 & 3 & @xmath38 & 5 & & v&&&&v&&&v & & & & + 3 & 4 & @xmath39 & 6 & & & v&&v&&v&&&v & & & + 3 & 5 & @xmath40 & 7 & & & & v&&v&&v&&&v & & + 3 & 6 & @xmath41 & 8 & & & & & & & v&&&&&v & + 4 & 4 & @xmath42 & 9 & & & & & v&&&&&v & & & + 4 & 5 & @xmath43 & 10 & & & & & & v&&&v&&v & & + 4 & 6 & @xmath44 & 11 & & & & & & & v&&&v&&v & + 4 & 7 & @xmath45 & 12 & & & & & & & & v&&&v&&v + 4 & 8 & @xmath46 & 13 & & & & & & & & & & & & v & + why finite ? as a consequence of discretization , the fock states are denumerable and orthonormal . since @xmath7 has only positive eigenvalues and since each particle has a lowest possible value of @xmath47 , the number of particles in a fock state is limited for any fixed value of the _ harmonic resolution _ @xmath48 , defined by eq.([eq:2.1 ] ) . one recognizes this best by way of example . the lowest possible value @xmath49 allows only one @xmath0-pairs ( a single gluon can not be in a color singlet ) . for @xmath50 , the fock space contains in addition pairs of gluons , @xmath0-pairs plus a gluon , and two @xmath0-pairs , as illustrated in table [ tab:3 ] . for @xmath51 , one has at most 8 particles , namely 4 @xmath52-pairs . in the table all 13 fock - space sectors possible for @xmath53 are enumerated by @xmath54 . note that the above classification does not change when @xmath55 is increased further : one just adds more complicated fock - space sectors . their number grows quadratically with @xmath55 and has the value @xmath56 . the single particle momenta can take all possible values subject to the constraints in eq.([eq:2.1 ] ) . each of the above fock space sectors contains infinitely many individual fock states , since the transverse momenta can take either sign . their number can be regulated by some convenient `` fock - space regularization '' : a fock state with @xmath57 particles is included only if the single particle four - momenta satisfy @xcite @xmath58 . however , it was not realized in the past @xcite , that fock - space regularization is almost irrelevant in the continuum theory . one needs a better alternative , and we propose to work with `` vertex regularization '' . at each vertex , a particle with four - momentum @xmath59 is scattered into two particles with respective four - momentum @xmath60 and @xmath61 . one can _ regulate the interaction _ by a condition on the off - shell mass @xmath62 , @xmath63 in terms of the regulator mass scale @xmath64 . this innocent step will turn out as a rather efficient tool for analyzing the divergences , see below . to get more insight , the relevant momenta are parametrized as @xmath65 , @xmath66 and @xmath67p^+,[1-z]\vec p_{\!\perp}-\vec l_{\!\perp},p_2 ^ -)$ ] . the off - shell mass becomes then simply @xmath68 and the mass scale governs how much the particles can go off their equilibrium values @xmath69 and @xmath70 . for @xmath71 , the phase space is reduced to a point , and consequently the interaction is reduced to zero : the hamiltonian matrix ( or the integral equation ) is diagonal , and has the spectrum of the free theory irrespective of the matrix dimension governed by @xmath72 . once the fock space is defined , one can calculate the matrix elements of the ( light - cone ) hamiltonian @xcite . it is written here as a sum of three parts , @xmath73 . the kinetic energy @xmath74 survives the limit of the coupling constant @xmath75 going to zero . it is diagonal in fock - space representation , its eigenvalue is the _ free invariant mass squared _ of the particular fock state . the _ vertex interaction _ @xmath76 is the relativistic interaction _ per se_. it is linear in @xmath77 and changes the particle number by 1 . matrix elements of @xmath78 which change the particle number by 3 ( as in the instant form ) vanish kinematically : the vacuum _ does not fluctuate _ , see for example @xcite . the _ instantaneous interactions @xmath79 are consequences of working in the light - cone gauge , are _ gauge artifacts _ , and are proportional to @xmath80 . they either conserve particle number ( seagull interactions ) , or change it by 2 ( fork interactions ) . for the moment we will drop the instantaneous interaction @xmath81 , but we will restore it later . most of the hamiltonian matrix elements are then zero . only those where the particle number is changed by 1 are potentially non - zero , as displayed in table [ tab:3 ] . the diagonal blocks are also essentially zero matrices , except for the kinetic energies which reside on the diagonal . it is a long standing problem of field theory of knowing how such a matrix can have bound states . the key problem is that the dimension of the hamiltonian matrix increases exponentially . suppose , the regularization procedure allows 10 discrete momentum states in each direction , _ i.e. _ in the one longitudinal and the two transverse directions of @xmath82 . a particle has then roughly @xmath83 degrees of freedom . a fock - space sector with @xmath57 constituent particles has thus @xmath84 different fock states , since they are subject to the constraints , eq.([eq:2.1 ] ) . sector 13 alone with its 8 particles has the dimension of about @xmath2 fock states . the hamiltonian method applied to gauge theory therefore faces a formidable matrix diagonalization problem . sooner or later , the matrix dimension exceeds imagination , and other than in 1 + 1 dimensions one has to develop new tools by matter of principle . one would like to reduce the full problem to an effective interaction between a quark and an antiquark , @xmath85 the derivation of such an effective interaction , is the subject of this work . effective interactions are a well known tool in many - body physics @xcite . in field theory the method is known as the tamm - dancoff - approach , since it was applied first to yukawa theory for describing the nucleon - nucleon interaction by tamm @xcite and by dancoff @xcite . let us review it in short . in a field theory , the fock - space sectors @xmath86 appear in a most natural way , see table [ tab:3 ] . as explained above , the hamiltonian matrix can be understood as a matrix of block matrices , whose rows and columns are enumerated by @xmath87 in accord with the fock - space sectors like in table [ tab:3 ] . correspondingly , one can rewrite eq.([eq:2.2 ] ) as a block matrix equation : @xmath88 one recalls that each sector contains many individual fock states with different values of @xmath89 , @xmath90 and @xmath91 . the rows and columns of the matrix can always be split into two parts . one speaks of the @xmath92-space @xmath93 with @xmath94 , and of the rest , the @xmath95-space @xmath96 . eq.([eq:319 ] ) can then be rewritten conveniently as a @xmath97 block matrix equation @xmath98 rewrite the second equation as @xmath99 , and observe that the quadratic matrix @xmath100 could be inverted to express the q - space wavefunction @xmath101 in terms of the @xmath92-space wavefunction @xmath102 . but here is the problem : the eigenvalue @xmath103 is unknown at this point . one therefore solves first an other problem : one introduces _ the starting point energy _ @xmath104 as a redundant parameter at disposal , and _ defines the @xmath105-space resolvent _ or propagator as the inverse of the block matrix @xmath106 , thus @xmath107 inserting this into eq.([eq:321 ] ) defines an effective interaction @xmath108 which by construction acts only in the @xmath109-space @xmath110 the effective interaction is thus well defined : it is the original block matrix @xmath111 plus a part where the system is scattered virtually into the @xmath105-space , propagating there ( @xmath112 ) by impact of the true interaction , and scattered back into the @xmath92-space . every value of @xmath104 defines a different hamiltonian and a different spectrum . varying @xmath104 one generates a set of _ energy functions _ whenever one finds a solution to the _ fix point equation _ @xmath114 one has found one of the true eigenvalues and eigenfunctions of @xmath28 , by construction . one should emphasize that one can find all eigenvalues of the full hamiltonian @xmath28 , irrespective of how small one chooses the @xmath92-space . explicit examples for that can be found in @xcite . it looks as if one has mapped a difficult problem , the diagonalization of a huge matrix onto a simpler problem , the diagonalization of a much smaller matrix . but this is only true in a restricted sense . one has to invert a matrix . the numerical inversion of a matrix takes about the same effort as its diagonalization . in addition , one has to vary @xmath104 and solve the fix point equation ( [ eq:350 ] ) . the numerical work is thus larger than the direct diagonalization . the advantage of working with an effective interaction is of an analytical nature because the resolvents can be approximated systematically . the two resolvents @xmath115 defined once with and once without the non - diagonal interaction @xmath116 , are identically related by @xmath117 , or by the infinite series of perturbation theory . the point is , of course , that the kinetic energy @xmath74 is a diagonal operator which can be trivially inverted to get the unperturbed resolvent @xmath118 . in practice @xcite , one has never gone beyond the first non - trivial order of perturbation theory for formulating the effective interaction , namely @xmath119 . this , of course , is highly unsatisfactory . the identification of the @xmath0-sector with the @xmath109-space and the related effective interaction appears as the most natural thing to do . but in practice , the tamm - dancoff - approach ( tda ) has two serious defects : ( 1 ) the approach is technically useful only if one truncates the perturbation series to its very first term . this destroys lorentz and gauge invariance . ( 2 ) if one identifies @xmath104 with the eigenvalue as one should , the effective interaction develops a non - integrable singularity , both in the original instant form @xcite and in the front form @xcite . in order to force the approach to work , one has to replace the eigenvalue @xmath104 by the kinetic energy in the @xmath109-space @xmath120 @xcite . in gauge theory the fock space sectors appear in the most natural way , and in dlcq the number of fock - space sectors is finite from the outset as shown in table [ tab:3 ] . the block matrix is sparse due to the nature of the underlying hamiltonian , most of the block matrices are zero matrices . one should make use of that ! the tamm - dancoff approach , particularly the step from eqs.([eq:321 ] ) and ( [ eq:322 ] ) to eq.([eq:345 ] ) , can be interpreted as the reduction of a block matrix dimension from 2 to 1 . but there is no need for identifying the @xmath109-space with the lowest sector . nothing prevents us from choosing the last sector as the @xmath95-space : the same steps as above reduce then the block matrix dimension from @xmath121 to @xmath122 . the effective interaction acts in the now smaller space . this procedure can be iterated . the price to pay is that one has to deal with ` resolvents of resolvents ' , or with iterated resolvents . ultimately , one arrives at block matrix dimension 1 where the procedure stops : the effective interaction in the fock - space sector with only one quark and one antiquark is defined unambiguously . in order to make progress one needs to readjust notation @xcite . suppose , in the course of this reduction , one has arrived at block matrix dimension @xmath57 , with @xmath123 . denote the corresponding effective interaction @xmath124 . since one starts from the full hamiltonian in the last sector @xmath121 , one has to convene that @xmath125 . the eigenvalue problem corresponding to eq.([eq:345 ] ) reads then @xmath126 observe that @xmath127 and @xmath128 refer here to sector numbers . now , in analogy to eq.([eq:332 ] ) , define @xmath129 note that this definition requires only the inversion of the sector hamiltonian @xmath130 , but not of the total hamiltonian . the effective interaction in the ( @xmath131)-space becomes then @xmath132 this holds for every block matrix element @xmath133 . to get the corresponding eigenvalue equation one substitutes @xmath57 by @xmath134 everywhere in eq.([eq:406 ] ) . everything proceeds like in section [ sec:2 ] , including the fixed point equation @xmath135 . but one has achieved much more : eq.([eq:414 ] ) is a _ recursion relation _ which holds for all @xmath94 ! [ t ] the method of iterated resolvents was applied in @xcite to the paradigmatic case of a @xmath136-matrix , including explicit numerical examples . applying the method to the block matrix structure of qcd , as displayed in table [ tab:3 ] , is particularly easy and transparent . by definition , the last sector contains only the diagonal kinetic energy , thus @xmath137 . its resolvent is calculated trivially . then @xmath138 can be constructed unambiguously , followed by @xmath139 , and so on , until one arrives at sector 1 . grouping them in a different order , one finds for the sectors with one @xmath0-pair : @xmath140 } \ , \label{eq:610}\\ h_{q\bar q\,g } = \qquad h _ 3 & = & t_3 + \phantom{[}v g _ 6 v + v g _ 6 v g _ 5 v g _ 6 v \phantom { ] } + v g _ 4 v \ , \label{eq:620 } \\ h_{q\bar q\,gg } = \qquad h _ 6 & = & t_6 + \phantom{[}v g _ { 10 } v+ v g _ { 10 } v g _ 9 v g _ { 10 } v \phantom { ] } + v g _ 7 v \ , % \label { eq:621 } \\ h_{q\bar q\,ggg } = \qquad h _ { 10 } & = & t_{10 } + \left [ v g _ { 15 } v + v g _ { 15 } v g _ { 14 } v g _ { 15}v\right ] + v g _ { 11 } v \ . \label { eq:622 } \end{aligned}\ ] ] the quark - gluon content of the respective sectors is added here for an easier identification . the square bracket in the last expression indicates how the effective interaction looks for a higher @xmath55 ; for @xmath51 the content of the square bracket has to be set to zero . correspondingly , one obtains for the sectors with two @xmath141-pairs : @xmath142 } \ , % \label { eq:631 } \\ h_{q\bar q q\bar q\ , g } = \qquad h _ 7 & = & t_7 + \phantom{[}v g _ { 11 } v+ v g _ { 11 } v g _ { 10 } v g _ { 11 } v \phantom { ] } + v g _ 8 v \ , % \label { eq:632 } \\ h_{q\bar q q\bar q\ , gg } = \qquad h _ { 11 } & = & t_{11 } + \left [ v g _ { 16 } v + v g _ { 16 } v g _ { 15 } v g _ { 16}v\right ] + v g _ { 12 } v \ , % \label { eq:633 } \end{aligned}\ ] ] and for those with with three @xmath141-pairs @xmath143 } \ , % \label { eq:641 } \\ h_{q\bar q q\bar q q\bar q\ , g } = \qquad h _ { 12 } & = & t_{12 } + \left [ v g _ { 17 } v + v g _ { 17 } v g _ { 16 } v g _ { 12}v\right ] + v g _ { 13 } v \ . % \label { eq:642 } \end{aligned}\ ] ] in either case , their structure is different from the one in the pure glue sectors which is @xmath144 } \ , \label{eq:643 } \\ h_{gg\ , g } = \qquad h _ 5 & = & t_5 + v g _ { 6 } v + \phantom{[}v g _ { 9 } v \phantom { ] } \ , % \label { eq:644 } \\ h_{gg\ , gg } = \qquad h _ 9 & = & t_{9 } + v g _ { 10 } v + \left [ v g _ { 14 } v \right ] \ . \label{eq:645}\end{aligned}\ ] ] having arrived at this point , one can restore first the limit @xmath145 , simply by suppressing the square brackets in the above relations . second , the instantaneous interaction @xmath81 omitted thus far can be restored _ ex post_. due to the peculiar structure of the gauge theory hamiltonian displayed in table [ tab:3 ] , the vertex interaction appears only in even pairs , typically in the combination @xmath146 . it is plausible that by simply substituting @xmath147 restores the full hamiltonian @xmath148 , as was checked explicitly in @xcite . the full effective hamiltonian @xmath149 is given by eq.([eq:610 ] ) and displayed diagrammatically in fig . [ fig:6_1 ] . what does it mean physically ? the first term is the kinetic energy ( @xmath150 ) in the @xmath0-space . in the second term ( @xmath151 ) , the vertex interaction @xmath78 creates a gluon and scatters the system from the @xmath0- into the @xmath152-space . as indicated in fig . [ fig:6_1 ] by the vertical line with the subscript ` 3 ' , the three particles propagate there in ` 3'-space under impact of the full interaction before the gluon is absorbed . the gluon can be absorbed either by the antiquark or by the quark , represented by the two respective graphs @xmath153 and @xmath154 in the figure . the third term ( @xmath155 ) , finally , in eq.([eq:610 ] ) describes the virtual annihilation of the @xmath0-pair into two gluons , and is represented by the graph @xmath156 in fig . [ fig:6_1 ] . it generates an interaction between different quark flavors . as a net result the interaction scatters a quark with helicity @xmath157 and four - momentum @xmath158 into a state with @xmath159 and four - momentum @xmath160 , and correspondingly the antiquark . in the continuum limit , the resolvents are replaced by propagators and the eigenvalue problem @xmath161 becomes again an integral equation but one with a more transparent structure : @xmath162 \langle x,\vec k_{\!\perp } ; \lambda_{q } , \lambda_{\bar q } \vert \psi _ b\rangle & & \nonumber\\ + \sum _ { \lambda_q^\prime,\lambda_{\bar q}^\prime } \!\int_d\!dx^\prime d^2 \vec k_{\!\perp}^\prime\ , \langle x,\vec k_{\!\perp } ; \lambda_{q } , \lambda_{\bar q } \vert u+u_a\vert x^\prime,\vec k_{\!\perp}^\prime ; \lambda_{q}^\prime , \lambda_{\bar q}^\prime\rangle\ , \langle x^\prime,\vec k_{\!\perp}^\prime ; \lambda_{q}^\prime,\lambda_{\bar q}^\prime \vert \psi _ b\rangle & & \ .\label{eq:445}\end{aligned}\ ] ] the effective quark mass @xmath163 includes the impact of diagram @xmath154 . the domain @xmath164 restricts integration in line with regularization ( see below ) . the effective interaction @xmath116 includes all fine and hyperfine interactions . the flavor - changing annihilation interaction @xmath165 is probably of less importance in a first assault and will be disregarded in the sequel . the eigenvalues refer to the invariant mass @xmath166 of a physical state and the corresponding wavefunction @xmath167 gives the probability amplitudes for finding in that state a flavored quark with momentum fraction @xmath89 , intrinsic transverse momentum @xmath82 and helicity @xmath168 . both are boost - invariant quantities . color plays a subordinate role because the fock states can be made color - neutral . the eigenfunctions @xmath169 represent the normalized projections of the full eigenfunction @xmath170 onto the fock states @xmath171 . their knowledge is sufficient to retrieve all desired fock - space components of the total wavefunction @xmath170 . let us work that out explicitly by asking for the probability amplitude to find a @xmath172- or a @xmath173-state in an eigenstate @xmath170 of the full hamiltonian . the key is the upwards recursion relation eq.([eq:410 ] ) . obviously , one can express the higher fock - space components @xmath174 as functionals of @xmath175 by a finite series of quadratures , _ i.e. _ of matrix multiplications or of momentum - space integrations . one needs not solve another eigenvalue problem . the first two equations of the recursive set in eq.([eq:410 ] ) are @xmath176 the sector hamiltonians @xmath177 have to be substituted from eqs.([eq:620 ] ) and ( [ eq:643 ] ) . in taking block matrix elements of them , the formal expressions are simplified considerably since many of the hamiltonian blocks in table [ tab:3 ] are zero . one thus gets simply @xmath178 and therefore @xmath179 . substituting this into eq.([eq:448 ] ) gives @xmath180 . these findings can be summarized more readable as @xmath181 correspondingly , one can write down the higher fock - space wavefunctions as functionals of only @xmath182 . of course , one has to readjust the overall normalization of the wavefunction , and this depends on how many fock spaces one decides to include . finally , it should be emphasized again that the finite number of terms contrasts strongly to the infinite number of terms in perturbative series , because the iterated resolvents re - sum the series to all orders in closed form . the above results are all exact . aiming for the higher fock space amplitudes is not an esoteric problem . one needs them for calculating the structure functions , for example . they are the probability amplitudes for finding in a hadron a quark with longitudinal momentum fraction @xmath89 , irrespective of its transverse momentum , and to get them , one must sum over all fock - space amplitudes @xmath183 , see for example @xcite . [ t ] let us discuss in some greater detail the structure of the sector hamiltonians , particularly the effective interaction in the @xmath184-space as given by eq.([eq:620 ] ) . the corresponding graphs are displayed diagrammatically in figs . [ fig:6_2 ] and [ fig:6_3 ] . those in fig . [ fig:6_2 ] differ from those in fig . [ fig:6_1 ] only by an additional gluon . the gluon does not change quantum numbers under impact of the interaction and acts like a spectator . therefore , the graphs in fig . [ fig:6_2 ] will be referred to as the ` spectator interaction ' @xmath185 . in the graphs of fig . [ fig:6_3 ] , the gluons are scattered by the interaction , and correspondingly these graphs will be referred to as the ` participant interaction ' @xmath186 . the analogous separation into spectators and participants can be made in all quark - pair - glue sector hamiltonians : @xmath187 the hamiltonian is additive in spectators and participants . the spectator interaction @xmath188 therefore can be associated with its own resolvent @xmath189 , @xmath190 with an exact relation to the full resolvents @xmath191 , _ i.e. _ @xmath192 . equivalently , one can write it as an infinite series @xmath193 , as usual . the difference to the usual perturbative series in sec . [ sec:2 ] is , that there the ` unperturbed propagator ' @xmath194 refers to the system without interactions while here the ` unperturbed propagators ' @xmath189 contain the interaction in the well defined form of @xmath195 . one therefore deals here with a ` perturbation theory in medium ' . note that the present series implies also different physics : the system is not scattered into other sectors but stays in sector @xmath57 . this is reflected in the operator identity , which is obtained straightforwardly from eqs.([eq:626 ] ) and ( [ eq:641 ] ) : @xmath196 taking the inverse gives @xmath197 , thus @xmath198 since @xmath199 and @xmath189 commute , one ends up with @xmath200 . in all of the above effective interactions the square matrices @xmath199 are sandwiched between a quark - pair - glue resolvent @xmath201 and the vertex @xmath78 , @xmath202 with @xmath203 being an effective vertex . the index @xmath57 needs not to be carried on . one can thus rewrite systematically eqs.([eq:610])-([eq:622 ] ) . working backwards yields @xmath204 instead of being similar , the quark - pair - glue sector hamiltonians @xmath205 are now all identical . note that , in the strict sense , they are equal only in the continuum limit . for any finite @xmath55 , there is a last sector which is creating edge effects . the resolvents @xmath206 , @xmath207 and @xmath208 carry no bar . in the pure glue sectors , the distinction between participants and spectators makes no sense . thus far the approach is formally exact . in the sequel , this rigor is given up in favor of transparency . for calculating the effective interaction @xmath116 one needs the resolvents @xmath209 . for calculating @xmath209 , one needs @xmath210 and @xmath211 , and so one . does this ever come to an end ? a careful analysis shows that there is a simple relation between them since all @xmath212 are all _ bona fide _ hamiltonians . suppose we have solved the eigenvalue problem in the @xmath0-space , @xmath213 the eigenvalues are enumerated by @xmath214 . the eigenfunctions @xmath215 are a complete set . despite working in the continuum limit , we continue to use summation symbols for the sake of a more compact notation . suppose further that an @xmath104 was found which has the same value as the lowest eigenvalue @xmath216 . the substitution @xmath217 will hence forward be done without explicitly mentioning . next , ask for the eigenvalues and eigenfunctions in the @xmath152-space : @xmath218 it is important to note that one needs not to perform another calculation , becuase the result is already known . by construction , the gluon is a free particle which moves relative to the meson subject to momentum conservation . the eigenfunction is a product state @xmath219 . parameterizing the gluon s four - momentum as @xmath220 implies that the gluon moves with momentum fraction @xmath221 and transverse momentum @xmath222 . it also implies that the meson moves with momentum fraction @xmath223 and transverse momentum @xmath224 . since the gluons are massless , the eigenvalues are therefore @xmath225 knowing the eigenvalues and eigenfunctions , one can calculate the exact resolvent . a few identical rewritings give @xmath226\,,\qquad{\rm with } \label{eq:6.69}\\ a&= & \sum _ { b , s } \vert\psi_{b , s}\rangle\,{y(m_b^2-m^2 ) \over q^2+y(m_b^2-m^2)}\ , \langle \psi_{b , s } \vert \,,\\ \overline g _ 3 ( q;\bar q;g ) & = & -{y\over q^2 } \ , \quad{\rm and}\quad q^2 = ( y^2 m ^2 + \vec q_{\!\perp}^{\,2}){1\over 1-y } \,.\label{eq:6.68}\end{aligned}\ ] ] since the operator @xmath227 by definition can not become a dirac-@xmath228 function @xmath229 , one can drop @xmath227 , as an approximation . the same type of approximation parametrizes the two gluon propagator @xmath230 in terms of the glue ball mass @xmath231 . let us digress for a short discussion . one should note first that the above approximation can be _ controlled e posteriori _ : once the eigenfunctions in the @xmath0-sector are available numerically , one can check whether the exact definition of the resolvent eq.([eq:6.69 ] ) is peaked like a @xmath228-function , with a residue predicted by eq.([eq:6.68 ] ) . second , the above notation is suggestive for @xmath232 being related to the single - particle four - momentum transfer along the quark line in graph @xmath153 in figure [ fig:6_1 ] . in fact one gets @xmath233 $ ] , for sufficiently small @xmath221 . the momentum transfer is the same as in eq.([eq:6.68 ] ) if one replaces the ` current mass ' @xmath234 by the ` constituent mass ' @xmath235 . apart from that the exact resolvent agrees with the perturbative propagator : in the solution , the particles propagate like free particles . third , and most importantly , instead of having resolvents of resolvents , the hierarchy of iterated resolvents is broken . _ ex post _ , this justifies the _ ad - hoc _ procedures in the tamm - dancoff approach , both in the original work @xcite and in its light - cone adaption @xcite . last but not least , one should emphasize that the complicated operator @xmath227 in eq.([eq:6.69 ] ) can perhaps be omitted when dealing with the low energy part of the spectrum . this becomes inadequate if not false at sufficiently high excitations , at sufficiently large @xmath104 , where one might be in the region of ` overlapping resonances ' @xcite . this requires different techniques , possibly those based on random matrix models @xcite . it is here where the _ concept of a temperature _ will possibly make its way into the hamiltonian bound - state problem . [ t ] next , how can one approximate the vertex functions @xmath199 ? with eq.([eq:6530 ] ) one has @xmath236 for @xmath237 , this reads @xmath238 ) ^ { -{1\over2 } } $ ] . expanding up to the first non - trivial term in the expansion gives @xmath239 for being consistent one must set @xmath240 . unfortunately , this was not realized in the previous work @xcite . the vertex function @xmath241 identifies itself as the conventional vertex function . the altogether 5 diagrams of light - cone qcd had been calculated before by thorn @xcite and by perry @xcite . because they are not available in the form we need them in the sequel , we have recalculated them and present them elsewhere @xcite . in the nomenclature of @xcite , the mass term @xmath242 cancels almost completely with the two vertex corrections @xmath243 and @xmath244 . in the sequel we consider explicitly only the two vacuum polarization diagrams @xmath245 and @xmath246 , as given in figs . [ fig : g_1 ] and [ fig : g_2 ] . parameterizing the particle momenta as @xmath247 we calculate first the propagators in the @xmath248 and @xmath249 sectors , @xmath250 they depend on @xmath232 , the momentum transfer along the quark line . the vacuum polarization diagrams are then calculable according to the rules of light - cone perturbation theory @xcite and yield @xmath251 the dependence on @xmath232 is caused by the propagators . the term with @xmath252 in @xmath246 cancels with @xmath253 , like in the previous work @xcite . these authors have however not kept track of the @xmath232-dependence . dropping this term replaces @xmath246 with @xmath254 . the cut - off functions @xmath255 are defined by vertex regularization in eq.([eq : i1 ] ) and ( [ eq : i2 ] ) @xmath256 the limiting transverse momentum @xmath257 describes a semi - circle in the appropriate units , @xmath258 \nonumber\\ \quad{\rm with}\quad b&=&{\lambda^2 + 2m_1(m_1+m_2)\over \lambda^2+(m_1+m_2)^2 } \quad{\rm and}\quad c^2={\lambda^2\over\lambda^2+(m_1+m_2)^2 } { \lambda^2 + 4m_1m_2\over\lambda^2+(m_1+m_2)^2 } \,.\end{aligned}\ ] ] the semi - circle is limited by its intersections at @xmath259 and @xmath260 , with @xmath261 for sufficiently large @xmath64 and equal masses @xmath262 , the limits become @xmath263 . for the purpose of regularization , the gluons are endorsed with a small mass @xmath264 , and thus vertex regularization regulates all divergences on the light cone . one gets straightforwardly @xmath265 this is approximated by replacing the term @xmath266 by its maximum at @xmath267 , that is by @xmath268 . the same type of approximation is done for @xmath246 , yielding finally @xmath269 in either case , the terms depending on @xmath270 or @xmath271 have been dropped since they are small . the effective coupling constant @xmath272 , @xmath273 depends therefore on the momentum transfer . finally , the effective quark and gluon masses become @xmath274 they are obtained straightforwardly by light - cone perturbation theory @xcite combined with vertex regularization . the diagram for the effective quark mass @xmath275 is given in fig . [ fig : g_3 ] . there is a corresponding diagram for the effective gluon mass @xmath276 . collecting terms and restricting to @xmath277 colors , the effective interaction as defined in eq.([eq:445 ] ) becomes without the flavor changing interaction @xmath278 @xmath279\langle x,\vec k_{\!\perp } ; \lambda_{q } , \lambda_{\bar q } \vert \psi\rangle & & \nonumber\\ - { 1\over 3\pi^2 } \sum _ { \lambda_q^\prime,\lambda_{\bar q}^\prime } \!\int_d\!dx^\prime d^2 \vec k_{\!\perp}^\prime\ , \,{\overline\alpha_s(q ) \over q ^2 } \,\langle \lambda_q,\lambda_{\bar q}\vert s(q)\vert \lambda_q^\prime,\lambda_{\bar q}^\prime\rangle \ \langle x^\prime,\vec k_{\!\perp}^\prime ; \lambda_q^\prime,\lambda_{\bar q}^\prime \vert \psi\rangle & & \,.\label{eq : i65}\end{aligned}\ ] ] one recalls that the four - momentum transfer along the quark ( or anti - quark ) line is @xmath280 . as we are supposed to do , we restore the instantaneous interaction @xmath81 in the @xmath0-space when deriving the above result , thus @xmath281 . both the instantaneous interaction @xmath81 and the vertex interaction @xmath282 have a non - integrable singularity , _ cf _ @xcite , which cancels each other such that only the integrable coulomb singularity remains . the latter manifests itself in eq.([eq : i65 ] ) in the `` coulomb denominator '' @xmath283 . the spinor factor @xmath284 represents the familiar current - current coupling which describes all fine and hyperfine interactions @xmath285 \over \sqrt{x x^\prime } } \ , { \left [ \overline u ( k_{\bar q},\lambda_{\bar q } ) \,\gamma_\mu\ , u(k_{\bar q}^\prime,\lambda_{\bar q}^\prime)\right ] \over \sqrt{(1 - x ) ( 1- x ^\prime ) } } \,.\end{aligned}\ ] ] one can simplify the expression technically by suppressing in the latter all momentum dependence . evaluating the spinors at equilibrium , _ i.e. _ at vanishing transverse momenta and at momentum fractions @xmath286 and @xmath287 , gives @xcite @xmath288 \over \sqrt{x_{q}x_{q}^\prime } } { \left[\overline u(k_{\bar q},\lambda_{\bar q})\,\gamma_\mu\ , u(k_{\bar q}^\prime,\lambda_{\bar q}^\prime)\right]\over \sqrt{x_{\bar q } x_{\bar q}^\prime } } \bigg\vert_{\rm equilibrium } = 4(m_q+m_{\bar q})^2\ , \delta_{\lambda_{q},\lambda_{q}^\prime } \delta_{\lambda_{\bar q},\lambda_{\bar q}^\prime } \,.\end{aligned}\ ] ] omitting the spin - dependence of the interaction implies to restrict oneselve to the central part of the interaction . in this approximation eq.([eq : i65 ] ) becomes @xmath289\langle x,\vec k_{\!\perp}\vert \psi\rangle - { 4(m_q+m_{\bar q})^2\over 3\pi^2 } \int_d\!\!\!dx^\prime d^2 \vec k_{\!\perp}^\prime \,{\overline\alpha_s(q ) \over q ^2 } \ \langle x^\prime,\vec k_{\!\perp}^\prime \vert \psi\rangle \,.\label{eq : i68}\ ] ] it is this equation , finally , which we are going to analyze by the renormalization group . [ t ] lr @xmath290 & 6 mev + @xmath291 & 10 mev + @xmath292 & 200 mev + @xmath293 & 1.40 gev + @xmath294 & 4.70 gev + @xmath295 & 180 gev + @xmath296 & 0.2 + @xmath297 & 100 mev + @xmath298 & 13 mev + in order to do calculations , one has to specify 10 `` input parameters '' . as specified in fig . [ tab : quarks ] , these are the six flavor masses @xmath299 @xcite , the coupling constant @xmath300 , and the two regularization parameters @xmath64 and @xmath264 . to be concrete , these parameters are specified in fig . [ tab : quarks ] , and subject to future change . the eigenvalues of the effective equation ( [ eq : i68 ] ) depend on all of them , _ i.e. _ @xmath301 . while their dependence on the masses and coupling constant is not unexpected , they should not depend on @xmath64 , since that is an arbitrary and unphysical parameter . one therefore must require that the eigenvalues satisfy the renormalization group equations @xmath302 since the lagrangian ` bare ' parameters themselves are arbitrary , one can try to let them be functions of @xmath64 in order to satisfy eq.([eq : i69 ] ) : @xmath303 , @xmath304 , and @xmath305 . the numbers in fig . [ tab : quarks ] can be viewed as such functions at the particular value @xmath306 mev . if one finds one single set of functions which makes all eigenvalues @xmath307 independent of @xmath64 , one has solved the problem . it is not really necessary that @xmath308 . in the present work , the eigenvalues of the full light - cone hamiltonian @xmath309 in eq.([eq:2.2 ] ) are conceptually identical with the eigenvalues of the effective hamiltonian , eq.([eq : i68 ] ) . one must therefore analyze the latter . the dependence on the cut - off resides in three well separated locations : ( 1 ) in the domain of integration @xmath164 , ( 2 ) in the effective coupling constant @xmath310 , and ( 3 ) in the effective mass parameters @xmath311 and @xmath312 . as an advantage of vertex regularization , one can decouple the scales in @xmath164 and in the functions @xmath313 and @xmath314 . in the sequel we disregard the dependence on @xmath164 . if @xmath313 and @xmath314 do not dependent on @xmath64 , the eigenvalues can not depend on @xmath64 , either . it therefore suffices to require @xmath315 the variations of @xmath316 have to be performed of course at a fixed value of @xmath232 . solving the corresponding differential equations for @xmath317 and @xmath318 yields the required functions . there are two regimes which are particularly interesting . one is the _ asymptotic regime _ @xmath319 in which eqs.([eq : i62 ] ) and ( [ eq : i63 ] ) become @xmath320 solving eq.([eq : i70 ] ) in this regime yields _ asymptotic freedom _ in the front form , as shown first by thorn @xcite and later by perry _ the asymptotic regime is particularly important for high energy scattering . but in a bound state calculation , the very large momentum transfers are less important than the very small ones , because of the coulomb singularity @xmath321 , see eqs.([eq : i65 ] ) and ( [ eq : i68 ] ) . it provides the binding . in the sequel we therefore emphasize the _ bound state regime _ near @xmath321 , @xmath322 to our recollection , such a regime has not been analyzed by the renormalization group thus far . we shall differentiate between the two light quarks with mass @xmath323 ( u , d ) and the four heavy quarks with mass @xmath324 ( s , c , b , t ) . for convenience we collect some mass parameters consistent with @xcite in fig . [ tab : quarks ] . in the present context , ` light ' and ` heavy ' is defined operationally by @xmath325 consider first the effective gluon mass . in a gauge theory the vector bosons should have zero mass @xmath326 . this in turn defines @xmath264 @xmath327 according to eq.([eq : i64 ] ) . expanding the logarithm for the former gives @xmath328 with the parameter values of fig . [ tab : quarks ] , the numbers in the parenthesis go like @xmath329 . it is thus save to neglect the impact of the light quark masses . this yields @xmath330 mev as quoted in the figure and removes the regulating gluon mass as a free parameter from the theory . in the vicinity of @xmath321 , the effective coupling constant goes like @xmath331 the last term is due to expanding the coefficient of @xmath232 up to second order . treating differently the heavy and the light quarks again and substituting the gluon mass gives @xmath332 near @xmath321 the renormalization group equation for the effective coupling constant ( [ eq : i70 ] ) gives @xmath333+{\alpha^2\lambda\over \pi q^2}\,\delta\lambda = 0\ , , \qquad{\rm thus}\quad \delta(\alpha\lambda)=0 \,.\label{eq : i81}\ ] ] quite unexpectedly , @xmath334 turns out as a renormalization group invariant . the coupling constant `` runs '' like @xmath335 it is admitted to call this `` asymptotic freedom in the bound state regime '' . the subscript @xmath336 denotes the renormalization point . a particular `` renormalization point '' is the set of numbers in fig . [ tab : quarks ] . finally , with the same approximations , the effective quark mass becomes according to eq.([eq : i64 ] ) @xmath337 renormalizing the quark mass according to eq.([eq : i71 ] ) and introducing @xmath338 as an integration variable , thus @xmath339)}$ ] , gives @xmath340 one notes that the mass ratios @xmath341 are renormalization group invariants , as they should . in the bound state regime , the kernel of the integral equation ( [ eq : i68 ] ) becomes thus @xmath342 its fourier transforms behave like @xmath343 , _ i.e. _ like the superposition of a coulomb and a linear potential governed by a `` string constant @xmath344 '' , see for example @xcite . the theory `` confines '' the two coefficients @xmath345 are related to each other by the renormalization group . the string constant becomes a parameter of the theory which must be determined by experiment . the value @xmath346 mev consistent with fig . [ tab : quarks ] is probably not very realistic . although the present approach has yet to be applied to hadrons , the calculational scheme is rather different from the one taken by wilson and collaborators @xcite , see also @xcite and @xcite , mostly because the present approach emphasizes so strongly the region around @xmath347 . a detailed comparison should be given separately when dealing with renormalizing the scales connected to the integration domain @xmath164 in eqs . ( [ eq : i65 ] ) and ( [ eq : i68 ] ) . it is more than likely that the similarity transform of wegner and collaborators @xcite will be there more than useful . in order to do calculations , one must give 8 parameters : the bare coupling constant , the renormalization scale and the six bare flavor masses @xmath348 . as always in the sequel , we understand these parameters as given at the renormalization point and drop the subscript @xmath336 . it is not relevant in this context whether one gives the bare or the dressed masses , since they are related by the coupling constant in a simple way , eq.([eq : i79 ] ) . these are seven dimensionful and one dimensionless parameter . we like to think of them in an other way . the relation between @xmath64 and @xmath344 is linear , see eq.([eq : i85 ] ) . because of that one can introduce @xmath349 as independent variables and drop any explicit reference to @xmath64 as the parameter of regularization . rather the string constant @xmath344 sets the scale of the problem . the rest are dimensionless parameters , the coupling constant or mass ratios , which are introduced to cope with the plenitude of experimental numbers . it looks as if the effective theory has one parameter more than the lagrangian , but this is not quite true . in a lagrangian theory , the absolute values of the spectral eigenvalues @xmath307 are physically meaningless . one can multiply the lagrangian and thus the hamiltonian with an arbitrary number without changing the physics . in a way , renormalization theory fixes that number by explicitly introducing a scale . finally , one should emphasize that one is not completely free in choosing the scale : the above formalism holds only for at least one ` heavy ' quark , with a mass much larger than @xmath344 . although there is a long way until one reaches the 180 gev of the top , one may ask what happens to the above when one transgresses this limit . if one repeats the above analysis for @xmath321 but in the regime @xmath350 the formalism changes . the situation is similar to the asymptotic regime : the coefficient of @xmath351 is roughly @xmath352 , thus unstable with respect to the renormalization group . as a consequence it fades away . the result of a renormalization group analysis depends on the scale at which it is performed . in conclusion , we are afraid to pretend that the spectrum of toponium will be that one of a de - confining system , with an ionization threshold like positronium . the present work has four important ingredients : ( 1 ) the method of iterated resolvents is applied to a gauge theory hamiltonian in the front form ; ( 2 ) in the solution the full quark - pair - gluon propagators can be well approximated by free propagators with the quarks carrying the constituent mass @xmath353 ; ( 3 ) vertex as opposed to fock space regularization is more efficient to regulate all divergences in the light - cone formalism ; ( 4 ) after a renormalization group analysis , the spectrum of the hamiltonian is independent of the two regularization parameters and in the confining regime a new mass scale appears which must by determined by experiment . after the dust has settled , one seems to get a glimpse of an idea how confinement could work in practice . one should emphasize that in this non - perturbative approach there is no fock space truncation . one calculates lorentz scalars , the eigenvalues of the invariant mass - squared , in a frame independent way . even the kernel of the interaction @xmath354 is expressed in terms of an invariant , @xmath355 , which is the momentum transfer of the quark . the present work adopts systematically the point of view that all properties of a lagrangian gauge field theory are contained in the canonical front - form hamiltonian , which is calculated in the light - cone gauge without the zero modes . periodic boundary conditions allow one to construct explicitly a hamiltonian matrix . after regularization , one is confronted with the diagonalization of a large but finite matrix . the binding of the particles in a field theory is provided by the virtual scattering into the higher fock spaces and the tamm - dancoff approach accounts for that , however it accounts for that only in lowest order . for large values of the coupling one loses control . these difficulties are overcome in the the present work after re - analyzing the theory of effective interactions . by inverting the hierarchy , it was possible to develop a new formalism , the method of iterated resolvents , and using this method , one can map the large hamiltonian matrix onto a matrix which is smaller . finally , the periodic boundary conditions are relaxed by going over to the continuum limit . having found the eigenfunctions of the effective interaction , one can retrieve all other many - body amplitudes in a self - consistent way . we saw that the effective interaction has two parts : the flavor conserving interaction @xmath116 and flavor changing interaction @xmath278 . their diagrammatic representations look like second order diagrams of perturbation theory , but represent a re - summation of perturbative graphs to all orders . particularly , @xmath116 bears great similarity with a one - gluon - exchange interaction . in analyzing this effective interaction one uses the apparent self - similarity in a gauge theory . ultimately , this aspect allows the ` breaking of the hierarchy ' : in the solution , the particles propagate like free particles . all many - body aspects reside in the vertex coupling functions , in which the well known divergences of field theory appear typically as @xmath356 . this can become large either for large transverse momentum @xmath357or for small longitudinal momentum fraction @xmath89 . in order to remove these divergences , one needs _ two regulator scales_. regulating the theory on the level of the fundamental interaction , we introduce vertex regularization associated with a mass scale @xmath64 and a small kinematical gluon mass @xmath264 . insisting that the gauge bosons have vanishing dynamical mass , relates @xmath264 to @xmath64 and removes it from the formalism . the appearance of a new scale @xmath344 provides the room for an interesting speculation . one knows empirically that the invariant mass square of the light mesons like the pion is linear in the quark masses . for dimensional reasons one needs a parameter of dimension mass to account for that . one believes that this parameter is the @xmath0-vacuum condensate @xmath358 , _ i.e. _ @xmath359 . the vacuum condensate is extremely difficult to calculate and usually is taken from experiment . with the appearance of the additional scale @xmath344 , which can play that role , one can have @xmath360 . also @xmath344 has to be fitted to experiment , but its role within the theory is well defined . w. lucha , f.f . schberl , and d. gromes , phys . rept . * 200 * , 127 ( 1991 ) ; and references therein . mackenzie , _ status of lattice qcd , in : ithaka 1993 , proceedings of ` lepton and photon interactions ' , p.634 ; hep - ph 9311242 ; and references therein . _ davies , k. hornbostel , g.p . lepage , a.j . lidsay , j. shigemitsu , and j. sloan , hep - lat/9506026 , phys.rev . * d52 * , 6519 ( 1995 ) . s. godfroy and n. isgur , phys.rev . * d32 * , 189 ( 1985 ) . see for example : yu.l . dokshitzer , v.a . khoze , a.h . mueller and s.i . troyan , , ( editions frontires , gif - sur - yvette , 1991 ) . s.j . brodsky , h.c . pauli , and s.s . pinsky , `` quantum chromodynamics and other field theories on the light cone '' , heidelberg preprint mpih - v1 - 1997 , stanford preprint slac pub 7484 , apr . 1997 , 203 pp , hep - th/9707455 , submitted to physics reports . verbaarschot , h.a . weidenmller , and m. zirnbauer , phys.lett . * c129 * , 367 ( 1985 ) . j.j.m . verbaarschot , phys.rev.lett . * 72 * , 2531 ( 1994 ) . t. wettig , a. schfer , and h.a . weidenmller , phys.lett . * b367 * , 28 ( 1996 ) . thorn , phys.rev . * d19 * , 639 ( 1979 ) . perry , a. harindranath , and w.m . zhang , phys.lett . * b300 * , 8 ( 1993 ) . j. raufeisen , and h.c . pauli , `` the one - loop vertex corrections on the light cone '' , to be published ( 1997 ) .

we study a new kind of phase ordering phenomenon in coarse - grained depth models of the hill - valley profile of fluctuating surfaces with zero overall tilt , and for hard - core particles sliding on such surfaces under gravity . we find that several such systems approach an ordered state with large scale fluctuations which make them qualitatively different from conventional phase ordered states . we consider surfaces in the edwards - wilkinson ( ew ) , kardar - parisi - zhang ( kpz ) and golubovic - bruinsma - das sarma - tamborenea ( gbdt ) universality classes . for ew and kpz surfaces , coarse - grained depth models of the surface profile exhibit coarsening to an ordered steady state in which the order parameter has a broad distribution even in the thermodynamic limit , the distribution of particle cluster sizes decays as a power - law ( with an exponent ) , and the 2-point spatial correlation function has a cusp ( with an exponent ) at small values of the argument . the latter feature indicates a deviation from the porod law which holds customarily , in coarsening with scalar order parameters . we present several numerical and exact analytical results for the coarsening process and the steady state . for linear surface models with dynamical exponent , we show that for , for , and there are logarithmic corrections for , implying for the ew surface and for the gbdt surface . within the independent interval approximation we show that . we also study the dynamics of hard - core particles sliding locally downwards on these fluctuating one - dimensional surfaces and find that the surface fluctuations lead to large - scale clustering of the particles . we find a surface - fluctuation driven coarsening of initially randomly arranged particles ; the coarsening length scale grows as . the scaled density - density correlation function of the sliding particles shows a cusp with exponent , and for the ew and kpz surfaces . the particles on the gbdt surface show conventional coarsening ( porod ) behavior with . 0.5 cm pacs numbers : 05.70.ln , 05.40.-a , 02.50.-r , 64.75.+g # 1#2#3 = 0.75 2

cond-mat0102521

phase ordering dynamics describes the way in which domains of an ordered state develop when an initially disordered system is placed in an environment which promotes ordering . for instance , when a simple ferromagnet or alloy is quenched rapidly from very high to very low temperatures @xmath14 , domains of equilibrium low-@xmath14 ordered phases form and grow to macroscopic sizes . a quantitative description of the ordering process is provided by the time development of the two - point correlation function ; asymptotically , it is a function only of the separation scaled by a length which increases with time , typically as a power law @xcite . new phenomena and effects can arise when we deal with phase ordering in systems which are approaching _ nonequilibrium _ steady states . in this paper , we study a coupled - field nonequilibrium system in which one field evolves autonomously and influences the dynamics of the other . the system shows phase ordering of a new sort , whose principal characteristic is that fluctuations are very strong and do not damp down in the thermodynamic limit hence the term fluctuation - dominated phase ordering ( fdpo ) . in usual phase ordering systems such as ferromagnetic ising model , if one considers a finite system and waits for infinite time , then the system reaches a state with magnetization per site very close to the two possible values of the spontaneous magnetization , @xmath15 or @xmath16 , with very infrequent transitions between the two . this is reflected in a probability distribution for the order parameter which is sharply peaked at these two values , with the width of the peaks approaching zero in the thermodynamic limit ( fig . [ fdpo](a ) ) . by contrast , in the fdpo steady state , the system continually shows strong fluctuations in time without , however , losing macroscopic order . accordingly , the order parameter shows strong variations in time , reflected eventually in a probability distribution which remains broad even in the thermodynamic limit ( fig . [ fdpo](b ) ) . the physical system we study consists of an independently stochastically fluctuating surface of zero average slope , on which reside particles which tend to slide downwards guided by the local slopes of the surface . somewhat surprisingly , a state with uniform particle density is unstable towards large scale clustering under the action of surface fluctuations . eventually it is driven to a phase - ordered state with macroscopic inhomogeneities of the density , of the fdpo sort . besides exhibiting a broad order parameter distribution , this state shows unusual scaling of two - point correlation functions and cluster distributions . it turns out that much of the physics of this type of ordering is also captured by a simpler model involving a coarse - grained characterization of the surface alone , and we study this as well . a brief account of some of our results has appeared in @xcite . in the remainder of the introduction , we first discuss the characteristics of fdpo _ vis a vis _ normal phase - ordered states . we then discuss , in a qualitative way , the occurrence of fdpo in the surface - driven models under study . the layout of the rest of the paper is as follows . in section [ fdpo_cd ] , we define and study the coarsening and steady states of three different coarse - grained depth models of the fluctuating surfaces . in section [ understandfdpo ] , we demonstrate the existence of a power - law in the cluster size distribution , and show how it can give rise to fdpo . in section [ fdpo_sp ] , we discuss ordering of sliding particles on fluctuating surfaces . in section [ robust ] , we explore the robustness of fdpo with respect to changes in various rates defining the nonequilibrium process . finally , in section [ conclusion ] we summarize our principal results , and discuss the possible occurrence of fdpo in models of other physical systems . with the aim of bringing out the features of fluctuation dominated phase ordering ( fdpo ) in nonequilibrium systems , let us recall some familiar facts about phase ordered states in equilibrium statistical systems . we first discuss different characterizations of spontaneous ordering , following the paper of griffiths @xcite on the magnetization of idealized ferromagnets . we follow this with a discussion of fluctuations about the ordered state . * ( a ) * in the absence of a conservation law , the magnetization @xmath17 is an indicator of the ordering : m= 1 l^d _ n s_n [ defm11 ] where @xmath18 is the linear size , @xmath19 is the dimension and @xmath20 is spin at site @xmath21 . in the thermodynamic limit , the thermal average of the absolute value m_1= lim_l |m| ( nonconserved ) [ defm1 ] with boltzmann - gibbs weights for configurations provides an unequivocal measure of the order . this is because in the low - temperature ordered phase , the probability @xmath22 of occurrence of magnetization @xmath17 is peaked at @xmath23 and @xmath16 ; the peak widths approach zero in the thermodynamic limit @xmath24 , so that the average value @xmath25 coincides with the peak value @xmath15 ( fig . [ fdpo]@xmath26 ) . for the conserved order parameter case , the value of the magnetization is a constant and is same in both the disordered and ordered phase . one therefore needs a quantity that is sensitive to the difference between order and disorder . the simplest such quantity is the lowest nonzero fourier mode of the density @xcite |q| = 1 l|_n e^2i n / l(1+s_n)2| [ defq*1 ] where @xmath27 denotes the average magnetization in the @xmath28-dimensional plane @xmath21 oriented perpendicular to the @xmath29 direction . the modulus in eq . [ defq*1 ] above leads to the same value for all states which can be reached from each other by a translational shift . in the low-@xmath14 ordered phase , @xmath30 is expected to be a sharply peaked function , with peak widths vanishing in the thermodynamic limit . then the mean value @xmath31 defined by q_1= lim_l |q| ( conserved ) [ defq * ] serves as an order parameter . a disordered state corresponds to @xmath32 , while a perfectly ordered state with @xmath33 in half of the system and @xmath34 in the other half corresponds to @xmath35 * ( b ) * another characterization of the order is obtained from the asymptotic value of the 2-point spatial correlation function @xmath36 . at large separations @xmath37 , @xmath38 is expected to decouple : s_o s_o+r= s_o s_o+r = m_c^2 . [ defmc ] a finite value of @xmath39 indicates that the system has long - range order . a value @xmath40 would indicate a perfectly ordered pure phase without any droplets of the other species ( like the @xmath41 state of an ising ferromagnet ) , while @xmath42 would indicate that the phase has an admixture of droplets of the other species ( like the state of an ising ferromagnet for @xmath43 ) . in a finite system , @xmath44 is a function only of the scaled variable @xmath45 in the asymptotic scaling limit @xmath46 ( see also * ( d ) * below ) . an operational way to find the value of @xmath39 is then to read off the intercept ( @xmath47 ) in a plot of @xmath44 versus @xmath45 ; it gives @xmath48 in the @xmath24 limit . in equilibrium systems of the type discussed above , @xmath25 ( defined in eq . [ defm1 ] ) and @xmath39 coincide . * ( c ) * with a conserved scalar order parameter , the low-@xmath14 state is phase - separated , with each phase occupying a macroscopically large region , and separated from the other phase by an interface of width @xmath49 . the interfacial region is quite distinct from either phase , and on the scale of system size , it is structureless and sharp . * ( d ) * customarily in phase - ordered steady states , the spatial correlation function @xmath38 has a scaling form in @xmath50 , for @xmath51 where @xmath18 is the size of the system . in the limit @xmath52 , @xmath38 follows the form @xcite c(r ) m_c^2(1 - 2|r / l| ) ( |r / l| 0 ) [ porodr ] the origin of the linear fall in eq . [ porodr ] is easy to understand in systems where phases are separated by sharp boundaries on the scale of the system size , as in * ( c ) * above : a spatial averaging of @xmath53 produces @xmath54 with probability @xmath55 ( within a phase ) and @xmath56 with probability @xmath50 ( across phases ) . the linear drop with @xmath50 implies that the structure factor @xmath57 , which is the fourier transform of @xmath38 , is given , for large wave - vectors @xmath58 ) , by : ~1 ^d+1 . [ porod ] this form of the decay of the structure factor for scalar order parameters is known as the porod law . it is worth remarking that the forms eqs . [ porodr ] and [ porod ] also describe the behaviour of the two - point correlation function in an infinite system undergoing phase ordering starting from an initially disordered state . in such a case , @xmath18 denotes the coarsening time - dependent length scale which is the characteristic size of an ordered domain . * ( e ) * for usual phase - ordered systems , spatial fluctuations are negligible in the limit of the system size going to infinity . hence the averages of 1-point and 2-point functions over an ensemble of configurations are well represented by a spatial average for a single configuration in a large system . the phase ordering of interest in this paper occurs in certain types of nonequilibrium systems , and the resulting steady state differs qualitatively from the ordered state of equilibrium systems and other types of nonequilibrium systems considered earlier @xcite . the primary difference lies in the effects of fluctuations . customarily , fluctuations lead to large variations of the order parameter which scale sublinearly with the volume , and so are negligible in the thermodynamic limit . fluctuation effects are much stronger here , and lead to variations of the order parameter in time , without , however , losing the fact of ordering . below we discuss how the properties * ( a)-(e ) * discussed above are modified . * ( a ) * nonzero values of the averages @xmath59 and @xmath31 ( eqs . [ defm1 ] and [ defq * ] ) continue to indicate the existence of order , but no longer provide an unequivocal measure of the order parameter . this is because the probability distributions @xmath22 and @xmath30 remain broad even in the limit @xmath60 ( as shown schematically in fig [ fdpo]@xmath61 ) . * ( b ) * the measure @xmath62 of long - range order is nonzero , and its value can be found from the intercept @xmath63 . however , the value of @xmath39 is , in general , quite different from @xmath25 . * ( c ) * as with usual ordered states , the regions of pure phases are of the order of system size @xmath18 . but in contrast to the usual situation , there need not be a well - defined interfacial region , distinct from either phase . rather , the region between the two largest phase stretches is typically a finite fraction of the system size , and has a lot of structure ; this region itself contains stretches of pure phases separated by further such regions , and the pattern repeats . representative spin configurations @xmath64 for the two cases are depicted schematically in fig . [ cusp1 ] . this nested structure is consistent with a power - law distribution of cluster sizes , and thus of a critical state . the crucial extra feature of the fdpo state is that the largest clusters occupy a finite fraction of the total volume , and it is this which leads to a finite value of @xmath39 as in ( b ) above . representative spin configurations @xmath64 for the two cases are depicted schematically in fig . [ cusp1 ] . * ( d ) * the ensemble - averaged spatial correlation function @xmath38 continues to show a scaling form in @xmath50 . however , in contrast to eq . [ porod ] it exhibits a cusp ( fig . [ cusp1 ] ) at small values of @xmath50 : c(r ) m_c^2(1 - b|r l|^ ) . [ scale ] this implies that the scaled structure factor varies as ~1 ^d+ [ cusp ] with @xmath65 . this represents a marked deviation from the porod law ( eq . [ porod ] ) . we will demonstrate in some cases that this deviation is related to the power - law distribution of clusters in the interfacial region separating the domains of pure phases , as discussed in * ( c ) * above . * ( e ) * the spatial average of 1-point functions ( @xmath17 or @xmath66 ) and the 2-point function @xmath44 as a function of @xmath50 in a single configuration of a large system typically do not represent the answers obtained by averaging over an ensemble of configurations . this reflects the occurrence of macroscopic fluctuations . having described the general nature of fluctuation - dominated phase ordering , we now discuss the model systems that we have studied and which show fdpo . we consider physical processes defined on a fluctuating surface with zero average slope . the surface is assumed to have no overhangs , and so is characterized by a single - valued local height variable @xmath67 at position @xmath29 at time @xmath68 as shown in fig . the evolution of the height profile is taken to be governed by a stochastic equation . the height - height correlation function has a scaling form @xcite for large separations of space and time : ^2~|x - x|^2f(|t - t| |x - x|^z ) . [ hhcorr ] here @xmath69 is a scaling function , and @xmath70 and @xmath2 are the roughness and dynamical exponents , respectively . a common value of these exponents and scaling function for several different models of surface fluctuations indicate a common universality class for such models . in this paper we will study one - dimensional surfaces belonging to three such universality classes of surface growth . similar studies of two - dimensional surfaces @xcite show that similar fluctuation - dominated phase - ordered states arise in these cases as well . before turning to the physical model of particles sliding on such fluctuating surfaces , we address the notion of phase ordering in coarse - grained depth models associated with these surface fluctuations . in fig . [ sur1 ] we show the function @xmath71 which take values @xmath72 , @xmath73 and @xmath74 depending on whether the height is below , above or at the same level as some reference height @xmath75 . explicitly , we have @xmath76 . different definitions of @xmath77 define variants of the model ; these are studied in section ii . starting from initially flat surfaces , we study the coarsening of of up - spin or down spin phases , which arise from the evolution of surface profiles . with the passage of time , the surface gets rougher up to some length scale @xmath78 . the profile has hills and valleys , the base lengths of which are of the order of @xmath78 , implying domains of like - valued @xmath79 whose size is of the same order . once the steady state is reached , there are landscape arrangements of the order of the system size @xmath18 which occur on a time scale @xmath80 . however , these landscape fluctuations do not destroy long - range order , but cause large fluctuations in its value . now let us turn to the problem of hard - core particles sliding locally downwards under gravity on these fluctuating surfaces . figure [ par1 ] depicts the evolution of particles falling to the valley bottoms under gravity . when a local valley forms in a region ( fig . 3(a ) @xmath81 fig . 3(b ) ) , particles in that region tend to fall in and cluster together . the point is that particles stay together even when there is a small reverse fluctuation ( valley @xmath82 hill as in fig . [ par1](b ) @xmath82 ( c ) ) ; declustering occurs only if there is a rearrangement on length scales larger than the size of the valley . the combination of random surface fluctuations and the external force due to gravity drive the system towards large - scale clustering . results of our numerical studies show that in the coarsening regime , the typical scale of ordering in the particle - hole system is comparable to the length scale over which surface rearrangements take place . further , the steady state of the particle system exhibits uncommonly large fluctuations , reflecting the existence of similar fluctuations in the underlying coarse - grained depth models of the hill - valley profile . similar effects are seen in 1-point and 2-point correlation functions . the dynamics of surface fluctuations can be modelled by langevin - type equations for the height field @xmath67 . the evolution equations for the one - dimensional edwards - wilkinson ( ew ) @xcite , kardar - parisi - zhang ( kpz ) @xcite , and golubovic - bruinsma - das sarma - tamborenea ( gbdt ) @xcite surface fluctuations are respectively @xmath83 where @xmath84 is a white noise with @xmath85 and @xmath86 , and @xmath87 , @xmath88 and @xmath89 are constants . in one dimension , the ew and kpz models can be simulated using lattice gas models whose large - distance large - time scaling properties coincide with those of the corresponding continuum theories . the lattice gas is composed of @xmath90-valued variables @xmath91 on a @xmath92 lattice with periodic boundary conditions , where the @xmath93 spins occupy the links between sites . the values @xmath94 or @xmath73 represent the local slopes of the surface ( denoted by @xmath95 or @xmath96 , respectively ) . the dynamics of the interface is that of the single - step model @xcite , with stochastic corner flips involving exchange of adjacent @xmath97 s ; thus , @xmath98 with rate @xmath99 , while @xmath100 with rate @xmath101 . for symmetric surface fluctuations ( @xmath102 ) , the behavior at large length and time scale is described by the continuum ew model . for @xmath103 , the surface evolution belongs to the kpz class . corresponding to the configuration @xmath104 we have the height profile @xmath105 with @xmath106 . for simulating a surface fluctuating via a gbdt process , we used a solid - on - solid model with depositing particles piling up on top of each other . the height @xmath107 at site @xmath108 is the height of the pile of particles at that site . during each micro - step a particle is deposited randomly on a site @xmath108 . if the new height @xmath107 at @xmath108 , is greater than @xmath109 and @xmath110 , then with equal probability ( @xmath111 ) three things are attempted the deposited particle can remain at site @xmath108 , or can move to the neighboring sites @xmath112 or @xmath113 . it actually completes the left or right move only if there is an increase in the coordination - ordination number of the particles @xcite . let us imagine a process of coarse - graining which eliminates fine fluctuations of the height profile , and replaces the height field @xmath107 at site @xmath108 by a variable @xmath114 which is + 1 , -1 or 0 depending on whether the surface profile at site @xmath108 is below , above or exactly coincident with a certain reference level , which is the same at all @xmath108 . the aim is to have a coarse - grained construction of locations of large valleys and hills . our procedure depends on the choice of the reference level , and we have explored three choices ( the cd1 , cd2 and cd3 models ) which are discussed below . in model cd1 , the reference level is set by the initial condition , which corresponds to an initially flat interface : @xmath115 . the coarse - grained depth function is then s(x , t ) = -sgn[h(x , t ) ] . [ cd1 ] with the passage of time , the surface becomes rougher , so that @xmath67 develops hills and valleys with respect to the @xmath74 level . as the base lengths of the hills and valleys grow in size , there is a growth of the domains of the variable @xmath71 . we are able to characterize the coarsening behaviour of this model analytically in some cases . in a finite system , at long enough times the surface moves arbitrarily far away from its initial location . thus the steady state of the cd1 model is trivial all @xmath114 are 1 , or all are -1 , with probability one . this clearly happens because the reference level in the cd1 model is fixed in space . this leads us to examine models ( cd2 and cd3 ) where the reference level moves along with the surface , so that we may expect nontrivial steady state properties . in model cd2 , the coarse - grained depth function s_i = - sgn[h_i ] [ cd2 ] where @xmath106 as defined in the earlier section . note that at all times @xmath68 , the origin is pinned so that @xmath116 . the height function of the continuum version of the cd2 model is related to that of cd1 as : @xmath117 . the function @xmath114 is @xmath72 , @xmath73 or @xmath74 accordingly as the height @xmath107 at site @xmath108 is below , above or at the zero level . a stretch of like @xmath114 s @xmath118 represents a valley with respect to the zero level . the time evolution of the cd2 model variables @xmath119 is induced by the underlying dynamics of the bond variables @xmath120 defined in the previous subsection . this model was studied by us in @xcite . finally model cd3 , is defined as follows : @xmath107 is constructed from @xmath97 s exactly as described for the cd2 model , but then one defines s_i = -sgn[h_i - h(t ) ] [ cd3 ] where @xmath121 is the instantaneous average height which fluctuates with time . this definition was used earlier by kim _ et al _ @xcite who were studying domain growth in an evolving kpz surface . each of the cd models defined above has its own merits and limitations . we will see below that the cd1 model proves to be analytically tractable ( for gaussian surface fluctuations ) in the coarsening regime , while for the cd2 model several exact results can be derived in the steady state . of the three models , the cd3 model most resembles the model of sliding hard - core particles on the surface that is studied in section [ fdpo_sp ] below . in this section , our primary focus is on coarsening properties of a class of cd1 models . to this end , we will focus on the equal time correlation function @xmath122sgn[h(x , t)]\rangle . \label{corr1}\ ] ] we consider only linear interfaces evolving from a flat initial condition @xmath123 according to the langevin equation , @xmath124 where @xmath125 is a gaussian white noise with @xmath126 and @xmath127 . the dyanmic exponent @xmath2 specifies the relaxation mechanism . for example , @xmath128 corresponds to edwards - wilkinson ( ew ) interface and @xmath129 corresponds to golubovic - bruinsma - das sarma - tamborenea ( gbdt ) interface . since @xmath130 is a gaussian noise and the evolution equation ( [ interface1 ] ) is linear , the height field @xmath67 is a gaussian process . for gaussian processes , it is straightforward to evaluate the correlation function in eq . ( [ corr1 ] ) exactly and one finds , @xmath131 , \label{arcsine}\ ] ] where @xmath132 is given by , @xmath133 now the normalized height correlation function @xmath132 can be easily computed for linear interfaces evolving via eq . ( [ interface1 ] ) by taking the fourier transform of eq . ( [ interface1 ] ) . from eq . ( [ interface1 ] ) , assuming flat initial condition , the fourier transform , @xmath134 is given exactly by , @xmath135 inverting this fourier transform we get , @xmath136 where the scaling function @xmath137 is given by , @xmath138 using this exact expression of @xmath132 in eq . ( [ arcsine ] ) , we get the exact correlation function for arbitrary linear interface model parametrized by the dynamic exponent @xmath2 . it is also evident that @xmath139 is a single function of the scaled distance , @xmath140 . the small distance behavior of the scaling function can be easily derived from the small argument asymptotics of the integral in eq . ( [ scale2 ] ) . let us first consider the ew interface with @xmath128 . in this case the integral in eq . ( [ scale2 ] ) can be done ( by putting a factor @xmath141 in the exponential , i.e. , writing @xmath142 and then differentiating with respect to @xmath141 and then integrating back with respect to @xmath141 upto @xmath143 ) and we get , @xmath144 a change of variable , @xmath145 gives a more compact expression , @xmath146 integration by parts yields the desired short distance behaviour , @xmath147 putting this back in eq . ( 10 ) and expanding the arcsine , we get , @xmath148 thus the correlation function has a square - root cusp at the origin for the @xmath128 cd1 model . one can similarly do the small distance analysis for arbitrary @xmath149 . we find that for general @xmath2 , @xmath150 where @xmath151 is a @xmath2-dependent constant and the cusp exponent @xmath152 is given by , @xmath153 for @xmath154 , we find additional logarithmic corrections , @xmath155 where @xmath156 . thus our exact results indicate that @xmath157 is a critical value . for @xmath158 , one recovers the linear cusp in the correlation function at short distances ( and hence porod s law ) indicating sharp interfaces between domains as in the usual phase ordering systems . but for @xmath159 , one gets a @xmath2-dependent cusp exponent signalling anomalous phase ordering dominated by strong fluctuations and a significant deviation from porod s law . the value @xmath160 is the one across which a morphological transition has been shown to occur in gaussian surfaces @xcite , in the context of spatial persistence of fluctuating surfaces . unlike the cd1 model , we have not been able to analytically characterize the coarsening properties of the cd2 or cd3 models , in which the reference level moves with time . however the coarsening properties in both cd2 and cd3 models can be studied numerically . below we present the numerical results for the equal time correlation function @xmath161 for the cd3 model in three different cases where the underlying surface is evolving respectively by the ew , kpz and gbdt dynamics . the initial condition chosen was @xmath162 at odd bonds and @xmath73 at even bond locations , ensuring that the height profile was globally flat . we used a lattice with a number @xmath163 of bonds and equal number of sites . at time @xmath164 correlations gradually develop as the @xmath79-spin domains grow . in figs . [ ewcd3 ] , [ kpzcd3 ] , and [ dtcd3 ] we show the data for @xmath161 as a function of @xmath68 ( insets of the respective figures ) , and how they collapse on to a single curve @xmath165 in each case , on scaling @xmath37 by a @xmath68-dependent length scale @xmath78 . for each of the three cases , we see that @xmath166 , where the dynamical exponent @xmath167 , @xmath168 , and @xmath169 , respectively for the ew , kpz and gbdt surfaces . notice that the scaling curves for ew and kpz surfaces have a cusp at small values of the argument @xmath170 , and the cusp exponent ( eq . [ scale ] , [ cusp ] ) @xmath11 for both . for the gbdt surface there is no cusp , and @xmath171 . we note that these results for the cd3 model are consistent with the analytical results in eq . ( [ e17 ] ) of the cd1 model . the fact that the correlation function has a scaling form in @xmath172 , with a nonzero intercept implies that at infinite time the system would reach an ordered steady state , as the value of @xmath173 at any fixed @xmath37 ( no matter how large ) approaches the value of the intercept at large enough time . the intercepts of all the three curves in figs . [ ewcd3 ] , [ kpzcd3 ] , and [ dtcd3 ] have the value @xmath8 implying that @xmath40 for the cd3 model . in a finite system , as time passes the surface diffuses away from its @xmath174 location . as discussed above , this leads to a trivial steady state in the cd1 models , corresponding to all @xmath175 ( or all @xmath176 ) with probability one . we need the reference level to keep up with the surface in order to probe the steady state aspects of coarse - grained surface fluctuations . this is accomplished in the cd2 and cd3 models . in both the cd2 and cd3 models we will see below that the cluster size distribution of the @xmath114 variables varies as a power law @xmath177 in the steady state . the order parameters have a broad distribution and the scaled 2-point function has a cusp . it is well known that for both ew and kpz surfaces in @xmath92 , the steady states have random local slopes @xcite , i.e. the steady state probability distribution of the height profile is p(\{h } ) = p_o e^- . [ rw ] this leads to a mapping of each surface configuration in the cd2 and cd3 models to a random walk ( rw ) trajectory . the correspondence is as follows : @xmath178 or @xmath73 can be interpreted as the rightward or leftward rw step at the @xmath108th time instant . then in the cd2 model , @xmath179 or @xmath74 depending on whether the walker is to the right , to the left or at the origin after the @xmath108th step . in the cd3 model , the reference point for demarcating left ( @xmath180 ) and right ( @xmath181 ) is the average of displacements ( heights ) , and can be fixed only after the full trajectory is specified ; then with respect to @xmath182 , the value of the position of the walker at every @xmath108th instant gets specified and hence also the @xmath114 spins . for the cd2 model with ew or kpz dynamics , exact results for different properties in the steady state can be derived , because the surface profiles map on to random walks . periodic boundary conditions imply that the rw starts at time @xmath74 from the origin and comes back to the origin after @xmath18 time steps . evidently , the lengths of clusters of @xmath183 spins ( or @xmath184 spins ) represent times between successive returns to the origin . thus @xmath185 , the probability distribution of the cluster sizes @xmath186 , for the cd2 model is exactly the well - known distribution ( @xmath187 ) for rw return times to the origin , which behaves as @xmath188 ( for large @xmath186 ) with a cutoff at @xmath189 . thus @xmath190 in this model . for the cd3 model , the variable reference point makes it difficult to make exact statements , but we expect that the cluster size distribution at large lengths @xmath186 will still be given as @xmath191 . the numerically determined @xmath185 s for the cd2 and cd3 models are plotted in fig . [ plcd2cd3 ] , and they show the expected power - law decay . we note that the power - law distribution of the intervals between successive returns is related to the spatial persistence of fluctuating interfaces@xcite . for linear interfaces evolving via eq . ( [ interface1 ] ) , the corresponding spatial persistence exponents were computed recently in ref . @xcite . from these results we infer that for cd models with gbdt dynamics , the cluster size distribution in the steady state also has a power law distribution , @xmath192 in @xmath193-dimensions . the distributions of the order parameters for each of the cd2 and cd3 models are broad . for the cd2 model , an appropriate ( nonconserved ) order parameter is average value @xmath25 of modulus of @xmath194 ( see eq . [ defm1 ] ) , which for the rw represents the excess time a walker spends on one side of the origin over the other side . in order to respect periodic boundary conditions , we need to restrict the ensemble of rws to those which return to the origin after @xmath18 steps . the full probability distribution of @xmath17 over this ensemble is known from the equidistribution theorem on sojourn times of a rw @xcite : prob(m ) = 1/2 , m , [ probm1 ] i.e. every allowed value of @xmath17 is equally likely . this implies @xmath195 and @xmath196 . for the cd3 model , most often half of the surface profile is above the average height level and half below it . as a consequence , we find numerically that the distribution of cluster sizes @xmath185 decays sharply beyond @xmath197 . this resembles the sliding hard - core particles and hence the conserved order parameter @xmath31 is more suitable to describe the ordering in this model than @xmath25 . we monitor the average value @xmath31 of @xmath198 ( eq . [ defq * ] ) , where @xmath199 . this order parameter has a value @xmath74 for a disordered configuration and a value @xmath200 for a fully phase separated configuration with two domains of @xmath201 and @xmath202 spins , each of length @xmath197 . the numerical value of the distribution @xmath203 of @xmath204 is shown in fig . [ probq1cd3 ] , and the average value @xmath31 in the limit of large system size numerically approaches the value @xmath205 . it is apparent from fig . [ probq1cd3 ] that @xmath203 is broad , and is larger for larger @xmath204 . the width , which remains finite in the thermodynamic limit , signifies that large - scale fluctuations occur frequently in the system . finally we turn to the 2-point spatial correlation functions in the steady state of cd2 and cd3 models . the growing length scale @xmath78 as @xmath206 is limited by the system size @xmath18 . in fig . [ stcd2cd3 ] we show the scaling of data for @xmath38 in the steady state as a function of @xmath45 for an ew surface . both the curves show a cusp at small values of @xmath45 , with cusp exponent @xmath11 . since successive rw returns to the origin are independent events , the calculation in section [ understandfdpo]a below based on independence of intervals , is exact for the cd2 model . [ iia3 ] holds , and we conclude that the correlation function cusp exponent @xmath207 exactly for the cd2 model . this also implies the result that @xmath1 even in the coarsening regime for the cd2 model with ew and kpz surfaces . this is because at any time @xmath68 , regions of a coarsening system which get equilibrated are of length @xmath208 . now the correlation function @xmath209 is obtained by spatial averaging over the system , and hence equivalently averaging over an ensemble of several steady state configurations of subsystem size @xmath210 . thus the exact result for @xmath152 in the steady state carries over to the coarsening regime . we have seen in the previous section that the distribution of like - spin clusters follows a slow power law decay in the cd models . we will demonstrate below that on the basis of this power law , we may understand both ( i ) the cusp in the 2-point function , as well as ( ii ) ordered phases which occupy a finite fraction of system size . we now show analytically within the independent interval approximation ( iia ) @xcite that the cusp exponent @xmath152 and the power law exponent @xmath0 are related . within this scheme , the joint probability of having @xmath21 successive intervals is treated as the product of the distribution of single intervals . in our case , the intervals are successive clusters of particles and holes , which occur with probability @xmath185 . defining the laplace transform @xmath211 , and @xmath212 analogously , we have @xcite @xmath213 where @xmath214 is the mean cluster size . in usual applications of the iia , the interval distribution @xmath185 has a finite first moment @xmath215 independent of @xmath18 . but that is not the case here , as @xmath185 decays as a slow power law @xmath216 for @xmath217 , with the function @xmath218 denoting that the largest possible value of @xmath186 is @xmath18 . this implies that @xmath219 for large enough @xmath18 . considering @xmath79 in the range @xmath220 , we may expand @xmath221 ; then to leading order , the right hand side of eq . ( [ iia1 ] ) becomes @xmath222 , implying @xmath223 . this leads to @xmath224 this has the same scaling form as eq . ( [ scale ] ) . matching the cusp singularity in eqs . ( [ scale ] ) and ( [ iia2 ] ) , we get @xmath225 we recall ( see section [ cdst ] ) that the assumption of independent intervals which underlies the iia in fact holds exactly for the cd2 model , and eq . [ iia3 ] implies that @xmath1 in the steady state and the coarsening regime for the cd2 model . for other models like the cd3 model , or the sliding particle models we will encounter in the subsequent sections , the iia gives insight into the origin of the cusp from the power laws , although it is not exact . we now turn to our claim ( ii ) , that the very same distribution which gives rise to power - law distributed broad boundaries with a collection of small clusters , also gives rise to large clusters of size @xmath226 of ` up ' or ` down ' spins , which form the pure phases . for the cd2 model , we numerically studied the sizes of the largest cluster @xmath227 for systems of different sizes @xmath18 ; we show them in fig . [ largestl ] . the full distribution @xmath228 scales as a function of @xmath229 . the average value is @xmath230 . we also find a similar scaling of the distribution @xmath231 , for the second largest clusters of size @xmath232 , and @xmath233 ( see fig . [ largestl ] ) . some understanding of the fact that the size of largest clusters are of order @xmath18 can be reached by considering the statistics of extreme values . applied to our case , if @xmath234 cluster lengths are drawn at random from a distribution of lengths given by @xmath235 , then the probability distribution @xmath236 that the largest cluster is of length @xmath29 goes as @xmath237 @xcite . the latter distribution peaks at @xmath238 . in the cd2 problem @xmath190 . now , in a system of length @xmath18 we have on an average @xmath239 clusters . if we make the approximate replacement of @xmath234 by this average number @xmath239 , we immediately get @xmath240 . this explains how , although the average cluster sizes are of order @xmath241 , there are always clusters with sizes of order @xmath18 . this is reminiscent of the behaviour of the largest loops in a random walk @xcite . further we found the contribution to magnetization coming from the largest clusters in the system and compared them with the total magnetization of the system , configuration by configuration . in fig . [ largemag ] , we show scatter plots of @xmath242 which is the magnetization obtained from summing the spins of the largest cluster , @xmath243 which is obtained by summing spins of largest and the second largest cluster , and @xmath244 by summing those down to the third largest cluster against the total magnetization @xmath245 . the convergence of the scatter plots towards the @xmath246 line , shows that the few largest clusters give a major contribution to the magnetization of the system . each of these large clusters is a pure phase with magnetization @xmath8 , and thus gives rise to @xmath247 in the curves in fig [ stcd2cd3 ] . in this section we consider the physical problem of hard - core particles sliding locally downwards on the fluctuating surfaces discussed in the previous sections . we find that the downward gravitational force combined with local surface fluctuations lead to large scale clustering of the hard - core particles . the phase - separated state which arises mirrors the hill - valley profile of the underlying surface . for example , the particles on ew and kpz surfaces show fdpo with the cluster distribution , 1-point function , and 2-point function behaving as in their cd model counterparts . on the other hand , particles on the gbdt surface show conventional ordering . let us first define a sliding particle ( sp ) model on a one - dimensional lattice . this is a lattice model whose behaviour resembles that depicted in fig . the particles are represented by @xmath90-valued ising variables @xmath248 on a one - dimensional lattice with periodic boundary conditions , where @xmath249 spins occupy lattice sites . the @xmath250 variables occupy the bond locations and represent the surface degrees of freedom as described in section [ defcd ] for the cd2 model , and their dynamics involves independent evolution via rates @xmath251 and @xmath252 as discussed earlier . for the particles , @xmath253 represents the occupation of site @xmath108 . a particle and a hole on adjacent sites ( @xmath108,@xmath113 ) exchange with rates that depend on the intervening local slope @xmath254 ; thus the moves @xmath255 and @xmath256 occur at rate @xmath257 , while the inverse moves occur with rate @xmath258 . the asymmetry of the rates reflects the fact that it is easier to move downwards along the gravitational field . for most of our studies we consider the strong - field ( @xmath259 ) limit for the particle system . we set @xmath260 . the dynamics conserves @xmath261 and @xmath262 ; we work in the sector where both vanish . this corresponds to a @xmath263 filled system of particles on a surface with zero average tilt . for the ew surface , we took @xmath264 , while for the kpz surface we took @xmath265 and @xmath266 . in section [ robust ] , we discuss departures from these conditions and explore the robustness of fdpo to these changes . for the gbdt surface , the evolution of which was described in section [ surfevol ] , a chosen particle moves to its right or left with equal probability ( @xmath267 ) if there is locally a non - increasing height gradient . thus again the rate of update of the particles is same as that of the surface . the problem can be specified at a coarse - grained mesoscopic level by the continuum equations for the density field @xmath268 corresponding to the discrete variable @xmath269 for the particles . since the particle density is conserved , the starting point is the continuity equation @xmath270 , where @xmath271 is the local current . under the hydrodynamic assumption , the systematic part of the above current is @xmath272 , since for viscous dynamics , the speed is proportional to the local field , in this case the local gradient of height . moreover there is a diffusive part @xmath273 which is driven by local density inhomogeneities , and a noisy part @xmath274 which arises from the stochasticity . the noise @xmath275 is a gaussian white noise . the total density can be written as @xmath276 , where @xmath277 is the average density and @xmath278 is the fluctuating part . this implies finally that the density fluctuation @xmath278 evolves via the following equation : @xmath279 using the well - known mapping in @xmath92 between the density field @xmath278 and the height field @xmath280 of the corresponding interface problem @xcite , one has the relation @xmath281 . this implies from eq . [ particleeq ] that the lowest order term in the evolution equation of @xmath280 is proportional to @xmath282 . this linear first - order gradient term is the result of the gravitational field which acts on the particles . the evolution of the field @xmath67 is given by eq . [ surfaceeq ] . thus a continuum approach to the problem of the sliding particles requires analysis of the semi - autonomous set of nonlinear equations [ surfaceeq ] and [ particleeq ] as one of the fields evolves independently but influences the evolution of the other . the problem belongs to the general class of semiautonomous systems , like the advection of a passive scalar in a fluid system @xcite . the sp model is a special case of the lahiri - ramaswamy ( lr ) model @xcite of driven lattices such as sedimenting colloidal crystals . the general lr model has two - way linear couplings between the @xmath283 and @xmath284 fields , and its phase diagram has recently been discussed in @xcite . the sp model of interest here has autonomous evolution of the @xmath285 , and corresponds to the lr critical line which separates a wave - carrying phase @xcite from a strongly phase separated state @xcite . further , in a model of growing binary films considered in @xcite , in the limit where the height profile evolves independently , the problem gets mapped to noninteracting domain walls ( if annihilation is neglected ) rolling down slopes of independently growing surfaces . the latter problem becomes similar to ours , on thinking of the domain walls as particles . but the fact that they are noninteracting in contrast to the hard - core particles may introduce other physical effects into the problem . we start with a surface in steady state , and allow an initially randomly arranged assembly of sliding particles to evolve on it . in an initial short - time relaxation , particles slide down to the bottom of local minima . after this , the density distribution evolves owing to the rearrangement of the stochastically evolving surface , whose local slopes guide particle motion . we found in numerical simulations that the surface fluctuations actually drive the system towards large scale clustering of particles . this can be seen as follows . after time @xmath68 , the base lengths of coarse - grained valleys of length @xmath286 would have overturned , where @xmath2 is the dynamical exponent of the surface . we thus expect that the latter length scale sets the scale of particle clustering at time @xmath68 . to test this we monitored the equal time correlation function @xmath287 by monte - carlo simulation . we found that it has a scaling form c = f(r / l(t ) ) with l ~t^1/z [ splt ] in accord with the arguments given above . the data for @xmath209 for the particles on ew , kpz and gbdt surfaces are shown to collapse in figs . [ ewsp ] , [ kpzsp ] and [ dtsp ] , respectively . evidently , eq . [ splt ] holds quite well for all three surfaces , despite the widely different values of @xmath2 for the three . the onset of scaling will be discussed further in section [ robust ] , where we discuss the effect of varying the ratio of rates of relative updates of the particles and the surface . to determine the short distance behavior of the decay of @xmath161 as a function of @xmath288 , we evaluated the structure factor @xmath289 for @xmath161 . for any finite @xmath78 , we may write = c_o(r ) + c_s(r / l ) , [ analyticcr ] where @xmath290 is the analytic part which decays over small distances @xmath37 , while @xmath165 is the nonanalytic part which scales as a function of @xmath170 . we are primarily interested in @xmath165 , and so need to subtract the appropriate @xmath291 from @xmath161 . in terms of the scaled variable @xmath292 , @xmath291 contributes only to @xmath293 , in the limit @xmath294 . in that limit we write @xmath295 , and determine @xmath291 by seeing which value gives the longest power - law stretch for @xmath296 , as judged by eye . in fig . [ sksub ] we show @xmath297 for a late time , obtained without any subtraction and after subtraction of @xmath298 with @xmath299 . the power law decay as @xmath300 , stretches over a substantially larger range in the latter case , corresponding to a real space decay with a cusp exponent @xmath152 . a nonzero value of @xmath301 implies that @xmath42 , as @xmath39 is given by @xmath302 . this indicates that the particle - rich phase has some holes and _ vice versa_. in fig . [ sk3 ] we show , @xmath297 corresponding to the three different surfaces at @xmath303 . we find that for the ew surface @xmath11 , for the kpz surface it is @xmath304 , and for the gbdt surface it is @xmath305 . thus there is a deviation from the porod law behavior for the ew and kpz surface fluctuations , and no such deviation for the gbdt surface . in all three cases , we see that the behaviour of the 2-point functions in the particle system resembles the corresponding correlation functions of the cd model for the underlying surface . in the kpz case , the value of the exponent @xmath306 is different from its value @xmath307 in the cd model counterpart . for the ew and gbdt surfaces , the values of @xmath152 are @xmath308 and @xmath309 respectively as in the corresponding cd models . since the sp model corresponding to the gbdt surface does not exhibit anomalous behaviour of the scaled two - point correlation function which is a signature of fdpo , we do not consider it further in our subsequent discussion of the steady state . we first study 1-point functions in order to characterize the steady state . as the system phase separates , a suitable quantity to study is the magnitude of the fourier components of the density profile @xmath310 where @xmath311 and @xmath312 . a signature of an ordered state is that in the thermodynamic limit , the average values @xmath313 go to zero for all @xmath314 , except at @xmath315 . we monitored these averages for the system of sliding particles , with the average @xmath316 performed over the ensemble of steady state configurations . in figs . [ qkewh ] and [ qkkpz ] we show the values of @xmath317 as a function of @xmath314 for various system sizes @xmath18 , for the ew and kpz surfaces respectively . in both the cases , for all @xmath318 the value of @xmath313 falls with increasing @xmath18 indicating that @xmath319 in the thermodynamic limit , for any fixed , finite @xmath314 . but for @xmath320 , we see that the value of @xmath321 approaches a constant . the sharpening of the curves near @xmath315 implies an ordered steady state . the above behavior of @xmath313 as a function of @xmath314 suggests that we take the value @xmath322 , ( corresponding to @xmath33 ) as a measure of the extent of phase separation . we have used @xmath323 as the order parameter earlier also for the cd3 model , and note that it has been used earlier in other studies of phase - separated systems @xcite . here we find that @xmath324 and @xmath325 for particles on the ew and and kpz surfaces respectively . the latter values being nonzero indicates that the steady state is ordered . at the same time , the values being less than @xmath326 indicates that the states deviates substantially from a phase separated state with two completely ordered domains . to have a full characterization of the fluctuations which dominate the ordered state , one should actually evaluate the probability distributions of all the @xmath66 s , e.g. @xmath327 , @xmath328 , @xmath329 , @xmath330 . we show below ( in fig . [ probq1 ] ) one of these distributions , namely that of @xmath327 , for an ew surface . we find that the distribution @xmath203 remain broad ( with root - mean - square deviation being @xmath331 ) even as @xmath24 , indicating again the dominance of large scale fluctuations . it is instructive to monitor the variation of @xmath204 as a function of time @xmath68 , for different system sizes . for an ew surface ( fig . [ q1tl ] ) the value of @xmath204 shows strong excursions about its average value , consistent with the broad distribution shown in fig . [ probq1 ] . the temporal separation period of these fluctuations of the order parameter increases roughly as @xmath332 , but their amplitude is independent of @xmath18 . consequently @xmath203 approaches an @xmath18-independent form as @xmath24 . a temporally oscillatory order parameter also has been found earlier in a model for comparative learning @xcite . however the temporal behaviour in our case is quite different from the almost periodic fluctuation in the latter model , as the fourier spectrum of the time series in @xmath333 in our case follows a broad power - law . we have not pursued a detailed study of the temporal behaviour any further . the fluctuation of @xmath204 in fig . [ q1tl ] , gives rise to an interesting question : does the system become disordered and lose the phase ordering property when the value @xmath204 falls to low values ? the answer is no , as is very clearly brought out in fig . [ q123 t ] in which @xmath334 , @xmath335 , @xmath336 have been plotted simultaneously as a function of time @xmath68 for a single evolution of the system . we observe that a dip in @xmath204 is accompanied by a simultaneous rise in the value of either @xmath335 or @xmath336 . this implies that whenever the system loses a single large cluster ( making @xmath204 small ) either two or three such clusters appear in its place ( making the values of @xmath335 and @xmath336 go up ) . thus the system remains far from the disordered state , and always has a few large particle clusters which are of macroscopic size @xmath226 . a numerical study showed that the average size of the largest particle cluster @xmath337 . we have seen above that in the sp models , the order parameter has a broad distribution just as in their cd model counterparts . we observe further that the particle and hole cluster size distributions in the steady state of the sp model decay as a power - law : @xmath338 . in fig . [ plcrspew ] for the ew surface , we find that the particle ( denoted by symbols ) and hole ( denoted by lines ) distributions coincide , with @xmath339 . by contrast , fig . [ plcrspkpz ] for the kpz surface shows that the particle and hole distributions are not identical . this is because with asymmetric rates ( @xmath103 ) , the surface has an overall motion in one direction , such that the downward motion of the particles and the upward motion of the holes , due to gravity , are no longer symmetrical . we checked that the distributions for particles and holes get interchanged if the rates @xmath251 and @xmath252 are interchanged . the exponent for the decay of both the particle and hole distributions is @xmath340 . finally we note that the 2-point correlation functions in the steady state of the sp model exhibit a scaling form in @xmath45 and have the same cusp exponents as in the coarsening regime ( with @xmath341 being replaced by @xmath18 ) . for the ew surface , the scaling curve shown in the inset of fig . [ plcrspew ] exhibits a cusp with @xmath11 . the corresponding curve for the kpz surface , shown in the inset of fig . [ plcrspkpz ] also exhibits a cusp , with @xmath342 . the fact that @xmath343 in these curves , as for those in the coarsening regime of the sp model , is indicative of the fact that the pure phase which are particle rich also have holes in them . in this respect , the pure phases differ from their cd model counterparts . we have seen above that the fdpo of the sliding particles in the sp model is qualitatively of the same type as in the cd models for the underlying surfaces . we measured the average overlap @xmath344 to get a quantitative estimate of the extent of correlation between the sliding particles ( holes ) and the valleys ( hills ) of the underlying surface . we found that it is nonzero as we expected , e.g. for the ew surface @xmath345 and @xmath346 corresponding to @xmath114 being defined within cd2 and cd3 models . the overlap is greater in case of cd3 model , since the domains are most often smaller than @xmath197 and this matches with the fact that particles clusters are also of size @xmath347 . on the other hand , domains in the cd2 model can be almost as big as @xmath18 . for the kpz surface , @xmath345 corresponding to the overlap between particles and the coarse grained depth variables @xmath64 s of the cd3 model . we did several numerical tests to check the robustness of the fluctuation dominated ordered state for the sliding particle ( sp ) problem . \(i ) we explored the effect of varying the ratio @xmath348 , the relative rate at which the particles get updated as compared to the surface . \(ii ) we allowed the possibility of a small but finite rate ( @xmath349 ) of the particles to hop uphill on a local @xmath97 slope . \(iii ) we made the overall slope nonzero , in the case of the kpz surface . we found that fdpo stays with ( i ) and ( ii ) , while it is lost with ( iii ) . for the ew surface , with @xmath350 ( i.e. the surface moving @xmath351 times slower than the sliding particles ) , we found that @xmath31 remains close to but slightly larger than @xmath352 , the value for @xmath353 . we checked the correlation function @xmath209 in the coarsening regime , and found that it has a cusp as a function of @xmath170 with the exponent @xmath11 . for @xmath354 ( i.e. the surface moving @xmath351 times faster ) , we found @xmath355 . the latter value indicates a lesser degree of ordering and this is also mirrored in the 2-point function @xmath209 : the collapse of the data as a function of @xmath170 occurs beyond a time which is greater than that for @xmath356 , i.e. the scaling regime sets in much later . nevertheless , at large enough times , the cusp exponent is unchanged ( @xmath357 ) . figure [ skrates ] ( lower curves ) shows the log - log plot of @xmath296 versus @xmath358 for the three rates @xmath354 , @xmath8 and @xmath359 . all of them have slopes @xmath360 , which indicate @xmath11 . a similar evaluation of @xmath161 was also done for the kpz surface , and is also shown in fig . [ skrates ] ( upper curves ) . the observed slope of @xmath361 implies that the cusp exponent remains @xmath306 for all of them . we conclude that the variation of update rates affects the degree of ordering but not the asymptotic scaling properties , as indicated by the constancy of the cusp exponent @xmath152 . so far we have considered uphill hopping rate to be strictly zero , i.e. @xmath362 . by allowing for @xmath349 , i.e. allowing for upward motion of particles , we saw that the fdpo persists , so long as @xmath363 . in fig . [ sk_upneq0 ] , we show @xmath364 as a function of @xmath365 , at a large time @xmath68 for ew and kpz surfaces respectively , when the ratio @xmath366 . we find that the slopes are @xmath360 and @xmath361 in the two cases indicating that the values of the cusp exponents are still @xmath367 and @xmath306 respectively , for the two surfaces . this points to the universality of the value @xmath367 ( ew ) and @xmath306 ( kpz ) over a range of models with different values of @xmath368 , and also with respect to varying @xmath369 . we also investigated the effect of having an overall tilt of the kpz surface . this leads to an overall movement of the transverse surface fluctuations , which are the analogues of kinematic waves in particle systems @xcite . in the presence of such a wave , the profile of hills and valleys of the surface sweep across the system at finite speed , and the particles do not get enough time to cluster . consequently the phenomenon of fdpo is completely destroyed . in fig . [ kinekpzspcr ] we show @xmath370 as a function of @xmath37 ( there is no scaling by @xmath78 ) for several @xmath68 . the curves are independent of @xmath68 , in the absence of coarsening towards a phase ordered state . in this paper we have discussed the possibility of phase ordering of a sort which is dominated by strong fluctuations . in steady state , these fluctuations lead to variations of the order parameter of order unity , but the system stays ordered in the sense that with probability one , a finite fraction of the system is occupied by a single phase . the value of this fraction fluctuates in time , leading to a broad probability distribution of the order parameter . we demonstrated these features in two types of models having to do with surface fluctuations the first , a coarse - grained depth ( cd ) model where we could establish these properties analytically , and the second a model of sliding particles ( sp ) on the surface in question . for these models we found that besides ( a ) the broad probability distribution of the order parameter ( which we may take to be the defining characteristic of fdpo ) , the steady state was also characterised by ( b ) power laws of cluster size distributions and ( c ) cusps in the scaled two - point correlation function , associated with the breakdown of the porod law . the connection between ( b ) and ( c ) was elucidated using the independent interval approximation . further , an extremal statistics argument showed that the largest cluster drawn from the power law distribution is of the order of the system size ; this implies a macroscopic ordered region , so that within our models , properties ( a ) and ( b ) are connected . there are several open questions . does fluctuation - dominated phase ordering occur in other , completely different types of systems as well ? are properties ( b ) and ( c ) necessarily concomitant with the defining property ( a ) of fdpo ? can one characterize quantitatively the dynamical behaviour in the fdpo steady state ? our model of particles sliding on a fluctuating surface relates to several physical systems of interest . first , it describes a new mechanism of large scale clustering in vibrated granular media , provided the vibrations are random both in space and time . second , it describes a special case ( the passive scalar limit ) of a crystal driven through a dissipative medium , for instance a sedimenting colloidal crystal @xcite . finally , related models describe the formation of domains in growing binary films @xcite . it would be interesting to see if ideas related to fdpo play a role in any of these systems . it would also be interesting to examine fluctuating phase - ordered states in other nonequilibrium systems from the point of view of fdpo . for instance , in a study of jamming in the bus - route model studied in @xcite , the largest empty stretch in front of a bus was found to be of order @xmath18 , and it is argued that such a stretch survives for a time which is proportional to @xmath371 for a nonvanishing rate of arrival of the passengers . these features are reminescent of the behaviour of the cd and sp models derived from the edwards - wilkinson model discussed above . however , more work is required to make a clear statement about fdpo in the bus - route model . in general , fluctuation - dominated phase ordering is evidently a possibililty that should be kept in mind when discussing new situations involving phase ordering in nonequilibrium systems , both in theory and in experiment .

in a dirty metal , electron - electron interactions in the spin - triplet channel lead to singular corrections to a variety of physical quantities . we show that these singularities herald the emergence of ferromagnetism . we calculate the effective action for the magnetic moment of weakly - interacting electrons in a dirty metal and show that a state with finite ferromagnetic moment minimizes this effective action . the saddle - point approximation is exact in an appropriate large- limit . we discuss the physics of the ferromagnetic state with particular regard to thermal fluctuations and localization effects .

cond-mat0302503

the presence in a solid of randomly - distributed imperfections causes its electrons to propagate diffusively rather than ballistically . as a result , when two electrons interact , they tarry in each other s vicinity , and become more strongly correlated @xcite . as shown by finkelstein @xcite and castellani , _ et al . _ @xcite , this effect leads to an instability of the diffusive fermi liquid fixed point ( see also refs . ) which manifests itself through divergent corrections to a variety of physical quantities . the instability exists for arbitrarily weak coupling and for arbitrarily weak disorder . its simplest incarnation is in an electron system in a weak magnetic field @xmath1 which suppresses quantum interference effects and relegates them to subleading order in the dimensionless resistance@xcite @xmath2 ( measured in units of @xmath3 ) while interactions make a contribution at leading order . since the relevant interaction is in the spin - triplet particle - hole channel , the @xmath4 factor must be tuned to zero if @xmath5 so that there is no zeeman splitting , which would cut off the divergences . though this instability was identitified nearly 20 years ago , there is , at present , no consensus on its outcome . the recent discovery of evidence for a metal - insulator transition in two - dimensional electron gases ( 2degs ) in si - mosfets @xcite , gaas @xcite heterostructures , and si - ge @xcite has led to a re - examination of the problem of interacting electrons in a random potential . the interesting phenomena seen in these experiments occur in a regime in which interactions are strong ( large @xmath6 ) and quantum interference effects are likely to be important because the conductivity is on the order of the quantum limit @xmath7 and @xmath8 . though a number of interesting theories @xcite have been proposed , none have provided a complete explanation of the experimental data , which is nt surprising , given that the presumably simpler problem of weakly - interacting electrons in a weak random potential with @xmath5 , @xmath9 remains unsolved . it is this simpler problem which is the subject of this paper though it is unresolved , the instability caused by the triplet interaction is , in fact , a crime with only one plausible suspect : ferromagnetism . since the diffusive fermi liquid is unstable , the system must settle into some other phase at low temperature ; since the instability is in the spin - triplet channel , it is presumably one in which the spins are ordered . the basic interaction in this problem is exchange , which is ferromagnetic . there is no available mechanism which generates antiferromagnetic interactions . to see this , note that the instability occurs when the conductance is large , while superexchange occurs in or near an insulating state ( since it arises when the kinetic energy is treated as a perturbation ) there would be rkky interactions between localized spins , such as magnetic impurities , if we had any , but we do nt . in the absence of any antiferromagnetic interactions , we rule out spin density - waves ( sdws)@xcite , spin glass order , and random singlet phases@xcite . the latter two furthermore require large effective disorder while the instability of interest here is manifest for small effective disorder@xcite . it has been suggested that the instability of the triplet interaction signals the formation of ` local moments'@xcite . in the weak - interaction , weak - disorder limit , however , the instability of the diffusive fermi liquid only develops at large length scales , so the resulting magnetic moments are anything but ` local ' . local moments might be expected when the energy cost for flipping a spin at short length scales is very small . this might occur near the stoner transition of the clean system @xcite , where the critical susceptibility , @xmath10 , is of curie form for a mean - field exponent @xmath11 . there is also a substantial local moment regime with curie susceptibility in the large-@xmath12 hubbard model ( because of the weakness of superexchange for large @xmath12 ) and doped semiconductors ( because of competing ferromagnetic and antiferromagnetic interactions ) . however , these scenarios are not relevant to the presumably simpler , though still non - trivial , problem of the weakly - interacting , weakly - disordered limit . this leaves us with one remaining possibility and , indeed , a number of authors have proposed that the singularity of the triplet interaction channel leads to ferromagnetism . the basic underlying physics is simply that diffusing electrons interact more strongly because they move more slowly and , as a result , the interaction energy gained by ( at least partially ) polarizing the system _ always _ outweighs the single - particle energy cost , in contrast to a clean system , where the interaction energy gain is a constant @xmath12 which must exceed the single - particle energy cost @xmath13 , i.e. the stoner criterion ( @xmath14 is the density of states at the fermi energy ) . andreev and kamenev @xcite proposed such a mechanism for finite - size systems by computing the disorder enhancement of the interaction matrix element for a state with two electrons in single - particle states separated by small energies . heuristically , one can say that the two wavefunctions are concentrated in the same minima of the potential . chamon and mucciolo @xcite found a ferromagnetic saddle - point solution of the one - loop effective action of finkelstein s theory . their calculation generalizes stoner s by including quantum corrections resulting from interactions between diffusion modes . belitz and kirkpatrick @xcite advanced the same hypothesis when they noted that the phase transition which they expected to be driven by the growth of the triplet interaction is of the same form as a transition to ferromagnetism in a disordered system . in this paper , we calculate the effective action for the magetization in a weakly - interacting electron system in a slightly dirty metal . the effective action is calculated to lowest order in the dimensionless resistance @xmath2 . it exhibits the physics described above and is clearly unstable to the formation of a ferromagnetic moment . in principle , the emergence of some other type of order could intervene ( this effect would have to be non - perturbative in @xmath2 ) before ferromagnetism develops , but the arguments given above rule out this possibility . thus , we conclude that in the limit of small @xmath2 , the singularity of the spin - triplet interaction channel leads to ferromagnetism . our results substantiate the conclusions of previous authors @xcite . the basic philosophy is simple : since the rg equations tell us that the interaction in the triplet channel grows in importance as the energy scale is lowered , we try to satisfy this term in the action first by allowing the magnetization , @xmath15 , to develop a non - zero expectation value . in order to do this , we compute an effective action @xmath16 $ ] for the magnetization @xmath15 taking finkelstein s ` non - linear @xmath17-model ' for interacting electrons in a disordered system as the point of departure . since our goal is to calculate an effective action @xmath16 $ ] to lowest order in the resistance @xmath2 , finkelstein s @xmath17-model is merely a convenient way of organizing the pertinent diagrams . we believe that it is intuitively appealing to follow finkelstein and work directly with a theory which retains only diffusion modes and drops the superfluous high - energy electronic degrees of freedom , but this is a matter of convenience , and it is a option which we can take regardless of whether new terms are generated at the two - loop or higher level ( i.e. regardless of whether the non - linear terms in the theory are really related by symmetry , as they would be in a @xmath17-model ) . our derivation emphasizes the similarity between , on the one hand , the rg equation for the triplet - channel interaction in a dirty metal and the consequent development of ferromagnetism with , on the other hand , the rg equation for the cooper - channel interaction in a clean fermi liquid and the consequent development of superconductivity . we believe that this illuminates the inevitability of our conclusion . once ferromagnetic order has developed , the system becomes essentially one of spinless electrons whose diffusion is decoupled from the spin waves of the system . in this regime , the suppression of the resistivity which accompanies the growth of the triplet interaction parameter thus , our results indicate that a metallic state , if it exists in two dimensions , does not owe its existence to the growth of the triplet interaction parameter . in two dimensions , our results should be interpreted as follows . as we lower the temperature , the system develops local ferromagnetic order as we cross the mean - field transition temperature @xmath18 . the spin density is controlled by classical thermal fluctuations which prevent order and lead to a correlation length @xmath19 at low - temperatures . in the regime @xmath20 , we have a system in which charge diffuses while the spins are gapped . the localization physics of this system is essentially the same as that of spinless electrons , so the system leaves the diffusive regime and eventually becomes localized . thus , our small @xmath2 analysis can not be used all the way to zero temperature . the trend is towards a ferromagnetic insulator , but the onset of insulating behavior can thwart the development of ferromagnetism . recent experiments @xcite performed near the metal - insulator transition indicate an enhancement of the spin susceptibility which is cut off by the onset of insulating behavior . the metal - insulator transition is outside of the regime of validity of our calculation , but this is , nevertheless , qualitatively consistent with our picture . in @xmath21 dimensions , when the interaction strength is larger than a threshhold value of @xmath22 , there is a ferromagnetic transition at a finite temperature @xmath23 . if the resistance is smaller than a critical value @xmath24 , then the resistance will remain finite down to zero temperature , where the system is in a metallic ferromagnetic state . we begin with a system of interacting electrons moving in a quenched random potential @xmath25 in @xmath26 . the imaginary - time action is @xmath27 where @xmath28 and @xmath29\end{gathered}\ ] ] for later convenience , we have written the second term in momentum space and have split the interaction into nearly forward scattering , in which electrons with momenta and spins @xmath30 , @xmath31 are scattered to @xmath32 , @xmath33 , where @xmath34 , @xmath35 and exchange scattering , in which @xmath36 , @xmath37 . if the microscopic electron - electron interaction is @xmath38 , then the associated bare couping constants are @xmath39 , @xmath40 . as finkelstein noted , these are subject to finite fermi liquid renormalization in the ballistic regime , even before they begin to flow in the diffusive regime . we assume that the system is in a weak magnetic field , so that we can ignore cooper scattering , for which @xmath41 , @xmath42 . in order to perform the average over the quenched random potential , we replicate the theory and then average the replicated functional integral over the probability distribution @xmath43 . now , we have @xmath44\\ t{\sum_{n , m , l;a}}\int \biggl[{\gamma_1}\ , { \psi_{n,\alpha}^{a\,\dagger}}({\bf k}+{\bf q } ) { \psi^a_{m,\alpha}}({\bf k } ) { \psi_{l,\beta}^{a\,\dagger}}({\bf p}-{\bf q } ) { \psi^a_{l+m - n,\beta}}({\bf p } ) + { \gamma_2}\ , { \psi_{n,\alpha}^{a\,\dagger}}({\bf k}+{\bf q } ) { \psi^a_{m,\alpha}}({\bf p } ) { \psi_{l,\beta}^{a\,\dagger}}({\bf p}-{\bf q } ) { \psi^a_{l+m - n,\beta}}({\bf k } ) \biggr]\end{gathered}\ ] ] where @xmath45 is a replica index ; eventually , we must take the replica limit , @xmath46 . @xmath14 is the single - particle density - of - states at the fermi energy . the resulting effective action now contains four quartic terms : one disorder term and two interaction terms . note that the disorder term is non - local in time since it only depends on two matsubara frequencies , and it couples different replicas . the interaction terms depend on three matsubara frequencies , so they are local in time , and they do not couple different replicas . we decouple these quartic terms with three hubbard - stratonovich fields , @xmath47 , @xmath48 , @xmath49 . it will be useful to think of @xmath50 both as matrices with indices @xmath51 and also as functions of a single frequency @xmath52 ; they have only a single replica index , however , because interactions do not mix different replicas . note that @xmath53 and @xmath54 are essentially the spin and charge density . integrating out the fermions , we can write the partition function as @xmath55}}\ ] ] where @xmath56 $ ] is given by @xmath57 = \sum\int\biggl[-\text{tr}\,\ln\left(i{\epsilon_n}+ \frac{1}{2m}{\nabla^2}+\mu + \frac{i}{2\tau}q + i{\gamma_1 } \ , y + { \gamma_2 } \ , x \right ) + \frac{\pi n_f}{4\tau}\text{tr}\left({q^2}\right)\\ + \ , \frac{1}{2}\,{\gamma_1}\,\text{tr}\left({y^2}\right)\ , + \ , \frac{1}{2}\,{\gamma_2}\,\text{tr}\left({x^2}\right ) \biggr ] \label{eq : h - s - action}\end{gathered}\ ] ] in this expression , we use ` tr ' to mean the trace over all of the indices which are not explicitly written . matsubara indices are treated as ordinary matrix indices except that their summations come with factors of @xmath58 . the saddle - point equations are @xmath59 @xmath60 @xmath61 the momentum integrals on the right - hand - sides of ( [ eq : s - p - eqn - x ] ) and ( [ eq : s - p - eqn - y ] ) are simply @xmath62 , according to ( [ eq : s - p - eqn - q ] ) . hence , they have the solution @xmath63 note that the right - hand sides of the second and third equations of ( [ eq : s - p - eqn - x ] ) and ( [ eq : s - p - eqn - y ] ) vanish for @xmath64 because the saddle - point @xmath65 is diagonal in matusubara frequency while they vanish for @xmath66 because of cancellation between positive and negative frequencies . as we will see later , for @xmath67 sufficiently large , @xmath68 becomes unstable and there is an @xmath69 solution of ( [ eq : s - p - eqn - x ] ) . this is the stoner instability . upon including low - energy fluctuations of @xmath65 diffusion modes an @xmath69 solution arises even for @xmath67 small , as we will see shortly . the low - energy fluctuations about this saddle - point are transverse fluctuations of @xmath65 , which can be parametrized by @xmath70 : @xmath71 the diagonal blocks of @xmath72 correspond to matsubara indices @xmath73 and @xmath74 while the upper right block corresponds to @xmath75 ; the lower left block , to @xmath76 . these fluctuations correspond to diffusion of charge and spin . the longitudinal fluctuations of @xmath65 as well as fluctuations of @xmath15 , @xmath54 are gapped at the classical level ( i.e. tree - level ) . in order to derive an effective field theory for these diffusion modes , we shift @xmath77 in ( [ eq : h - s - action ] ) @xmath57 = \sum\int\biggl[-\text{tr}\,\ln\left(i{\epsilon_n}+ \frac{1}{2m}{\nabla^2}+\mu + \frac{i}{2\tau}q\right ) + \frac{\pi n_f}{4\tau}\text{tr}\left({q^2}\right ) - \pi{n_f}{\gamma_1}\text{tr}\left(qy\right ) + i\pi{n_f } { \gamma_2}\text{tr}\left(qx\right)\\ + \,\frac{1}{2}\left(1 + 2\pi\left(\tau t{\sum_m}\right){n_f}{\gamma_1}\right ) \,{\gamma_1}\,\text{tr}\left({y^2}\right ) \,+\ , \frac{1}{2}\left(1 - 2\pi\left(\tau t{\sum_m}\right){n_f}{\gamma_2}\right ) \,{\gamma_2}\,\text{tr}\left({x^2}\right ) \biggr ] \label{eq : h - s - action2}\end{gathered}\ ] ] formally , the factors of @xmath78 in the final two terms are infinite as a result of the unrestricted matsubara sum . however , upon integrating out the massive longitudinal modes of @xmath65 , we see that they are the first terms in the series @xmath79 , which is simply @xmath80 . thus , we replace the factors of @xmath78 by @xmath80 in ( [ eq : h - s - action2 ] ) . alternatively , we could expand the @xmath81 about the diffusive saddle - point and keep only terms up to second order in @xmath15 and @xmath54 . we would then obtain the same expressions , namely ( [ eq : h - s - action2 ] ) with the factors of @xmath78 replaced by @xmath80 . the coefficient of @xmath82 is now @xmath83 , which becomes negative for @xmath84 : the stoner instability . finkelstein actually expanded the @xmath81 only to linear order in @xmath15 and @xmath54 . for small @xmath85 , @xmath86 , there is no difference , but the stoner instabilty is missed . we now expand the @xmath81 about the diffusive saddle - point . thus , we require that @xmath65 is a constrained field of the form given in ( [ eq : q - param ] ) . @xmath87 = \pi{n_f}\int d^d x\ ; \biggl\{d\,\text{tr}{\left(\nabla q\right)^2 } - 4iz \text{tr}\left(\hat{\epsilon}q\right)\\ - { \gamma_1}\text{tr}\left(qy\right ) + i { \gamma_2}\text{tr}\left(qx\right)\\ + \,\frac{1}{2}\left(1+{n_f}{\gamma_1}\right ) \,\frac{\gamma_1}{\pi{n_f}}\,\text{tr}\left({y^2}\right)\\ + \ , \frac{1}{2}\left(1-{n_f}{\gamma_2}\right ) \,\frac{\gamma_2}{\pi{n_f}}\,\text{tr}\left({x^2}\right ) \biggr\}\end{gathered}\ ] ] the diffusion constant @xmath88 is given by @xmath89 . following finkelstein , we integrate out @xmath50 and obtain an effective action for the diffusion modes . @xmath90 = \pi{n_f}\int d^d x\ ; \biggl\{d\,\text{tr}{\left(\nabla q\right)^2 } - 4i \text{tr}\left(\hat{\epsilon}q\right)\\ -\pi{n_f}{\tilde \gamma_1 } q^{aa}_{{n_1}{n_2},\alpha\alpha } q^{aa}_{{n_3}{n_4},\beta\beta } \delta_{n_1-n_2+n_3-n_4}\\ + \pi{n_f}{\tilde \gamma_2 } q^{aa}_{{n_1}{n_2}\alpha\beta } q^{aa}_{{n_3}{n_4}\beta\alpha } \delta_{n_1+n_2-n_3-n_4 } \biggl\}\end{gathered}\ ] ] note that ordinary matrix multiplication rules reflect the non - locality in time of the first term ( the ` disorder term ' ) in this action . the @xmath91 interaction terms do not involve matrix multiplication and are , consequently , local in time . in this expression , @xmath92 . these corrections to @xmath93 follow from our retention of the @xmath82 and @xmath94 terms which finkelstein drops . for @xmath93 small , @xmath95 , and finkelstein s effective action is recovered . initially , the coefficient of the @xmath96 term is @xmath97 , as in ( [ eq : almost - sigma - model ] ) , but quantum corrections cause it to flow , so we have followed finkelstein in introducing the coupling @xmath98 . the one - loop rg equations associated with this action are @xmath99 where @xmath2 is the dimensionless resistance , @xmath100 ; @xmath101 ; @xmath102 ; and we have set @xmath103 in order to avoid clutter . for sufficiently large interaction strength , the correct starting point is presumably a wigner crystal , which corresponds to a radically different saddle - point of the effective action ( [ eq : h - s - action ] ) . thus , even though these rg equations were computed to all orders in @xmath104 for @xmath2 small , they should be understood as holding only for the vicinity of the diffusive saddle - point . the rg equations for @xmath98 , @xmath105 , @xmath67 can be combined to give : @xmath106 hence , @xmath107 is a constant . for long - ranged coulomb interactions , the ward identity implies that this constant is zero : @xmath108 . more generally , we can introduce @xmath109 , @xmath110 , so that @xmath111 . since @xmath98 grows under the rg flow ( [ eq : rgeqns ] ) , the final term can be dropped and we have @xmath112 . in the case of spinless electrons , @xmath98 does not grow under the corresponding rg flow , so we can not make this simplification ; we will later discuss the similar case of spin - polarized electrons . making the above substitution , we find that the rg equations for @xmath113 and @xmath98 are : @xmath114 thus , @xmath113 and @xmath98 grow as the scale is increased . eventually , this growth causes these equations to break down . if we ignore the flow of @xmath2 , then the breakdown scale is @xmath115 ; in fact , the scale is somewhat lower because @xmath2 flows . as noted in ref . , the breakdown of these equations is reminiscent of the situation in a clean fermi liquid , where the cooper channel interaction flows according to @xmath116 . the breakdown of the latter at a scale @xmath117 signals the onset of superconductivity or , at least , pairing . unlike in the case of a clean fermi liquid with a four - fermion cooper channel interaction , @xmath98 diverges as well in our problem . however , the divergence of @xmath98 is a by - product of the divergence of @xmath113 ; in the cooper pairing problem , if the interaction which drives pairing could also contribute to mass renormalization , we would have the same situation as in the finkelstein problem . indeed , this is what happens when pairing is driven by a gauge field . the development of pairing not only cuts off the flow of the interaction , but also the flow of the effective mass @xcite . similarly , we expect that the physics which resolves the divergence of @xmath113 will also resolve the divergence of @xmath98 . to see how the breakdown of ( [ eq : rgeqns ] ) leads to ferromagnetism , let us go back to the action ( [ eq : almost - sigma - model ] ) . instead of integrating out @xmath15 and @xmath54 at the paramagnetic diffusive saddle point , let s integrate out @xmath118 and @xmath54 in order to obtain an effective action @xmath16 $ ] for the magnetization @xmath15 . the spin polarization @xmath15 acts as a zeeman field . a constant field aligned along the @xmath119 direction , @xmath120 ( taken to be replica - diagonal ) , gives a gap to @xmath121 and @xmath122 but leaves @xmath123 and @xmath124 unaffected . to see this , note that a constant field couples only to the diagonal matusbara indices @xmath125 . expanding @xmath126 and @xmath127 and retaining only the leading terms , which are quadratic , we find that @xmath128\ ] ] an imaginary gap @xmath129 in matsubara formalism corresponds to a real gap in real - time . @xmath121 and @xmath122 have propagators : @xmath130 when we integrate out @xmath121 and @xmath122 , infrared divergences will be cutoff by @xmath15 , as the single - particle gap cuts off divergences in bcs theory . since @xmath123 and @xmath124 are unaffected by @xmath15 to this order , they simply contribute an unimportant constant to @xmath16 $ ] . we will integrate out @xmath118 to one loop to obtain @xmath16 $ ] . this is not quite the same as a calculation to lowest order in @xmath2 , but the basic physics is illustrated most simply in a one - loop calculation , so we will present it first . to maintain further simplicity , let us consider the case @xmath131 ; the general case is a straightforward but more complicated extension , to which we return later . in the next subsection , we will present a more careful calculation which includes the effects of @xmath105 and includes all of the terms which contribute to lowest order in @xmath2 . at one loop , we only need to retain terms in ( [ eq : almost - sigma - model ] ) up to quadratic order in @xmath118 @xmath132 = \pi{n_f}\sum \int \biggl\ { d\ , { \rm tr}\left(\nabla{v_+^\dagger } \ , \nabla { v^{}_+}\right ) + z |{\omega_{nm}}| { v_{nm,+}^\dagger } v_{mn,+}^ { } + \frac{1}{2 } \left(1-{n_f}{\gamma_2}\right ) \frac{\gamma_2}{\pi{n_f}}\,{\rm tr}\left({x^2}\right ) \\ + i{\gamma_2}\left [ { \rm tr}\left(v^{}_{nm,+ } x(n - m)\right ) + { \rm c.c.}\right ] + i { \gamma_2 } \left [ { \rm tr } \left(v^{}_{ml,+ } v^\dagger_{ln,+ } x(n - m)\right ) + { \rm c.c.}\right ] + \left({v_+}\rightarrow{v_-}\ , , \ : x\rightarrow -x\right ) \biggr\}\end{gathered}\ ] ] here , we have written @xmath133 , @xmath134 , @xmath135 , and @xmath136 . we now shift @xmath15 in order to remove the bi - linear coupling between @xmath15 and @xmath118 . this is the simplest way to proceed with a one - loop computation of @xmath16 $ ] . to go beyond one - loop , it is more convenient to shift @xmath118 , as we will see later . performing the shift @xmath137 for @xmath138 and @xmath139 for @xmath140 , where @xmath141 , we have : @xmath142 = \pi{n_f}\sum \int \biggl\ { d\ , { \rm tr}\left(\nabla{v^\dagger } \ , \nabla v\right ) + z |{\omega_{nm}}| { v_{nm;\,a}^\dagger } v^{}_{mn;\,a } - { \gamma_2 } v^\dagger_{nm;\,a } v^{}_{m'n';\,a}\,\delta_{n - m , n'-m'}\\ + i { \gamma_2 } \left [ { \rm tr } \left(v^{}_{ml;\,a } v^\dagger_{ln;\,b } { \sigma^z_{ab}}\,x(n - m)\right ) + { \rm c.c.}\right ] + \frac{1}{2 } \left(1-{n_f}{\gamma_2}\right)\ , \frac{\gamma_2}{\pi{n_f}}\,{\rm tr}\left({x^2}\right ) \biggr\ } \label{eq : imp - terms}\end{gathered}\ ] ] a @xmath143 term is also generated as a result of this shift , but it does nt contribute to @xmath16 $ ] at one loop , so we drop it . the @xmath118 propagator which follows from ( [ eq : imp - terms ] ) is now @xmath144 where @xmath145 and @xmath146 is as defined after ( [ eq : rgeqns ] ) . -treated as a classical background field can be viewed as a sum over one - loop diagrams with @xmath15 insertions ( the thick lines on the right).,width=312 ] integrating out @xmath118 , we obtain all one - loop diagrams of ( [ eq : imp - terms ] ) , with @xmath15 treated as a classical background field or , simply , a @xmath147 expression containing the @xmath118 propagators in the @xmath15 background field . later , we will return to the possibility of other saddle - points , but for now , let us assume , in the name of simplicity , that @xmath15 is static , translationally - invariant , and replica - symmetric . for such @xmath15 , the effective action takes the form : @xmath148 = { \sum}\int \frac{1}{2}\left(1-{n_f}{\gamma_2}\right)\ , { \gamma_2}\ , { \rm tr}\left({x^2}\right)\\ + { \rm tr}\ln\biggl(\left[z|\omega_{nm}|+d{q^2}\pm i { \gamma_2}{x}\right ] \delta_{nn'}\,\delta_{mm'}\,\delta_{aa'}\,\delta_{bb'}\\ - \ : { \gamma_2}\ : \delta_{n - m - n'+m'}\ , \delta_{aa'}\,\delta_{bb'}\,\delta_{ab}\biggr)\end{gathered}\ ] ] this may be rewritten in terms of the propagators @xmath149 as @xmath150 = { \sum}\int \frac{1}{2}\left(1-{n_f}{\gamma_2}\right)\ , { \gamma_2}\ , { \rm tr}\left({x^2}\right ) + { \rm tr}\ln\biggl [ \left({d^{x+}_0}\:-\ : { \gamma_2}\,{d^{x+}_0}\,{d^{x+}_2}\,\delta_{ab}\right ) \left({d^{x-}_0}\:-\ : { \gamma_2}\,{d^{x-}_0}\,{d^{x-}_2}\,\delta_{ab}\right ) \biggr]\end{gathered}\ ] ] when we take the product of the two factors on the right - hand side of ( [ eq : s - eff - x-2 ] ) , the @xmath151 term has a free summation over the replica index @xmath152 which vanishes in the replica limit @xmath46 for replica - symmetric solutions @xmath153 . dropping this term and using the relation @xmath154 , we have @xmath155 = { \sum}\int \biggl\ { \frac{1}{2}\left(1-{n_f}{\gamma_2}\right)\ , { \gamma_2}\ , { \rm tr}\left({x^2}\right)\\ - { \rm tr}\ln\left({d^{x+}_2}{d^{x-}_2 } - { d^{x+}_0}{d^{x-}_0 } \right ) \biggr\}\end{gathered}\ ] ] evaluating the integral , we find @xmath156 = \beta { l^2}\biggl [ \frac{1}{2}\left(1-{n_f}{\gamma_2}\right)\ , { \gamma_2}\ , { x^2}\,-\\ t\,\pi{n_f}{\gamma_2 ^ 2}\ , \left(\frac{1}{z } - \frac{1}{z_2}\right)\,{x^2 } \left(\ln\left(\frac{\lambda}{{\gamma_2}x}\right)-1\right ) \biggr]\end{gathered}\ ] ] where @xmath157 is the area of the system . as a result of the logarithm on the second line , the minimum of this effective action is at non - zero @xmath15 for arbitrarily small @xmath86 . we will discuss this non - trivial solution and the saddle - point equation ( or gap equation ) which it satisfies in more detail in the next section . at this stage , the important lesson is that , at the one - loop level , the weakly - interacting diffusive electron gas is always unstable to the development of non - zero @xmath15 , i.e. ferromagnetism . let us examine more closely the saddle - point equation for @xmath15 in the simplest setting , that of section [ sec : one - loop - s ] . if we vary equation ( [ eq : s - eff - x-2 ] ) with respect to @xmath15 ( before performing the momentum integral and matsubara sum ) , then we obtain the saddle - point equation : @xmath158\cr & = & \pi{n_f}{\gamma_2 } \sum\int\biggl [ \frac{1}{{\left[{z_2}|\omega_{nm}|+d{q^2}\right]^2 } + { \gamma_2 ^ 2}{x^2 } } \:-\ : \frac{1}{{\left[z|\omega_{nm}|+d{q^2}\right]^2 } + { \gamma_2 ^ 2}{x^2 } } \biggr]\cr & = & t \,\pi{n_f}{\gamma_2}\ , \left(\frac{1}{z } - \frac{1}{z_2}\right ) \ln\left(\frac{\lambda}{{\gamma_2}x}\right)\end{aligned}\ ] ] the integrals on the right - hand - side of the second line are , for @xmath68 , precisely the integrals which determine the contributions of diagrams ( d ) and ( e ) to the rg equation for @xmath86 . we will further discuss this point below . the right - hand - side of the saddle - point equation is logarithmically divergent in the @xmath159 limit . hence , there is always a solution @xmath160 we have taken @xmath161 as the ultraviolet cutoff and @xmath162 . this is the non - trivial minimum of the action to which we referred at the end of section [ sec : one - loop - s ] . the structure of this saddle - point equation is clearly reminiscent of the bcs gap equation . the integrals on the right - hand - side are logarithmically divergent as expected for a marginally relevant interaction . a symmetry - breaking order parameter must develop in order to cut them off . we can submit our saddle - point solution to an important check by verifying its consistency with the rg equations . the gap to spin - flip excitations , @xmath163 is an observable quantity , and should be independent of the ultraviolet cutoff , @xmath164 . even though the cutoff is taken to be a physical scale @xmath165 , we are free to integrate out short - wavelength fluctuations in order to lower the cutoff while simultaneously renormalizing the couplings @xmath85 , @xmath86 , @xmath2 , and @xmath98 . this must occur in such a way that the saddle - point solution , @xmath163 , remains unchanged . for this reason , we regard the one - loop calculation of section [ sec : one - loop - s ] as incomplete . it is consistent with an rg equation which only includes diagrams ( d ) and ( e ) of figure [ fig : five - diagrams ] for the flow of @xmath86 . diagrams ( a ) , ( b ) , and ( c ) are not incorporated , and neither is the flow of @xmath85 , @xmath2 , @xmath98 . . solid lines are @xmath118 propagators . dotted lines are bare @xmath15 propagators.,width=312 ] though the preceeding one - loop calculation captures the basic physics of the problem , it does not capture all of the order @xmath2 physics . as we have just noted , the effective action ( [ eq : s - eff - x-1 ] ) only includes terms corresponding to two of the five diagrams in figure [ fig : five - diagrams ] namely , ( d ) and ( e ) which contribute to ( [ eq : rgeqns ] ) . the apparent contradiction stems from the fact that the physics of a one - loop rg equation might only appear at higher - loops in a saddle - point calculation . this is surprising at first glance , but can be understood at a technical level by noting that a contribution to the effective action of the form @xmath166 is equally important at the saddle - point as a contribution of the form @xmath167 since the extra logarithm compensates for the extra factor of @xmath2 . a simple example of a model in which this occurs is given in the appendix . the effects of @xmath85 on the flow of @xmath86 , @xmath2 , @xmath98 enter in a two - loop calculation of the effective action . consider the diagrams of figure [ fig : gamma_1-renorm ] . these are two - loop diagrams which , when combined with the one - loop diagram of the previous section , have the effect of shifting in the effective action of the previous section the values of @xmath86 , @xmath2 , and @xmath98 from their bare values . these diagrams can receive contributions from non - spin - flip diffusion propagators which are not cut off by the magnetization , @xmath15 . they are charge and @xmath168 diffusion modes . since we expect them to be a part of the low - energy theory , we can choose not to integrate them out or to integrate out modes above a cutoff of our choosing . the natural choice is to integrate out these modes above @xmath163 so that the resulting effective field theory for @xmath15 , @xmath123 , @xmath124 contains only excitations with energies less than @xmath163 . ( we could alternatively imagine performing the calculation at finite - temperature so that @xmath58 is effectively the cutoff for both spin - flip and non - spin - flip propagators . such a calculation could be used to obtain the transition temperature @xmath23 for the putative transition . ) this effects the following replacements : @xmath169 where @xmath170 $ ] which renormalize @xmath86 , @xmath2 , and @xmath119 . the first diagram renormalizes @xmath86 while the second renormalizes @xmath2 and @xmath119 . the quadratic vertices are @xmath85 interaction vertices.,width=312 ] diagrams ( a)-(c ) of figure [ fig : five - diagrams ] can not contribute directly to @xmath16 $ ] for static @xmath15 because the external @xmath118 lines can not carry zero frequency . the matsubara indices @xmath171 , @xmath172 of @xmath173 necessarily satisfy @xmath174 , @xmath175 , so their difference can not vanish . for the same reason , these diagrams can not be linked with ( d ) and ( e ) , in the way demonstrated in the toy model in the appendix . it is possible that contributions related to diagrams ( a)-(c ) enter @xmath16 $ ] through even higher - loop contributions or , perhaps , when fluctuations in @xmath15 are included . however , this touches upon the unresolved issue of the renormalizability of the theory at higher - loops . in the absence of a clearer understanding of such higher - loop contributions to @xmath16 $ ] , we conclude that the one - loop effective action ( [ eqn : s_eff - x-4 ] ) with the replacement ( [ eqn : two - loop - contribution ] ) is the most systematic result which we can obtain at present . we expect that going beyond one - loop in @xmath118 or including the effects of fluctuations of @xmath15 will quantitatively but not qualitatively modify our result . in section [ sec : one - loop - s ] , we removed the bi - linear coupling between @xmath15 and @xmath118 by shifting @xmath15 . it may seem more natural to remove the bi - linear coupling between @xmath15 and @xmath118 by performing the shift @xmath176 . if we then integrate out @xmath118 , we will not find a non - trivial replica - symmetric saddle - point ( though there may be replica - symmetry - breaking saddle - points ) , and the resulting action will not contain any of the physics of the one - loop rg equations . in order to recover this physics , we must now go to one - loop in @xmath16 $ ] . going to one - loop in @xmath16 $ ] , we recover the saddle - point of our initial action of section [ sec : one - loop - s ] . the relevant diagrams are diagrams ( d ) and ( e ) of figure [ fig : five - diagrams ] , but the solid lines are shrunk to a point when these diagrams are viewed as @xmath97-loop diagrams in @xmath16 $ ] . diagrams corresponding to ( a ) , ( b ) , and ( c ) have dotted lines attached to the solid external legs ( via the bi - linear coupling between @xmath15 and @xmath118 ) ; however , these vanish in the static limit for the same reason as above . thus , at this level , there is no advantage in proceeding this way , and there is a definite disadvantage in that one extra step is necessary . however , it is possible that a higher - loop calculation is simpler in this approach . the one - loop effective action which we computed in section [ sec : one - loop - s ] is actually exact in a generalization of the model . the most naive large-@xmath0 version of the model is not particularly simple . if we simply introduce @xmath0 flavors of electrons @xmath177 , @xmath178 , then the @xmath65-matrix has two flavor indices , @xmath179 , and the large-@xmath0 limit allows us to drop all non - planar diagrams . however , we do not know how to sum all of the planar diagrams . instead , we consider the following generalization : we suppose that there are @xmath0 flavors of diffuson fields @xmath180 , @xmath178 . this generalization does not have a simple interpretation in terms of the original electron variables , but it is technically simple . the action ( [ eq : imp - terms ] ) is generalized to : @xmath142 = \pi{n_f}\sum \int \biggl\ { d\ , { \rm tr}\left(\nabla{v^\dagger } \ , \nabla v\right ) + z |{\omega_{nm}}| { v_{nm;\,a\alpha}^\dagger } v^{}_{mn;\,a\alpha } - { \gamma_2 } v^\dagger_{nm;\,a\alpha } v^{}_{m'n';\,a\alpha}\ , \delta_{n - m , n'-m'}\\ + i { \gamma_2 } \left [ { \rm tr } \left(v^{}_{ml;\,a\alpha } v^\dagger_{ln;\,b\alpha } { \sigma^z_{ab}}\,x(n - m)\right ) + { \rm c.c.}\right ] + \frac{n}{2 } \left(1-{n_f}{\gamma_2}\right)\ , \frac{\gamma_2}{\pi{n_f}}\,{\rm tr}\left({x^2}\right ) - \frac{1}{n}\,{\gamma_2 } { \sum_\beta}\left [ { \rm tr } \left(v^{}_{\alpha } v^\dagger_{\alpha } v^{}_{\beta}\right ) + { \rm c.c.}\right ] \biggr\ } \label{eq : imp - terms - large - n}\end{gathered}\ ] ] note that there is still only a single @xmath15 field . the explicit factor of @xmath0 in front of the penultimate term and the factor of @xmath181 in front of the final term ensures that every term in the action is @xmath182 . for this action , the diagram which we calculated in section [ sec : one - loop - s ] is the sole @xmath182 diagram . all others are @xmath183 or smaller . furthermore , our effective action is multiplied by a coefficient @xmath0 . thus , in the @xmath184 limit , our one - loop saddle - point calculation is exact . at first glance , it may seem almost contradictory to take an @xmath184 limit while simultaneously taking a replica limit @xmath185 . however , the flavor index structure in the action ( [ eq : imp - terms - large - n ] ) is very different from the replica index structure . the latter is chosen to ensure that closed fermion loops vanish while the latter selects a preferred set of the remaining diagrams . hence , the two limits are consistent . having concluded that the divergence of the triplet interaction amplitude leads to ferromagnetism , the obvious next question is `` so then what ? '' to answer this , we must examine the diffusion of electrons in a ferromagnetic metal , and the interaction between diffusion modes and spin waves . in the ferromagnetic state , we can expand @xmath15 about its perfectly - ordered ground state , which we take to be along the @xmath186-direction @xmath187 we can also perform a gradient expansion of @xmath188 $ ] . the result is essentially dictated by symmetry : @xmath189 = \int \left ( \epsilon_{ij } { m_i}{\partial_\tau}{m_j } + k\left(\nabla{m_i}\right)^2 + o\left({\left(\nabla{m}\right)^4}\right)\right)\ ] ] where @xmath190 are the @xmath191 and @xmath192 components of @xmath193 . the constant @xmath194 is the spin stiffness of the ferromagnet . we now reinstate the coupling of the order parameter to the diffusion modes of the system , as in ( [ eq : almost - sigma - model ] ) or ( [ eq : v - x - action1 ] ) . @xmath195 = \pi{n_f}\sum \int \biggl\ { d\ , { \rm tr}\left(\nabla{v^\dagger } \ , \nabla { v^{}}\right ) + z |{\omega_{nm}}| { v_{nm}^\dagger } v_{mn}^{}\\ + i { \gamma_2}\,{x_0 } \left[{\rm tr}\left({\sigma_z}v^{}_{nm } v^{\dagger}_{mn}\right ) + { \rm c.c.}\right]\\ + i { \gamma_2}\ , \left[{m_i}(\omega_{n - m})\ : { \rm tr}\left({\sigma_i}v^{}_{nm } \right ) + { \rm c.c.}\right]\biggr\}\\ + { \sum_n}\int \left ( \epsilon_{ij } i{\omega_n}{m_i}{m_j } + k\left(\nabla{m_i}\right)^2 \right)\end{gathered}\ ] ] where @xmath196 are pauli matrices and we have dropped higher - order terms in @xmath197 . a purist may object that the @xmath118 s should not appear in the action for @xmath197 since the dynamics for the order parameter was obtained by integrating out the @xmath118 s . if we now calculate with this action , we run the risk of double - counting the degrees of freedom of the system . in order to avoid this , we must either calculate only to tree - level in @xmath118 ( but allow loops in @xmath197 ) or introduce counter - terms which cancel the final line of ( [ eq : v - m - action1 ] ) . in the second approach , by computing loops in @xmath118 , we will cancel the @xmath197 counterterms . the physics of this model is not completely manifest in the above form ( [ eq : v - m - action1 ] ) because a uniform rotation of the ferromagnetic order parameter appears to be an excitation since there are non - derivative terms in @xmath197 when , in fact , it is nt . this can be made manifest by rotating the quantization axis of the diffusing electrons so that it is aligned along the local direction of the order parameter . this is done by enacting the variable change @xmath198 the action is now @xmath199 = { n_f}\sum \int \biggl\ { d\ , { \rm tr}\left(\nabla{v^\dagger } \ , \nabla { v^{}}\right ) + z |{\omega_{nm}}| { v_{nm}^\dagger } v_{mn}^{}\\ + i { x_0}{\gamma_2}\ , \left[{\rm tr}\left({\sigma_z}v^ { } v^{\dagger } \right ) + { \rm c.c.}\right]\\ d\ , { \rm tr}\left(\nabla{v_{nm}^\dagger } \ : \nabla\!\left(\epsilon_{ij } { m_i}(\omega_{n - l } ) \left[{\sigma_j},{v_{ml}^ { } } \right]\right)\right ) + { \rm c.c.}\biggr\ } \\ + { \sum_n}\int \left ( \epsilon_{ij } i{\omega_n}{m_i}{m_j } + k\left(\nabla{m_i}\right)^2 \right)\end{gathered}\ ] ] as noted earlier , there is a gap for spin - flip diffusion modes , as given by the second line of ( [ eq : v - m - action2 ] ) . thus , we can ignore these modes at low energies and focus on charge diffusion and @xmath168 diffusion ; @xmath200 are carried by spin waves @xmath201 . from the structure of the interaction term on the third line of ( [ eq : v - m - action2 ] ) , we see that @xmath201 can not couple directly to charge or @xmath168 diffusion alone , as dictated by rotational invariance . a charge or @xmath168 diffusion mode , upon emitting or absorbing a spin wave , must then become a spin - flip diffusion mode . thus , this coupling can be ignored at energies below the gap for spin - flip diffusion modes . even in the absence of such a gap , the coupling would have dimension @xmath202 in @xmath203-dimensional space and be strongly irrelevant , as expected for interactions with goldstone modes which must occur through derivative couplings such as that on the third line of ( [ eq : v - m - action2 ] ) . thus , the low - energy physics of the ferromagnetic diffusive metal state is that of ferromagnetic spin waves decoupled from diffusion in a zeeman field . the latter has been well - studied @xcite , and is similar to the spinless electron problem . interactions suppress the conductivity , as may be seen from the rg equations : @xmath204 @xmath205 is the part of the interaction between @xmath206 diffusion modes which includes a contribution from both the direct and exchange channels , @xmath207 ; its bare value is @xmath208 . in two dimensions , these rg equations describe a flow towards insulating behavior . however , there is a metal - insulator transition in @xmath21 dimensions through which the system can be tuned by varying the bare resistivity through a critical value of @xmath22 . we examine the situation in some more detail in the next section . the preceeding discussion is strictly true at zero temperature . at finite temperatures , thermal fluctuations will destroy two - dimensional long - range ferromagnetic order , but some signature of the zero - temperature ferromagnetic phase persists . if a finite external magnetic field is applied , the magnetization is a scaling function of temperature and magnetic field . a large - n calculation for an @xmath209 magnet theory gives the following expression for the scaling function for the susceptibility@xcite : @xmath210 \label{eqn : suscept - scaling}\ ] ] where @xmath211 and @xmath212 are the rescaled spin stiffness and magnetic field respectively , @xmath213 is a root of the equation @xmath214 , and @xmath215 is the zero temperature magnetization . the susceptibility is given by @xmath216 , which diverges exponentially as @xmath217 . in the limit of weak interactions and weak disorder , there is a substantial temperature regime in which there is ferromagnetic order with a long but finite correlation length and the resistivity is small . if the dimensionless resistivity , @xmath2 , is taken to be @xmath218 at temperature @xmath219 k , and @xmath220 , the average interspace between electrons measured in the unit of effective bohr radius , is taken to be 0.1 , then the susceptibility is large below @xmath221 mk . however , the resistivity becomes @xmath183 at temperatures @xmath222 . while ferromagnetism sets in at very low temperatures , insulating behavior is completely academic . in the temperature interval between these two scales , we expect the susceptibility to be of the above form ( [ eqn : suscept - scaling ] ) . when @xmath6 is larger , the separation between these scales decreases , but we should also be more circumspect about using this weak - coupling theory . of course , if the magnetic field is removed , we pass from the unitary ensemble to the orthogonal one , and weak - localization effects should drive the system insulating much more rapidly . in @xmath21 dimensions , the renormalization group equations are modified by the appearance of tree - level dimensions associated with the spatial dimensionality : @xmath223 by varying @xmath86 , we can tune the system through a paramagnet - ferromagnet transition . for @xmath2 , @xmath113 small , the flows are into the fixed line @xmath224 . this fixed line terminates at @xmath225 . for @xmath226 , there is a metal - insulator transition at @xmath227 . in the vicinity of this transition , @xmath113 is relevant . if it flows to sufficiently large values , then ferromagnetism will develop . however , if @xmath2 flows to zero before this happens , then the system will remain paramagnetic . both possibilities can occur as soon as @xmath113 becomes nonzero , so the ferromagnet - paramagnet boundary which begins on the @xmath224 axis at @xmath225 terminates at @xmath228 . ferromagnetism , if it occurs , will have a finite onset temperature . at lower temperatures , the conductivity may be metallic or insulating , depending on @xmath2 . in the former case , we have a ferromagnetic metal . in the latter case , the system becomes insulating at lower temperatures . it is possible that the ferromagnetic order is destroyed by the onset of insulating behavior . this can not be resolved by the methods in this paper ; thus , we have left the phase boundary of the paramagnetic insulating state intentionally vague . ( in fact , it may not be paramagnetic , but might instead be a spin glass or some other non - ferromagnetic phase . ) dimensions . pm is the paramagnetic metal phase ; fm , the ferromagnetic metal phase ; fi , the ferromagnetic insulator . pi is a non - ferromagnetic insulator : possibly paramagnetic , but perhaps a spin glass.,width=312 ] in this paper , we have used a straightforward and absolutely standard technique albeit in a somewhat complicated and technically - involved context to argue that the instability of the weakly - interacting electron system in a weak random potential signals the development of ferromagnetism . since there is a single instability occurring in the weak - interaction limit and it occurs while the resistivity is small , we are afforded a great deal of control over this problem . the instability bespeaks of the relative importance of one term in the action compared to the others . we need only satisfy this term which we do by allowing for the development of a ferromagnetic moment without worrying about other terms , since none of them are relevant . the demonstrable absence of a competing instability is the main justification for a simple mean - field theory . this may be formalized in a large-@xmath0 generalization of the problem . further credence is given by the physical arguments of the introductory section which rule out any other resolution of the divergence associated with the triplet amplitude . this is , in fact , a remarkable state of affairs , given that there is no comparable controlled theory for itinerant electron ferromagnetism in clean metals . in a clean metal , the stoner instability occurs at strong coupling , where there are other competing instabilities and no unambiguous way of deciding between them . it is striking how little is known about the phase diagram of the 2deg . even at zero magnetic field , the phase diagram as a function of @xmath6 and disorder strength is only known in the small and large @xmath6 limits for vanishing disorder . intermediate @xmath6 is poorly understood , as is finite disorder . we believe that we have , in this paper , extended _ terra cognita _ to include the small @xmath6 , weak disorder region . we emphasize that our result does not shed too much light on the metal - insulator transition which is observed in strongly - interacting 2degs . the development of ferromagnetism does not lead to a metallic state in two dimensions . once ferromagnetism has developed , weak localization ( which we have neglected in this paper ) and interaction corrections both drive the system insulating . there is no contradiction with experiments , however , since the transition seen there is in the strong - interaction , high - resistivity ( @xmath229 ) regime where our result is not expected to be valid . on the other hand , our result _ does _ rule out theories which conjecture that the divergence of the triplet interaction leads to a metallic state _ at zero - temperature _ in 2d . we would like to thank c. chamon , a. kamenev , e. mucciolo , l. radzihovsky , r. shankar , and t. senthil for discussions . this work was supported by the national science foundation under grant no . dmr-9983544 and by the a.p . sloan foundation . when we transform an action by introducing hubbard - stratonovich fields and integrating out the original variables of the action , the result can be an action which looks very different from the original one . however , the coupling constants of the two actions are related , and the corresponding rg equations should be as well . in the simplest cases , this is completely transparent . consider , for instance , a system of @xmath0 flavors of spinless fermions in one - dimension . suppose that their density is commensurate with the underlying crystal lattice and that their interaction respects the @xmath230 symmetry which mixes the fermions : @xmath231 here , @xmath45 . @xmath232 annihilate fermions of flavor @xmath233 at , respectively , the right and left fermi points . this is the @xmath230 gross - neveu model . let us consider its large-@xmath0 limit . first , consider the rg equation for @xmath4 . there is a single one - loop diagram , depicted on the left of fig . [ fig : gn - one - loop ] . all other diagrams are smaller by powers of @xmath181 . the resulting rg equation is : @xmath234 if we introduce a hubbard - stratonovich field @xmath235 which decouples the interaction @xmath236 then we can integrate out @xmath237 : @xmath238 where we have combined @xmath232 into a two - component spinor on which @xmath239 act . @xmath235 cuts off the infrared logarithmic divergence of the @xmath240 term . @xmath241 where @xmath164 is the ultraviolet cutoff . the second line suggests that @xmath242 renormalizes @xmath243 . thus , the effective action manifestly embodies the physics contained in the rg equation ( [ eq : gn - rgeqn ] ) . this is also clear from the close kinship between the one - loop diagram which gives the @xmath244 in ( [ eq : delta - action ] ) and the one - loop diagram which yields the rg equation ( [ eq : gn - rgeqn ] ) . however , strictly speaking , the formal relationship between the rg equation and the effective action ( [ eq : delta - action2 ] ) follows from the cutoff - independence of physically - measurable quantities , @xmath235 being the simplest . from ( [ eq : delta - action2 ] ) , we see that the saddle - point value of @xmath235 is @xmath245 differentiating both sides with respect to @xmath164 and requiring that @xmath246 vanish , we find : @xmath247 which is equivalent to the rg equation ( [ eq : gn - rgeqn ] ) . now , consider @xmath251 , where @xmath235 is introduced to decouple the @xmath4 interaction , as before . we can no longer calculate it exactly . however , if we assume that @xmath252 , we can calculate it to lowest non - trivial order in @xmath253 . the lowest order diagram by which it contributes is the second one of fig . [ fig:1 - 3-loop - toy ] . we integrate out @xmath237 completely and integrate out all @xmath254 fluctuations above some cutoff @xmath255 . we are free to choose @xmath255 as we like , of course , but the natural choice is @xmath256 so that our effective action is for excitations with energies of order @xmath235 . the resulting effective action is @xmath257 the last term looks strange . it looks like an @xmath258 renormalization of @xmath243 , which would imply a contribution of order @xmath259 to @xmath260 . also , this term is proportional to @xmath261 rather than @xmath262 as we might have expected from the second diagram of fig . [ fig : gn - one - loop ] and the rg equation ( [ eq : gn-rgeqn ] ) . the problem is that we are looking at @xmath263 , which is not observable , rather than the vacuum expectation value of @xmath235 , which is . when we compute @xmath235 , the two difficulties noted above resolve each other . minimizing the effective action , we find that , to lowest non - trivial order in v : @xmath264 or @xmath265 requiring @xmath164-independence of @xmath235 , we recover the rg equation for @xmath4 and also the rg equation for @xmath253 which we did nt compute . the lesson is that a one - loop contribution to the rg equation for a coupling constant might manifest itself in an effective action only through a higher - loop ( in the above example , three - loop ) diagram . in the interacting electron problem in a random potential , diagrams ( a ) , ( b ) , and ( c ) could enter @xmath16 $ ] through three - loop diagrams . note that finkelstein uses @xmath4 to denote resistance , as did chamon , ludwig , and nayak , in keeping with his notation . however , the more widespread usage is for @xmath4 to denote conductance . to avoid confusion , we adopt the notation of castellani , di castro , lee , and ma : @xmath2 denotes resistance . s. v. kravchenko _ et al . _ , phys . rev . b * 51 * , 7038 ( 1995 ) ; s. v. kravchenko _ et al . _ , lett . * 77 * , 4938 ( 1996 ) ; d. simonian , s. v. kravchenko , m. p. sarachik , and v. m. pudalov , phys . * 79 * , 2304 ( 1997 ) . when the bare interaction is weak , disorder does not allow the former , an sdw , and its necessary concomitant , a cdw , to form since the energy scale at which the interaction becomes strong is small compared to the inverse of the elastic scattering time . on the other hand , one can imagine that strong interactions could lead to the formation of an sdw in the ballistic regime , i.e. at length scales shorter than the mean - free path . at longer scales , sdw order would be disrupted by impurities , but the correlation length could be long if the impurity density is low . in the problem which we consider , however , interactions are assumed to be weak in the ballistic regime . they only become strong in the diffusive regime , where sdw order will not develop . what we really mean by this is that spin glass order can not be important _ at this scale_. however , local ferromagnetism , which develops at the scale at which the triplet interaction becomes strong , could be disrupted at longer scales ( at which the effective disorder strength is large ) so that spin glass order e.g. a cluster spin glass forms at the longest length scales . furthermore , in an insulating state , caused by wigner crystallization and/or anderson localization , a spin glass could form since there will be competition between different multi - spin exchange process which lead to effective spin - spin interactions of different signs .

we present a study of the morphological fractions and color magnitude relation in the most distant x - ray selected galaxy cluster currently known , xmmxcs j at , using a combination of optical imaging data obtained with the _ hubble space telescope _ advanced camera for surveys , and infrared data from the multi - object infrared camera and spectrograph , mounted on the 8.2 m _ subaru _ telescope . we find that the morphological mix of the cluster galaxy population is similar to clusters at . within the central 0.5 mpc , approximately% of the galaxies identified as likely cluster members are ellipticals or s0s ; and% are spirals or irregulars . therefore early type galaxies were already entrenched as the dominant galaxy population in at least some clusters approximately gyr after the big bang . we measure the color magnitude relations for the early type galaxies , finding that the slope in the relation is consistent with that measured in the coma cluster , some gyr earlier , although the uncertainty is large . in contrast , the measured intrinsic scatter about the color magnitude relation is more than three times the value measured in coma , after conversion to rest frame . from comparison with stellar population synthesis models , the intrinsic scatter measurements imply mean luminosity weighted ages for the early type galaxies in j of gyr , corresponding to the major epoch of star formation coming to an end at . we find that the cluster exhibits evidence of the ` downsizing ' phenomenon : the fraction of faint cluster members on the red sequence expressed using the dwarf - to - giant ratio ( dgr ) is within a radius of . this is consistent with extrapolation of the redshift evolution of the dgr seen in cluster samples at . in contrast to observations of some other clusters , we find a lack of very bright galaxies within the cluster .

0903.1731

the galaxy populations of clusters are dominated by galaxies of early morphological type , ellipticals and s0s , which form a tight color magnitude relation ( cmr ) or ` red sequence ' . for decades , this relation was interpreted as evidence that the stars in early type galaxies in clusters are uniformly old , being formed at redshift @xmath15 , with the slope of the relation being primarily due to a mass - metallicity relation @xcite . the alternative interpretation of the cmr as being predominantly an age sequence was conclusively ruled out by observations of clusters at @xmath16 , which showed that the slope of the cmr evolves little with redshift @xcite . a number of studies of clusters conducted with the _ hubble space telescope _ have shown that the cmr of elliptical galaxies remains well established at progressively higher redshifts ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , at least up to @xmath17 @xcite . the only study to date of the cmr in a cluster at @xmath18 is consistent with this picture @xcite . measurements of the intrinsic scatter about the cmr can be used to constrain the ages of the constituent stellar populations of cluster early type galaxies , and indicate that major epoch of star formation in clusters was completed at @xmath19 ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? studies of the fundamental plane of elliptical galaxies up to high redshift similarly indicate that the bulk of stellar populations in elliptical galaxies were formed at @xmath19 ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? these observations are consistent with a simple formation scenario for elliptical galaxies , in which they formed the bulk of their stellar mass in a single event at high redshift , and evolved passively thereafter . however , the latest semi - analytic models of galaxy formation , constructed within the hierarchical formation paradigm , are also able to successfully reproduce the old ages of stellar populations in elliptical galaxies @xcite . recent observations over a wide range in redshift in both clusters and the field have revealed that star formation appears to be completed earlier in higher mass galaxies than in low mass galaxies , a phenomenon dubbed ` downsizing ' @xcite . in color - magnitude diagrams of high redshift clusters , this effect manifests itself as a deficit of faint galaxies on the red sequence in comparison to clusters observed at lower redshift ( e.g. * ? ? ? * ; * ? ? ? * ) , and may have an environmental dependence in the sense that the faint end of the cmr is populated at earlier times in denser regions , such as in the cluster cores , compared to low density regions such as groups ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? this could be explained qualitatively by a scenario in which the faint end of the red sequence is being built up by the transformation of star forming , spiral galaxies into passively evolving s0s as redshift increases . up to @xmath20 , studies of magnitude - limited samples of cluster galaxies have revealed that the elliptical galaxy fraction within clusters is found to remain roughly constant , while the fraction of s0 galaxies decreases , with a corresponding increase in the fraction of spiral and irregular galaxies @xcite . when considering stellar mass - limited samples , it has been found that the fraction of massive , early type galaxies as a whole remains roughly constant , at least up to @xmath21 @xcite . this suggests that the much stronger evolution seen in magnitude - limited samples is dominated by fainter , lower mass ( sub-@xmath22 in the galaxy stellar mass function ) galaxies . it should be noted that at the current time the evidence for downsizing within the centers of rich clusters is still somewhat contentious ( @xcite ; @xcite ) . in this paper , we present a study of the galaxy morphologies and color magnitude relation in xmmxcs j@xmath0 at @xmath1 @xcite , the most distant x - ray selected galaxy cluster currently known . to date , only five other spectroscopically confirmed clusters are known at @xmath23 , xmmu j@xmath24 at @xmath25 , discovered serendipitously with _ xmm - newton _ @xcite ; three clusters at @xmath26 from the irac shallow cluster survey ( iscs ; * ? ? ? * ; * ? ? ? * ) ; and the @xmath27 cluster j@xmath28 , discovered by the _ spitzer _ adaptation of the red - sequence cluster survey ( sparcs ; * ? ? ? j@xmath0 was discovered as part of the ongoing optical follow up campaign to the _ xmm _ cluster survey ( xcs ; * ? ? ? * ) , which has the primary aim of constraining the cosmological parameters through measuring the evolution of the cluster mass function with redshift . predictions for the expected cosmological constraints that are expected to be achieved by the full survey can be found in @xcite . j@xmath0 has x - ray luminosity @xmath29 ergs s@xmath30 , temperature @xmath31 kev @xcite , and velocity dispersion @xmath32 km s@xmath30 @xcite . there is mild evidence that the cluster velocity distribution is bimodal , which , if confirmed by further spectroscopic observations , would indicate that the cluster is undergoing a merger close to the line of sight @xcite . the structure of this paper is as follows . we begin by describing the observations and data reduction in [ s_observationsdatareduction ] . in [ s_analysis ] , we describe our photometric measurements , morphological classification , and photometric redshift selection of the cluster members . we present the morphological fractions , fits to the color magnitude relation , and inferred ages for the stellar populations of the early type galaxies in [ s_results ] . finally , we discuss our findings in comparison with previous work in [ s_discussion ] . we assume a concordance cosmology of @xmath33 , @xmath34 , and @xmath35 km s@xmath30 mpc@xmath30 throughout , where @xmath36 is the energy density associated with a cosmological constant . for our study of the galaxy populations of j2215.9@xmath371738 , we use a combination of optical data in the @xmath38 and @xmath39 bands obtained with the _ hubble space telescope _ ( _ hst _ ) advanced camera for surveys ( acs ; * ? ? ? * ) , and near infrared data in the @xmath40 and @xmath41 bands taken using the multi - object infrared camera and spectrograph ( moircs ; * ? ? ? * ) on the 8.2 m _ telescope at the summit of mauna kea , hawaii . as shown in fig . [ f_passbandsplot ] , at the cluster redshift of @xmath1 the optical data lies blueward of the 4000 break , and therefore near infrared data are essential to provide a color which is able to separate the cluster red sequence from bluer , star forming galaxies . the @xmath6 color at the cluster redshift of @xmath1 roughly corresponds to rest frame @xmath8 , while @xmath42 is roughly equivalent to rest frame @xmath43 . [ f_rgbimage ] presents a color - composite ( @xmath39 , @xmath40 , @xmath41 ) image of the cluster . it is apparent from fig . [ f_rgbimage ] that j@xmath0 lacks an obvious brightest cluster galaxy ( bcg ) , in contrast with the only other similarly high redshift x - ray selected cluster currently known , xmmu j@xmath24 at @xmath25 ( * ? ? ? * see also @xcite ) . similarly , the galaxy distribution is much less compact in j@xmath0 than in j@xmath24 . we now describe the observations and data reduction performed for each instrument in turn . j@xmath0 was observed using the acs wide field channel ( wfc ) as part of a program designed to place constraints on the dark energy through observations of high redshift type ia supernovae . the field of view of the acs wfc is approximately @xmath44 , and the pixel scale of the detectors is @xmath45 per pixel . a detailed description of the acs observations is presented in @xcite . a total of 3320 sec of exposure was obtained in @xmath38 across 12 frames . in @xmath39 , 16935 sec of exposure was obtained in 64 frames . in both cases , the final stacked images were created using the ` multidrizzle ` ` pyraf ` task . j2215.9@xmath371738 was observed with moircs in photometric conditions on ut 2007 august 08 . the observations were carried out in service mode . the field of view of moircs is @xmath46 , which is imaged at a resolution of @xmath47 per pixel on to a pair of @xmath48 hawaii-2 detector arrays . unfortunately , one of the detector arrays was inoperable at the time our observations were carried out . our observations were designed to place the target cluster at the center of the second , working detector array . we performed imaging in the @xmath40 and @xmath41 bands , using a 9-point circular dither pattern of radius 25@xmath49 to ensure good sky subtraction . we obtained a total of 1485 sec of integration in @xmath40 , for an individual frame time of 165 seconds at each dither position . in the @xmath41 band , the total integration time was 1242 sec , achieved by co - adding @xmath50 sec exposures at each dither position . the observations were obtained at airmass @xmath51 . the seeing was excellent , at @xmath52 in both @xmath40 and @xmath41 . the data were reduced using the ` mcsred ` package for the ` iraf`environment in the standard manner : the individual dither frames were flat fielded , sky subtracted , corrected for distortion caused by the optical design of moircs , and registered to a common pixel coordinate system . the final science images were created from the @xmath53-clipped mean of the dither frames . as most of our acs data were obtained before ut 2006 july 04 , we adopted ab photometric zeropoints in the @xmath38 and @xmath39 acs bands of 25.678 and 24.867 respectively . after this date , the acs photometric zeropoints changed due to a change in the temperature of ccds : frames obtained after this change occurred were adjusted to match the above zeropoints . during the moircs observations , we obtained imaging of several photometric standards from the ukirt faint standards list . the zeropoints we determined in the @xmath40 and @xmath41 bands are [email protected] and [email protected] on the vega system respectively . for the near infrared data we adopt conversions to the ab magnitude system @xcite of @xmath55 and @xmath56 @xcite . all magnitudes quoted throughout the rest of this paper are on the ab system , unless stated otherwise . to measure accurate colors for the galaxies in j2215.9@xmath371738 , we need to perform measurements through a consistent set of photometric apertures . we used the ` sextractor ` package @xcite in dual image mode to achieve this , where objects are detected and apertures are defined using the first image , while photometry is performed on the second image . we chose to use the moircs @xmath41 image as the detection image . as the acs and moircs imaging data have different angular and pixel resolution , some additional processing was required prior to performing the photometry . firstly , we trimmed the acs and moircs images to cover a common area of dimensions @xmath57 , centered on the cluster x - ray position . this area encloses the overlapping acs coverage from the individual dither frames , rejecting higher noise regions at the edges covered by fewer acs pointings . we then scaled the moircs images from their native pixel scale of 0.117@xmath49 per pixel to the acs pixel scale of 0.05@xmath49 per pixel , interpolating using a 3rd order spline polynomial . the moircs images were then registered to the acs images to pixel accuracy . finally , the acs images were matched to the 0.5@xmath49 angular resolution of the moircs images by convolution with a gaussian filter with @xmath58 pixels . to estimate the photometric errors in each band , we constructed rms images for use as weight images by ` sextractor ` . for the moircs images , we used the rms images produced by the ` mcsred ` data reduction package - these are simply the standard deviation at each pixel derived from the 9 individual dither frames . for the acs images , the rms images were constructed from the drizzled science images and the exposure time maps output by the ` multidrizzle ` task , taking into account the photon statistics and the detector read noise . the drizzling process used to make the final acs science images correlates the pixel - to - pixel noise , which , if ignored , would lead to underestimated photometric errors . we corrected for this effect following the methodology outlined in the appendix of @xcite . the effect of the interpolation operations on the photometry from the moircs images was tested by cross matching catalogs constructed at the native moircs pixel scale , with catalogs made from the images resampled to match the acs pixel scale . objects were matched between each catalog within a 0.5@xmath49 radius and the @xmath59 aperture magnitudes were compared . we found that the median magnitude difference between the two catalogs was @xmath60 for @xmath61 . we decided to treat the scatter between the catalogs as an additional source of uncertainty : we fitted an exponential model to the scatter as a function of magnitude , applied this model to the @xmath62 aperture magnitudes in the catalog constructed at the acs pixel resolution , and added the results in quadrature to the photometric errors output by ` sextractor ` . the increase in the @xmath40 and @xmath41 aperture magnitude uncertainties resulting from this procedure is small , @xmath63 mag . at @xmath64 . to classify objects as stars and galaxies , we constructed a second set of catalogs in the manner described above , with the exception that the acs images were not matched to the 0.5@xmath49 resolution of the moircs images , and instead were read into ` sextractor ` at their native @xmath65 angular resolution . we used the neural - network based star - galaxy classification provided by ` sextractor ` ( ` class_star ` ) to determine the object types . we classify all objects with ` class_star`@xmath66 as measured in the @xmath39 image as galaxies ; all other objects are assumed to be stars . we chose to carry out the classification in the @xmath39 band , because the @xmath39 image is significantly deeper than the @xmath38 image , reaching an approximate @xmath67 limiting magnitude of @xmath68 for galaxies , compared to @xmath69 for the @xmath38 image ( estimated from the photometric uncertainties ) . we corrected the photometry for the effect of galactic extinction using the dust maps and software of @xcite . in all that follows , ` sextractor ` ` mag_auto ` magnitudes were adopted as measurements of total galaxy magnitudes , and colors were measured in 1.0@xmath49 diameter circular apertures , equal to twice the seeing of the moircs images . this is comparable to the typical diameter of the cluster member galaxies . note that we do not attempt to correct the photometry of objects blended as a consequence of performing the object detection in the @xmath41-band . only a small number of such objects are seen , perhaps due to the galaxy distribution in j@xmath0 being less compact than other similarly high redshift clusters , as noted in [ s_observationsdatareduction ] above . as shown in [ s_resultsmorphologies ] , this has a negligible effect on our results . morphological classification was performed on all 201 galaxies with magnitudes brighter than @xmath70 , by visual inspection of the @xmath39-band imaging data . this is @xmath71 magnitudes brighter than the approximate @xmath67 magnitude limit of the @xmath39-band image . due to the high redshift of j@xmath0 , the morphological classification was carried out in the rest frame @xmath72-band , rather than the traditional @xmath73-band ( e.g. * ? ? ? furthermore , the typical diameter of the cluster member galaxies in the acs imaging is @xmath74 pixels , making detailed morphological classification on the traditional hubble system difficult . for these reasons , we chose to place the galaxies into four broad morphological bins : elliptical ( e ; corresponding to e1@xmath37e9 on the traditional hubble system ) , lenticular ( s0 ) , spiral ( sp ; sa@xmath37sd , sba@xmath37sbd on the traditional hubble system ) or irregular ( irr ) . the classification was carried out on the complete sample by a team of five classifiers ( cac , mh , bh , sas , jps ) . each galaxy in the sample was examined by means of a 150 pixel on a side postage stamp image , centered on the galaxy in question . no further information on the galaxy to be classified ( e.g. , position , redshift , color or magnitude ) was provided during the classification process , in order to ensure that the classification was determined strictly on the evidence of the imaging data alone , guarding against potential biases . the same training set of galaxy images , constructed from the @xcite study of galaxy morphologies in @xmath75 clusters , was provided to each classifier . we found there was generally good agreement on the morphological classification of the individual galaxies among the five classifiers : majority agreement ( i.e. @xmath76% of the votes cast by the classifiers were for a particular morphology ) was reached for 85% of the sample . agreement by @xmath77 of the classifiers was reached for 67% of the sample . we estimated the rms scatter between the morphological fractions determined by each classifier and adopted this as our estimate of the morphological classification error , used in [ s_resultsmorphologies ] below . for the elliptical galaxy fraction , we measure a scatter of @xmath78 ; for the s0 galaxy fraction , we obtain @xmath79 ; and for the late type galaxy fraction ( sp+irr ) , we obtain @xmath80 . the uncertainty in the early type galaxy fraction ( e+s0 ) is @xmath81 . these uncertainties are comparable in size to those obtained by @xcite . note that in this study we have not taken into account the potential impact of surface brightness dimming , the effect of which can impair the ability to distinguish between , e.g. ellipticals / s0s , and s0/sa galaxies . @xcite conducted simulations in order to quantify the size of this effect in their acs study of the morphology density relation in @xmath20 clusters , and found that the effect of surface brightness dimming was successfully mitigated by their chosen exposure times ( typically @xmath82 sec in @xmath39 ) . we therefore expect that the impact of surface brightness dimming on our results is alleviated to some extent by our long exposure time , @xmath83% longer than the typical exposure time used by @xcite , and by the broad morphological bins we have adopted . secure spectroscopic redshifts have been obtained for a total of 58 galaxies within the @xmath84 field covered by the acs and moircs observations . the majority of these redshifts come from observations of j@xmath0 using the deep imaging multi - object spectrograph ( deimos ; * ? ? ? * ) on the 10 m keck ii telescope , and the focal reducer and low dispersion spectrograph ( fors2 ; * ? ? ? * ) on the 8 m very large telescope ( vlt ) , and are described in @xcite . observations through a further three deimos slit masks have been obtained subsequent to this work ; one being observed in september 2007 , and two in september 2008 . details of the 2007 and 2008 keck observations will be presented in s. a. stanford et al . ( 2009 , in preparation ) . of the 58 galaxies with spectroscopic redshifts , 24 are cluster members , and so we chose to use the available four band ( @xmath38 , @xmath39 , @xmath40 , @xmath41 ) photometry to determine cluster membership using photometric redshifts , for all galaxies in the @xmath41-selected catalog described in [ s_analysisphotometry ] above . we used the spectral template fitting code ` eazy ` @xcite to estimate photometric redshifts . ` eazy ` has been designed to be especially useful in cases where spectroscopic redshifts are only available for a biased subset of galaxies . this is the case for our dataset , as color , magnitude criteria were used to preferentially select likely cluster members as spectroscopic targets . furthermore , the limited size of our spectroscopic subsample makes the use of ` training set'-type techniques inappropriate , such as , e.g. neural - network based redshift estimates @xcite , or iterative correction of empirical spectral templates @xcite . we used the default set of five spectral templates included with ` eazy ` , which is constructed by applying the non - negative matrix factorization technique of @xcite to a set of 3,000 stellar population models from the pegase library @xcite , and a catalog of synthetic galaxy photometry derived from the semi - analytic simulation of @xcite . we chose to use the option within ` eazy ` to fit a linear combination of all of the templates to our galaxy catalog , over a redshift range of @xmath85 , using the @xmath59 aperture magnitudes as the input photometry . finally , we chose to use the @xmath86magnitude based bayesian redshift prior included with ` eazy ` , which is derived from the synthetic , semi - analytic model galaxy photometry used to define the template set , and adopted the maximum likelihood redshift estimate after the application of this prior as the redshift estimate @xmath87 for each galaxy . we checked the accuracy of the photometric redshift estimates produced by ` eazy ` by comparison with the spectroscopic subsample . given the wavelength coverage of our available photometric bands , we would not expect to obtain reliable photometric redshifts for galaxies with redshifts below @xmath13 , as photometric redshift techniques rely on identifying strong spectral breaks in order to assign redshift estimates , and the @xmath38-band only drops below 4000 in the rest frame at @xmath88 . we therefore performed the check on the photometric redshift accuracy on the 36 galaxies with spectroscopic redshifts @xmath89 . we found that the typical scatter @xmath90 in the photometric redshift residuals @xmath91 was @xmath92 , where @xmath90 was estimated using a biweight scale estimator ( e.g. * ? ? ? * ) , which is robust to outliers . the photometric redshifts appear to be almost unbiased in this redshift range , as the median @xmath93 ; however , a small number of objects have overestimated photometric redshifts , as can be seen from the negative tail of the distribution shown in fig . [ f_zpzshistogram ] . a cut of @xmath94 was used to define cluster membership , where @xmath95 is the cluster redshift of @xmath1 - i.e. galaxies with @xmath96 were considered to be members of the cluster . we used the spectroscopic subsample to assess the amount of contamination from galaxies with spurious photometric redshifts ( i.e. with @xmath97 , @xmath98 ) ; contamination from interlopers , defined as galaxies with velocities outside of the range @xmath99 around the cluster redshift , where @xmath100 is the line of sight velocity dispersion measured in @xcite ( i.e. @xmath101 or @xmath102 , but @xmath103 ) ; and missing spectroscopically identified members ( i.e. with @xmath104 , @xmath105 and @xmath106 ) . initially , using only the cut in @xmath87 to define membership , a total of 13 galaxies ( 22% of the spectroscopic subsample ) were identified as having spurious photometric redshifts , all with @xmath107 . however , galaxies with unreliable photometric redshift estimates have flat or multi - modal posterior redshift probability distributions , and can be easily identified by their @xmath108 value ( introduced by * ? ? ? * and labelled ` odds ` in the output from ` eazy ` ) , which is the fraction of the integrated probability that lies within @xmath109 of the photometric redshift estimate @xmath87 ( in the case of ` eazy ` , @xmath110 ) . galaxies with lower @xmath108 values have broader redshift probability distributions , and correspondingly less reliable photometric redshift estimates . applying a cut of @xmath111 reduces the number of spurious photometric redshifts in the spectroscopic subsample to two galaxies ( 3.4% of the spectroscopic subsample ) . we therefore chose to use this cut in @xmath108 in all that follows . a more conservative cut of @xmath112 reduces the number of galaxies with spurious photometric redshifts to just a single galaxy , but this comes at the cost of a smaller sample size and increased error bars on the derived color magnitude relation parameters ( see [ s_resultscmrs ] ) . the number of spectroscopically confirmed non cluster members contaminating the photometrically selected sample was found to be three ( 10.3% of the sample ) . however , four spectroscopically confirmed cluster members were missed by the photometric redshift selection ( 16.7% of the sample of 24 confirmed cluster members ) . using a broader cut in @xmath87 than we have applied naturally reduces the fraction of missed spectroscopic members , but at the cost of significantly increasing the contamination from spectroscopically confirmed interlopers . we show in [ s_resultsmorphologies ] and [ s_resultscmrs ] that our measurements are robust to reasonable changes of the parameters governing the photometric redshift selection . we further refined our photometric membership selection by excluding from the sample all galaxies with @xmath41-band magnitudes brighter than the bcg , i.e. @xmath113 . although j@xmath0 lacks an obvious bcg , it possesses several candidates that are of similar brightness ( see * ? ? ? * ) , and this additional cut helps by removing seven bright , blue ( @xmath114 ) , lower redshift galaxies from the sample . these objects are probably at @xmath115 , as indicated from their photometric redshifts , and one of these objects is a known interloper with spectroscopic redshift @xmath116 . finally , we removed from the sample all known interlopers with spectroscopic redshifts , and added back in the four spectroscopically confirmed members missed by the photometric redshift selection . the final sample of photometrically selected members within the acs / moircs imaging area contains 64 galaxies . the colors , magnitudes , spectroscopic and photometric redshifts , and morphologies of the sample are recorded in table [ t_galaxyphotometry ] . fig . [ f_morphologies ] shows postage stamp images of all the galaxies with morphological classifications ( i.e. , all galaxies with @xmath70 ) , located within 0.75 mpc of the cluster x - ray position , and selected as cluster members using the photometric criteria described in [ s_photozs ] above . in addition to the @xmath39-band data from which the morphology of each object was determined , we also show a corresponding postage stamp taken from the @xmath41-band image that was used to perform the object detection ( see [ s_analysisphotometry ] ) . only four objects in this sample were found to be blended in the @xmath41-selected catalog - these are ids 824 , 610 , 696 and 1047 , and are noted with a pair of morphologies in fig . [ f_morphologies ] and table . [ t_galaxyphotometry ] . within a radius of 0.5 mpc of the cluster x - ray position , there are a total of 39 galaxies with morphological classifications . the fraction of elliptical galaxies in this subsample is @xmath117 ; the fraction of lenticular galaxies is @xmath118 ; and the fraction of late type galaxies is @xmath119 . the combined fraction of early type galaxies is @xmath120 . in calculating the errors on the morphological fractions quoted above , we add in quadrature the classification error ( see [ s_analysismorphologies ] above ) , the poisson uncertainty , and the uncertainty in cluster membership due to the photometric redshift selection . to account for this latter effect , we perform 1000 monte - carlo simulations in which the photometric redshift of every galaxy is replaced by a random variate drawn from its photometric redshift probability distribution , as output by ` eazy ` , and apply the algorithm described in [ s_photozs ] to select cluster members . we find that the dominant source of error is the classification error ; the photometric redshift selection increases the size of the uncertainties by at most @xmath121 4% for a given morphological bin . to further check that the results are robust to the choice of parameters used to define the photometric redshift selection , we also calculated the morphological fractions for different photometric redshift cuts . for a more conservative choice of @xmath112 , we obtain consistent results of @xmath122 ; @xmath123 ; and @xmath124 , for a sample of 34 galaxies within a radius of 0.5 mpc . similarly , for a more stringent photometric redshift cut of @xmath125 with @xmath111 , we obtain @xmath126 ; @xmath123 ; and @xmath127 , for a sample of 31 galaxies within the same selection radius . considering only the 19 galaxies with spectroscopic redshifts located within 0.5 mpc of the cluster x - ray position , we obtain results consistent with those above , although with larger uncertainties . for this sample we find @xmath128 ; @xmath129 ; and @xmath130 . given the large uncertainties , we are unable to examine the variation of the morphological fraction with radius . for example , extending the radius within which the morphological fractions are calculated to 0.75 mpc does not change our results , as it only increases the sample size by a further three galaxies . the object identified as the brightest cluster galaxy ( bcg ) in j@xmath0 , i d 688 , has a clear s0 morphology as seen in fig . [ f_morphologies ] . it is the brightest cluster member in the @xmath41-band ( although there are in fact several other candidates of similar brightness ) , and has a spectrum consistent with a luminous red galaxy ( lrg ) template ( see fig . 1 of * ? ? . however , it lies @xmath131 kpc away from the cluster x - ray position . although this would be unusual for clusters at low redshift , we note that the two brightest members of the @xmath132 cluster rdcs j@xmath133 also have s0 morphologies , and furthermore , are located at clustercentric distances @xmath134 kpc @xcite . therefore it is possible that j@xmath0 is not untypical of clusters at this epoch in this regard , given the small number of objects currently known at @xmath14 . a comparison of the stellar masses of bcgs in a sample of @xmath135 clusters , including j@xmath0 , with the latest semi - analytic models of galaxy formation is presented in @xcite . we present the @xmath6 and @xmath42 color magnitude diagrams ( cmds ) with fitted color magnitude relations ( [ s_resultscmrs ] , below ) in fig . [ f_zjcmd ] and fig . [ f_zkcmd ] respectively . panel ( a ) of each figure shows all galaxies in the combined acs / moircs imaging area ; panel ( b ) shows only members selected according to the photometric criteria described in [ s_photozs ] above . in each of these plots we show the approximate @xmath67 limiting magnitude for galaxies as a blue vertical dashed line ( estimated from the size of the photometric errors ) ; the corresponding @xmath67-limit in color as a diagonal dashed blue line ; and the approximate limit of the morphological classification caused by the @xmath136 limit ( see [ s_analysismorphologies ] above ) as a diagonal dashed green line . along the top axis of each plot , we show the magnitude relative to the expected characteristic magnitude ( @xmath22 ; labelled as @xmath137 or @xmath138 appropriately for each band in the plots ) of the galaxy luminosity function at @xmath1 . the value of @xmath22 was estimated by passively evolving the value of @xmath139 found for clusters at @xmath140 by @xcite to @xmath1 , transformed to the passbands used in this paper . the evolution correction was performed by adopting a @xcite single burst of star formation model , beginning at redshift @xmath141 , decaying exponentially with characteristic time scale @xmath142 gyr , with solar metallicity , and @xcite initial mass function ( imf ) . this model acceptably reproduces the observed evolution of @xmath139 over the redshift range @xmath143 for the data in @xcite . it is immediately apparent from figs . [ f_zjcmd ] and [ f_zkcmd ] that the bright end of the color magnitude relation in j@xmath0 is unpopulated : the brightest galaxies have @xmath144 , approximately as bright as the expected value of @xmath22 . in contrast , the only other @xmath145 x - ray selected cluster currently known , xmmu j@xmath24 at @xmath25 , has four spectroscopically confirmed members with @xmath41 magnitudes in the range @xmath146 , within a 0.2 mpc radius of the cluster center @xcite . this corresponds to a difference in absolute magnitude of @xmath147 mag between the bcgs ( see also * ? ? ? * ) , neglecting the correction for passive evolution , as the difference in lookback time between the two clusters is only 0.2 gyr . given the small number of systems observed to date at such high redshifts , it is unclear which of these objects is more typical of the general cluster population at this epoch . we note that the brightest galaxies in the compact cluster associated with the @xmath148 galaxy gdds-12 - 5859 have similar @xmath149-band magnitudes to the brightest members of j@xmath0 @xcite . however , further spectroscopic observations of this object are required to confirm its nature as a gravitationally bound system at @xmath150 . inspection of figs . [ f_zjcmd ] and [ f_zkcmd ] shows that limiting the cmr fitting to a purely morphologically selected sample of e+s0 galaxies would restrict the total magnitude range in comparison to the magnitude limit of the complete sample . we therefore chose to perform the analysis on two samples - a morphologically selected sample of e+s0 galaxies ( hereafter referred to as the ` morphologically selected sample ' ) , and a sample defined without this restriction , spanning a larger magnitude range ( hereafter referred to as the ` photometrically selected sample ' ) . in both cases , we only include galaxies that pass the photometric membership selection criteria described in [ s_photozs ] above , to reduce the contamination from non - cluster member galaxies significantly . for the photometrically selected sample , we applied an additional color cut of @xmath151 , @xmath152 as appropriate for the cmr being fitted , in order to avoid biasing the fits by the inclusion of a handful of faint galaxies much bluer than the red sequence . an additional benefit of performing the analysis on a sample not selected on the basis of morphology is to examine the sensitivity of the results to the morphological classification . systematic misclassification of some galaxies , for example face - on s0s as ellipticals , or flattened late type galaxies as s0s , would increase the apparent scatter about the cmr and result in the inference of younger galaxy ages ( see , e.g. , * ? ? ? the color magnitude relations were fitted to the color magnitude data for each sample , defined in [ s_resultscmds ] above , using a robust biweight linear least squares method . we estimated the errors on the fits in a monte - carlo fashion : we generated 1,000 realizations of the data , replacing the color of each galaxy with a random variate , assuming that the color errors have a gaussian distribution , and fitted the cmr for each of these simulated datasets . the 1@xmath153 errors in the slope and intercept of the measured cmr were taken to be the standard deviations of the slopes and intercepts measured for the 1,000 simulated datasets . for the morphologically selected sample of early type ( e+s0 ) galaxies , we obtained the following relations : @xmath154 @xmath155 for the photometrically selected sample , we obtained similar results : @xmath156 @xmath157 both the slopes and intercepts measured for the morphologically and photometrically selected samples agree at better than @xmath158 for both the fits to the @xmath6 and @xmath42 cmrs , with the agreement being slightly better between the two samples in @xmath6 . the uncertainties on the cmr slopes and intercepts are relatively large , due to the large scatter about the cmrs that can be seen in figs . [ f_zjcmd ] and [ f_zkcmd ] . as figs . [ f_zjcmd ] and [ f_zkcmd ] show , there are only ten spectroscopically confirmed members with early type morphologies located within 0.5 mpc of the cluster x - ray position with @xmath159 . we found that fitting the cmr for this purely spectroscopically selected sample gave significantly worse constraints on the cmr than the samples supplemented with photometric redshifts ; for example , the slope of the @xmath6 cmr is found to be @xmath160 . this is driven by one object ( i d 526 ) , which has a clear elliptical morphology ( see fig . [ f_morphologies ] ) , but is significantly bluer than the other spectroscopically selected early type galaxies . for this reason , we do not consider the purely spectroscopically selected galaxy sample further in this paper . the observed scatter about the cmr was measured using a biweight scale estimate of the color residuals , with the error being estimated from 1,000 bootstrap samples . the intrinsic scatter @xmath161 in the cmr was estimated by subtracting in quadrature the rms of the color errors of the galaxy sample to which the cmr was fitted . for the morphologically selected sample of early type ( e+s0 ) galaxies , we measured @xmath162 , and @xmath163 . for the photometrically selected sample , we found @xmath164 , and @xmath165 . the measurements of intrinsic scatter are therefore in excellent agreement , at better than @xmath166 , between the morphologically and photometrically selected samples . this suggests that systematic morphological classification errors of the type described in [ s_resultscmds ] are not a problem at the level of precision achieveable with our galaxy sample . we examined the dependence of our results on the photometric redshift selection in a number of ways . adopting a conservative cut of @xmath112 , we found no significant differences with the results reported above : the differences between the cmr slope and intercept values in each band and for each sample are significant at the @xmath167 level . restricting the photometric redshift selection to only galaxies with @xmath168 , with @xmath111 , we find likewise . in all cases the measurements of intrinsic scatter are robust to the changes in the cmr fit parameters ; we find that the differences across all samples and in all bands are significant at the @xmath169 level . as a final check on the robustness of the results to the photometric redshift selection , we performed 500 monte - carlo simulations in which , in addition to the photometry of each object in the catalog being randomized appropriately according to the size of the photometric errors , the photometric redshifts were also re - estimated using ` eazy ` , and the algorithm used to select cluster members described in [ s_photozs ] was applied . for each of these simulations , the cmr fit parameters and scatter measurements were performed in the manner described above . again , we found no significant differences from the above results , with the mean slope and intercept of the monte - carlo simulations being within @xmath158 of the results for both the photometrically and morphologically selected samples in both @xmath6 and @xmath42 , and the internal scatter measurements being in agreement across all samples at better than the @xmath169 level . table [ t_fitresults ] presents a summary of the measured fit parameters for the @xmath6 and @xmath42 cmrs , for both the morphologically and photometrically selected samples . the intrinsic scatter about the cmr can be used to estimate the major epoch of star formation for early type galaxies , given the assumption of a particular stellar population model and that the cmr is predominantly a sequence in metallicity and not age ( e.g. * ? ? ? this latter assumption was proven to be sound by observations of the cmr in distant clusters ( e.g. * ? ? ? * ) . in actual fact , recent observations have shown that the cmr is primarily a sequence in stellar mass , in that the brightest galaxies tend to be more massive , metal rich and older than the galaxies at the faint end of the cmr , with the scatter in the cmr at the faint end being primarily due to differences in age ( e.g. * ? ? ? we modelled the cmr scatter in the manner described by ( * ? ? ? * see also @xcite ) . we assumed as our baseline galaxy model a composite stellar population ( csp ) of solar metallicity , with @xcite initial mass function ( imf ) , formed in a single burst of star formation , decaying exponentially with characteristic time scale @xmath142 gyr . to check the dependence of our results on the stellar models , we used csps derived from ( * ? ? ? * hereafter bc03 , using the padova 1994 stellar tracks ) , and ( * ? ? ? * hereafter m05 ) . one major difference between these two families of models is that the treatment of thermally pulsing asymptotic giant branch ( tp - agb ) stars is different , with the m05 models being shown to infer younger ages for galaxies in the hubble ultra deep field ( udf ) in comparison to the bc03 models @xcite . we estimate the expected cmr scatter in the following way for each combination of csp model and photometric bands . we assume that galaxies are born in a single burst of star formation at a time @xmath170 between the epoch of recombination and time @xmath171 , where the value of @xmath171 is varied in steps between zero and the epoch at which the cluster is observed ( i.e. , up to a maximum of @xmath172 gyr in our adopted cosmology ) . for each value of @xmath171 , we construct a sample of 10,000 simulated galaxies with ages assigned randomly from a uniform distribution with range @xmath173 , and estimate the scatter in the color distributions using a biweight scale estimate . the value of @xmath171 corresponding to the observed intrinsic cmr scatter is then used to infer estimates of the minimum and mean luminosity weighted galaxy age assuming a particular csp . [ f_scatterfits ] shows the expected intrinsic scatter @xmath161 around the cmr for the bc03 ( red ) and m05 ( blue ) models as a function of minimum age ( dashed lines ) or mean luminosity weighted age ( solid lines ) for the morphologically selected sample . adopting the bc03 model , the measured intrinsic scatter around the @xmath6 cmr indicates that the minimum age of the stellar populations of the red sequence galaxies in j@xmath0 is @xmath174 gyr ( corresponding to a minimum redshift of formation of @xmath175 ) . the corresponding mean luminosity weighted age is @xmath176 gyr ( i.e. , @xmath177 ) . the ages inferred from the m05 model are younger ; the minimum age being @xmath178 gyr ( @xmath179 ) , and the mean luminosity weighted age being @xmath180 gyr ( @xmath181 ) . however , the difference between the results for the two different models is not statistically significant . in @xmath42 , the minimum age is @xmath182 gyr ( @xmath183 ) , and the mean luminosity weighted age is @xmath184 gyr ( @xmath185 ) , assuming the bc03 model this is in excellent agreement with the age inferred from the scatter about the @xmath6 cmr , though of course the colors are not completely independent . for the m05 model , we again find lower ages : the minimum age is @xmath186 gyr ( @xmath187 ) , and the mean luminosity weighted age is @xmath188 gyr ( @xmath189 ) . it is interesting to note that the measured scatter about the @xmath42 cmr is near the limit of the allowed scatter in the m05 model ( see fig . [ f_scatterfits ] ) , although the uncertainty in the measurement is large . in fact , in the case of the photometrically selected sample , the measured scatter is larger than the maximum scatter expected in the m05 model . this could be a consequence of the greater weight given to tp - agb stars in the m05 models relative to the bc03 models , which begins to have a significant effect at the @xmath190-band and redder wavelengths ( see fig . 18 of * ? ? ? the observed @xmath41-band samples the rest frame @xmath191-band at the redshift of the cluster . as the measurements of the intrinsic scatter in @xmath6 and @xmath42 are consistent between the morphologically and photometrically selected samples ( see [ s_resultscmrs ] above ) , the inferred ages derived from the photometrically selected sample are consistent with those quoted above ; however , the uncertainties are typically @xmath192% smaller due to the increased sample size . we provide a summary of these results in table [ t_ages ] . some studies of clusters at high redshift have revealed a deficit in the number of faint red sequence galaxies in comparison to observations of clusters at low redshift ( e.g. * ? ? ? * ; * ? ? ? * ; * ? ? ? this has been attributed to ` downsizing ' , where star formation is observed to have terminated at earlier epochs in more massive galaxies @xcite . in the case of j@xmath0 , [ f_zjcmd ] and [ f_zkcmd ] show an apparent paucity of galaxies on the red sequence at faint magnitudes ( @xmath193 ) , which may be indicative of this phenomenon . we quantify this deficit using the dwarf - to - giant ratio ( dgr ) . following @xcite , we define giants as galaxies with absolute @xmath194-band magnitudes brighter than @xmath195 , and dwarfs as galaxies fainter than this limit , but brighter than @xmath196 . these limits correspond to values at @xmath197 after correction for the effect of passive evolution . in performing the passive evolution correction , we adopt the same @xcite simple stellar population model as used by @xcite , with solar metallicity , @xcite imf , and formation redshift @xmath198 . the dividing magnitude between dwarfs and giants is equivalent to @xmath199 in the observed frame of j@xmath0 after this correction is applied , and the corresponding faint magnitude limit is @xmath200 ( see [ s_discussionu - vscatter ] below for details of the adopted conversion between @xmath194 and @xmath40 magnitudes ) . thus , the faint magnitude limit is @xmath201 mag . fainter than the 5@xmath153 limiting magnitude of our observations , and so the limits that we obtain on the dgr may be slightly underestimated in comparison to @xcite . we estimated the dgr using the sample of photometrically selected cluster members defined in [ s_photozs ] and listed in table [ t_galaxyphotometry ] . following @xcite , using all of the members within @xmath202 ( @xmath203 mpc for j@xmath0 ; @xcite ) , we find that the @xmath204 . in estimating the dgr of galaxies on the red sequence , we only consider the @xmath6 cmr , as unlike the @xmath42 cmr , it is well suited to comparison with other studies at low redshift . applying a color cut of @xmath205 mag . around the fit to the @xmath6 cmr derived from the morphologically selected sample , we find that the red sequence @xmath206 . the uncertainties are estimated assuming poissonian errors , added in quadrature to the uncertainty due to the photometric redshift selection , calculated using the full ` eazy ` photometric redshift probability distribution for each galaxy by the same technique described in [ s_resultsmorphologies ] above . the uncertainty due to the photometric redshift selection typically increases the size of the error bars on the dgr measurements by @xmath207% . we checked the sensitivity of these results to the parameters governing the photometric redshift selection . restricting the membership selection to galaxies with @xmath208 or @xmath125 does not affect these results significantly , the dgr measurements in each case being well within @xmath169 of the above results . extending the selection radius to @xmath209 mpc does not change these results significantly : we estimate that the @xmath210 if all the photometrically selected members are included ; if we include only those members within @xmath211 mag . of the red sequence , then we find that the @xmath212 . the morphological fractions we find for j@xmath0 are in good agreement with studies of clusters at @xmath2 , some @xmath147 gyr later than the epoch at which our target cluster is observed . @xcite studied the morphology density relation in a sample of six clusters at @xmath2 . assuming that @xmath213 is roughly equivalent to the magnitude limit of @xmath214 adopted by @xcite , as at the redshift of j@xmath0 the @xmath40-band samples the rest frame @xmath194-band , then the depth of our study is sufficient to make a comparison , as all the galaxies in our sample brighter than this limit have morphological classifications ( see fig . [ f_zjcmd ] ) . as @xcite quote their results in bins of local galaxy surface density @xmath215 , we need to estimate the median local galaxy surface density for our sample in order to compare results . following @xcite , we calculate @xmath215 for each galaxy by counting the 10 nearest neighbors with @xmath216 , and dividing by the rectangular area enclosed . we find the median local galaxy surface density within a radius of 0.5 mpc of the cluster x - ray position is @xmath217 mpc@xmath218 . in their second highest density bin with @xmath219 mpc@xmath218 , @xcite find @xmath220 , in good agreement with the results of our study . a study of the morphology density relation at @xmath2 was also conducted by the acs guaranteed time observations ( gto ) team @xcite , who quote morphological fractions within @xmath221 for the individual objects that make up their sample . the magnitude limit of the @xcite study is @xmath70 , as adopted in this work . for j@xmath0 , the @xmath221 radius is @xmath222 mpc @xcite , i.e. approximately the same as the 0.5 mpc radius in which we have calculated the morphological fractions . the mean morphological fractions within @xmath221 for the @xcite sample ( @xmath54 the standard error on the mean ) are @xmath223 ; @xmath224 ; @xmath225 ; and @xmath226 . the results we obtain for j@xmath0 are in agreement at the @xmath166 level for all morphological bins . a larger cluster sample would be needed to search for possible evolutionary effects between @xmath227 and @xmath2 . the mean redshift we find for the major epoch of star formation in the early type galaxies in j@xmath0 of @xmath10 ( see [ s_discussionscatter ] ) is similar to that found for elliptical galaxies in other high redshift clusters . @xcite found a mean formation redshift of @xmath228 for the elliptical galaxies within 0.5 mpc radius of the centers of the two lynx clusters at @xmath229 , from modelling the cmr scatter using single burst bc03 models , in good agreement with our result . similarly , @xcite conducted a study of the @xmath230 cmr in the only other @xmath145 x - ray selected cluster currently known , xmmu j@xmath24 , finding @xmath228 for galaxies , selected using a color cut , within 0.2 mpc of the cluster center . for well studied clusters at lower redshift , slightly smaller formation redshifts have been measured from the scatter about the cmr : @xcite estimated @xmath231 from the @xmath232 e+s0 cmr of rdcs j@xmath233 at @xmath234 , a result also found by @xcite from the @xmath230 cmr ; @xcite found @xmath235 from the cmr of elliptical galaxies in rdcs j@xmath133 at @xmath132 ; and @xcite estimate that the elliptical galaxies in the @xmath236 clusters ms @xmath237 and rx j@xmath238 completed their major epoch of star formation at @xmath239 . all of these studies used bc03 stellar population synthesis models in arriving at their results . to compare our measurements of color magnitude relation parameters with other studies at lower redshift , we derived linear conversions between observed colors and magnitudes and @xmath8 , @xmath194 ( vega ) in the rest frame of the coma cluster , using @xmath240 gyr @xcite models of several metallicities ( @xmath241 , @xmath242 , @xmath243 , @xmath244 ) , with a range of ages ( corresponding to @xmath245 , as appropriate for the early type population , [ s_discussionscatter ] above ) . the method used is the same as that described in appendix ii of @xcite , except that we convert magnitudes to @xmath194 apparent ( vega ) at the distance of the coma cluster . the following transformations were used : @xmath246 @xmath247 note that in the following discussion , we only consider the @xmath6 cmr of j@xmath0 , as the @xmath42 cmr ( roughly equivalent to rest frame @xmath43 ) is a poor match to other studies in the literature . as the cmr fitting parameters for j@xmath0 derived from the morphologically and photometrically selected samples are consistent , we only quote transformed quantities corresponding to the morphologically selected sample below . [ f_u - vscatterevolution ] shows the evolution of the intrinsic scatter @xmath248 about the cmr , as traced by studies of several clusters at different redshifts . some of the studies shown in this plot measured @xmath161 through passbands roughly equivalent to @xmath8 at their respective redshifts , and so the correction to @xmath249 using the models is small . these are the @xcite study of the coma cluster ; the work by @xcite at @xmath250 ; and the two @xmath236 clusters observed by @xcite . at @xmath135 , the acs guaranteed time observations ( gto ) team measured @xmath161 in @xmath232 , which is more closely matched to rest frame @xmath251 , and so the transformation of these results to @xmath249 is larger and more uncertain . both the @xcite study of rdcs j@xmath233 at @xmath234 , and the @xcite study of the lynx clusters at @xmath229 , show similar values of @xmath161 to those measured at @xmath252 by @xcite , and @xmath236 by @xcite , although the uncertainties are large . the general trend in fig . [ f_u - vscatterevolution ] is of increasing scatter about the cmr as redshift increases , as expected if the stellar populations were formed at much higher redshifts than observed and have evolved passively thereafter . the intrinsic scatter about the red sequence observed in j@xmath0 at @xmath1 is @xmath253 , more than three times as large as that observed in the coma cluster @xcite , some @xmath7 gyr earlier , and one and a half times that observed in ms @xmath237 and rx j@xmath238 at @xmath236 , @xmath254 gyr earlier . fig . [ f_u - vslopeevolution ] shows values of the absolute @xmath249 cmr slope for the same comparison sample of clusters from the literature . in contrast to the measurements of intrinsic scatter about the cmr , we see no evidence for evolution in the slope over the last @xmath7 gyr . for j@xmath0 , the value of the cmr slope is @xmath255 , after conversion to @xmath249 . this agrees with the slope measured for the coma cluster @xcite at better than the @xmath169 level , though the uncertainty is large . the lack of evolution of the cmr slope is consistent with the cmr remaining as primarily a sequence in metallicity up to the highest redshifts so far observed . to compare the evolution of the cmr intercept , we used the cmr of coma as measured by @xcite to calibrate the cmr as a metallicity sequence , and transformed the cmrs of j@xmath0 and the literature comparison sample , such that the intercept was evaluated at the @xmath194-band magnitude corresponding to solar metallicity . the metallicity calibration of the cmr of coma was performed by calculating @xmath194-band ( vega ) magnitudes corresponding to the @xmath8 colors of @xmath256 gyr bc03 models of several metallicities , assuming the slope and intercept of the coma cmr as measured by @xcite , and formation redshift @xmath257 @xcite . under these assumptions , solar metallicity in the coma cmr corresponds to @xmath258 . we calculated the corresponding @xmath194-band magnitudes and transformed cmr intercepts for the other clusters in the comparison sample assuming passive evolution of this model . errors in the transformed values were estimated using a monte - carlo method , taking into account the uncertainties in the measured cmrs , and the color and magnitude transformations to @xmath259 , @xmath194 ( vega ) . [ f_u - vsolarmetallicityevolution ] shows the results of this exercise , with the expected passive evolution track of the baseline stellar population model overplotted , for several different formation redshifts . the cmr intercept evaluated at solar metallicity for j@xmath0 is consistent with @xmath260 , which is slightly lower than the formation redshift inferred from the intrinsic scatter ( [ s_discussionscatter ] , above ) , but in good agreement given the large uncertainties . this is consistent with the literature comparison sample : there is some scatter , which could indicate a range of star formation histories for the galaxy population in different clusters , but the uncertainties are large . photometric calibration uncertainties may also contribute to the scatter ; see , for example , the discussion in the @xcite study of the lynx clusters at @xmath229 . we show the effect of changing the stellar population model from bc03 to m05 on the expected passive evolution of the colors of solar metallicity galaxies in fig . [ f_u - vsolarmetallicityevolutionmaraston ] . lower formation redshifts are expected if this family of models is assumed , with some of the literature comparison sample being expected to have formed most of their stars at @xmath261 . in the case of j@xmath0 , [ f_u - vsolarmetallicityevolutionmaraston ] implies @xmath262 . again , this is in good agreement with the age derived from the intrinsic scatter about the cmr for the m05 models . the dwarf - to - giant ratio measured for j@xmath0 appears to be significantly lower than has been measured in clusters at lower redshifts . @xcite quote a red sequence luminous - to - faint ratio within @xmath263 for the coma cluster equivalent to a dgr of @xmath264 , nearly ten times the value of the red sequence dgr in j@xmath0 ( @xmath11 ) . although the dgr we measure is likely to be slightly underestimated , given that our magnitude limit is @xmath201 mag brighter than the faint magnitude limit adopted by @xcite , it seems very unlikely that this would be enough to account for the deficit of faint red sequence galaxies relative to coma . several studies at lower redshift have parametrized the evolution of the red sequence dgr with redshift . recently , @xcite compiled several measurements of the dgr in clusters at @xmath265 , including studies by @xcite ; @xcite ; and @xcite , transformed on to a common system defined by @xcite , as we have adopted in performing our estimate of the dgr ( described in [ s_resultsdgr ] above ) . @xcite found that the evolution of the dgr with redshift of their composite sample is @xmath266 . at the redshift of j@xmath0 , the expected @xmath267 from this relation , in good agreement with our estimate of the dgr . we conclude that j@xmath0 appears to show a significant deficit of faint red sequence galaxies compared to clusters in the local universe , and that the measured dgr is of similar magnitude to that expected from extrapolation of the dgr - redshift relation measured at @xmath13 . however , it is known that the cluster - to - cluster scatter of the dgr is considerable @xcite , and study of a larger sample of clusters , ideally with large numbers of spectroscopically confirmed members , at similar redshift to j@xmath0 is required in order to confirm whether or not this holds for the general population at this epoch . we have conducted a study of the morphological fractions and color magnitude relation in the galaxy cluster xmmxcs j@xmath0 at @xmath1 . this is the first such study of an x - ray selected cluster at @xmath268 . we found : 1 . the brightest members of j@xmath0 have @xmath41 magnitudes corresponding to the expected value of @xmath269 in the galaxy luminosity function . this is significantly fainter than the brightest galaxies in the only other @xmath145 x - ray selected cluster studied to date , xmmu j@xmath24 at @xmath25 @xcite . the morphological fractions are @xmath270 , @xmath118 , @xmath119 , similar to other clusters at @xmath2 . thus , the dominant component of the galaxy population observed in clusters at low redshift was already in place @xmath5 gyr after the big bang . 3 . after transformation from @xmath6 to @xmath249 , the slope of the color magnitude relation is consistent with that of the coma cluster , implying little evolution over the last @xmath7 gyr , though the measurement uncertainty is large . in contrast , the intrinsic scatter about the color magnitude relation is more than three times the value measured in the coma cluster , after conversion from @xmath6 to @xmath249 . 4 . from comparison with stellar population models , the intrinsic scatter about the color magnitude relation implies mean luminosity weighted ages for the stellar populations of the early type galaxies in j@xmath0 of @xmath9 gyr , corresponding to the main epoch of star formation in these galaxies coming to an end at @xmath10 . comparison of the intercept of the color magnitude relation with passive evolution of the same stellar population models , calibrated relative to the coma cluster , yields consistent results . 5 . j@xmath0 shows evidence of the ` downsizing ' phenomenon : the red sequence dwarf - to - giant ratio for the cluster is @xmath11 within a radius of @xmath12 , although this is likely to be underestimated slightly in comparison to other studies due to the depth of our photometry . this is consistent with extrapolation of the redshift evolution of the dgr measured from cluster samples at @xmath13 within the large uncertainties . we thank the referee for a number of suggestions that improved the clarity of this paper . this work is based in part on data collected at the subaru telescope , which is operated by the national observatory of japan , and _ xmm_-newton , an esa science mission funded by contributions from esa member states and from nasa . we thank ichi tanaka for the development of the ` mcsred ` package used to reduce the moircs data . we acknowledge financial support from the south african national research foundation , the uk science and technology facilities council , and the university of sussex physics & astronomy department . financial support for this work was also provided by nasa through program go-10496 from the space telescope science institute , which is operated by aura , inc . , under nasa contract nas 5 - 26555 . this work was also supported in part by the director , office of science , office of high energy and nuclear physics , of the u.s . department of energy under contract no . ac02 - 05ch11231 , as well as a jsps core - to - core program `` international research network for dark energy '' and by jsps research grant 20040003 . this work was performed under the auspices of the u.s . department of energy by lawrence livermore national laboratory in part under contract w-7405-eng-48 and in part under contract de - ac52 - 07na27344 . 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 fortunate to have the opportunity to conduct observations from this mountain . cccccccc @xmath6 & 22.5 & 19 & @xmath271 & @xmath272 & @xmath273 & @xmath274 & @xmath275 + @xmath42 & 21.5 & 16 & @xmath276 & @xmath277 & @xmath278 & @xmath279 & @xmath280 + @xmath6 & 24.0 & 34 & @xmath281 & @xmath282 & @xmath283 & @xmath284 & @xmath285 + @xmath42 & 23.0 & 36 & @xmath286 & @xmath287 & @xmath288 & @xmath289 & @xmath290 + cccccc @xmath6 & bc03 & @xmath174 & @xmath291 & @xmath292 & @xmath293 + @xmath6 & m05 & @xmath178 & @xmath294 & @xmath295 & @xmath296 + @xmath42 & bc03 & @xmath182 & @xmath297 & @xmath298 & @xmath299 + @xmath42 & m05 & @xmath186 & @xmath300 & @xmath301 & @xmath302 + @xmath6 & bc03 & @xmath303 & @xmath304 & @xmath305 & @xmath306 + @xmath6 & m05 & @xmath307 & @xmath308 & @xmath309 & @xmath310 + @xmath42 & bc03 & @xmath311 & @xmath312 & @xmath313 & @xmath314 + @xmath42 & m05 & & & & +

we present a new sample of eclipsing binary stars in the large magellanic cloud ( lmc ) , expanding on a previous sample of 611 objects and a new sample of eclipsing binary stars in the small magellanic cloud ( smc ) , that were identified in the light curve database of the macho project . we perform a cross correlation with the ogle - ii lmc sample , finding 1236 matches . a cross correlation with the ogle - ii smc sample finds 698 matches . we then compare the lmc subsamples corresponding to center and the periphery of the lmc and find only minor differences between the two populations . these samples are sufficiently large and complete that statistical studies of the binary star populations are possible .

0711.1617

eclipsing binary stars ( ebs ) are important for astrophysical research in many ways . they may be used to obtain accurate estimates of star masses and radii ( * ? ? ? * and references therein ) . precise determination of stellar parameters can in turn be used to put theories of stellar structure and evolution to a stringent test by comparing measured parameters with theoretical predictions ( * ? ? ? * ; * ? ? ? * and references therein ) . ebs may also be used for distance determination and this use goes back several decades ; its history is reviewed by @xcite . since @xcite used an estimate of the parallax to @xmath2 aurig to infer the surface brightness of both its components , it has been known that a good photometric light curve plus a double line spectroscopic orbit admits a simple geometric relationship between the surface brightnesses of the stars and the distance to the eb ; @xcite however had no way at the time to make the reverse `` surface brightness to distance '' inference and his paper does no mention this possibility . after @xcite , other early papers @xcite used parallaxes obtained independently to estimate surface brightnesses , but , as remarked by @xcite , these pioneers surely knew of the potential of this technique to estimate distances . modern analyses of ebs have usually focussed on this technique ( e.g. @xcite ) . the method affords high precision due to its purely geometrical nature and has been applied by a number of authors to determine the distance to the large magellanic cloud ( lmc ) using hv2274 @xcite , hv982 @xcite , eros 1044 @xcite and hv 5936 @xcite ; an attempt to use ebs to determine the distance to m31 is currently under way @xcite and the direct project is attempting to measure the distance to m31 and m33 via ebs and cepheids @xcite ; other recent examples include @xcite who present a sample of detached binaries in the lmc for distance determination and @xcite who present the first determination of the distance and properties of an eb in m31 ; @xcite presents a sample of ebs with total eclipses in the lmc suitable for spectroscopic studies . in general it is important that distances be determined using a large sample of ebs to minimize the impact of systematic errors . a recent collection of references on extragalactic binaries can be found in @xcite . large - scale surveys to detect gravitational microlensing events have identified and collected light curves for large numbers of variable stars in the bulge of the milky way and in the magellanic clouds . eclipsing binary stars comprise a significant fraction of these collections . the macho collaboration has presented a sample of 611 ebs in the lmc with preliminary analyses of their orbits @xcite . a catalogue of @xmath3 ebs in the lmc found in the macho database has been just published by @xcite ; this catalogue was compiled by analyzing a list of @xmath4 stars classified as possible ebs in the macho database ; a cross correlation between these @xmath4 stars and our sample finds just @xmath5 matches , thus at least about @xmath6 ebs in our catalogue are new identifications . the @xmath4 classified as possible ebs were found in regions of parameter space such as color , magnitude , and period , where one does not expect to find pulsating variables and therefore the detected variability of these stars was _ tentatively _ ascribed to eclipses . regions where pulsating variables could exist were not considered while making this preliminary classification and ebs there were therefore not included in the list . in our search we did not rely primarily on cuts in parameter space and and we did not exclude a priori regions of this space where pulsating variables are present ; therefore we were able to classify many ebs in these regions , that were not included in the preliminary classification . the ogle collaboration ] has introduced a sample of @xmath7 ebs in the lmc @xcite and of @xmath8 ebs in the small magellanic cloud ( smc : * ? ? ? * ) . both samples were selected from their catalogue of variable stars in the magellanic clouds @xcite compiled from observations taken during the second part of the project ( ogle ii : * ? ? ? * ) and reduced via difference image analysis ( dia : * ? ? ? * ) . an earlier sample of @xmath9 ebs in the bar of the lmc was presented by the eros collaboration@xcite . other large variable star data sets are being produced by surveys not specifically designed to detect gravitational microlensing , such as the all sky automated survey ( asas : * ? ? ? ] the availability of large samples of ebs ( and the even larger ones that can be found by future surveys such as pan - starrs and lsst ) can have an important impact on stellar astrophysics . this impact can arise in two qualitatively different approaches . first , a large catalogue allows the discerning researcher to select carefully a few ebs for detailed follow - up study ; the distance estimation described above is an example of this . second , statistical analyses of an entire population become possible when a large collection is assembled ; such analyses of ebs have not previously been possible . to fulfill this promise there are challenges to overcome , including finding ebs in large data sets and automating their analysis . with regard to the first task , the _ discovery _ problem is complicated by the fact that ebs do not have clear relationships between their parameters ( period , luminosity , colors ) as do the major classes of pulsating variables . this makes it difficult to find them via simple and well understood cuts in parameter space . the first step toward automated discovery is thus to have a large sample of data on which to experiment with search techniques . this non - trivial exercise in mining large data sets can be useful for future surveys that are not necessarily aimed at binary star research . an example is given by @xcite and @xcite who employ an artificial neural network to identify ebs in the ogle - ii lmc and smc samples , but more needs to be done . with regard to _ analysis _ of ebs , the traditional approach has been to carefully analyze individual systems with the help of dedicated computer codes such as the wilson - devinney code ( wd : * ? ? ? * ; * ? ? ? this becomes impracticable when many thousands of stars are involved and an automated approach is required . the light curves in a previous sample of 1459 ebs in the smc found by ogle - ii @xcite were systematically solved by @xcite using an automated version of the wd code ; the asas collaboration has developed an automated classification algorithm for variable stars based on fourier decomposition @xcite ; @xcite found and analyzed 10000 bulge ebs from ogle - ii using debil ] , an eb analysis code that allows automated solutions of large eb data sets and works best for detached ebs ; a genetic algorithm based approach to finding good initial parameters for wd is described in @xcite . this paper is the first of a series of papers aimed at describing the eb samples in the macho database and is organized as follows : section [ sec : sample ] introduces the lmc and smc samples ; section [ sec : diagrams ] describes the color magnitude diagram ( cmd ) and the color period diagram , pointing out significant features in them , section [ sec : comparison ] compares the lmc and smc samples , section [ sec : crossogle ] describes the results of the cross correlation with the ogle lmc and smc samples , and section [ sec : online ] reports where and in what form the data presented in the paper can be accessed on line . the macho project was an astronomical survey whose primary aim was to detect gravitational microlensing events of background sources by compact objects in the halo of the milky way . the gravitational background sources were located in the lmc , smc and the bulge of the milky way ; more details on the detection of microlensing events can be found in @xcite and references therein . observations were carried out from july 1992 to december 1999 with the dedicated @xmath10 telescope of mount stromlo , australia , using a @xmath11 mosaic of @xmath12 ccd in two bandpasses simultaneously . these are called macho blue " , hereafter indicated with @xmath13 , with a bandpass of @xmath14 and macho red " , hereafter indicated with @xmath15 , with a bandpass of @xmath16 ; these widths are between the half - response points as estimated from figure @xmath17 of @xcite . the bandpasses and the transformations to standard johnson @xmath18 and cousins @xmath19 bands are described in detail in @xcite ; see in particular their figure @xmath17 for the instrumental throughput of the two macho bands . each macho object is identified by its field number ( @xmath20 for the lmc , @xmath21 for the smc ) , its tile number ( which can overlap more than one field ) , and its sequence number in the tile . these form the so called macho field.tile.sequence ( fts ) , which is used in this paper to label ebs . note that , since some overlap exists between fields , one star may have two or more fts identifiers . the lmc sample we present comprises @xmath0 ebs selected by a variety of methods which we describe in this section ; the sample includes the 611 ebs described in @xcite . the lmc magnitudes quoted in this paper have been obtained by using the following transformation : @xmath22 from now on we will use the symbols @xmath18 , @xmath19 , and @xmath23 to refer to standard magnitudes obtained from @xmath13 and @xmath15 via eq . [ eq : machocal ] for the lmc and eq . [ eq : smcmachocal ] for the smc , and _ not corrected for reddening _ ; we will also use @xmath13 and @xmath15 to indicate instrumental magnitudes . the observations number in several hundreds in both bandpasses for most light curves ; figure [ fig : pointhist ] shows histograms of the number of light curve points of the ebs in both bands ; the @xmath13 band has on average more observations than the @xmath15 band because one half of one of the red ccds was out of commission during part of the project . the central fields of the lmc were observed more often and the periphery less often as shown by the three peaks in the distribution where the first peak corresponds to the lmc periphery and the other two correspond to the center . this section describes the techniques employed to identify ebs both in the lmc and in the smc ; the results we quote are relative to the lmc . from now on we will always use the term _ unfolded light curve _ to indicate a set of time ordered observations and will reserve the term _ light curve _ to indicate a set of time ordered observations _ folded _ ( or _ phased _ ) around a period , omitting for brevity the adjectives `` folded '' and `` phased '' ; we will also use the terms eb , system , and object interchangeably . all sources in the survey were subjected to a test for variability @xcite and a large number of variable sources were identified @xcite . this first test starts by first eliminating the @xmath24 most extreme photometric data points ; the resulting unfolded light curve is fitted to constant brightness and its @xmath25/dof is calculated . the elimination of the most extreme points is expected to reduce the influence of noise and yield a good fit for a constant source , but not for a truly variable one . we consider the source variable if the @xmath25/dof thus computed can occur by chance with probability @xmath26 or less . sources that were flagged for variability were tested for periodicity . periods were found using the supersmoother algorithm ( * ? ? ? * first published by @xcite ) . the algorithm folds the unfolded light curve around trial periods and selects those periods in which the smoothed light curve matches the data best in a statistical sense ; we have selected the best 15 possible periods ranked by the smoothness of the light curve . periods were found separately for the red and blue unfolded light curves . the period selected as the best one by the program turned out to be `` correct '' in @xmath27 of the ebs for at least one band ; for @xmath28 of the ebs the second best period turned out to be correct for at least one band and only for @xmath29 of the ebs did the procedure fail to find a good period . in these cases `` correctness '' was determined by direct visual inspection . lccccc red & @xmath0 & @xmath30 & @xmath31 & @xmath32 & @xmath33 + & & @xmath34 & @xmath35 & @xmath36 & @xmath37 + blue & @xmath0 & @xmath38 & @xmath39 & @xmath40 & @xmath41 + & & @xmath42 & @xmath35 & @xmath36 & @xmath28 + the supersmoother program can fail in two manners when fitting ebs . first when one eclipse ( the secondary ) is very shallow , supersmoother may not recognize it and yield a period twice the correct one . second , when the two eclipses have nearly equal depth supersmoother may confuse the secondary and the primary eclipses yielding a period half the correct one . the first failure happened , for one or both bands , in about @xmath36 of ebs , whereas the second happened in about @xmath35 of ebs . these cases are easy to correct upon visual inspection . for @xmath43 ebs supersmoother gave a period which was some other multiple of the correct one for at least one band ; these were fixed upon visual inspection . for the remaining ebs we tried folding the light curves around the other periods selected by supersmoother and managed to identify the correct period for most of them . in @xmath44 cases in which there were ogle - ii counterparts we adopted ogle periods since , though differing in some cases by less than @xmath26 from the periods found by supersmoother , they gave a much better light curve . we found @xmath45 stars in which the secondary eclipse was not evident , either because it was shallow or because the light curve was noisy , but with an ogle - ii counterpart in which it was clearly visible ; these stars have not been included in the catalogue . the periods in the two bands differ on average by @xmath46 . these results are summarized in table [ tab : period ] . the search for variable objects in the lmc gave @xmath47 objects of which @xmath48 were found to be periodic . to find ebs in this sample we considered a variety of properties of light curves . the techniques we employed are : 1 . look at the number of photometric excursions ( `` dips '' ) in the light curve . an eb is expected to show two `` dips '' in an entire period corresponding to the two eclipses as opposed to a cepheid or an rrlyr star for which only one is expected . the number of photometric excursions was calculated by supersmoother by counting the number of times the smoothed light curve crosses the mean : we selected stars with two excursions in both bands . we additionally imposed a cut on light curve amplitudes : calling @xmath49 and @xmath50 the amplitudes of the blue and red light curves respectively , as computed by supersmoother , we imposed @xmath51 . this amplitude cut was imposed to help in eliminating rrlyrs from the sample , since , considering a population of several thousands probable rrlyrs found in the macho database , we found that , on average , @xmath52 ; imposing a cut @xmath51 should therefore filter out many rrlyr . in our sample just @xmath53 ebs ( @xmath54 ) pass this cut ; we then removed the amplitude cut and look at the number of photometric excursions alone we found that @xmath55 ebs ( @xmath56 ) pass this relaxed cut . 2 . look at the ratio of number of points five standard deviations _ away _ ( @xmath57 ) from the median to the number of points five standard deviations _ below _ the median ( @xmath58 ) . this ratio is expected to be @xmath59 for a typical single variable star . for an eb we expect this ratio to approach @xmath60 as the signal to noise in the photometry increases , as most `` outlier '' points are due to eclipses . we imposed a cut @xmath61 in both bands and found that @xmath62 ebs pass it ( @xmath63 ) , whereas for the overall variable star dataset the figure is @xmath64 out of @xmath65 or @xmath66 . 3 . use a decision tree . we applied the decision tree program described in @xcite , which was run on all the variable objects on the catalogue and gave for each the probability that it was an eb , an rrlyr , a cepheid , a long period variable or an unknown object . in all @xmath67 objects were found most likely by the decision tree to be ebs but only @xmath68 were found to be real on visual inspection and included in the sample . 4 . use a similarity technique . we finally tested the sample with a technique described in @xcite aimed at finding `` outliers '' in large data sets of variable star light curves . the technique aims at finding objects whose light curves are most dissimilar , in a statistical sense , from an `` average '' light curve built out of all the light curve in the data set . this is accomplished by looking at all the pairs of light curves to find their mutual similarity as defined in @xcite and , for each light curve , by then combining these measures , to find its overall similarity to the rest of the sample . light curves with low measure of similarity are flagged as outliers . this approach was useful in finding misfolded lightcurves since it found many objects for which the period for one band gave a badly folded light curve but the period for the other band gave a good folding . this happened for @xmath69 ebs , despite these periods differing on average by just @xmath70 ; in this case we selected the period that gave the good light curve for both bands . since all the techniques we used give some false positives , each candidate was also visually inspected before inclusion in the sample ; we paid closer attention to those stars which could more easily be classified as ebs without being so , like ellipsoidal variables ( see subsection [ subsec : ell ] ) and cepheids and rrlyr mistakenly folded around a period twice their real one . cccccc eb & @xmath0 & @xmath55 & @xmath53 & @xmath62 & @xmath68 + & & @xmath71 & @xmath72 & @xmath34 & @xmath73 + all variable sources & @xmath48 & @xmath74 & @xmath75 & @xmath64 & @xmath76 + & & @xmath77 & @xmath78 & @xmath66 & @xmath77 + lc @xmath79 dips and @xmath61 & @xmath80 + @xmath79 dips , @xmath81 , and @xmath61 & @xmath82 + @xmath79 dips and decision tree & @xmath83 + @xmath61 , and decision tree & @xmath84 + @xmath79 dips , @xmath81 , and decision tree & @xmath85 + @xmath79 dips , @xmath61 , and decision tree & @xmath86 + @xmath79 dips , @xmath81 , @xmath61 , and decision tree & @xmath87 + these results of our search techniques are summarized in table [ tab : cuts ] , and table [ tab : morethanonecut ] shows the number of ebs that pass more than one cut . as the numbers show these heuristic tests are far from perfect and tend to give too many candidates ; however we feel that the large dimension of our sample can allow the determination of more stringent tests , an absolute necessity for the analysis of future surveys . our search gave @xmath88 objects observed in more than one field : these duplicates have different macho field numbers , but typically the same tile number . in this case we summed the numbers of observations in both bands for each field and chose the one which had the highest total number of observations : the object is identified by that corresponding fts only . figure [ fig : perhist ] shows a logarithmic histogram of the period distribution . note that the periods range from a fraction of a day to several hundreds of days . figure [ fig : maghist ] shows a histogram of the distribution of median @xmath18 , @xmath19 and @xmath23 . note that the magnitudes range in values from @xmath89 to @xmath90 both in @xmath18 and @xmath19 with a peak around @xmath91 . the average photometric error for the lmc is @xmath92 in both instrumental bands ; the average error for a light curve as a function of median relative magnitude is in standard magnitudes is shown in figure [ fig : magerr ] . for the lmc sample we estimated the distribution the root mean square ( rms ) of the residuals @xmath93 of the observations around a theoretical light curve as a function of median relative magnitude where @xmath94 is the value of the observed magnitude orbital phase @xmath95 , and @xmath96 is the theoretical value at the same phase . theoretical values were obtained by fitting the light curves using the jktebop ] code @xcite . the jktebop code is based on the ebop code @xcite , which implements the model by @xcite with some modifications ; jktebop in turn adds several modifications and extensions to the original ebop code that make it easier to use , especially when fitting a large number of light curves . before fitting we eliminated outlying points by taking averages of all points in boxes containing from @xmath97 to @xmath98 points along a light curve and discarding the points more than @xmath79 standard deviations away from these averages . we obtained starting values for the model parameters by running the debil code @xcite and using the values it computed ; the limb darkening values for the @xmath18 and @xmath19 bands were taken from @xcite . we fixed the ratio of the masses , @xmath99 , to @xmath17 , @xmath100 and @xmath101 and did the fit in each case taking in the end the best result . finally we selected light curves with @xmath102 to show in figure [ fig : magrms ] . out of @xmath103 ebs in the lmc the program converged in @xmath104 cases in the @xmath19 band and in @xmath105 cases in the @xmath18 band ; we found @xmath106 fits with @xmath102 in the @xmath19 band and @xmath107 in the @xmath18 band . we point out that those fits were made only with the aim of obtaining a good theoretical light curve for as many observed light curves as possible in a fast and automated manner , so that a residuals distribution could be calculated . in particular we did _ not _ attempt to accurately determine astrophysical parameters for our ebs . this is also the reason why we discarded points at just @xmath79 standard deviations away from the moving averages , which could result in eliminating potentially interesting information for some ebs ; such objects are obviously deserving of more in depth study which we did not attempt here . while in general our fits were good in the case of largely separated , undistorted systems , they were often bad for close , strongly distorted ones , which is to be expected since jktebop is not meant to be used for such systems ; also in several cases our fits were bad because the scatter of the observed values was larger than the observational errors , which suggests that other physical phenomena , such as pulsation of one or both components , are present . the rms distributions of the residuals vs. median magnitudes @xmath18 and @xmath19 are shown in figure [ fig : magrms ] . figures [ fig : lcs1 ] , [ fig : lcs2 ] , [ fig : lcs3 ] , [ fig : lcs4 ] , [ fig : lcs5 ] , and [ fig : lcs6 ] show some examples of light curves . the panels on the left show the original light curves , those on the right show the light curves with the outlying points removed , the error bars , the theoretical light curves from the fit and the residuals the ebs shown are meant to be representative of the sample , this is why some examples of bad fits are included . comparing the panels on the left with the panels on the right for these figures gives an idea of the effect of removing outlying points . in particular the figures suggest that for those objects with good fits the procedure resulted in the elimination of truly outlying points ; these ebs are mostly detached with undistorted components . for objects with bad fits , which mostly comprise ebs with close and strongly distorted components , the situation is less clear . for example the system labelled 1.3442.172 shown in figure [ fig : lcs2 ] exhibits some points in the @xmath19 band , at secondary eclipse around phase @xmath108 , that run almost parallel to the main light curve but at a higher magnitude . such points may or may not be physically significant ; some of these are removed by our procedure , but the fit is nevertheless bad . the systems labelled 1.3804.164 in figure [ fig : lcs4 ] and 1.4055.98 in figure [ fig : lcs5 ] show a large and step - like scatter band , the reason of which is , we think , intrinsic variability of the secondary component , as suggested by the fact that the band becomes much narrower at primary eclipse but not at secondary eclipse ; this interpretation is also suggested by the fact that the residuals show an oscillating behavior as a function of phase . in both 1.4055.98 and 36.5943.658 in figure [ fig : lcs6 ] the scatter band is much larger than the observational error which explains their bad fits . these examples show that the samples contain many ebs which could be deserving of more careful study which we did not attempt here . the properties of these ebs are summarized in table [ tab : lcs ] ; of the @xmath109 ebs shown @xmath110 have two photometric excursions in both bands and @xmath111 , @xmath112 are found by the decision tree , @xmath113 have @xmath61 , and @xmath79 have a counterpart in the ogle - ii sample : 1.3442.172 ( counterpart ogle050149.20 - 691945.5 in field lmc@xmath114sc15 ) and 1.4055.98 ( counterpart ogle050542.06 - 684732.8 in field lmc@xmath114sc13 ) . cccccccl 65.8581.67 & 05:33:09.458 & -65:30:29.13 & 0.19 & 18.93 & 18.13 & 0.8 & shortest period in sample + 9.5000.790 & 05:11:20.548 & -70:21:00.16 & 0.24 & 19.70 & 19.23 & 0.47 & very short period + 1.3442.172 & 05:01:49.005 & -69:19:45.60 & 1.02 & 17.21 & 17.26 & -0.05 & fairly typical eb + 10.4035.145 & 05:05:02.233 & -70:06:13.27 & 2.53 & 17.14 & 17.19 & -0.05 & fairly typical eb + 68.10843.699 & 05:47:10.738 & -67:58:55.74 & 3.07 & 13.56 & 13.99 & -0.43 & bluest in sample + 62.7240.102 & 05:24:55.090 & -66:11:55.28 & 4.17 & 18.30 & 18.30 & 0.00 & very high eccentric orbit + 1.3804.164 & 05:03:36.536 & -69:23:32.27 & 4.19 & 16.88 & 16.90 & -0.02 & algol type + 41.2459.43 & 04:55:43.321 & -70:18:00.44 & 13.18 & 16.74 & 16.82 & -0.08 & highest eccentric orbit + 1.4055.98 & 05:05:42.201 & -68:47:33.44 & 24.51 & 17.32 & 17.11 & 0.21 & fairly typical eb + 12.10443.34 & 05:44:47.185 & -70:27:26.65 & 319.37 & 17.35 & 16.30 & 1.05 & reddest in sample + 62.6514.2213 & 05:20:32.499 & -66:13:17.92 & 417.60 & 16.88 & 16.23 & 0.65 & long period + 36.5943.658 & 05:17:15.478 & -71:57:45.69 & 633.70 & 16.35 & 15.54 & 0.81 & longest period in sample the information in table [ tab : lcs ] is also available in its entirety via the link to the machine - readable version above . the ebs are arranged by ascending period . units of right ascension are hours , minutes , and seconds , and units of declination are degrees , arcminutes , and arcseconds . ellipsoidal variability occurs in a close binary system when one ( or both ) component(s ) is ( are ) tidally distorted by the companion . if the binary system is detached , as most systems in our samples are , the distorted stars assume the asymmetric , egg like shape of the roche equipotential surface whereas in the case of contact system the shape of the common equipotential surface is more reminiscent of a dumbbell . the light curve of an ellipsoidal variable system reveals a continuously varying profile , with two maxima and two minima per period , with the minima often having different depth , whereas the maxima are usually equal . the main reason for this variability is that , as the stars rotate , their projected areas on the sky vary , reaching a maximum at the two quadratures and a minimum at the two conjunctions ; the measured flux thus varies in the same way during a period . more information on ellipsoidal variables can be found in @xcite a large sample of ellipsoidal variables in the lmc has been released by the ogle collaboration @xcite ; an analysis of ellipsoidal variables found in the macho database has been published by @xcite . a binary system can present both eclipses and ellipsoidal variability but in many cases it may not be possible to clearly recognize an eclipse from visual inspection ; this poses a problem for the compilation of eb catalogues since the light curves of ebs and non eclipsing ellipsoidal variables can be easily confused . we looked for possible contamination by ellipsoidal variables in our sample and found @xmath115 systems exhibiting ellipsoidal variability which we then visually checked more carefully than other stars which were clearly ebs . we also attempted a less subjective approach by fitting these systems with the ebop program @xcite following the prescriptions of @xcite ; however since ebop is not designed for analyzing such distorted systems the final decision about whether or not to include a star exhibiting ellipsoidal variability in the sample was taken upon visual inspection . figure [ fig : ell ] shows two examples of eb systems with pronounced ellipsoidal variability , basic data on these systems are given in table [ tab : ell ] . cccccccl 1.3934.140 & 05:04:23.977 & -68:49:21.65 & 85.88 & 17.68 & 17.03 & 0.65 & no + 58.6147.42 & 05:18:03.677 & -66:27:20.17 & 105.08 & 17.78 & 17.06 & 0.72 & no + 15.10916.25 & 05:47:21.831 & -71:10:46.49 & 355.28 & 17.04 & 16.05 & 0.99 & yes + 14.9588.6 & 05:39:28.498 & -71:00:46.01 & 411.04 & 14.75 & 13.95 & 0.80 & yes + the smc sample comprises @xmath1 ebs selected via the same techniques as the lmc ebs and confirmed by visual inspection ; the general considerations of the preceding subsection regarding search for variability apply here as well . the sky coverage in the smc corresponds to macho fields 206 , 207 , 208 , 211 , 212 and 213 ; field center coordinates for these fields are given in table [ tab : smcfields ] . cccc 206&1:05:21.70&-72:26:58.3&(j2000.0 ) + 207&0:57:16.58&-72:34:57.0&(j2000.0 ) + 208&0:48:03.19&-72:34:20.9&(j2000.0 ) + 211&0:58:27.40&-73:04:55.3&(j2000.0 ) + 212&0:49:10.27&-73:13:32.9&(j2000.0 ) + 213&0:40:18.91&-73:08:49.5&(j2000.0 ) + magnitudes quoted for the smc have been obtained by using transformations which differ slightly in the zero point from the lmc ones due to the larger exposure times in the smc @xcite ; they are reported in eq . [ eq : smcmachocal ] . @xmath116 the smc search gave @xmath117 duplicates and again we chose the field which had the highest total number of observations and the object is identified by that corresponding fts only . figure [ fig : smcpointhist ] shows the histogram of the number of observations in both bands for the ebs in the sample . figure . [ fig : smcperhist ] shows a logarithmic histogram of the period distribution . figure [ fig : smcmaghist ] shows the histograms of the magnitudes for both bands as well as for color . magnitudes range in values from @xmath118 to @xmath90 both in @xmath18 and @xmath19 bands , with a peak around @xmath119 . the average photometric error for the smc is again @xmath92 in both instrumental bands ; the error as a function of standard magnitude is shown in figure [ fig : smcmagerr ] . figures [ fig : smclcs1 ] , [ fig : smclcs2 ] , [ fig : smclcs3 ] , and [ fig : smclcs4 ] show some examples of light curves ; their properties are summarized in table [ tab : smclcs ] . the `` bump '' shown by the star labelled 207.16374.39 is probably due to star spots : we were able to roughly reproduce it by appropriately choosing spots on the components and fitting the light curve using the phoebe@xcite software package . cccccccl 211.16529.5 & 00:59:31.368 & -73:26:56.04 & 0.28 & 16.00 & 15.43 & 0.57 & shortest period in sample + 207.16652.1084 & 01:00:43.656 & -72:51:51.48 & 0.36 & 19.67 & 19.59 & 0.08 & very short period + 206.16883.214 & 01:04:25.272 & -72:37:48.36 & 1.14 & 18.16 & 18.18 & -0.02 & fairly typical eb + 207.16656.93 & 01:00:56.345 & -72:36:44.40 & 1.23 & 18.00 & 17.96 & 0.04 & highly eccentric orbit + 206.16717.285 & 01:01:41.400 & -72:19:25.68 & 1.65 & 18.35 & 18.32 & 0.03 & fairly typical eb + 208.15740.63 & 00:46:13.949 & -72:52:37.03 & 1.74 & 16.13 & 16.24 & -0.11 & bluest in sample + 212.15903.2269 & 00:49:18.192 & -73:21:55.44 & 2.42 & 17.51 & 17.51 & 0.00 & highly eccentric orbit + 207.16315.289 & 00:55:48.168 & -72:29:33.36 & 3.34 & 17.93 & 17.94 & -0.01 & highly eccentric orbit + 211.16195.61 & 00:53:59.729 & -72:56:56.13 & 4.73 & 16.44 & 16.43 & 0.01 & fairly typical eb + 208.15912.323 & 00:49:10.440 & -72:46:37.56 & 120.51 & 18.84 & 18.35 & 0.49 & highly eccentric orbit + 208.16.58 & 00:49:28.392 & -72:49:40.80 & 137.89 & 17.17 & 16.50 & 0.67 & reddest in sample + 207.16374.39 & 00:56:25.872 & -72:22:15.96 & 186.34 & 16.30 & 16.30 & 0.00 & long period + 212.15673.13 & 00:45:46.824 & -73:31:32.52 & 200.27 & 16.00 & 15.35 & 0.65 & long period + 211.16418.53 & 00:57:5.304 & -73:15:10.44 & 234.64 & 17.23 & 16.82 & 0.41 & highly eccentric orbit + 206.17005.6 & 01:06:10.224 & -72:06:24.48 & 371.89 & 15.69 & 15.16 & 0.53 & highly eccentric orbit + 211.16310.174 & 00:55:30.024 & -72:52:49.44 & 1559.81 & 18.23 & 17.61 & 0.62 & longest period in sample + the information in table [ tab : smclcs ] is also available in its entirety via the link to the machine - readable version above . the ebs are arranged by ascending period . units of right ascension are hours , minutes , and seconds , and units of declination are degrees , arcminutes , and arcseconds . figure [ fig : cmd ] shows the cmd for the @xmath0 ebs in the lmc sample ; the lower magnitude limit is @xmath120 . we estimated the reddening by using the lmc extinction map described in the lmc photometric survey of @xcite . the extinction catalog produced by the survey is available for query on line and we retrieved the values of @xmath121 specified in table [ tab : lmcextinction ] , based on the _ hot _ stars only found by the survey ( @xmath122 ) in a radius of @xmath123 ( the maximum allowed ) around the positions specified in table [ tab : lmcextinction ] , which sample the eb position distribution . ccc 06:06:00 & -69:05:00 & 0.48 + 06:06:00 & -72:43:00 & 0.97 + 05:40:00 & -65:30:00 & 0.75 + 05:40:00 & -69:05:00 & 0.80 + 05:40:00 & -72:30:00 & 0.78 + 05:20:00 & -65:30:00 & 0.53 + 05:20:00 & -69:05:00 & 0.50 + 05:20:00 & -72:30:00 & 0.56 + 05:00:00 & -65:30:00 & 0.50 + 05:00:00 & -69:05:00 & 0.46 + 05:00:00 & -72:43:00 & 0.45 + 04:40:00 & -69:05:00 & 0.62 + 04:40:00 & -72:30:00 & 0.97 + from the values of table [ tab : lmcextinction ] we derive a mean value for @xmath121 of @xmath124 which we use to characterize the average lmc @xmath18 extinction . we use the reddening vector @xmath125 from ( * ? ? ? * and references therein ) and find @xmath126 , more than a factor of two and a half larger than the value @xmath127 found by @xcite . this is likely due to the fact that the reddening for the ebs are likely to be along lines of sight toward young , hot stars in the @xcite catalog which are derived to have higher @xmath121 , while the @xcite @xmath121 was derived from observations of rr lyrae stars . as the cmd shows the sample is made up mostly of bright early type stars : from the range in magnitudes and assuming an lmc distance modulus of @xmath128 @xcite we see that in most cases at least one component is of spectral type b or a @xcite . we employ the term _ young star region _ to describe the main feature on the left part of the cmd , rather than main sequence because our sample contains some bright and short lived stars that may not be burning hydrogen in their core while still being on the blue part of the cmd ; likewise we employ the term _ evolved star region _ to indicate the feature on the red part of the cmd . we define the young star region as @xmath129 and the evolved star region as @xmath130 : with these definitions we find @xmath131 ebs in the young star region and @xmath132 ebs in the evolved star region . we used a simple cut to separate the young star region from the evolved star region rather than a more precise one because our cut can be easily seen both in the cmd and in the color period diagrams . an interesting feature of the cmd is the lack of a clear gap between the young star region and the evolved star region ; there is instead a fairly continuous transition , with a higher number of systems filling the hertzsprung gap that would be expected from cmds of single stars . this may indicate that these systems are composed of a more massive and hence more evolved and redder star and a less massive , less evolved , bluer one . the color period diagram is shown in the left panel of figure [ fig : smcvslmc ] ; this diagram clearly shows the young ebs with periods @xmath133 and a second population ( @xmath134 objects ) of long period , evolved ebs with periods @xmath135 and @xmath130 . several interesting features emerge in figure [ fig : smcvslmc ] : there is a paucity of long period objects on the young star region of the cmd , and a corresponding lack of short period objects with very red colors . furthermore , the red ( @xmath130 ) population shows a positive correlation between period and color . there are virtually no long period , blue objects or short period , red objects in this group . in contrast , the young star region shows no such correlation . this structure in the color period diagram is a consequence of ( i ) kepler s third law , ( ii ) the probability that an eb is favorably oriented in space to allow eclipses to be detected ( @xmath136 ) where @xmath137 and @xmath138 are the radii of the primary and secondary stars , respectively and @xmath139 is the semi major axis , and ( iii ) that objects usually evolve in this diagram at constant period , from blue to red . long period binary stars have large semi - major axes . when both stars are on the young star region , their relatively small radii yield a relatively low probability that they will eclipse when seen from our vantage point . when one of the pair evolves away from the young star region , one of these radii ( typically @xmath137 ) will increase . the consequence of this is an increase in the probability that eclipses will be detected . this accounts for the presence of red , long period stars and the absence of corresponding young progenitors . the situation is different for short period systems . these are relatively likely to be detected because of their small semi - major axes , and are prominent on the blue side . as one of these stars evolves and expands rapidly , it may engulf the companion and enter a stage of _ common envelope evolution _ in which the expanding star overflows the _ second _ lagrangian point @xmath140 @xcite ; leading to the disappearance of eclipses . common envelope system differ from contact binaries @xcite in which two young stars overflow their first lagrangian point ( @xmath141 ) and their roche equipotential surface assumes a dumbbell shape . contact systems usually show ellipsoidal variation and , if the orbital inclination is large enough , also eclipses ; ebs of the w uma type belong to this category . @xcite and @xcite provide more information on common envelope binaries . the correlation between period and color among red objects reflects the general correlation between radius and color for the evolved partner . the left panel of figure [ fig : smcvslmc ] reveals a population of @xmath142 ebs with low periods ( @xmath143 ) and high color ( @xmath144 ) . these objects are probably foreground galactic ebs composed of late type stars . this interpretation is suggested by several factors . first , due to the large angular extent of the lmc , there is foreground contamination in the lmc macho fields @xcite ; in particular the feature marked `` h '' in the cmd of their figure @xmath17 indicates foreground galactic disk stars and is centered at @xmath145 as is our presumptive foreground population . second , both the short period of these ebs , and the shape of their light curves which are either detached or mildly distorted , strongly suggests that the stars making up this population are small , late type stars ; this is further borne out by their color , again typical of a solar like star ; finally , the cmd of this population , shown in figure [ fig : cmdforeground ] clearly shows what appears to be a turnoff feature at @xmath146 , with few evolved objects ( @xmath147 ) . it is interesting to note that the overall shape of this population in the color period diagram shows , on a smaller scale , the same features of the lmc diagram ; a main sequence is clearly visible in figure [ fig : cmdforeground ] and the evolved ebs show the same color period correlation of their lmc counterparts in figure [ fig : smcvslmc ] . figure [ fig : smccmd ] shows the cmd for @xmath148 ebs in the smc sample out of the @xmath1 in the sample ; one eb which has valid data only in the @xmath15 band is not shown because it was not possible to determine the standard magnitudes via eq . [ eq : smcmachocal ] . the general remarks made for the lmc cmd apply here as well and we used a reddening vector with the same inclination as the lmc . the figure clearly shows the young star region which is composed of @xmath149 ebs ( @xmath150 ) whereas there are just @xmath151 evolved ebs . we estimate the smc reddening from @xcite ; from their figure @xmath152 we infer a mean @xmath153 for their hotter smc population , relevant to our sample which is composed mostly of early type hot stars on the young star region . we use the same reddening vector as the lmc , @xmath125 , and find a mean @xmath154 . the color period diagram is shown in the right panel of figure [ fig : smcvslmc ] ; the general considerations made for the lmc color period diagram apply here as well . the diagram shows @xmath113 ebs with low period ( @xmath155 ) and red color ( @xmath156 ) which are probably foreground objects . the much smaller number of foreground objects in the smc is probably due both to its smaller angular size and to its higher galactic latitude ( @xmath157 compared to @xmath158 for the lmc ) . we can further test the hypothesis that these two short period , red populations in the lmc and smc are due to foreground objects by comparing the ratio of their numbers to the ratio of the areas of the lmc and smc , which we can estimate from figures [ fig : pos ] and [ fig : smcpos ] ; if these objects are indeed foreground these two ratios should be roughly equal . from the figures we can estimate the sky area of the lmc as @xmath159 and the sky area of the smc as @xmath160 ; their ratio is thus similar to the ratio of the numbers of ebs in the two populations as expected . although the basic features of the cmd and the color period diagram are the same for lmc and smc , there are some differences , shown by figure [ fig : smcvslmc ] and table [ tab : smcvslmc ] . ccccccc lmc & @xmath0 & @xmath161 & @xmath162 & @xmath163 & @xmath164 & @xmath165 + smc & @xmath148 & @xmath166 & @xmath167 & @xmath168 & @xmath169 & @xmath170 + the fraction of blue ( @xmath129 ) ebs is higher in the smc than in the lmc . more striking , the fraction of blue long period ( @xmath171 , @xmath135 ) ebs is much higher in the smc than in the lmc . to further investigate the differences between the two samples we carried out a kolmogorov - smirnov ( ks ) test on the distributions of the absolute magnitudes @xmath172 and @xmath173 , their difference @xmath174 , and the periods @xmath175 of the two samples . we adopted a distance modulus of of @xmath176 for the smc @xcite and @xmath177 for the lmc @xcite . magnitudes and colors were dereddened using @xmath178 , @xmath179 for the lmc and @xmath180 , @xmath181 for the smc before subtracting the distance moduli . the empirical cumulative distribution functions ( ecdfs ) of these quantities are shown in figure [ fig : ecdf ] . a ks test confirms that the distributions of @xmath175 and @xmath174 are different at @xmath182 confidence level ; for the distributions of @xmath183 and @xmath184 the ks test gives a probability of @xmath185 and @xmath186 respectively for them being different . the @xmath174 plot shows that ebs in the smc tend to be bluer , not surprising given that a higher percentage of objects belong to the young star cmd region in the smc than in the lmc . the period distributions show that the smc ebs have on average shorter periods , again not surprising given the much higher percentage of evolved systems in the lmc than in the smc and the fact that evolved systems have on average higher periods than the young systems ( as shown by the color period diagrams in figure [ fig : smcvslmc ] ) . the ebs in our samples have been cross correlated with the corresponding ogle - ii samples . stars in the samples were identified if their right ascension ( ra ) and declination ( dec ) differed by less than @xmath187 and if their periods differed by @xmath188 . we used a very large search radius to be conservative . the astrometric precision for both surveys is typically @xmath189 , but a few stars which had much larger differences in ra and/or dec turned out to be matches upon inspection of their periods : in particular we found in the lmc @xmath190 matches with a position difference bigger than @xmath191 and @xmath79 matches with a position difference bigger than @xmath192 . however most of the matches were within narrower radii : for the lmc roughly half of the matches ( @xmath193 out of @xmath194 ) were found within of @xmath195 , compatible with the astrometric precision of both macho and ogle surveys ; almost all of them ( @xmath196 ) were within @xmath197 . we tested the robustness of our method of finding matches by investigating the probability for two periodic objects with a period difference of @xmath188 to be within @xmath187 of each other . to do this we selected a random sample of @xmath198 objects out of the @xmath199 periodic ones found by macho in the lmc and we counted the frequency of pairs of objects with both periods from the red and the blue lightcurves ( as found by supersmoother ) differing by @xmath188 and positions within @xmath187 . most of the matches we found were due to the same object being observed in different tiles and only in one case did we find a possibly genuine match ; we thus conclude that the probability of two objects being erroneously classified as a match is @xmath200 and therefore our method of finding matches is robust . our search produced @xmath194 matches in the lmc . the macho and ogle - ii periods agree to high accuracy as shown in figure [ fig : crossper ] ; usually much better than the @xmath26 cut we imposed . histograms of position differences are shown by figure [ fig : diffposhist ] . both panels show the entire span of the differences ; the differences in right ascension range from @xmath201 to @xmath202 but the left panel shows this range multiplied by the cosine of the declination ( @xmath203 ) which gives a range from @xmath204 to @xmath205 . the sky coverage of the two surveys was different : ogle - ii , from which the sample was derived , covered about @xmath206 in the central region of the lmc whereas the sky coverage of macho was larger . figure [ fig : pos ] shows the positions of the ebs in both catalogues and figure [ fig : machofields ] shows the corresponding macho field number ; the fields at the center of the lmc are drawn with continuous lines , the ones at the periphery with dashed lines . in view of the different sky coverage of the macho and ogle - ii surveys , it is interesting to analyze separately the stars in macho fields covering the center of the lmc ( which roughly correspond to the ogle - ii sky coverage ) and the stars in the macho fields at the periphery . figure . [ fig : machofields ] shows that the fields in the center are 1 , 2 , 3 , 5 , 6 , 7 , 8 , 10 , 11 , 12 , 13 , 14 , 15 , 18 , 19 , 47 , 77 , 78 , 79 , 80 , 81 and 82 . there are @xmath207 ebs in the center and @xmath208 in the periphery . figure [ fig : cmdcenter ] shows the cmd and the color period diagram for the center and the periphery of the lmc respectively ; figure [ fig : perhistcenter ] shows the histograms for the magnitudes , the color and the period . the figures reveal several differences between the two samples ; to check these we performed a ks test for @xmath18 , @xmath23 and @xmath175 and found that at high confidence level ( @xmath182 ) the distributions are different . these differences are statistically significant but not large enough to be considered astrophysically important . in particular the difference in period distribution could be attributed to the lower sampling in the outer fields , making it less likely to detect the longer period ebs . the sampling might also be the cause of the periphery having a higher percentage of bright ebs , since these are easier to find with fewer epochs . we finally investigated why we did not find more matches with ogle - ii . the most important reason , we think , is the fact that the techniques employed in assembling the samples are different : the ogle team built their samples via neural networks @xcite . the two surveys have roughly comparable limiting magnitudes , @xmath209 ; nevertheless we checked how the performance of the two surveys varied with magnitude . we compared the distribution of the @xmath18 magnitudes for the @xmath207 ebs in the central region of the lmc in our sample with those of @xmath210 matches and the @xmath211 ebs from ogle - ii without macho counterparts ( these two numbers do not add up to @xmath7 , the size of the ogle - ii sample , because for some ebs the @xmath18 magnitude was not reported ) . the histograms of these three distributions are shown in figure [ fig : v_macho_and_ogle ] . the figure shows that ogle - ii ebs with macho counterparts , peaking at @xmath212 like the macho sample , are on average brighter than the ones without macho counterparts , which peak at @xmath213 . the shape of the @xmath18 distribution of ogle - ii ebs with macho counterparts much more closely resembles the macho @xmath18 distribution ; a ks test gives a probability of the two distributions being the same of @xmath214 . the distributions of macho @xmath18 and ogle - ii @xmath18 without macho counterparts , as well as those of these two ogle - ii populations are , on the other hand , shown to be different at @xmath182 confidence level . both distributions vanish at @xmath215 , showing that their limiting magnitudes are comparable . we then studied the distributions of the periods : figure [ fig : per_macho_and_ogle ] shows that the ogle - ii ebs without macho counterparts have on average longer periods than macho ebs in the center of the lmc and than ogle - ii ebs with a macho counterpart and the difference is statistically significant in both cases ; the period distributions of the macho eb of the lmc center and of the macho ogle - ii matches are statistically different as well . therefore , ogle - ii finds a higher proportion of fainter objects than macho and this does play a role in not finding an higher number of ogle - ii counterparts to our sample . we finally studied the distribution of the macho - ogle matches as a function of @xmath18 . we first counted the number of ogle - ii lmc ebs in the macho fields , finding @xmath216 of them , out of a total of @xmath7 ; of these @xmath216 ebs , @xmath194 were the matches described above and @xmath217 did not have a macho counterpart ( this last number is smaller than @xmath211 , the total number of ogle ebs without macho counterpart , because we are now only considering ogle ebs in macho fields ) . we then studied the distribution , as a function of @xmath18 , of the ogle ebs in the macho fields that both had and did not have a macho counterpart and for which the @xmath18 magnitude was reported : there were @xmath210 of the former and @xmath218 of the latter . we finally performed the inverse calculation , by first counting the number of macho lmc ebs in the ogle - ii fields and finding @xmath219 of them ; of these @xmath220 had an ogle - ii counterpart and @xmath221 did not ; we studied the distribution of both these populations as a function of @xmath18 . the ogle - ii field boundaries were estimated by taking the coordinates of the most extreme ebs in each ogle - ii field . these findings are summarized in table [ tab : v_matches ] . lc ogle - ii lmc ebs & @xmath7 + ogle - ii lmc ebs in macho fields & @xmath216 + ogle - macho matches & @xmath194 + ogle - ii lmc ebs without macho counterpart & @xmath217 + ogle - macho matches with reported @xmath18 & @xmath210 + ogle - ii lmc ebs without macho counterpart with reported @xmath18 & @xmath218 + macho ebs in ogle fields & @xmath219 + macho ebs in ogle fields with ogle counterpart & @xmath220 + macho ebs in ogle fields without ogle counterpart & @xmath221 + the distributions of both the ogle - ii ebs _ with _ macho counterpart and of the macho ebs _ with _ ogle - ii counterparts are shown in figure [ fig : v_matches ] ; the figure shows the fraction of matches in magnitude bins of @xmath222 ; the bin centers range from @xmath223 to @xmath224 . the error bars are estimated by assuming that the matches in each magnitude bin follow a binomial distribution with probability @xmath225 where @xmath226 is the number of matches in each magnitude bin and @xmath227 is the total number of ebs ; the error in the expected fraction of matches is then given by eq [ eq : errorbar ] ; in both distributions of figure . [ fig : v_matches ] the error bar for the brightest magnitude bin is not shown since in both cases there is only one match , rendering eq . [ eq : errorbar ] meaningless . the figure shows that the fraction of matches increases for brighter magnitudes as expected ; the fall at @xmath228 in the distribution of ogle matches is probably due to small number statistic as evidenced by the large error bar . @xmath229 the same general considerations apply to the smc sample : the search , performed with the same criteria as the lmc , produced @xmath230 matches . figure [ fig : smccrossper ] shows the percentage difference of the macho and ogle - ii periods vs macho period for the matches ; figure [ fig : smcdiffposhist ] shows the histogram of the differences in ra and dec . unlike the lmc , the sky coverage of the two surveys was approximately the same . figure [ fig : smcpos ] shows the positions of the macho and ogle ebs on the sky . we again investigated why we did not find more matches with ogle - ii . looking at figure [ fig : smcpos ] it is evident that one of the reasons is the somewhat different sky coverage of the two surveys ( macho fields 207 , 208 and 211 are only partially covered by ogle ) , but we also looked for other possible explanations . again , the most likely explanation is the ways in which the samples were assembled , but we also considered the differences in the distributions of magnitudes and periods . as for the lmc , we compared the distribution of the @xmath18 magnitudes for the @xmath148 ebs in our sample which have a valid @xmath18 with those of @xmath231 matches and the @xmath232 ebs from ogle - ii without macho counterparts ( again these two numbers do not add up to @xmath8 , the size of the ogle - ii sample , because for some ebs the @xmath18 magnitude was not reported ) . the histograms of these three distributions are shown in figure [ fig : smcv_macho_and_ogle ] . the figure shows that ogle - ii ebs with macho counterparts , peaking at @xmath233 like the macho sample , are on average brighter than the ones without macho counterparts , which is more spread out and roughly constant between @xmath234 . a ks test shows that these three distributions are different at @xmath235 confidence level . the behavior at high magnitudes is different for macho and ogle - ii : the figure suggests a magnitude limit of @xmath120 for macho and @xmath236 for ogle - ii ; for @xmath237 macho finds many more ebs than ogle - ii . we then studied the distributions of the periods : figure [ fig : smcper_macho_and_ogle ] shows that ogle - ii ebs with macho counterparts have periods that cluster more in the @xmath238 range , as do the macho ebs and a ks test gives a probability @xmath239 for the two distribution to be the same . on the other hand ogle - ii ebs without macho counterparts have a larger spread of period values , with more objects having @xmath240 and with small `` bumps '' in the distributions at @xmath241 and @xmath242 and a ks test shows it to be different from both macho and ogle - ii macho matches at @xmath182 confidence level . the @xmath18 distributions of both the ogle - ii ebs without macho counterpart and of the macho ebs without ogle - ii counterparts are shown if figure [ fig : v_matches_smc ] ; the figure shows the fraction of matches in magnitude bins of @xmath222 with centers ranging from @xmath223 to @xmath243 ; the error bars are estimated again by eq . [ eq : errorbar ] . we conclude that differences in sky coverage and in techniques used in assembling the samples , as well as different magnitude limits and in general different behavior at high magnitude of the two surveys all play a role in not finding an higher number of ogle - ii counterparts to our sample ; the fact that the smc was less observed than the lmc by macho may also explain why we find fewer long period objects , while the fact that the exposure time for the smc was double that of the lmc @xcite may explain why we find many faint objects . the data presented in this paper can be accessed on line at the astronomical journal website and are mirrored at the harvard university initiative in innovative computing ( iic ) /time series center . at both sites the data consist of a summary table for each cloud and light curves for all the ebs in the samples . light curve files contain the _ unfolded _ data in macho magnitudes ; these same files can also be retrieved from the macho website . finally for the lmc , the input and output files used in the jktebop fits are provided . this work uses public domain data from the macho project whose work was performed under the joint auspices of the u.s . department of energy , national nuclear security administration by the university of california , lawrence livermore national laboratory under contract no . w-7405-eng-48 , the national science foundation through the center for particle astrophysics of the university of california under cooperative agreement ast-8809616 , and the mount stromlo and siding spring observatory , part of the australian national university . khc s work is performed under the auspices of the u.s . department of energy by lawrence livermore national laboratory in part under contract w-7405-eng-48 and in part under contract de - ac52 - 07na27344 . this work uses public domain data obtained by the ogle project . we are grateful to julia kregenow for finding part of the eclipsing binary stars in the sample , to john rice for providing the original reference for the supersmoother algorithm and to peter eggleton for pointing out the difference between contact and common envelope binaries and claude lacy for providing the ebop program . we thank the referee for many helpful suggestions . lf acknowledges the kind hospitality of the institute of geophysics and planetary physics at lawrence livermore national laboratory and of the harvard - smithsonian center for astrophysics where part of the work was done . ribas , i. , jordi , c. , vilardell , f. , guinan , e. f. , hilditch , r. w. , fitzpatrick , e. l. , valls - gabaud , d. , & gimenez , a. 2003 , 25th iau meeting , properties and distances of eclipsing binaries in m31 extragalactic binaries , 25th meeting of the iau , july 2003 , sydney , australia

we present an analysis of the host properties of 85224 emission - line galaxies selected from the sloan digital sky survey . we show that seyferts and liners form clearly separated branches on the standard optical diagnostic diagrams . we derive a new empirical classification scheme which cleanly separates star - forming galaxies , composite agn- galaxies , seyferts and liners and we study the host galaxy properties of these different classes of objects . liners are older , more massive , less dusty and more concentrated , and they and have higher velocity dispersions and lower [ oiii ] luminosities than seyfert galaxies . seyferts and liners are most strongly distinguished by their [ oiii ] luminosities . we then consider the quantity l[oiii]/ , which is an indicator of the black hole accretion rate relative to the eddington rate . remarkably , we find that at _ fixed _ l[oiii]/ , all differences between seyfert and liner host properties disappear . liners and seyferts form a continuous sequence , with liners dominant at low and seyferts dominant at high . these results suggest that the majority of liners are agn and that the seyfert / liner dichotomy is analogous to the high / low - state transition for x - ray binary systems . we apply theoretical photo - ionization models and show that pure liners require a harder ionizing radiation field with lower ionization parameter than seyfert galaxies , consistent with the low and high x - ray binary states .

astro-ph0605681

the majority of nearby agn and agn candidates have nuclear optical spectra that are dominated by emission lines of low ionization species such as [ oi ] @xmath26300 , and @xcite . this class of agn was first defined by @xcite as liners ( low ionization narrow emission - line regions ) . liners have lower luminosities than seyfert galaxies or quasars and are therefore often referred to as low - luminosity active galactic nuclei ( llagn ) . liner emission is extremely common in the nuclei of galaxies ; up to @xmath3 of all galaxies have nuclear spectra typical of liners @xcite . despite the prevalence of liners in galaxies and decades of study , the power source of liners is still under debate . some liner galaxies have double - peaked broad balmer lines @xcite , while others have compact radio cores @xcite , evidence for hard x - ray spectra @xcite , and/or uv variability @xcite . these observations provide circumstantial evidence for an agn power source for the liner emission . other possible power sources include fast shocks @xcite , photoionization by hot stars @xcite , or photoionization by an old , metal - rich stellar population @xcite . liner emission has been observed in extranuclear regions associated with large - scale outflows and related shocks @xcite , or regions shocked by radio jets @xcite . recent investigations into the stellar populations of liner host galaxies have yielded important insight into their nature . @xcite and @xcite detected stellar wind lines in uv spectra of weak liners . similar features have been detected in some seyfert galaxies from nuclear starbursts that are a few myr old @xcite . @xcite searched the hst stis spectra of 28 liners for wolf - rayet features . they found no wolf - rayet features and little evidence of young stars in liners with strong . in liners with low intermediate - age stars contribute significantly to the stellar continuum . @xcite found that while massive stars do not contribute significantly to liner spectra , high order balmer absorption lines are detected in @xmath4% of liners that have relatively weak emission . these results indicate that the current liner classification scheme encompasses two or more types of galaxies , or galaxies at different stages in evolution . liners are commonly classified using their optical emission - line ratios . the first optical classification classification scheme to segregate liners from other spectral types was proposed by @xcite . this scheme uses line ratios of the low ionization species [ ] @xmath5 and @xmath6 compared to the high ionization species [ ] @xmath7 to separate liners from seyferts . this scheme requires the use of additional diagnostics to remove star - forming galaxies . the most common method to remove star - forming galaxies is based on the @xcite empirical diagnostic diagrams using the optical line ratios , , , and . the baldwin et al . scheme was revised by @xcite and @xcite . an alternative scheme was proposed by @xcite that includes an additional division between `` pure '' liners and liner/ `` transition '' objects using the ratio . transition objects have line ratios that are intermediate between the two classes . this division is arbitrary because there was no clear division in between transition or true liner classes . the first purely theoretical classification scheme to distinguish between agn , liners , and region - like objects was derived by @xcite . they used a combination of modern stellar population synthesis , photoionization , and shock models to derive a `` maximum starburst line '' and an `` extreme mixing line '' for separation of the three types of objects . kewley et al . concluded that previous liner classification schemes include starburst - seyfert composites in the liners class , as well as bona - fide liners . recently , @xcite shifted the kewley et al . extreme starburst line to make a semi - empirical fit to the outer bound of @xmath8 sdss spectra . this outer bound defines the region where composite starburst - agn objects are expected to lie on the diagnostic diagrams . although much progress has been made in the optical classification of the ionizing source in galaxies , none of these classification schemes has been able to divide cleanly between seyfert , liner and composite / transition types . the primary reason for this problem is the lack of a sufficiently large sample in which empirical boundaries between the different galaxy classes can be observed . in this paper , we use 85,224 galaxies from the sloan digital sky survey ( sdss ) data release 4 ( dr4 ) emission - line catalog ( described in section [ sample ] ) to develop a new semi - empirical classification scheme for seyferts , liners , and composite objects ( section [ class ] ) . we use this new classification scheme to investigate the host properties of agn in section [ host ] . our results are discussed in section [ discussion ] and we present our conclusions in section [ conclusions ] . throughout this paper , we adopt the flat @xmath9-dominated cosmology as measured by the wmap experiment ( @xmath10 , @xmath11 ; @xcite ) . our sample was selected from the 567486-galaxy dr4 sample according to the following criteria : 1 . signal - to - noise ratio s / n@xmath12 in the strong emission - lines , @xmath7 , , @xmath13 , and @xmath14 . 2 . redshifts between @xmath15 . the s / n criterion is required for accurate classification of the galaxies into star forming or agn dominated classes ( e.g. , @xcite ) . our lower redshift limit ensures that the galaxy properties derived from the fiber spectra are not dominated by the small fixed - size aperture . @xcite analysed the effect of a fixed size aperture on metallicity , star - formation rate , and reddening . they concluded that a minimum aperture covering fraction of @xmath16% is required for the spectral properties within the aperture to approximate the global values . for the @xmath17 fiber aperture of the sdss , a 20% covering fraction corresponds roughly to a redshift of @xmath18 . our upper redshift limit avoids incompleteness in the liner class . liners typically have lower luminosities than seyfert galaxies , and are therefore found at lower redshifts than seyferts in the magnitude - limited sdss survey . we will investigate incompleteness as a function of galaxy type in section [ new_class ] the resulting sample contains 85224 emission - line galaxies and does not include duplicates found in the original dr4 catalog . we use the publically available emission - line fluxes ( described in @xcite ) . these fluxes were calculated using a sophisticated technique that applies a least - squares fit of stellar population synthesis models and dust attenuation to the continuum . once the continuum has been removed , the emission - line fluxes were fit with gaussians , constraining the width and velocity separation of the balmer lines together , and similarly for the forbidden lines . we have corrected the emission - line fluxes for extinction using the balmer decrement and the @xcite reddening curve . we assume an @xmath19 and an intrinsic h@xmath20/h@xmath21 ratio of 2.85 for galaxies dominated by star formation and h@xmath22h@xmath23 for galaxies dominated by agn ( the balmer decrement for case b recombination at t@xmath24k and @xmath25 ; * ? ? ? a total of 5414 ( 6% ) of galaxies in our sample have balmer decrements less than the theoretical value . a balmer decrement less than the theoretical value can result from one or a combination of ( 1 ) intrinsically low reddening , ( 2 ) errors in the stellar absorption correction , and ( 3 ) errors in the line flux calibration and measurement . for the s / n of our data , the lowest e(b - v ) measurable is 0.01 . we therefore assign these 5414 galaxies an upper limit of e(b - v)@xmath26 . @xcite proposed a suite of three diagnostic diagrams to classify the dominant energy source in emission - line galaxies . these diagrams are commonly known as bpt diagrams and are based on the four optical line ratios , , , and . * ; hereafter ke01 ) used a combination of stellar population synthesis models and detailed self - consistent photoionization models to create a theoretical `` maximum starburst line '' on the the bpt diagrams . the maximum starburst line is determined by the upper limit of the theoretical pure stellar photoionization models . galaxies lying above this line are likely to be dominated by an agn . to rule out possible composite galaxies , ( * ? ? ? * ; hereafter ka03 ) modified the ke01 scheme to include an empirical line dividing pure star - forming galaxies from seyfert- composite objects whose spectra contain significant contributions from both agn and star formation . figure [ bpt]a shows the versus standard optical diagnostic diagram for our sample . the ke01 and ka03 classification lines are shown as solid and dashed lines . galaxies that lie below the dashed ka03 line are classed as region - like galaxies . star - forming galaxies form a tight sequence from low metallicities ( low , high ) to high metallicities ( high , low ) which we will refer to as the `` star - forming sequence '' . the agn mixing sequence begins at the high metallicity end of the star - forming sequence and extends towards high and values . galaxies that lie in between the two classification lines are on the agn- mixing sequence and are classed as composites . composite galaxies are likely to contain a metal - rich stellar population plus an agn . galaxies that lie above the ka03 line are classed as agn . figures [ bpt]b and [ bpt]c show the versus and versus diagrams for our sample . the ke01 classification line provides an upper bound to the star - forming sequence on these diagrams . galaxies that are classed as composites using the diagram ( figure [ bpt]a ) lie mostly within the star - forming sequence on the diagram ( figure [ bpt]b ) and about half - way into the star - forming sequence in the ( figure [ bpt]c ) diagram . the line ratio is more sensitive to the presence of low - level agn than or thanks primarily to the metallicity sensitivity of . the log ( ) line ratio is a linear function of the nebular metallicity until high metallicities where the log ( ) ratio saturates @xcite . this saturation point causes the star - forming sequence to be almost vertical at @xmath27 . any agn contribution shifts the towards higher values than this saturation level , allowing removal of galaxies with even small agn contributions . in this work , we classify pure star - forming galaxies as those that lie below the ka03 line on the vs diagnostic diagram . composite galaxies lie above the ka03 line and below the ke01 line . the optical spectra of composites can be due to either ( 1 ) a combination of star - formation and a seyfert nucleus , or ( 2 ) a combination of star - formation and liner emission . the narrow line emission from galaxies lying above the ke01 line is likely to be dominated by an agn . it is clear from the and diagrams ( figures [ bpt]b and [ bpt]c ) that galaxies containing agn lie on two branches . seyfert galaxies lie on the upper branch while liners lie on the lower branch , thanks to their low ionization line emission . we use the and diagrams to separate seyfert from liner galaxies . we use only galaxies with s / n@xmath28 in each of the strong lines to derive our new classification scheme . we define an empirical base point , @xmath29 ( blue circles in figure [ arcs ] ) and we define annuli with widths of 0.1 dex centered on @xmath29 . for the galaxies in each of these annuli with radii ( r ) for which the two agn branches are well - defined ( 0.5 - 1.5 dex from @xmath29 ) , we compute histograms of angle from the x - axis . we then step the 0.1 dex annuli through the data from @xmath29 in 0.02 dex increments and recalculate the minima of each histogram for each increment . a total of 84 histograms were created and minima between the seyfert and liner curves were found for all histograms . examples of these histograms are given in figure [ angle_hist ] . a. the large filled circle represents the empirical base point @xmath29 for the seyfert and liner sequences . concentric arcs of 0.1 dex ( red solid lines ) show the binning of our sample with radius . we calculate histograms of angle from the x - axis centered at @xmath29 for each bin . [ arcs],width=321 ] , width=321 ] to divide between the two agn branches , we take the line of best - fit through the histogram minima for each of the two diagnostic diagrams . this line provides an empirical division between seyferts and liners and will be referred to as the `` seyfert - liner classification line '' . adding this line to our previous two classification lines from ke01 and ka03 , we obtain a means to distinguish between star - forming galaxies , seyferts , liners , and composite galaxies . our new classification scheme is shown in figure [ class_diag ] , and is defined as follows : \1 . * star - forming galaxies * lie below and to the left of the ka03 classification line in the vs diagram and below and to the left of the ke01 line in the and diagrams : @xmath30/h \beta ) < 0.61 / ( \log([nii]/h \alpha ) - 0.05 ) + 1.3,\ ] ] @xmath30/h \beta ) < 0.72 / ( \log([sii]/ha ) - 0.32 ) + 1.30,\ ] ] and @xmath30/h \beta ) < 0.73 / ( \log([oi]/ha ) + 0.59 ) + 1.33\ ] ] \2 . * composite galaxies * lie between the ka03 and ke01 classification lines on the vs diagram : @xmath31/h \alpha ) - 0.05 ) + 1.3 < \log([oiii]/h \beta),\ ] ] @xmath31/ha ) - 0.47 ) + 1.19 > \log([oiii]/h \beta),\ ] ] \3 . * seyfert galaxies * lie above the ke01 classification line on the , , and diagnostic diagrams and above the seyfert - liner line on the and diagrams , i.e. @xmath31/ha ) - 0.47 ) + 1.19 < \log([oiii]/h \beta),\ ] ] @xmath32/ha ) - 0.32 ) + 1.30 < \log([oiii]/h \beta),\ ] ] @xmath33/ha ) + 0.59 ) + 1.33 < \log([oiii]/h \beta)\\ ( or\,\ , \log([oi]/ha ) > -0.59)\end{aligned}\ ] ] and @xmath34/ha ) + 0.76 < \log([oiii]/h \beta),\ ] ] @xmath35/ha ) + 1.30 < \log([oiii]/h \beta).\ ] ] \4 . * liners * lie above the ke01 classification line on the , , and diagnostic diagrams and below the seyfert - liners line on the and diagrams , i.e. @xmath31/ha ) - 0.47 ) + 1.19 < \log([oiii]/h \beta)\ ] ] @xmath32/ha ) - 0.32 ) + 1.30 < \log([oiii]/h \beta),\ ] ] @xmath30/h \beta ) < 1.89 \log([sii]/ha ) + 0.76,\ ] ] @xmath36/ha ) + 0.59 ) + 1.33 & < \log([oiii]/h \beta)\\ ( or \,\ , \log([oi]/ha ) > -0.59)\end{aligned}\ ] ] @xmath30/h \beta ) < 1.18 \log([oi]/ha ) + 1.30,\ ] ] \5 . * ambiguous galaxies * are those that are classified as one type of object in one or or two diagrams and classified as another type of object in the remaining diagram(s ) . in our scheme , ambiguous galaxies fall into one of two categories : ( a ) galaxies that lie in the seyfert region in either the or diagram and in the liner region in the remaining ( or ) diagram , or ( b ) galaxies that lie in the composite region ( below the ke01 line ) in the diagram but that lie above the ke01 line in either the or diagram . according to this scheme , our 85224-galaxy sample contains 63893 ( 75% ) star - forming galaxies , 2411 ( 3% ) seyferts , 6005 ( 7% ) liners , and 5870 ( 7% ) composites . the remaining galaxies are ambiguous galaxies ( 7045 ; 8% ) . using the classifications obtained in the previous section , we investigate other line diagnostic diagrams that may be able to separate the different classes in a simpler way . figure [ oiiioii_oiha ] shows the vs diagnostic diagram for the -region like seyferts , liners , and composites . we exclude ambiguous objects from this plot . liners and seyferts form two distinct groups on this diagram and both groups are easily separated from the region - like galaxies and composites . the reason for this clean separation is twofold ; is a sensitive diagnostic of the ionization parameter of the gas , while is sensitive to the hardness of the ionizing radiation field . the ionization parameter is a measure of the amount of ionization that a radiation field can drive as it moves through the nebula . seyfert galaxies have a higher ionization parameter than liners ( by definition ) or star - forming galaxies and therefore seyferts separate vertically ( ) from the other classes of objects . both seyfert galaxies and liners have hard power - law ionizing radiation fields and thus separate from star - forming galaxies in the horizontal ( ) direction . interestingly , composite galaxies lie within the star - forming sequence on this diagram , indicating that the hardness of the ionizing radiation field , and the ionization parameter of composites are likely to be dominated by their star - formation . to calculate an empirical separation between liners , seyferts , and @xmath37composites , we find the minima of a 2-dimensional histogram of and . these minima are fit with a least - squares line of best - fit . the resulting empirical separations are shown in blue . for comparison , the @xcite classification line is shown ( purple dashed line ) . clearly the heckman classification scheme would separate most liners from seyfert galaxies . this diagnostic diagram is a more simple method for separating liners , seyferts , and star - forming galaxies ( including composites ) . our separations are given by ; + * & composites : * @xmath38/[oii ] < -1.701 \log([oi]/h\alpha)-2.163\ ] ] * liners : * @xmath39/h\alpha)-2.163 < \log ( [ oiii]/[oii])\ ] ] @xmath38/[oii ] < 1.0 \log([oi]/h\alpha ) + 0.7\ ] ] * seyferts : * @xmath40/h\alpha)-2.163 < & \log ( [ oiii]/[oii])\\ 1.0 \log([oi]/h\alpha ) + 0.7 < & \log ( [ oiii]/[oii])\end{aligned}\ ] ] note that the vs diagram should not be used to separate galaxies from composite objects and it relies on accurate reddening correction between [ ] @xmath7 and @xmath41 . in figure [ class_comp ] we compare our new classification scheme with previous methods of classification . the solid green line shows where the ( * ? ? ? * ; hereafter h80 ) liner classification line lies in relation to our seyfert - liner line . h80 defined liners as having @xmath42 [ ] @xmath7 and @xmath43 [ oiii ] @xmath44 . the h80 line has a very similar slope to our seyfert - liner line and is close to the 0.1 dex error markers for our line . the h80 liner criteria are sometimes used in combination with additional line ratio criteria using and/or , , to remove star - forming galaxies ( e.g. , @xcite ) . using our `` extreme starburst line '' to remove galaxies dominated by their star formation ( including composites ) , we find that all galaxies that are classed as liners using the h80 criteria are also classed as liners using our scheme . conversely , approximately 56% of our liners are also classified as liners using the h80 scheme . the remaining galaxies lie above the h80 line but below our seyfert - liner dividing line . if we do not remove star - forming galaxies and we impose the h80 criteria ( @xmath42 [ ] @xmath7 and @xmath43 [ oiii ] @xmath44 ) , then 15% of galaxies that are classed as liners using the h80 criteria are also classed as liners in our scheme . the remaining 85% are galaxies ( 68% ) , composites ( 7% ) , and ambiguous galaxies ( 9% ) . alternatively , @xmath4578% of our liners are also classed as liners using the h80 scheme . an alternative classification scheme was proposed by vo87 for galaxies where is not measured . the vo87 scheme is shown in figure [ class_comp ] ( black dotted lines ) . if our classification is correct , galaxies classified as liners according to the vo87 scheme are either true liners , seyferts , or composites . of the galaxies classified as liners in the vo87 scheme , 67% are liners , 5% are seyferts , 4% are composites , 0.01% are region - like , and 23% are ambiguous objects . these ambiguous objects are likely to be composites or transition objects because they lie within the composite region in one or two of the bpt diagrams and in the agn region in the remaining diagrams . on the other hand , approximately 87% of our liners are also classified as liners using the vo87 method . * ; hereafter hfs97 ) defined a new classification scheme for the four different classes of objects using the nuclear emission - line ratios of 418 galaxies . their classification scheme is shown on figure [ class_comp ] as green dot - dashed lines . the hfs97 liners criterion for and begins on our seyfert - liner classification line . because of this liner criterion , 92% of galaxies classed as liners in the hfs97 scheme are also liners in our scheme . the remaining 8% are ambiguous objects that lie within our liner region on one or more diagrams and within the seyfert region on the remaining diagrams . to conclude , because @xmath4590 - 100% of galaxies classified using h80 ( with reliable removal of galaxies ) or hfs97 as liners remain liners in our scheme , previous studies of liners defined according to h80 or hfs97 should reflect the true properties of liners . in contrast , @xmath3 of objects in liner studies that have used the vo87 scheme may be agn- composites and seyferts . without pre - removal of galaxies , the h80 scheme includes a substantial fraction ( 68% ) of galaxies . the relative ratio of composite , seyfert , and true liners will depend on any additional selection criteria used in previous liners surveys ( such as luminosity or color selection ) . , width=321 ] at redshifts @xmath46 we can detect galaxies with @xmath47 irrespective of their stellar mass - to - light ratios . thus we only consider galaxies with masses greater than @xmath48 in this analysis . we note that the fraction of galaxies with masses less than @xmath48 is small ( 4% ) . in figure [ classfrac_z ] , we show how the fraction of non- galaxies classed as seyfert , liner and composite changes over the redshift range of our sample . ambiguous galaxies are not shown . the fraction of liners falls with redshift from @xmath49 to @xmath50 in the @xmath51 mass range , @xmath52 to @xmath50 in the @xmath53 mass range , and @xmath54 to @xmath50 in the @xmath55 mass range . this drop in liner fraction is similar to the slope that @xcite found for low luminosity agn ( log(l[oiii])@xmath56 ) . the drop occurs because weak emission lines become increasingly difficult to detect as the physical aperture subtended by the sdss fibre becomes larger . over the redshift range @xmath57 , the change in liner fraction is not very large , so aperture bias should not affect our conclusions substantially . , width=321 ] figure [ comp_effects ] shows the overall redshift distribution , g - band fiber covering fraction , and concentration index of each spectral type . liners have redshift distributions skewed towards lower redshifts than seyferts or composites , as already seen in figure [ classfrac_z ] . figure [ comp_effects]c indicates that the liners in our sample are more concentrated than seyferts or composites . liners have only a slightly lower mean g - band fiber covering fraction for liners ( @xmath58 ) than for seyfert galaxies ( @xmath59 ) figure [ comp_effects]b ) . because the difference in g - band covering fraction between seyferts and liners is small , we proceed with our analysis of the host properties of liners and seyferts . to explore the host properties as a function of distance from the star - forming sequence , we define an empirical linear distance ( in dex ) from the star - forming sequence for both the seyfert and liner branches on the vs diagnostic diagram shown in figure [ sfdist ] a base point ( red and blue circles in figure [ sfdist ] ) is defined for each branch . a peak radius of 1.55 dex from the base point is defined for seyferts . the peak radius is 1.65 dex for liners . the distance between the peak radius and the base point for each branch is divided into ten bins . the position of the base and peak points have been chosen to ensure that the first and last bins contain atleast 50 data points . note that because is defined in dex in log line - ratio space , it does not give the fraction of star - formation or agn emission in a galaxy . in figure [ syl ] we show the relative ratio of seyferts and liners as a function of . in this plot , we include composite seyfert+ and liner+ ( classified using our seyfert - liner dividing line ) as seyferts and liners respectively . only pure star - forming galaxies classified according to the ka03 classification line ( equation 1 ) and liner / seyfert ambiguous galaxies are excluded from figure [ syl ] . the composite - pure agn boundary corresponds roughly to @xmath60 . figure [ syl ] shows that the ratio of seyfert to liners remains roughly constant with distance from the star - forming sequence . this result is unexpected because both seyfert- and liner - composites exist and there is a large luminosity difference between seyferts and liners . one might expect that in seyferts- composites , the stellar emission would be overwhelmed by the luminous seyfert emission while in liners , stellar emission could dominate in many liner- composites . if this expectation were correct , there should be many more liner- composites than seyfert- composites . the fact that the seyfert to liner ratio is roughly independent of @xmath61 ( figure [ syl ] ) implies that star formation and agn are coupled such that more powerful agn also have stronger star formation . , width=321 ] , width=321 ] using our new classification scheme , we explore the host properties of the agn in our sample . the host properties of agn in the sdss were previously studied by @xcite . they used the [ oiii ] luminosity to discriminate between seyferts and low - luminosity agn . the following analysis extends this work by comparing the host properties of liners , seyferts and composites as a function of distance from the star - forming sequence . we use the 4000 break ( @xmath62 ) and the h@xmath63 balmer absorption line as indicators of the age of the stellar population . the 4000 break is created by absorption lines located around 4000 and was first defined by @xcite , and recently redefined by @xcite as : @xmath64 the strength of the 4000 break is influenced by temperature and metallicity . as the temperature of the stellar atmospheres decreases with age , the metal opacity strengthens and 4 becomes large . d4000 was calibrated empirically by @xcite and theoretically by @xcite . the theoretical calibrations predict that d4000 provides a reliable estimate of the galaxy age for galaxies with mean stellar ages less than a few gyr . in the presence of current or recent star formation , d4000 is small because the metal opacity in the atmospheres of o and b stars weakens . a complementary age indicator is the equivalent width ( ew ) of the absorption line . the absorption line is produced in the atmospheres of a to g stars and is useful for predicting the age of bursts that ended 4myr - 1 gyr ago @xcite . we use the index defined by @xcite : @xmath65 where @xmath66 is the flux within the ( 4091.00 - 4112.25 ) bandpass and @xmath67 is the pseudo - continuum flux within the bandpass . , width=321 ] , width=321 ] the evolution of d4000 and have been investigated by @xcite using high resolution spectral libraries and the @xcite stellar population synthesis models . they find that neither index depends strongly on metallicity until at least 1 gyr after the burst . in figure [ d4000_sfdist]a , we show the normalized distribution of d4000 for liners ( red ) . for comparison , we plot the d4000 distributions for seyferts ( black ) , composites ( blue ) , and region - like galaxies ( green ) . the stellar population of liners is older than the other galaxy types . seyferts and composites have a similar range of stellar ages , and region - like galaxies are dominated by relatively young star - forming regions , as expected . to avoid biasing our analysis against galaxies that contain significant star formation , in figure [ d4000_sfdist]b we give the normalized distribution of d4000 without defining composites as a separate class . rather , we use our seyfert - liner classification line to divide the composite galaxies into + liner and + seyfert classes . we include + liner and + seyfert in the liner and seyfert classes respectively . only pure star - forming galaxies classified according to the ka03 classification line ( equation 1 ) and liner / seyfert ambiguous galaxies are excluded . comparison of panels ( a ) and ( b ) shows that while the difference between seyferts and liners is somewhat reduced if the composites are included , there is still a very clear age offset between the two classes . in figure [ d4000_sfdist]c , we show how the stellar population evolves as a function of for seyferts , composites , liners , and ambiguous galaxies . figure [ d4000_sfdist]d gives the median d4000 at each 0.1 interval of . composites are included in the liner and seyfert classes as in figure [ d4000_sfdist]b . pure liners and seyferts correspond roughly to a of 0.5 . the correlation coefficients and corresponding significance of the correlation are given in table [ corr ] . clearly and d4000 are strongly correlated for both liners and seyferts . both types of objects have young stellar populations typical of active star - forming galaxies at low , by definition . both seyfert and liner hosts have older stellar populations at large . liners hosts have an older stellar population than seyferts at all but the smallest . figure [ hdelta_sfdist ] shows that a similar trend holds for . ) for liners ( red dashed ) , seyferts ( black solid ) , composites ( blue dotted ) and region - like galaxies ( green dot - dashed ) . ( b ) the distribution of @xmath68 for liners ( red - dashed ) , seyferts ( black solid ) , and region - like galaxies ( green dot - dashed ) where liners and seyferts include liner+ and seyfert+ composites . ( c ) @xmath68 for a uniform random sampling of the seyferts ( black ) and liners ( red ) , composites ( blue ) and ambiguous galaxies ( purple ) as a function of distance from the ke01 line . ( d ) the median @xmath68 for seyferts ( black ) and liners ( red ) as a function of distance from the ke01 line , including seyfert+ and liner+ galaxies . a distance of 1.0 indicates that the optical line ratios are dominated by ionizing radiation from the seyfert or liner nucleus . [ mass_sfdist],width=321 ] the sdss stellar masses were derived by @xcite , and @xcite using a combination of @xmath69-band luminosities and monte carlo stellar population synthesis fits to d4000 and h@xmath63 . the model fits provide powerful constraints on the star formation history and metallicity of each galaxy , thus providing a more reliable indicator of mass than assuming a simple mass - to - light ratio . @xcite compared @xmath70 sdss spectroscopically derived masses with ( a ) masses derived from population synthesis fits to the broadband sdss and 2mass colors , and ( b ) masses calculated from sdss velocity dispersions and effective radii . they concluded that the three methods for estimating mass agree to within @xmath71 dex over the @xmath72 range . the stellar mass distribution of sdss emission - line galaxies was studied previously by @xcite . they find a strong dependence between the fraction of galaxies containing agn and stellar mass . figure [ mass_sfdist]a confirms this result . the stellar mass distribution of star - forming galaxies ( green dot - dashed line ) is shifted substantially towards lower masses compared to the mass distributions of composites , seyferts , or liners . figure [ mass_sfdist]a shows that the stellar mass of liner galaxies is slightly higher than for seyfert galaxies . composite objects have a very similar mass distribution to seyfert galaxies but this may be a coincidence ; only a small fraction ( 14% ) of composites lie above the seyfert - liner dividing line on both and diagrams . the majority of composites ( 67% ) lie on the -liner sequence and therefore these galaxies may simply have stellar masses intermediate between region - like galaxies and liners . the median stellar mass of the composite galaxies ( @xmath73 ) is within 0.02 dex of the average of the median stellar mass for region - like and liner classes ( 10.52 ) . figure [ mass_sfdist]d shows that the stellar mass is roughly constant as a function of . there is only a small correlation between sf distance and stellar mass for seyferts and a stronger correlation for liners ( table [ corr ] ) . to investigate how the mass of galaxies at a given observed luminosity changes between the different spectral types , we use the median z - band mass - to - light ( m / l ) ratios . the m / l ratio of galaxies is influenced by the age of the stellar population and the attenuation by dust . , width=321 ] the sdss m / l ratios were derived by @xcite using monte carlo simulations of different star formation histories . they find that the m / l ratio correlates strongly with galaxy luminosity such that luminous galaxies have high m / l ratios while faint galaxies have a wide range of m / l ratios . @xcite used stellar evolution models to predict how m / l changes as a function of galaxy luminosity , and stellar population age . they show that the majority of luminous galaxies have m / l ratios consistent with having formed stars at constant rate over a hubble time , and that faint galaxies have formed most of their stars closer to the present . this effect is reflected in figure [ ml_sfdist]a where we plot the normalized distribution of m / l for star - forming galaxies , liners , seyferts and composites . seyferts , liners and composites have similar m / l distributions , with liners having only slightly larger m / l ratios on average than seyferts or composites . the median m / l ratio is 0.21 dex for liners , 0.18 dex for seyferts and 0.16 dex for composites ( the standard error of the median is small ; 0.002 - 0.004 dex ) . the m / l range of region - like galaxies is substantially lower than for the other galaxy types , including composites . the m / l range of region - like galaxies is consistent with their having had present star formation and substantial dust as expected . in region - like galaxies , the m / l ratio is correlated with metallicity . the relatively high m / l ratio of composites may therefore result from their metal - rich stellar population . the m / l ratio does not change as a function of ( figure [ ml_sfdist]d ) . the stellar velocity dispersion @xmath74 in a galaxy containing an agn is related to the black hole mass @xcite . figure [ vel_sfdist]a shows the @xmath74 distribution for each of the galaxy types and the relationship between @xmath74 and . liners have a broader distribution of than seyferts or composites . when composite liner+ and seyfert+ galaxies are included in the liner and seyfert classes ( figure [ vel_sfdist]b ) , it is clear that the distribution for liners has a larger tail towards high values . figure [ vel_sfdist]d also shows that is roughly constant with for seyfert galaxies , with only a slight increase in ( @xmath16 km / s ) , indicating that black hole mass is relatively constant along the seyfert branch but that black hole mass may be somewhat larger at high . we show that is strongly correlated with the optical line ratios , or for pure liners . ) for liners ( red dashed ) , seyferts ( black solid ) , composites ( blue dotted ) and region - like galaxies ( green dot - dashed ) . ( b ) the distribution of @xmath74 for liners ( red - dashed ) , seyferts ( black solid ) , and region - like galaxies ( green dot - dashed ) where liners and seyferts include liner+ and seyfert+ composites . ( c ) @xmath74 for a uniform random sampling of the seyferts ( black ) and liners ( red ) , composites ( blue ) and ambiguous galaxies ( purple ) as a function of distance from the ke01 line . ( d ) the median @xmath74 for seyferts ( black ) and liners ( red ) as a function of distance from the ke01 line , including seyfert+ and liner+ galaxies . a distance of 1.0 indicates that the optical line ratios are dominated by ionizing radiation from the seyfert or liner nucleus . [ vel_sfdist],width=321 ] in figure [ loiii_sfdist]a we show the extinction - corrected [ ] luminosity distribution for each spectral class . liners have a lower median [ ] luminosity than any of the other spectral types ; @xmath75)})=6.1 $ ] for liners and @xmath75)})=6.3 , 6.5 $ ] , and 7.3 for region - like , composites and seyferts respectively . seyferts have substantially larger l[oiii ] than the other spectral classes from the hard ionizing continuum produced by their agn . composite galaxies have a slightly higher median l [ ] than region - like galaxies and substantially higher median l[oii ] than liners . as discussed in section [ mass ] , the majority of composites lie on the liner branch , but 15% lie on the seyfert branch . the large l[oiii ] luminosity of these seyferts raises the median l[oiii ] for composites . the median [ ] luminosity rises as a function of for seyferts , the rise in l [ ] implies that the ionizing radiation field is harder for seyferts with high . @xcite found that the l[oiii ] and hard x - ray ( 3 - 20 kev ) luminosities of agn are correlated over four orders of magnitude . in liners , the median [ ] luminosity falls as a function of until @xmath76 after which it rises slowly . the drop in l [ ] between 0 - 0.6 occurs where + liner composites are included in the liner class . the star - formation in these + liner composites is contributing to the [ ] emission until 0.6 . luminosity ( ) for liners ( red dashed ) , seyferts ( black solid ) , composites ( blue dotted ) and region - like galaxies ( green dot - dashed ) . ( b ) the distribution of l[oiii ] for liners ( red - dashed ) , seyferts ( black solid ) , and region - like galaxies ( green dot - dashed ) where liners and seyferts include liner+ and seyfert+ composites . ( c ) l[oiii ] for a uniform random sampling of the seyferts ( black ) and liners ( red ) , composites ( blue ) and ambiguous galaxies ( purple ) as a function of distance from the ke01 line . ( d ) the median l[oiii ] for seyferts ( black ) and liners ( red ) as a function of distance from the ke01 line , including seyfert+ and liner+ galaxies . a distance of 1.0 indicates that the optical line ratios are dominated by ionizing radiation from the seyfert or liner nucleus . [ loiii_sfdist],width=321 ] in figure [ bd_sfdist ] we plot the balmer decrement for our sample . the median balmer decrement of liners is significantly smaller than for seyferts , region - like galaxies and composites . a substantial fraction ( 45% ) of liners have balmer decrements even less than the theoretical value of 3.1 , and 33% of liners have balmer decrements less than the theoretical value for galaxies dominated by star formation ( 2.86 ) . as discussed in section [ sample ] , a balmer decrement less than the theoretical value can result from a combination of : ( 1 ) intrinsically low reddening , ( 2 ) errors in the stellar absorption correction , and ( 3 ) errors in the line flux calibration and measurement . on the other hand , balmer decrements lower than 3.1 in agn can indicate a higher nebular temperature @xcite . liner spectra have strong stellar absorption and errors in the absorption correction could bias the balmer decrement for liners , particularly in spectra with low s / n . we note that many liners in our sample have low signal - to - noise ( s / n ) ratios ; 11% of seyferts have s / n()@xmath77 , while 70% of liners have s / n()@xmath78 . this low signal - to - noise may cause the low balmer decrement in some or perhaps most liners but it can not account for the difference for the entire liner sample because liners with high s / n()@xmath79 have a similar distribution of balmer decrements to those with lower s / n ratios ( figure [ bd_hist ] ) . @xcite found a similar result for the central few hundred parsecs of seyferts and liners and they suggest that seyfert galaxies contain nuclear regions that are more gas - rich ( and dusty ) than liner galaxies . our results confirm this conclusion . /h@xmath21 ) for liners ( red dashed ) , seyferts ( black solid ) , composites ( blue dotted ) and region - like galaxies ( green dot - dashed ) . ( b ) the distribution of h@xmath20/h@xmath21 for liners ( red - dashed ) , seyferts ( black solid ) , and region - like galaxies ( green dot - dashed ) where liners and seyferts include liner+ and seyfert+ composites . ( c ) h@xmath20/h@xmath21 for a uniform random sampling of the seyferts ( black ) and liners ( red ) , composites ( blue ) and ambiguous galaxies ( purple ) as a function of distance from the ke01 line . ( d ) the median h@xmath20/h@xmath21 for seyferts ( black ) and liners ( red ) as a function of distance from the ke01 line , including seyfert+ and liner+ galaxies . a distance of 1.0 indicates that the optical line ratios are dominated by ionizing radiation from the seyfert or liner nucleus . [ bd_sfdist],width=321 ] . even liners with high s / n have lower balmer decrements than the other spectral types . [ bd_hist],width=321 ] the balmer decrement is not strongly correlated with distance from the star - forming sequence for liners ( figure [ bd_sfdist]b ; @xmath80 ) , but it is mildly correlated with @xmath61 for seyferts ( @xmath81 with a formally zero probability of obtaining this value by chance ) in the sense that seyferts at large distances from the star - forming seqence contain @xmath82 dex less dust on average than seyferts that are close to the star - forming sequence . given that many of the the host properties of of seyferts and liners are substantially different ( d4000 , , , etc ) , it is important to understand whether this difference is a truly bimodal distribution or whether the host properties vary smoothly across the seyfert - liner dividing line . we define @xmath83 as the angle to the x - axis in the vs diagram centered on the blue base point in figure [ arcs ] . therefore liners have small @xmath83 while seyferts have large @xmath83 . figure [ angle ] shows how the host properties of agn change across the seyfert - liner boundary ( red dashed lines ) . the host properties clearly do not form a bimodal distribution ; they form a smooth sequence from seyfert values to liner values . the correlation coefficients and significance for each panel are given in table [ corr ] . the ( a proxy for the eddington ratio ) correlates most strongly with @xmath83 . we will discuss as a function of host properties in the following section . from the x - axis in the vs diagnostic diagram . the seyfert - liner boundary is shown as red dashed lines . seyferts lie to the right of the boundary while liners lie to the left of the boundary . [ angle],width=321 ] the accretion rate on the black hole is proportional to the bolometric luminosity of the agn . the stellar velocity dispersion can be used as an estimate of the black hole mass @xcite . in type i seyferts and quasars , the intrinsic [ ] luminosity scales with the agn bolometric luminosity ( see @xcite for a more detailed discussion ) . it follows that the quantity is proportional to @xmath1 . in figure [ loiiisig_sfdist ] we show that does not change as a function of distance from the star - forming sequence for seyfert galaxies . the ratio of liners is influenced by star - formation ( in ) for between 0 - 0.6 , as in figure [ loiii_sfdist ] . pure seyferts and liners ( @xmath85 ) have substantially different values of . figure [ loiiisig_sfdist]d indicates that is roughly constant with for seyferts . both and @xmath86 ( black hole mass ) rise with for seyferts , resulting in a constant . the rise in and @xmath86 with probably reflects the trend that older , more massive galaxies contain larger black holes . /@xmath0 ( corrected for extinction ) for liners ( red dashed ) , seyferts ( black solid ) , composites ( blue dotted ) and region - like galaxies ( green dot - dashed ) . ( b ) the distribution of l[oiii]/@xmath0 for liners ( red - dashed ) , seyferts ( black solid ) , and region - like galaxies ( green dot - dashed ) where liners and seyferts include liner+ and seyfert+ composites . ( c ) l[oiii]/@xmath0 for seyferts ( black ) and liners ( red ) , composites ( blue ) and ambiguous galaxies ( purple ) as a function of distance from the ke01 line . ( d ) l[oiii]/@xmath0 for seyferts ( black ) and liners ( red ) as a function of distance from the ke01 line , including seyfert+ and liner+ galaxies . a distance of 1.0 indicates that the optical line ratios are dominated by ionizing radiation from the seyfert or liner nucleus . [ loiiisig_sfdist],width=321 ] we show how the host galaxy properties of liners , seyferts and composites vary as a function of in figure [ loiiisig ] . remarkably , the relationship between and all of the host properties for both seyferts and liners forms one smooth sequence . for values exhibited by both seyferts and liners ( @xmath87@xmath88 ) , the median of the seyfert and liner distributions is almost identical for every host property . this result implies that ( 1 ) most ( if not all ) liners in our sample are agn , and that ( 2 ) is the major fundamental property that divides seyferts from liners and that host property differences are secondary . for example , liners are found in larger , older galaxies than seyferts . figure [ loiiisig ] shows that the black hole accretion rates are lower in larger , older galaxies and that mass and age do not need to be physically linked to the differences in the liner and seyfert emission line spectra . we check whether a similar effect occurs as a function of in figure [ loiiimulti ] . all of the relations between host properties and degrade when is used in place of , confirming that is the major property separating seyferts from liners . , width=321 ] , width=321 ] previous work has shown that that almost all normal galaxies contain black holes and black holes appear to evolve with their host galaxies @xcite . in section [ ledd ] we showed that although seyferts and liners display different host properties , these differences are largely resolved when the host properties are considered as a function of . thus the strongest difference between seyferts and liners is , or eddington ratio . remarkably , figure [ loiiisig ] implies the that if the eddington ratio of an agn is known , then its spectral class and broad host galaxy properties can be deduced . in figure [ angle ] , we show that the transition in eddington ratio from seyferts ( high ) to liners ( low ) is smooth and that the correlation between classification angle and is stronger than for any of the other host galaxy properties . this seyfert - liner transition may be analogous to the transition between high state and low state black hole accretion in x - ray binary systems ( e.g. , @xcite and references therein ) as suggested by @xcite . in x - ray binaries the transition from high to low states is believed to correspond to decreasing mass accretion rates and therefore the primary parameter in many x - ray binary models is the eddington ratio ( e.g. , @xcite ) . theory suggests that an unstable accretion disk may switch between high and low states @xcite . in this scenario , the energy released by viscous processes in the high state is radiated and the gas is relatively cool . in the low state the viscously released energy may be advected with the gas and radiates inefficiently . this advection dominated accretion flow ( adaf ) is stable to thermal and viscous instabilities @xcite . because the gas loses little energy through radiation in adaf models , the gas is extremely hot and the weak spectrum produced is harder than in the high state ( e.g. , @xcite ) . alternative scenarios to adaf models can also produce a weak hard spectrum ( * ? ? ? * ; * ? ? ? * e.g. , ) . our observed transition between seyfert and liners occurs at @xmath89 . assuming a bolometric correction of @xmath90}_{o}$ ] @xcite where @xmath91 is the observed [ ] luminosity , and assuming a mean extinction of @xmath92 , our seyfert - liner transition corresponds to an eddington ratio of @xmath93 . this transition @xmath94 is within the range of the observed ( and theoretical ) transition in x - ray binaries ; @xmath95 ( e.g. , @xcite and references therein ) . we investigate the broad , qualitative differences in radiation field between seyferts and liners using the theoretical agn photoionization models from mappings iii @xcite . we calculate a simple grid of photoionziation models using four power - law indices ( @xmath96 ) and four ionization parameters ( @xmath97 ) . the mappings models are self - consistent and include detailed dust physics and radiation pressure . we use a metallicity of 2and a hydrogen density of 1000 @xmath98 . figure [ models_oiiihb_oiha ] shows the position of our agn relative to the seyfert and liner branches on the vs diagnostic diagram . the seyfert branch can be reproduced by an ionizing radiation field with power - law ndex @xmath99 to -1.8 and a high ionization parameter ( @xmath100 to -1.0 ) . the liner branch spans a larger range of power - law index , requiring a harder radiation field than seyferts ( @xmath101 ) at log()@xmath102 . the liner branch requires a much lower ionization parameter than the seyfert branch ; the ionization parameter for liners ( @xmath103 ) is up to an order of magnitude lower than the ionization parameter typical for seyferts ( @xmath104 to -2.5 ) . figure [ models_oiiioii_oiha ] confirms this result and separates ionization parameter from power - law index over a broader range of ionization parameter . . the model grids are given for four ionization parameters ( @xmath105 ) and four power - law indices ( @xmath96 ) as marked . the radiation field in liners requires a lower ionization parameter than in seyfert galaxies and liners with strong require a harder radiation field than seyferts . [ models_oiiihb_oiha],width=321 ] . the model grids are given for four ionization parameters ( @xmath105 ) and four power - law indices ( @xmath96 ) as marked . the radiation field in liners requires a lower ionization parameter than in seyfert galaxies and liners with strong require a harder radiation field than seyferts . [ models_oiiioii_oiha],width=321 ] figures [ models_oiiihb_oiha ] and [ models_oiiioii_oiha ] indicate that the hardness of the ionizing radiation field increases as a function of distance from the star - forming sequence . we know that seyferts contain a young - intermediate age stellar population ( figures [ d4000_sfdist ] , [ hdelta_sfdist ] ) and that both seyfert- and liner - composites exist . given these results , one would expect that the stellar emission would be overwhelmed by the luminous seyfert emission in seyfert- composites but that stellar emission could dominate in many liner- composites . in this case , there should be many more liner- composites than seyfert- composites . the fact that the seyfert to liner ratio is independent of @xmath61 ( figure [ syl ] ) implies that star formation and agn are coupled in seyfert galaxies such that more powerful agn also have stronger star formation . further theoretical and observational work are required to confirm this connection between star formation and seyfert emission . we have analyzed the emission - line and host properties of 85224 galaxies from the sloan digital sky survey . we show that seyferts and liners form two separate branches on the standard optical diagnostic diagrams . we present a new optical classification scheme that successfully separates purely star - forming galaxies , seyferts , liners , and composite agn@xmath37star - forming galaxies . we show that vs diagnostic diagram easily discriminates between seyferts , liners and galaxies dominated by star - formation ( including composites ) . we use our new classification scheme to investigate the host properties of agn galaxies as a function of distance from the star - forming sequence . we find that : * the ratio of seyferts to liners is invariant with distance from the star - forming sequence . this result implies that star formation and agn power are coupled . * the stellar populations of seyferts and liners age with distance from the star - forming sequence . liners have a substantially older stellar population than the other spectral types . the youngest stellar population in liners has d4000 and values consistent with the oldest stellar population in seyferts on average . * the stellar mass and mass - to - light ratios of seyferts and liners are not strongly correlated with distance from the star - forming sequence . however , liners have slightly higher stellar masses and mass - to - light ratios on average than seyferts or composite objects . we use as an indicator of the accretion rate in eddington units . we compare the host properties of seyfert and liner galaxies as a function of and . we show that : * the strongest difference between seyfert and liner galaxies is . * when the host properties of seyferts and liners are considered in terms of , the differences in host properties is resolved and the host properties of seyferts and liners form a smooth sequence with . * when the host properties of seyferts and liners are considered in terms of , the smooth sequence seen with disappears . these results indicate that most ( if not all ) liners contain agn , and that the major fundamental difference between seyferts and liners is accretion rate . we use theoretical agn photoionization models to show that the liner branch requires a lower ionization parameter ( up to an order of magnitude ) than the seyfert branch , and that strong liners ( log()@xmath106 ) require a harder ionizing radiation field than galaxies on the seyfert branch . these results suggest that the transition between seyferts and liners is analogous to the high - state and low - state transition in x - ray binaries . l. j. kewley is supported by a hubble fellowship . this research has made use of nasa s astrophysics data system bibliographic services . funding for the creation and distribution of the sdss archive has been provided by the alfred p. sloan foundation , the participating institutions , the national aeronautics and space administration , the national science foundation , the u.s . department of energy , the japanese monbukagakusho , and the max planck society . the sdss web site is http://www.sdss.org/. the sdss is managed by the astrophysical research consortium ( arc ) for the participating institutions . the participating institutions are the university of chicago , fermilab , the institute for advanced study , the japan participation group , the johns hopkins university , los alamos national laboratory , the max - planck - institute for astronomy ( mpia ) , the max - planck - institute for astrophysics ( mpa ) , new mexico state university , university of pittsburgh , princeton university , the united states naval observatory , and the university of washington . , 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. , & tremaine , s. 2000 , apjl , 539 , l13 , g. , heckman , t. m. , tremonti , c. , brinchmann , j. , charlot , s. , white , s. d. m. , ridgway , s. e. , brinkmann , j. , fukugita , m. , hall , p. b. , ivezi ' c , . , richards , g. t. , & schneider , d. p. 2003 , mnras , 346 , 1055 , g. , heckman , t. m. , white , s. d. m. , charlot , s. , tremonti , c. , brinchmann , j. , bruzual , g. , peng , e. w. , seibert , m. , bernardi , m. , blanton , m. , brinkmann , j. , castander , f. , cs ' abai , i. , fukugita , m. , ivezic , z. , munn , j. a. , nichol , r. c. , padmanabhan , n. , thakar , a. r. , weinberg , d. h. , & york , d. 2003 , mnras , 341 , 33 , g. , heckman , t. m. , white , s. d. m. , charlot , s. , tremonti , c. , brinchmann , j. , bruzual , g. , peng , e. w. , seibert , m. , bernardi , m. , blanton , m. , brinkmann , j. , castander , f. , csbai , i. , fukugita , m. , ivezic , z. , munn , j. a. , nichol , r. c. , padmanabhan , n. , thakar , a. r. , weinberg , d. h. , & york , d. 2003 , mnras , 341 , 33 , d. e. 1989 , astrophysics of gaseous nebulae and active galactic nuclei ( research supported by the university of california , john simon guggenheim memorial foundation , university of minnesota , et al . mill valley , ca , university science books , 1989 , 422 p. ) , d. n. , verde , l. , peiris , h. v. , komatsu , e. , nolta , m. r. , bennett , c. l. , halpern , m. , hinshaw , g. , jarosik , n. , kogut , a. , limon , m. , meyer , s. s. , page , l. , tucker , g. s. , weiland , j. l. , wollack , e. , & wright , e. l. 2003 , apjs , 148 , 175 , 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 , apj , 574 , 740

we study the nonlinear rheology of a glass - forming binary mixture under the reversal of shear flow using molecular dynamics simulations and a schematic model of the mode - coupling theory of the glass transition ( mct ) . memory effects lead to a history - dependent response , as exemplified by the vanishing of a stress - overshoot phenomenon in the stress strain curves of the sheared liquid , and a change in the apparent elastic coefficients around states with zero stress . we investigate the various retarded contributions to the stress response at a given time schematically within mct . the connection of this macroscopic response to single - particle motion is demonstrated using molecular - dynamics simulation .

1211.6964

dense liquids exhibit slow relaxation processes in thermal equilibrium , characterized by a time scale @xmath0 that diverges as the liquid is driven to structural arrest into an amorphous solid @xcite . this opens a window rich in nonlinear phenomena under flow , both amenable to experiment and theoretical treatment . it is of interest , because the externally imposed flow strongly disturbs the equilibrium relaxation processes . more precisely , given a typical single - particle relaxation time scale @xmath1 , we consider shear flow with deformation rate @xmath2 such that @xmath3 , but @xmath4 . this nonlinear rheology has in particular been studied in colloidal dispersions and soft - matter systems , where the mesoscopic size of the constituent particles implies relaxation times of @xmath5 , easily probed by moderate flow in experiment . in the flowing steady state , shear thinning is the most prominent phenomenon : @xcite the apparent viscosity of the sheared fluid decreases rapidly with increasing shear rate @xmath2 . quite generally , the flowing steady state is characterized by a nonlinear relation between the macroscopic stress @xmath6 and the applied shear rate @xmath2 . going beyond the steady state , the behavior under start - up flow has received particular attention . as a steady shear flow of rate @xmath2 is suddenly switched on at @xmath7 , stresses @xmath6 increase from zero ( assuming the experiment was started in an equilibrated and hence stress - free configuration ) to their steady - state value @xmath8 . the resulting stress - strain curve @xmath9 , measured as a function of applied deformation @xmath10 for @xmath11 , exhibits a peculiar intermediate maximum , termed stress overshoot . this is typically found in amorphous metals @xcite , or ( for different reasons ) polymer melts @xcite . the maximum of the @xmath9 curve has been linked to material stability : it can be taken as a yield stress that separates reversible ( linear and nonlinear ) elastic deformation at small strains from irreversible plastic flow induced by larger deformations . more recently , stress overshoots have also been discussed in soft colloidal systems , underlining a generic mechanism applicable in a wide variety of system classes @xcite . as a transient phenomenon , the stress overshoot is susceptible to the history of sample preparation . molecular dynamics simulations have addressed this focusing on the shear - molten amorphous solid , where a systematic dependence was found on the waiting time @xmath12 between sample preparation and flow start - up @xcite , as long as the start - up configuration is not fully equilibrated , @xmath13 . in the engineering literature it is well established that the stress - strain curve depends on the deformation history of the sample . one particular example was found empirically by johann bauschinger in the 1880 s during experiments on various steels @xcite : these materials typically are softer under compressional load after having been subject to tension . this is , in the strict sense , called the bauschinger effect . as an abstraction , take the case where pre - shear is applied in one particular direction , and consider the modified stress strain curve measured if shear is then applied in the opposite direction . typically , the regime of elastic deformation shrinks , so that plastic rearrangements dominate much earlier than in an initially unstrained sample . reversing the shear rate , which we will call bauschinger effect also , in the following , has been discussed @xcite in the framework of the shear - transformation - zone theory ( stz ) , where one attributes internal degrees of freedom to the stzs , which carry the memory of the deformation history . recent simulation studies of the bauschinger effect @xcite in a system quenched into its low - temperature glassy state have been analyzed in terms of anisotropic elastic constants that arise due to the deformation history . it is a matter of current debate whether and how the rheological properties of amorphous materials deep in the glassy state ( here referred to as essentially the @xmath14 case ) , and at much higher temperaturs in the melt , are related . in athermal systems , concepts like stzs @xcite and avalanches of yielding events @xcite are widely recognized as useful to rationalize macroscopic behavior . but the range of applicability of these approaches is not well understood when dealing with liquids close to their glass transition , where thermal fluctuations dominate yielding mechanisms that may be qualitatively different from , say , athermal jamming @xcite . the rheology of dense liquids is the realm of a recent extension of the mode - coupling theory of the glass transition ( mct ) to sheared colloidal systems @xcite . intriguingly , some of the macroscopic phenomena are quite alike in both quasi - athermal systems and liquids driven by thermal noise . here we investigate , both by mode - coupling theory and computer simulation , the nonlinear rheology under flow reversal in the initially equilibrated liquid state , without referring to @xmath14 properties . the initial stress overshoot is shown to vanish if the flow is reversed in the steady state . hence , the bauschinger effect does not require an explanation associated with properties of the quenched glass or aging phenomena . we study its connection to the time - dependent memory of past flow that is typical for dense liquids , by investigating the effect of flow reversal applied at various times during the initial evolution from equilibrium to steady state . a clear separation between reversible elastic rearrangements and irreversible plastic flow emerges . we connect the macroscopic stress response to microscopic particle motion , quantified through the transient mean - squared displacements . the paper is organized as follows : section [ sec:1 ] contains the exposition of the theoretical approach , and section [ sec:2 ] gives an overview of the simulation . in sec . [ sec:3 ] we discuss the comparison between theory and simulation , and the insights thereby gained . section [ sec:4 ] concludes with an outlook . our theoretical approach is based on the mode - coupling theory of the glass transition ( mct ) @xcite , as extended to deal with nonlinear rheology in the integration - through transients ( itt ) formalism @xcite . itt - mct is a microscopic ( albeit approximate ) theory describing the dynamical correlation functions of the wave - vector dependent density fluctuations in glass - forming liquids . suitably generalized green - kubo relations provide the nonequilibrium transport coefficients in the nonlinear - response regime ; within mct , they are expressed as integrals over the density correlators . the coupling coefficients of the theory are fully determined by knowing the equilibrium static structure of the liquid . the theory then predicts an ideal glass transition to occur as a bifurcation transition : smooth changes in control parameters such as density or temperature bring about a discontinuous change in the long - time behavior of the density correlation functions . these decay to zero in the equilibrium liquid , but attain a finite long - time limit , the glass form factor , in the ideal glass . the theory compares semi - quantitatively to the linear rheology measured in colloidal hard sphere dispersions @xcite . since the full microscopic theory is rather cumbersome to solve , various simplifications have been devised . schematic models try to capture mathematically the essence of the bifurcation transition by reducing the number of correlation functions to a few or only one , at the cost of neglecting spatial information . the important correlations contained in the schematic models are temporal ones . memory effects are contained in generalized friction kernels depending on ( up to ) three times . causality is automatically enforced , and the flow history of the sample is contained in the friction - kernels via time- or strain - dependent elastic coefficients and structural relaxation functions . the latter describe plastic deformation , and their nonlinear equations of motion are the heart of the theory . the non - markovian equations of the schematic models contain a small number of fitting parameters that mimic the relation between the equilibrium structure ( encoding the interaction potential ) and the theory s coupling coefficients . smooth and regular variations of the parameters lead to dramatic elasto - plastic and viscoelastic changes in the structural relaxation and non - linear rheology . the interpretation of the effects in terms of spatially dependent particle rearrangements can not be achieved by the schematic model , and thus it remains unresolved to which degree force - chains , heterogeneities , or specific plastic rearrangement centers are described in a ( presumably ) space - averaged manner . a schematic mct model for the time - dependent rheology of colloidal suspensions was introduced in ref . [ ] ; it has been used to study large - amplitude oscillatory shear @xcite , mixed flows @xcite , and double - step strain @xcite , and generalizes a model used to analyze stationary flow curves @xcite . in the model , the nonlinear generalization of the green - kubo relation for the shear - stress tensor element @xmath15 reads , imposing spatially homogeneous but otherwise arbitrary simple shear flow with rate @xmath2 in the @xmath16-direction , with gradient in the @xmath17-direction , @xmath18)\,.\label{eqn : greenkubo}\ ] ] the generalized shear modulus @xmath19)$ ] depends on the full flow history encoded in the time - dependent shear rate @xmath20 . outside the steady state , it also depends on the two times corresponding to the underlying fluctuations separately . in the spirit of mct , we approximate the shear modulus in terms of the transient density correlation function @xmath21 , @xmath22 here , an elastic coefficient @xmath23 appears that was set to a constant in the original formulation of the model @xcite . it measures the strength of stress fluctuations caused by pair - density fluctuations . the microscopic theory implies a relation between @xmath21 and @xmath24 that involves an integration over wave vectors , including nontrivial weights that depend on time since density fluctuations are advected by shear . it is therefore plausible to extend the original schematic model by allowing for a prefactor @xmath25 that depends on time through the accumulated strain @xmath26 , @xmath27 symmetry dictates that the direction of the strain does not matter , leading to an even function @xmath28 . equation generalizes the time - dependent coupling first introduced for constant shear rates in ref . [ ] , where it was justified by a comparison with fully microscopic itt - mct calculations in two dimensions . the functional form is chosen simple enough for rapid numerical implementation and allows for a quantitative discussion of stress overshoots in model colloidal suspensions @xcite . the parameter @xmath29 sets the scale of the generalized shear modulus . we obtain it by matching it to the linear - response low - frequency plateau modulus @xmath30 of the simulation . the shear elastic constant @xmath31 itself often is called ` shear modulus ' , and can , e.g. , be measured in the low - frequency linear elastic modulus of the glass , @xmath32 , or as one of the two lam - coefficients of an isotropic solid @xcite . at a critical strain scale @xmath33 , the time - dependent modulus becomes negative and the stress - strain curve exhibits an overshoot with its peak at the critical strain . @xmath34 describes the decay of the overshoot . to reduce the number of free parameters , we keep it linearly proportional @xcite to @xmath35 . the time - dependent elastic coefficient given in eq . is carefully chosen to become negative in a small strain window relevant for the decay of the correlation functions . doing so , it allows to circumvent an obvious limitation of the schematic model : setting @xmath36 , eq . has a definite sign . in startup shear , cf . , this implies that @xmath37 varies monotonically with accumulated strain . the original schematic model hence does not describe stress overshoot phenomena . motivated by the full microscopic expression of itt - mct @xcite and the observation that the wave - vector dependent mct , even with further isotropy assumptions , produces a stress overshoot through a small negative dip in @xmath24 , cf . ref . [ ] , eq . is a simple way to incorporate the phenomenology . in the following , we take it as one established way to match the startup stress strain curves ( see below for a comparison to simulation ) , and discuss the consequences of the history - integral formulation of eq . under flow reversal . the schematic density correlator is the solution of a mori - zwanzig - like memory equation @xmath38 with initial condition @xmath39 . here @xmath40 denotes the initial decay rate of the density correlator , which corresponds to a reciprocal microscopic relaxation time . we take @xmath40 to be independent of the shear rate . to close this equation of motion , mct approximates the memory kernel @xmath41 through a nonlinear polynomial of the correlation functions . in contrast to a markovian approximation , this postulates that the kernel relaxes on the same time scale as the correlator , and that both need to be obtained self - consistently . conceptually , this is motivated by the assumption that stress fluctuations captured in @xmath41 arise from slow structural rearrangements , which themselves are captured in @xmath42 . the simplest schematic model that recovers the asymptotic behavior found for typical quiescent glass - forming liquids is the f@xmath43 model @xcite , whose extension to time - dependent shear reads @xmath44.\ ] ] the parameters @xmath45 and @xmath46 are the coupling coefficients driving the system through the glass transition . there is a line of bifurcation points @xmath47 , and we pick @xmath48 and @xmath49 with @xmath50 as in previous work . the parameter @xmath51 controls the distance to the glass transition : @xmath52 corresponds to liquid states , where the correlation function decays to zero in the quiescent system , on a time scale @xmath0 that diverges as @xmath53 . states with @xmath54 are idealized glass states , where the quiescent correlation function attains a positive long - time limit , @xmath55 . this value is called the non - ergodicity parameter or glass form factor . the elastic constant in the model then follows as @xmath56 . one of the predictions of itt - mct is that ( steady ) shear melts the glass ; the correlation functions even in the glass show shear - induced decay due to a loss of memory @xcite . in the schematic model , this is modeled by an ad - hoc strain - reduction function , @xmath57 symmetry again dictates that @xmath58 is an even function of the accumulated strain . the parameter @xmath59 is a critical strain and describes how fast the memory of the glass decays due to its deformation . in principle , the microscopic form of the mct equation suggests further combinations of the three times to appear in eq . , as exploited for example in refs . [ ] and [ ] . we neglect these terms here , to make contact with earlier schematic - model analyses of the nonlinear steady - state rheology of colloidal suspensions @xcite . in principle , the strain - reduction function @xmath58 could also contain negative regions , as arises in the stress - density coupling @xmath23 . this could cause the transient correlator @xmath42 to become negative for intermediate times as seen in computer simulations and microscopic itt - mct calculations in two dimensions @xcite . for simplicity , this is neglected . linear plot of the stress - strain relation @xmath9 at fixed density and shear rate for various flow histories : startup flow with constant shear rate ( red ) , flow reversal in the steady state ( s , purple ) , from the elastic regime ( el , blue ) , and from the point of the stress overshoot ( max , green ) . the linear increase with the quiescent elastic modulus @xmath31 in startup , and with @xmath60 after shear - reversal are indicated by dotted lines . characteristic stress strain points , whose transient modulus is discussed in the text , are marked by symbols : the red square gives the stationary stress . it corresponds to an integration along the line from a to b in fig . [ fig : timeplane ] . triangles mark stresses after flow reversal in the linear regime and correspond to an integration along the line c to e. discs mark stresses after flow reversal in the overshoot regime and are obtained via integrations along the line c to e in fig . [ fig : timeplane ] . ] the flow history we consider in this work is given by two intervals of constant shear rate : @xmath61 we mainly discuss a fixed value for @xmath2 , and different waiting times @xmath62 over which the `` pre - shear '' in the positive direction is applied . these times correspond to an accumulated strain @xmath63 . the amount of this pre - strain is decisive and determines the stress strain curve after switching . figure [ fig : hysteresis ] presents typical results obtained in the schematic model as unmodified stress - versus - strain curves ; it serves to introduce the flow history and characteristic parameters . after startup of shear at @xmath7 , the system first undergoes an elastic transient , characterized by a stress strain curve @xmath9 that is approximately linear , @xmath64 , given by the quiescent elastic shear modulus . at the end of this regime , the @xmath6-versus-@xmath65 curves becomes sublinear , until the position of the overshoot is reached at some strain @xmath66 , typically of the order of @xmath67 . at large strains , one reaches the steady state characterized by a constant @xmath68 independent of @xmath65 . for the flow reversal , we consider switching the shear rate at three typical points in either regime : @xmath69 inside the elastic transient , @xmath70 , and @xmath71 in the steady state . the @xmath72 values are marked in fig . [ fig : hysteresis ] . a value @xmath73 in the schematic model turns out to be sufficient so that no further @xmath72 dependence is observed for larger values besides a trivial shift . to ease the interpretation of the curves at different switching times , we choose @xmath69 such that the corresponding stress is equal to the steady - state one , @xmath74 . generally , after reversing the flow direction , the stress first decreases back to zero , defining a strain value @xmath75 . the stress strain curves in later figures will be compared in a @xmath76-versus-@xmath77 representation . while it does hide the `` hysteresis '' loop which can be recognized in fig . [ fig : hysteresis ] , and which is familiar in the engineering literature , it simplifies the quantitative analysis . all calculations in the schematic model are done with parameters that have been found typical in applications of predecessors of the model to large amplitude oscillatory shear @xcite , and especially to startup flow @xcite . the values are : @xmath78 , @xmath79 , @xmath80 , and @xmath81 ; see also fig . [ fig : stressstrain ] . the considered state in fig . [ fig : hysteresis ] , which will be analyzed in more detail in figs . [ fig:3erpanel_for_reversal_in_steadystate ] and [ fig:3erpanel_for_reversal_in_elastic_regime ] , is a glass very close to the glass transition in order to achieve a large separation of the structural dynamics from the short - time one ; @xmath82 . the shear rate @xmath83 , for which the bauschinger effect is studied , corresponds to a rather small peclet number , which in the model is given by pe@xmath84 , and thus takes the value @xmath85 . these two values provide access to the asymptotic regime of itt - mct for @xmath53 and pe@xmath86 . slightly different @xmath51 and @xmath2 values are used in the comparisons to the simulations in sec . [ sec:3 ] . figure [ fig : hysteresis ] contains a wealth of information , which will be discussed in sec . [ sec:3 ] together with the simulation results . we perform nonequilibrium molecular dynamics computer simulation for a glass - forming binary mixture . particles interact through a purely repulsive soft - sphere potential . denoting particle species by @xmath87 , the truncated lennard - jones potential due to weeks , chandler , and andersen @xcite reads @xmath88 & \text{$r < r_{c,\alpha\beta}$}\\ 0 & \text{else } \end{cases}\,,\ ] ] where @xmath89 is a cutoff for the interaction range . to ensure continuity of force and conservation of total energy in the nve ensemble , we apply a smoothing function , @xmath90 with @xmath91 $ ] with @xmath92 @xcite . we choose units of energy such that @xmath93 , and units of length such that @xmath94 . the unit of time is given by @xmath95 where @xmath96 are the masses of the particles . for the smaller particles , @xmath97 , and the mixture is additive , @xmath98 . this system , albeit with differing masses , has been previously studied in the quiescent state and identified as a model glass former in its equimolar composition @xcite . the system is coupled to a dissipative particle dynamics ( dpd ) thermostat : the equations of motion read @xcite [ eom ] @xmath99 where @xmath100 and @xmath101 are positions and momenta of the particles , and @xmath102 denotes conservative ( @xmath103 ) , dissipative ( @xmath104 ) , and random ( @xmath105 ) forces . with the interparticle separation @xmath106 between particles @xmath107 and @xmath108 of species @xmath109 and @xmath110 , @xmath111 here , @xmath112 is the relative velocity between the two particles , and @xmath113 their unit separation . the appearance of the relative velocity in the dissipative force is crucial to obtain local momentum conservation and galilean invariance @xcite . @xmath114 is a cutoff function , set to @xmath115 for @xmath116 and @xmath117 elsewhere . we have chosen @xmath118 in our simulations . @xmath119 is a parameter controlling the strength of friction forces ; we set @xmath120 . this parameter is not crucial for the results to be discussed @xcite . the @xmath121 are gaussian normal random variables . the dpd equations of motion , eq . , are integrated with a generalized velocity verlet algorithm @xcite . this algorithm integrates the reversible hamiltonian part of the equations with the velocity verlet scheme and then partially re - equilibrates the two - particle momenta , ensuring a boltzmann distribution in equilibrium . a time step of @xmath122 ( in the specified units of time ) is employed . two different neighbor lists are implemented , one for the force calculations , and one for the dpd thermostat . the simulation consists of @xmath123 particles in a three - dimensional box with volume @xmath124 and linear dimension @xmath125 , corresponding to a number density @xmath126 . at this density , no signs of crystallization or phase separation were observed in the studied temperature range from @xmath127 to @xmath128 . a glass transition point is estimated as @xmath129 according to mode - coupling theory . in the following we focus on the equilibrated fluid at @xmath128 . initial equilibration proceeded by using @xmath130 , assigning new velocities every 50 integration time steps . equilibration was checked by the decay of the incoherent intermediate scattering function at a wave number corresponding to a typical interparticle separation @xmath131 : runs were long enough to observe the decay of the correlation function to zero at long times . a set of @xmath132 independently equilibrated configurations served as initial configurations for the production runs employing the dpd thermostat . shear is applied in the @xmath16-direction with a gradient in @xmath17-direction and a positive shear rate @xmath2 initially . planar couette flow is imposed by periodic lees - edwards boundary conditions @xcite : the periodic image of a particle leaving or entering the simulation box in @xmath17-direction is displaced in @xmath16-direction according to the strain @xmath133 . the macroscopic response of the system to imposed shear flow is measured through the stress tensor . in the specified coordinate system , the dominant contribution at low shear rates is its @xmath134-component , given by the kirkwood formula @xcite , @xmath135\right\rangle\,.\ ] ] the first term is a kinematic contribution , while the second is the virial contribution incorporating the nonhydrodynamic forces exerted on each particle . angular brackets indicate canonical averages . in the following , we restrict the discussion of the bauschinger effect to one , rather high , exemplary shear rate , @xmath136 . we begin by a discussion of the properties of the computer - simulated system under startup of flow and in its steady state . qualitatively , they reproduce earlier results @xcite . here , they serve to determine the schematic model parameters prior to investigation of the bauschinger effect . plot of the stress - strain curves @xmath9 at fixed temperature for various shear rates . the dashed lines show the mct results for @xmath78 , @xmath137 , @xmath79 , @xmath138 , and @xmath139 . @xmath140 is @xmath141 for decreasing shear rates @xmath2 . ] figure [ fig : stressstrain ] shows the stress strain curves @xmath9 obtained after switch - on of steady shear at @xmath7 , for various shear rates @xmath2 at one particular temperature in the supercooled liquid . for small @xmath2 , the system behaves as a linear viscoelastic fluid , and the stress quickly reaches its steady state value , rising monotonically from zero . as the shear rate is increased , an intermediate overshoot in the @xmath6-versus-@xmath65 curve appears . depending on @xmath2 , the typical strain reached at the position of the overshoot is a few percent ; at the highest shear rates this value is about @xmath67 . it is in agreement with earlier observations @xcite , and one can argue that this typical strain corresponds to a typical size of nearest - neighbor cages that are being broken by shear @xcite . from the large-@xmath65 limit of the stress strain curves , one obtains the steady - state flow curve , @xmath142 . the values extracted from our simulation and the schematic - model fits exhibit shear - thinning : the stress increases sublinearly with the rate . note in fig . [ fig : stressstrain ] that the values of @xmath6 are given in simulation units , where @xmath143 . this confirms that the stresses we observe are dominated by thermal motion , since @xmath144 . this compares well to recent investigations of colloidal suspensions @xcite . the precise numerical value is found to differ among different systems , as it somewhat depends on the interaction potential and mixture parameters in experimental samples . linear plot of the stress - strain relation @xmath145 at fixed temperature and shear rate for various flow histories : starting from equilibrium ( eq , red line ) , after flow reversal in the steady state ( s , purple ) , from the elastic regime ( el , blue ) , and from the point of the stress overshoot ( max , green ) . reversal at @xmath73 is in the steady state as seen from comparing with reversal at @xmath146 ( black data ) . smooth lines result from the schematic model . ] the solid lines in fig . [ fig : stressstrain ] are fits with the schematic mct model including the time - dependent vertex for the green - kubo relation . these fits were optimized to reproduce the shape of the overshoot at the largest shear rate ( for which the bauschinger effect will be discussed below ) . further adjustments to improve agreement at the lower shear rates may well be possible , but note that the form of the overshoot is dictated by the empirical choice of @xmath25 . the parameters of the schematic mct model can be determined by fitting the simulation data in startup shear . using a fit procedure similar to the one outlined in ref . [ ] we arrive at the values listed in the caption of fig . [ fig : stressstrain ] . the parameters @xmath40 and @xmath51 determine the short- and long - time relaxation scales @xmath1 and @xmath0 in the model . they are fixed by a comparison of the quiescent equilibrium correlation functions at a typical intermediate wave number corresponding to the first peak of the static structure factor . linear - response fluctuations also allow to determine @xmath147 , corresponding to the plateau modulus @xmath31 and hence setting the units of stresses in the model . the parameters @xmath59 , @xmath35 , and @xmath148 are then tuned to the startup rheology data , taking the shear rate @xmath2 nominally from the simulation . once the parameters given in the caption of fig . [ fig : stressstrain ] have been fixed , the predictions of the schematic model regarding the modification of the stress strain curve under flow reversal are parameter - free . we now turn to the discussion of the bauschinger effect . figure [ fig : bauschinger ] shows the main result of the computer simulation , viz . the stress strain curves in the @xmath76-versus-@xmath77 representation , i.e. , shifted and inverted such that they all start from a stress - free state and display the behavior as a function of the additional strain imposed on this configuration . curves for various @xmath72 are shown , as discussed above . they all agree at large strains by construction , confirming that the steady - state stress does not depend on the shear history . comparing first the startup curve , @xmath149 , with the one for reversal in the steady - state , @xmath150 , the bauschinger effect becomes most apparent . the two curves differ in two main aspects : the linear slope at small deformations is lower , and the overshoot is gone in the curve corresponding to oppositely straining the ( steadily ) pre - sheared configuration . this holds for shear - reversal throughout the stationary state , as shown by the two curves for @xmath73 and @xmath146 included in fig . [ fig : bauschinger ] . both these differences arise because the system undergoes plastic deformations during pre - shear : confining the pre - shear to small strains , @xmath151 , so that flow reversal takes place inside the initial elastic - deformation regime , the overshoot is maintained almost unchanged , as is the effective elastic coefficient extracted from the initial rise of the stress strain curve . the cross - over between elasticity - dominated and plasticity - dominated pre - strain occurs gradually , as the curve for @xmath152 exemplifies . the solid lines in fig . [ fig : bauschinger ] represent the schematic - mct model calculations . the line corresponding to @xmath149 is a result of the fitting procedure described above . all the other theory curves then follow from the structure of the itt - mct equations . they describe the computer - simulation data extremely well , and capture both the decrease of the overshoot and the decrease of the effective elastic coefficient . velocity profiles @xmath153 as a function of position along the gradient direction @xmath17 for various times after application of a new flow direction . ( a ) starting from a quiescent configuration ( eq ) , shear in @xmath154 direction . ( b ) starting from the resulting steady state ( s ) at @xmath155 , shear in @xmath156 direction . insets mark the points for which profiles are shown along the stress - strain curves . a total @xmath157 independent configurations are averaged within a small time window to obtain the depicted graphics . ] velocity profiles @xmath153 as a function of position along the gradient direction @xmath17 for various times after application of a new flow direction . ( a ) starting from a quiescent configuration ( eq ) , shear in @xmath154 direction . ( b ) starting from the resulting steady state ( s ) at @xmath155 , shear in @xmath156 direction . insets mark the points for which profiles are shown along the stress - strain curves . a total @xmath157 independent configurations are averaged within a small time window to obtain the depicted graphics . ] before continuing the discussion , it is worth pointing out that none of the features we discuss appear to be explained by , or connected to shear banding or other inhomogeneous flow effects . to demonstrate this , we show in fig . [ fig : profiles ] the velocity profiles obtained from the simulation at various instants in time , both under forward shear and the subsequently reversed shear . to improve statistics , the instantaneous velocities were averaged over all the configurations and over a small time window of width @xmath158 . insets in the figure mark the points for which the velocity profiles @xmath153 are shown ; they all follow the linear behavior @xmath159 expected from homogeneous simple shear , to within statistical noise . similar results have also been obtained under startup shear @xcite . it appears that the small time needed to establish a linear velocity profile under lees - edwards boundary conditions is not relevant for our discussion . in fig . [ fig : profiles ] only the case for flow reversal in the steady state is shown , but qualitatively the same result is found for the velocity profiles obtained after switching the shear direction at the smaller @xmath70 or @xmath69 . stress overshoot phenomena have been linked to shear banding @xcite , but it appears that this mechanism is not relevant for our system . indeed , our results are compatible with a recent study suggesting a density - dependent critical pclet number for the appearance of shear bands @xcite . to understand the history dependence of the stress strain curves , let us split the integral in eq . into two parts . for @xmath160 , write @xmath161 with the summands @xmath162 in the last term , @xmath163 , the condition @xmath164 is fulfilled , so that the correlation and strain functions entering the integral are dependent on one - time only and describe the transient between an equilibrium and a stationary state . hence , this contribution is trivially related to the startup curve , @xmath165 ; see below in sect . [ sec:4c ] for details . to simplify the discussion , let us consider an ad - hoc generalization of the maxwell model of viscoelastic liquids to shear - thinning fluids . keeping the form of the green - kubo equation , we approximate the generalized shear modulus , eq . , by @xmath166 for the case of constant shear and constant @xmath23 , this form shows the same qualitative features as the shear - molten glass in schematic mct @xcite . to generalize this nonlinear maxwell model to unsteady flows is neither trivial nor unique , but assuming the rate of decorrelation to be set by the instantaneous shear rate , the above ansatz appears plausible , if somewhat crude @xcite . in the present case it even simplifies further because @xmath167 . the integrals determining @xmath37 and its two contributions @xmath168 and @xmath169 can then be solved analytically when inserting the form of @xmath25 given above . in order to quantitatively match the startup stress strain curve of the generalized maxwell model with the schematic model , we set @xmath170 . bauschinger effect as illustrated by a generalized maxwell model ( see text ) . solid lines are @xmath37 as a function of normalized time , @xmath171 , for the case @xmath172 corresponding to @xmath70 and @xmath173 corresponding to @xmath71 . dashed lines display the contributions @xmath174 ( upper , red ) and @xmath163 ( lower , blue ) . ] the response for large @xmath62 can now be understood , recalling that the integral appearing in @xmath174 is dominated by @xmath175 , where the integrand is still non - vanishing . the accumulated strain entering the correlation functions can then be written as @xmath176 . hence , @xmath168 becomes a function of @xmath177 only , and the precise time of switching the flow becomes irrelevant ( as it takes place in the steady state after startup ) . this is confirmed by numerical evaluation and also in the simulation , by varying @xmath71 . at @xmath178 , @xmath168 is nothing but the value of the startup curve ; as @xmath179 , this contribution vanishes since the functions in the integrand will generally decay . furthermore , at @xmath180 , a maximum occurs as the relevant strain @xmath181 when the integral is dominated by large @xmath182 . this maximum cancels out the corresponding minimum from @xmath169 , so that the stress overshoot is generically expected to vanish . this is shown in fig . [ fig : maxbausch ] for the case @xmath173 . for @xmath183 , however , the integral determining @xmath168 will have two relevant contributions , namely @xmath175 as well as @xmath184 , since the relevant strain at both these times equals @xmath59 . hence , @xmath174 in this case displays first a maximum and then a minimum , as exemplified in fig . [ fig : maxbausch ] for the case @xmath172 . the minimum is responsible for @xmath174 not to cancel the stress overshoot contained in @xmath163 . obviously , as @xmath185 , the contribution @xmath174 vanishes , so that perfect elastic recovery is obtained in this limit . while the generalized maxwell model simplifies the interpretation , the next sections show that broadly equivalent curves follow from the present schematic model , which provides a fundamental basis for this simplification . it exhibits a two step process , where the final decay recovers the properties of eq . . linear stress responses @xmath186 from simulation , fits of the schematic model , and maxwell model as labeled for various waiting strains @xmath187 . the inset gives the strain @xmath188 required for the stress to reduce to zero after shear reversal . to include the simulation point for @xmath189 , this value has been divided by 200 , using that the precise value of @xmath72 is irrelevant in the steady state.,title="fig : " ] we now turn to a discussion of the effective shear moduli , obtained as @xmath186 . they measure the effective elasticity remaining until the shear - driven relaxation sets in . figure [ fig : geffrelover ] displays the results from the schematic mct model ( red and blue stars ) together with the values extracted from the simulation data ( blue crosses ) . the agreement is quite satisfactory , noting that only the initial value at @xmath149 is a result of a fitting procedure . one observes a marked decrease of @xmath60 around @xmath190 , the position of the overshoot in the startup curve . hence , pre - shear indeed softens the material , but only by plastic deformation . also shown in fig . [ fig : geffrelover ] are the results from the generalized maxwell model ; qualitatively the same trend is seen . in the maxwell model , we could not find a closed expression to determine @xmath75 ; but the effective shear modulus defined as @xmath191 for @xmath192 can be evaluated analytically . it again follows the trend seen in the figure , albeit with an even lower limiting value at large @xmath72 . a missing information on the simulated stress - strain curve remains @xmath75 , the strain value where the stress - free state is achieved after strain reversal . its qualitative behavior can be understood easily . assuming that @xmath60 is unchanged for small @xmath72 , the relations @xmath193 and @xmath194 hold . as a result , @xmath195 . linearity holds even if @xmath60 changes , but the prefactor is somewhat different , as observed in the simulations . for large @xmath72 , @xmath196 , and the constant value in fig . [ fig : geffrelover ] is indeed roughly @xmath197 . view on the @xmath198-plane , with the three different time regimes for the density correlator @xmath42 . as the @xmath58-function depends only on the absolute value of @xmath2 , the one - time - solutions @xmath199 and @xmath200 are equal . ] to understand the history dependence of the stress strain curves within the schematic itt - mct model requires to analyze the time - dependence of the correlator which encodes plastic deformations , and to combine it with the one of the elastic coefficient @xmath23 , which encodes strain - dependent anelasticity . the structure of eq . and the time - dependence of the accumulated strain suggests to split the integration domain into three regimes , as shown in fig . [ fig : timeplane ] . it sketches the corresponding integration paths in the @xmath198plane leading to the stresses shown in fig . [ fig : hysteresis ] using eq . immediately , the strain follows from eq . ( remembering that @xmath201 ) : @xmath202 which identifies the three regimes . an integration for @xmath203 gives the startup stress strain curve , the integration for @xmath204 gives the stress contribution @xmath205 in eq . , and the integration for @xmath206 the contribution @xmath207 in eq . . recall that the correlators appearing in the mct model are transient correlation functions , i.e. , they are formed with the equilibrium distribution function always , although the full nonequilibrium dynamics appears in their propagator . a correlation function @xmath21 hence contains only information on the flow history in the interval @xmath208 $ ] . in particular , the correlation functions defined for @xmath209 are given by the startup transient solution @xmath210 for @xmath211 of the model ; the same holds true for those defined for @xmath164 , since there the shear rate is constant and the direction of shear can not enter by symmetry . because the strain in these two time regimes depends on the time difference only ( see eq . ) , the correlator @xmath212 also depends on the time difference only . this greatly reduces the complexity of eq . and allows the use of well - established integration schemes to solve @xmath213\ ; . \label{einzeiteneom}\end{aligned}\ ] ] here , @xmath214 abbreviates the time difference . the correlation function for @xmath204 however remains fully two - time dependent , as here the strain depends on @xmath215 . it is convenient to split the integral appearing in eq . explicitly at @xmath216 , @xmath217 here , @xmath218 is the same memory kernel also appearing in eq . . hence , in this formulation , one - time correlators or memory kernels previously calculated enter most succinctly . supplied with suitable initial conditions , eq . is a convenient way to deal with instantaneous shear - rate switches numerically @xcite . ( a ) shear moduli @xmath24 after shear reversal at @xmath219 in the steady state , shown for three different fixed times @xmath220 ( respectively strains @xmath65 ) , as functions of @xmath198 . curves marked with violet triangle , disc , and diamond symbols correspond to points marked with the same symbol in fig . [ fig : hysteresis ] . for times @xmath221 , all curves collapse by construction onto the startup result @xmath222 shown in red ( marked with a square ) . the inset shows the corresponding correlators @xmath223 ( they overlap on the resolution of the figure ) . ( b ) corresponding stress integrands @xmath224 on a linear @xmath198 axis . curves after shear reversal which would be negative for @xmath225 are mirrored to positive initial decay , where they overlap with the startup curve . ] evaluation of a stress @xmath226 requires integrations of the transient shear modulus along lines at constant @xmath220 in the plane shown in fig . [ fig : timeplane ] . the startup curve is obtained along lines of the type from a to b before the shear reversal , where all functions depend on @xmath198 . figures [ fig:3erpanel_for_reversal_in_steadystate ] and [ fig:3erpanel_for_reversal_in_elastic_regime ] include the transient density correlators , the transient shear moduli and the integrands of eq . for this case . as discussed in detail in refs . [ , , ] , the linear regime in the stress strain curve is connected to the plateau on intermediate times in @xmath227 and @xmath228 , and the stress - overshoot to the negative dip in @xmath228 at late times . reversing the direction of the shear at a late time when the stationary state was reached at @xmath219 , correlators and shear moduli change as shown in fig . [ fig:3erpanel_for_reversal_in_steadystate ] . three different final states are chosen along the reversing stress - strain curve : one in the elastic regime , one in the regime where the stress - maximum is now missing , and one in the late steady regime , where the steady state is reached again . obviously , the last curves agree with the startup curves ( except for the trivial minus sign in the integrand ) , as the same transient into a steady state is probed . here the integration path from c to d , where the shear rate takes one definite ( negative ) value , dominates and the time region , where the shear rate was different , affects the very final transients , only . moving the final time @xmath220 to correspond to a strain @xmath65 of around 10% , where the maximum in the startup curve appears , the transient shear modulus is affected in the negative region . there the strong variation in @xmath25 of eq . around @xmath35 becomes important . along the integration path from d to e , the memory of the @xmath229 shear enters , which causes the sign change in @xmath24 ; see the curve marked with a circle in panel ( a ) of fig . [ fig:3erpanel_for_reversal_in_steadystate ] . the integrand of the generalized green - kubo relation eq . varies around zero , see panel ( b ) of fig . [ fig:3erpanel_for_reversal_in_steadystate ] , in the region where the stress overshoot arose . this leads to a cancellation in the integral , and an absence of the maximum in the stress . shifting the final time @xmath220 to earlier values in the region , where the startup stress varies linearly with strain , causes the sign change in @xmath24 to move to earlier times @xmath182 as well ; see the curve marked with a triangle in panel ( a ) of fig . [ fig:3erpanel_for_reversal_in_steadystate ] . a stronger memory of the shear rate with opposite sign remains . as this arises especially during the final relaxation , the region in @xmath24 where stresses are anticorrelated , the overall integral over the transient shear modulus is smaller . as @xmath220 , respectively @xmath65 , lies in the region , where the linear increase in the stress occurs , this explains the softening of the effective elastic constant @xmath60 ; see the linear slope indicated in fig . [ fig : hysteresis ] at the violet triangle . importantly , in the itt - mct approach it arises not by a weakening of the plateau in the transient correlation functions , rather from the aftereffect of the previously stored stresses accumulated during the flow with opposite ( positive ) shear rate . it thus originates from the final relaxation and does not hold in the proper limit of infinitesimal strain . we chose @xmath230 to measure @xmath231 . like fig . [ fig:3erpanel_for_reversal_in_steadystate ] , but now for shear reversal at @xmath232 in the elastic regime . the curves end at a time @xmath233 , where the memory has not decayed to zero yet . this is as expected , because the system still responds elastic like before the early reversal at @xmath232 ; see the blue curve in fig . [ fig : hysteresis ] . ] the equivalent discussion holds for reversing the flow during the elastic regime of the startup stress - strain curve . this is indicated by a blue curve in fig . [ fig : hysteresis ] , and the corresponding transient functions are shown in fig . [ fig:3erpanel_for_reversal_in_elastic_regime ] . again a triangle , circle , and diamond marks final states on the reversing stress strain curve in the elastic , overshoot , and steady regime . the integration paths in the @xmath234-plane of fig . [ fig : timeplane ] do not change qualitatively , but the time @xmath62 of switching is now so short that the startup correlators along the line a to b have not decayed to zero . the system responds elastically , and the correlator @xmath210 and the modulus @xmath235 are still of the order of the plateau value , @xmath236 respectively @xmath31 . the latter value determines the linear increase of the startup stress - strain curve . the complete discussion of the changes in @xmath24 done in respect to fig . [ fig:3erpanel_for_reversal_in_steadystate ] carries over to fig . [ fig:3erpanel_for_reversal_in_elastic_regime ] with the sole difference that the two - time region d to e ( and d to e ) dominates at first . by continuity , it reproduces the stress @xmath237 , where the flow reversal took place . because the strain increases with @xmath215 in this region according to eq . , this contribution decays rapidly along the reversing stress - strain curve . it may be neglected beyond @xmath238 , where the initial elastic stress is destroyed . the remaining contribution along the lines c to d or c to d is equivalent to the startup flow along a to b , and thus after shear rate reversal in the elastic regime , the stress - strain curve agrees closely with the initial startup curve . in panel ( a ) , linear stress response @xmath186 as function of waiting strain @xmath187 in a fluid and a glass state ; for the latter , two different shear rates are shown . the quiescent elastic shear constant @xmath31 is observed in startup flow . + in panel ( b ) , the corresponding curves are given for the relative stress overshoots @xmath239.,title="fig : " ] the preceding discussion of the bauschinger effect at two typical waiting or reversal times @xmath62 corresponding to strains in the elastic and steady regime enables one to rationalize the stress - strain curves for all pre - shears @xmath72 . figure [ fig : geffrelovertheo ] summarizes the main findings for two different temperatures close to the glass transition in terms of the effective shear constant @xmath60 and the relative amplitude of the stress - overshoot @xmath239 . both change from their quiescent values as soon as @xmath72 becomes of the order of @xmath35 , where the transient shear modulus captures anticorrelated stress fluctuations . the variation of @xmath60 saturates when @xmath240 , because then stress correlations have become negligible ; then also the overshoot @xmath239 is gone . increasing the static correlations in the system , mimicked by @xmath51 in the schematic model , increases the elasticity . because bare peclet number pe@xmath86 are considered , no qualitative difference holds for a liquid and a glassy state . decreasing the shear rate in the glass , the discussed effects remain and just become somewhat smaller . in a fluid state , the linear response regime would be approached as soon as the dressed peclet or weissenberg number pe@xmath241 becomes of order unity . this situation where no overshoot arises @xcite and thus flow reversal only mildly affects the stress - strain curve is not shown here . to relate the macroscopic response discussed so far to the microscopic dynamics , we turn to a discussion of simulation results for the density correlation functions under shear . figure [ fig : fsqt_shear ] shows as exemplary cases the self - intermediate scattering functions ( tagged - particle density - correlation functions ) @xmath242 in the quiescent equilibrium ( @xmath243 , dashed ) , under startup flow ( @xmath244 , dash - dotted ) , and after flow reversal from the steady state ( @xmath245 using @xmath246 , solid lines ) . each curve has been sampled by averaging over 200 initial configurations , and we checked that stationarity is achieved comparing with reversal at @xmath189 . to avoid a discussion of shear - advection effects , wave vectors were chosen with zero component in the shear direction . the equilibrium curves exemplify the typical two - step relaxation found in glass - forming liquids : after an initial fast relaxation to an intermediate plateau , the final decay to zero is characterized by a large time scale @xmath247 . this final decay is much more stretched than exponential relaxation . the plateau value depends on the wave number @xmath248 , and for the correlation functions characterizing tagged - particle motion generically decreases with increasing @xmath248 . steady shear accelerates the dynamics , and induces relaxation from the plateau on a time scale set by @xmath249 ( with a @xmath248-dependent prefactor ) . this relaxation is much closer to exponential , and , for the transient correlation function shown here , even slightly compressed . these qualitative features confirm earlier observations @xcite . by symmetry , the same is true for the correlation functions probing the reversed - flow regime ( solid lines in the figure ) . however , as in the macroscopic response , a pronounced history dependence is seen : the intermediate plateau is much less pronounced , and the final shear - induced decay shows much more pronounced stretching , in particular at large @xmath248 . simulation results of the self intermediate scattering functions @xmath250 for the b - species in the two directions perpendicular to shear during startup and after shear reversal in the stationary state , for wave vectors @xmath251 $ ] in descending order . a total 200 independent simulational runs are averaged to obtain these graphics . the corresponding equilibrium functions are included ( dashed lines , label eq ) , shifted to coincide at short times . ] in discussing these dynamics , one has to emphasize the difference of these measurable quantities to the transient correlation functions used in mct . the theory defines correlation functions as averages over the equilibrium distribution ; they have the advantage to be better suited for subsequent approximations . in the simulation , such correlation functions can only be measured ( beyond the quiescent equilibrium ) under startup of steady shear from an equilibrated configuration . this is the case shown in fig . [ fig : fsqt_shear ] by the dash - dotted lines , which can be compared qualitatively in view of the lack of a @xmath248-dependence of the latter to correlators from the schematic model included in the inset of fig . [ fig:3erpanel_for_reversal_in_steadystate ] . for the case of flow reversal ( solid lines ) , the simulation implies averaging over the steady - state distribution corresponding to forward flow , and not the quiescent one . we observe that this quantity shows changes already in the relaxation around the plateau , and not just for features of the final relaxation . ( a ) mean - squared displacement of b particles in the vorticity direction . shown are the equilibrium curve for @xmath128 ( label eq ) , and transient msds @xmath252 as a function of @xmath253 , where @xmath182 is the time of initial flow start - up from equilibrium ( eq ) respectively flow reversal according to the cases discussed in fig . [ fig : stressstrain ] ( steady - state , s ; elastic transient , el ; strain corresponding to the overshoot , max ) . the equilibrium curve is shifted to coincide at short times . ( b ) logarithmic derivative @xmath254 for the cases shown in ( a ) . ] ( a ) mean - squared displacement of b particles in the vorticity direction . shown are the equilibrium curve for @xmath128 ( label eq ) , and transient msds @xmath252 as a function of @xmath253 , where @xmath182 is the time of initial flow start - up from equilibrium ( eq ) respectively flow reversal according to the cases discussed in fig . [ fig : stressstrain ] ( steady - state , s ; elastic transient , el ; strain corresponding to the overshoot , max ) . the equilibrium curve is shifted to coincide at short times . ( b ) logarithmic derivative @xmath254 for the cases shown in ( a ) . ] a related quantity that can be intuitively interpreted is the mean - squared displacement ( msd ) , @xmath255 , of a single particle . we choose randomly a particle of species b , and show the results in fig . [ fig : msd ] . here we show , in addition to the cases @xmath256 , @xmath244 , and @xmath257 discussed above , also the msds obtained for @xmath258 and @xmath259 . for simplicity , we only record the motion in the neutral direction ; the msd is then defined as @xmath260 to match the spatially averaged msd of the quiescent system at short times . @xmath261 is expected to become diffusive at long times , while the other components of the msd under shear attain superdiffusive asymptotes due to ( the high - density analog of ) taylor dispersion @xcite . it has been established by computer simulation and confocal microscopy studies @xcite that the stress - overshoot in startup flow is accompanied by superdiffusive particle motion on intermediate time scales . this is confirmed by the startup case shown in fig . [ fig : msd ] : while the equilibrium msd remains subdiffusive at all times , the transient msd obtained for @xmath244 follows the equilibrium one up to a time that corresponds to a few percent strain . then , it quickly crosses over to the diffusive regime obtained under steady - state flow . an effective exponent can be assigned to the msd curves by taking the logarithmic derivative , @xmath262 . this quantity is shown in the lower panel of fig . [ fig : msd ] . since the short - time motion of the simulation is ballistic , @xmath263 holds for @xmath264 ; in the plateau region , @xmath265 drops to values close to zero , and for long times , @xmath266 indicates diffusive motion . for times where the fast increase in the @xmath244 msd is seen , @xmath267 holds , in qualitative agreement with the simulation of ref . remarkably , the correlation between stress overshoot and superdiffusive motion of the individual particles holds beyond the case of startup . in fig . [ fig : msd ] , we observe superdiffusive msds for those cases , where the corresponding stress strain curve in fig . [ fig : bauschinger ] shows an overshoot ; for the case of flow reversal in the steady state , where the stress overshoot has vanished , also only subdiffusive motion is seen in the msd . this hints at a strong connection between the two phenomena . such a connection can be rationalized @xcite by invoking a generalized stokes - einstein relation : the mean - squared displacement is determined by a mori - zwanzig equation similar to eq . . the relevant memory kernel can be approximated by the generalized time - dependent shear modulus appearing in the green - kubo relation for the stress , eq . . although it does not provide a sufficient condition , one sees that a memory kernel with negative portions in the msd equations may lead to superdiffusive behavior . both stress - overshoot and superdiffusive msd would then arise in the same time window characterized by the strain @xmath35 . we have discussed the history - dependent response of sheared glass - forming liquids subject to shear flow whose direction is instantaneously reversed . starting from the equilibrated liquid , application of a constant shear rate @xmath2 causes the resulting stress @xmath6 to increase from zero to a steady - state value , undergoing an intermediate maximum , the stress overshoot . flow reversal , or equivalently , application of a constant shear rate to a system that is in steady state corresponding to the flow in opposite direction , results in a stress - versus - time curve that has no maximum . at the same time , the ( effective ) elastic shear modulus found from the initial rate of change of the stress is lowered , and the regime in applied strain shrinks over which the behavior is described by this initial linear elastic regime . these coincident phenomena can be seen as analogous to the bauschinger effect in simple shear flow . it should be pointed out that the original bauschinger effect @xcite was discussed for metals and steels under compressional and tensile load . however , given the generality of history - dependent flow phenomena in amorphous systems , one may expect that the underlying mechanisms are related to our discussion . we have presented a systematic investigation of the gradual change between the two cases of starting shear from a quiescent equilibrium and a sheared steady state that belongs to the opposite flow . reversing the flow direction after an initial deformation that remains in the linear elastic regime , no change in the stress overshoot or the elastic coefficients the position of the overshoot indeed , as envisaged already by bauschinger @xcite , marks the end of the reversible elastic regime : at this point , plastic deformation gradually takes over , and the effective elastic coefficients drop over a small window in applied pre - strain . our theoretical analysis follows the integration - through - transients ( itt ) scheme upon which the mode - coupling theory for colloidal rheology is based . the central physical mechanism expressed in our model is one of temporal history dependence : glass - forming liquids are visco - elastic , with a large time window @xmath0 over which the fate of past density fluctuations affects the present response . it is this history dependence that causes the bauschinger effect and similar pre - shear - dependent phenomena in the model , rather than spatial anisotropies induced by the pre - shear ( since no information on spatial variation of fluctuations is kept ) . this is in alignment with the microscopic mct and molecular - dynamics simulations of the sheared thermal fluid @xcite , where it was found that the steady state exhibits almost isotropic ( one - time ) pair correlation functions . this may be different in the athermal limit . it does , however , qualitatively explain the disappearance of the stress overshoot , and our findings for the tagged - particle correlation functions and mean - squared displacements . immediately , this discussion predicts , that inserting a waiting time between the opposing flows would cause the bauschinger effect to weaken and go away . any quiescent waiting period before shear - reversal would enable relaxation to proceed and would ultimately lead to startup curves from the quiescent state . when shear is started from the quiescent equilibrium configuration , nearest - neighbor cages need to be broken before the accumulated strain becomes effective in enhancing the relaxation dynamics . the critical strain is hence connected to a typical cage size ; some percent of the particle diameter following a lindemann - type criterion for melting . the sheared steady state is then characterized by less strong cages , and the directionality of the flow does not play a major role in this weakening effect . hence , subsequent shear in the opposite direction will lead to an earlier and more gradual relaxation of correlation functions ( cf . [ fig : fsqt_shear ] ) . the mean - squared displacements , observed to follow the unsheared equilibrium curve up to times corresponding to the critical strain in the case of startup shear , remain much closer to the sheared steady - state ones in the case of flow reversal from such a steady state . as a consequence , the sharp upturn observed for the startup case , connected to superdiffusive motion and a stress overshoot , vanishes . at present , the analysis of the initial stress overshoot after startup of steady shear is based on a schematic model that does have a number of fit parameters whose numerical ratio has to be determined phenomenologically . however , the model is rooted in an understanding of the microscopic mct @xcite . in simplifying the latter to arrive at a schematic model , one aims to keep those essential features of the solutions qualitatively alike that the model shall describe ; in this sense , the model used here is a minimal model containing stress - overshoot phenomena . it is important to note that the subsequent description of the fate of the stress overshoot under time - varying flow drops out naturally from the equations of motion . in this sense , the comparisons presented in figs . [ fig : stressstrain ] and [ fig : bauschinger ] are qualitatively different the first can be viewed as a model - motivated parametrization of the simulation data . the second the comparison with the data on the bauschinger effect entails a parameter - free theoretical prediction . we thank for funding by the deutsche forschungsgemeinschaft through research unit for 1394 , projects p3 and p8 . a. k. b. acknowledges funding through the german academic exchange service , daad - dlr programme . v. is funded through the helmholtz gesellschaft ( hgf , vh - ng 406 ) and the zukunftskolleg of the university of konstanz . 43ifxundefined [ 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 in @noop _ _ , , vol . , ( , , ) @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 * * , ( ) http://link.aps.org/doi/10.1103/physrevlett.106.055502 [ * * , ( ) ] @noop * * , ( ) @noop _ _ ( , ) @noop * * , ( ) @noop * * , ( )

we present novel tests of pre main - sequence models based on individual dynamical masses for the m7 binary lspm j1314 + 1320ab . joint analysis of our keck adaptive optics astrometric monitoring along with very long baseline array radio data from a companion paper yield component masses of ( ) and ( ) and a parallactic distance ofpc . we also derive component luminosities that are consistent with the system being coeval at an age ofmyr , according to bhac15 evolutionary models . the presence of lithium is consistent with model predictions , marking the first time the theoretical lithium depletion boundary has been tested with ultracool dwarfs of known mass . however , we find that the average evolutionary model - derived effective temperature (k ) is 180k hotter than we derive from a spectral type relation based on bt - settl models (k ) . we suggest that the dominant source of this discrepancy is model radii being too small by% . in a test that mimics the typical application of evolutionary models by observers , we derive masses on the h - r diagram using the luminosity and bt - settl temperature . the estimated masses are% ( 2.0 ) lower than we measure dynamically and would imply that this is a system of brown dwarfs , highlighting the large systematic errors possible when inferring masses from the h - r diagram . this is first time masses have been measured for ultracool (m6 ) dwarfs displaying spectral signatures of low gravity . based on features in the infrared , lspm j1314 + 1320ab appears higher gravity than typical pleiades and ab dor members , opposite the expectation given its younger age . the components of lspm j1314 + 1320ab are now the nearest , lowest mass pre main - sequence stars with direct mass measurements .

1605.07182

a major goal of stellar astrophysics is to understand the early evolution of stars , before they reach a stable equilibrium on the main sequence . in theory , the fundamental parameters of mass , composition , and angular momentum uniquely determine the course of all stellar evolution , including the pre main - sequence phase . binary stars are perhaps the most useful empirical calibrators available for testing stellar models , as their components share a common age and composition and dynamical masses can be derived from their orbital motion . while dozens of mass measurements have been obtained for pre main - sequence stars ( e.g. , see reviews from * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) , most of these are for stars more massive than 0.5 . steady progress has been made to push measurements to lower masses . however , there are still only a handful of masses measured for stars at or below the 0.20.3 peak in the initial mass function , leaving pre main - sequence models for a large fraction of stars poorly constrained . previous work to measure the masses of pre main - sequence stars has mostly focused on star forming regions , like the nearby taurus auriga and scorpius centaurus lupus crux complexes with ages of @xmath12110myr and distances of @xmath12150pc ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . over the last two decades a growing number of young stars much closer to the sun and with wider ranging ages ( @xmath128150myr ) have been identified . the proximity of these stars offers many benefits , including the possibility of spatially resolving binaries with smaller semimajor axes and correspondingly shorter orbital periods for dynamical mass determinations . lspm j1314 + 1320 was first identified as a star exhibiting proper motion @xmath13200 by @xcite . @xcite included it as a candidate late - type star in their high angular resolution survey using lucky imaging at @xmath14 and @xmath15 bands , where their sample was selected from objects with red @xmath16 colors in the lspm catalog @xcite . @xcite estimated a spectral type of m6 for lspm j1314 + 1320 from its @xmath16 color and discovered that it was a binary with a separation of 130mas and flux ratios of @xmath17mag and @xmath18mag . meanwhile , @xcite had identified lspm j1314 + 1320 as a potential nearby star with an estimated distance of 9.7pc , leading them to obtain spectra and astrometry that revealed a spectral type of m7 , h@xmath19 in emission , and a parallactic distance of @xmath20pc @xcite . they concluded that their original photometric distance estimate was much smaller than the parallactic distance due to unresolved binarity and/or extreme youth . the first definitive evidence for the youth of lspm j1314 + 1320ab came from analysis by @xcite of its published optical spectra and their own near - infrared ( nir ) spectrum , all spatially unresolved . these spectra display weak alkali lines , which are indicative of low surface gravity , and strong absorption . although @xcite had previously identified lspm j1314 + 1320ab as a likely new member of the ab doradus moving group ( @xmath12150myr ) , subsequent analysis and updated proper motion data has led to the conclusion that it can not be confidently associated with any known group @xcite . @xcite has also identified lspm j1314 + 1320ab as a source of bright ( @xmath121mjy ) , persistent radio emission with a flat spectrum across a wide range of frequencies ( 1.4322.5ghz ) , suggesting high levels of magnetic activity and a stable , large - scale magnetosphere @xcite . therefore , lspm j1314 + 1320ab is potentially useful as a benchmark not only for evolutionary models but also for models of stellar magnetism . we present here spatially resolved relative astrometric monitoring of lspm j1314 + 1320ab obtained with keck ao that allows us to determine its orbit and thereby its dynamical total mass ( ) . we combine our keck data with spatially resolved very long baseline array ( vlba ) absolute astrometry from a companion paper ( forbrich et al . 2016 , submitted ; hereinafter paper i ) to simultaneously model the orbital motion , proper motion , and parallax and thereby derive individual dynamical masses of @xmath211% precision . we confirm that lspm j1314 + 1320ab is unambiguously in the pre main - sequence phase of its evolution and use our dynamical masses to perform tests of models , examining properties such as luminosity , temperature , and lithium depletion . we have been monitoring the resolved orbital motion of both components of lspm j1314 + 1320ab using keck adaptive optics ( ao ) with the facility nir camera nirc2 . we used both direct imaging and non - redundant aperture masking to measure the binary s separation , position angle ( pa ) , and flux ratio in @xmath22 and @xmath23 bands . typical examples of our images and masking interferograms from each epoch are shown in figure [ fig : keck ] . analysis of masking data was done using a pipeline similar to that in previous papers containing nirc2 masking data ( e.g. , * ? ? ? * ; * ? ? ? * ) and is described in detail in section 2.2 of @xcite . for imaging data , we used the same methods described in our previous work ( e.g. , see * ? ? ? * ; * ? ? ? briefly , after a standard reduction of the images ( dark subtraction and flat fielding ) we fit a three - component two - dimensional gaussian model to the two binary components . for the last three epochs , when the binary separation was widest , we were able to use the starfinder package instead . starfinder iteratively solves for both the binary parameters and an image of the point - spread function . we then corrected our derived @xmath24 positions using the nirc2 distortion solution of @xcite , which has a pixel scale of @xmath25maspixel@xmath26 and a @xmath27 offset added to the orientation given in the nirc2 image headers . the last epoch was obtained after a major realignment of the keck ao system on 2015 apr 13 . therefore , for this epoch , we used an updated distortion solution from service et al . ( 2016 , submitted ) , which has a pixel scale of @xmath28maspixel@xmath26 and orientation of @xmath29 . table [ tbl : keck ] summarizes the resolved astrometry and flux ratios derived from our keck observations . as an estimate of our uncertainties , we used the rms of the best - fit binary parameters at a given epoch across the multiple images or masking interferograms . to vet these errors , we performed a standard 7-parameter orbit least - squares fit ( see section 3.1 of * ? ? ? * ) to our keck astrometry . we found a value of @xmath30 for 7 degrees of freedom ( dof ) , which has a probability of @xmath31 , so we concluded that our keck astrometric errors are reasonable . , it is the probability of obtaining a value for @xmath32 as high or higher than the observed value given the degrees of freedom . ] our @xmath23-band flux ratios are somewhat inconsistent with being constant across all epochs , with @xmath33 ( 6 dof ) . to achieve @xmath34 , a systematic error of 0.011mag added in quadrature is required . therefore , in the following analysis we use @xmath35mag , which is the weighted average after adding this error floor in quadrature to the individual values . we conservatively use the rms of the measurements as the uncertainty since it encompasses both the fitting errors and the potential for variability . finally , we note that our infrared flux ratios seem to be at odds with the optical flux ratios of @xmath17mag and @xmath36mag from @xcite . however , for another binary in their survey , lhs 1901ab , they also report very unequal flux ratios ( @xmath37mag and @xmath38mag ; * ? ? ? * ) while ao imaging gives flux ratios in @xmath39 of @xmath80.1mag . this suggests that the apparent discrepancy may simply be due to systematic errors in the flux ratios derived from lucky imaging , so we do not use them in our analysis . we obtained optical and near - infrared ( nir ) spectra of lspm j1314 + 1320ab as part of larger follow - up program of nearby bright m dwarfs from @xcite . in the optical we used the supernova integral field spectrograph ( snifs ; * ; * ? ? ? * ) on the university of hawaii 2.2-m telescope on maunakea , hawaii , on 2015 jan 9 ut . snifs provides simultaneous coverage from 32009700 at a resolution of @xmath40 . details of our snifs observations and reduction can be found in @xcite and @xcite , which we briefly summarize here . the pipeline detailed in @xcite performs dark , bias , and flat - field corrections , cleaned the data of bad pixels and cosmic rays , then fits and extracts the integral field unit spaxels into a one - dimensional spectrum . the @xcite reduction takes this spectrum and performs flux calibration and telluric correction based on white dwarf standards taken throughout the night and a model of the atmosphere above maunakea . the final reduced spectrum of lspm j1314 + 1320ab has a signal - to - noise ratio ( snr ) of @xmath13100 per pixel redward of 6000 . we obtained a nir spectrum of lspm j1314 + 1320ab using the spex spectrograph @xcite at the nasa infrared telescope facility ( irtf ) on maunakea , hawaii , on 2013 may 16 ut . our observations were taken in the short cross - dispersed ( sxd ) mode using the @xmath41 slit ( @xmath42 ) , yielding simultaneous coverage from 0.82.4 , with a small gap near 1.8 due to non - overlapping sxd orders . the target was placed at two positions along the slit ( a and b ) and observed in an abba pattern in order to subsequently subtract the sky background . six exposures were taken this way , yielding a snr @xmath43 per pixel in the @xmath44 and @xmath23 bands . to correct for telluric lines , we observed an a0v star immediately after the target . our spex spectrum was extracted using the spextool package @xcite , which performs flat - field correction , wavelength calibration , sky subtraction , and extraction of the one - dimensional spectrum . multiple exposures were combined using the ` xcombspec ` routine . a telluric correction spectrum was constructed from the a0v star and applied to the target spectrum using the ` xtellcor ` package @xcite . separate orders were stacked using the ` xcombspec ` tool , which also shifts the flux level in different orders to match each other . these corrections were 1% or less per order . we combined our keck / nirc2 relative astrometry of lspm j1314 + 1320ab with vlba absolute astrometry from paper i to perform a joint analysis of the orbit and parallax of the system . the vlba observations only detect one component of the binary at all epochs , and we identify it as the secondary component lspm j1314 + 1320b ( defined as the component that is fainter in our @xmath22- and @xmath23-band keck data ) . this is because the orbital motion seen in the vlba data is to the northeast over 20132014 , and our keck astrometry over the same time period indicates that the secondary was moving to the northeast relative to the primary . in addition to our keck astrometry here , we also used results from @xcite who measured a binary separation of @xmath45mas and pa of @xmath46 from their lucky imaging data on 2005 jun 15 ut . the combined data set contains a total of 7 epochs of resolved astrometry spanning 10.03yr and 9 epochs of absolute astrometry spanning 4.65yr . our astrometric model includes 15 total parameters , and six of these are orbit parameters shared between the keck and vlba data . the six common parameters are orbital period ( @xmath47 ) , eccentricity ( @xmath48 ) , inclination ( @xmath49 ) , argument of periastron ( @xmath50 ) , mean longitude at the reference epoch ( @xmath51 ) , and pa of the ascending node ( @xmath52 ) . we defined the reference epoch for our model as @xmath53jd ( i.e. , calendar date 2010.0 ) . the semimajor axes of the primary and secondary components about their barycenter are denoted @xmath54 and @xmath55 , respectively . since our keck data measure relative orbital motion , the keck semimajor axis parameter was @xmath56 , while the vlba semimajor axis parameter was simply @xmath55 . the remaining parameters are only used to model the vlba data : ra and dec at the reference epoch ( @xmath57 , @xmath58 ) ; proper motion in ra and dec ( @xmath59 , @xmath60 ) ; parallax ( @xmath61 ) ; and systematic error parameters in ra and dec ( @xmath62 , @xmath63 ) . these vlba error parameters are needed to model systematic errors in the absolute astrometric calibration , which are expected to be comparable from epoch to epoch . in order to properly mitigate our model from preferring extremely large values of these error parameters , we penalized the logarithm of the likelihood by @xmath64 , where @xmath10 is the quadrature sum of both error parameters . the parallax factor at each epoch was calculated from the jpl ephemeris de405 as described in equations 1 and 2 of @xcite . as in our previous work on joint analysis of relative and absolute astrometry @xcite , we used the affine invariant ensemble sampler ` emcee v2.1.0 ` @xcite to perform our markov chain monte carlo ( mcmc ) analysis . we used @xmath65 walkers ( chains ) with @xmath66 steps each , saving every 500th step for use in our analysis and removing the first 10% of all walkers as the burn - in portion . table [ tbl : mcmc ] summarizes the resulting posterior distributions and the priors on all of our parameters . we report both the best - fit parameters and the credible intervals that encompass 68.3% and 95.4% of each parameter s posterior values . the best - fit parameters give a total @xmath67 for 19 dof , @xmath68 , including both the keck data and the vlba data with the systematic errors added in quadrature to the nominal vlba measurement errors . figure [ fig : orbit ] displays this best - fit orbit and parallax solution alongside both the keck and vlba astrometry . table [ tbl : mcmc ] also gives credible intervals and best - fit values for a number of additional parameters that can be derived directly from our fitted parameters , e.g. , the distance computed from the absolute vlba parallax ( @xmath69 ) . most notably , we compute a precise total system mass from the keck orbit and vlba parallax , @xmath70 , and individual masses for the components by combining the vlba and keck orbits ( @xmath71 , @xmath72 ) . because of the very high precision of the vlba parallax ( @xmath73 ) , the uncertainties in these masses are dominated by the uncertainty in the orbital period ( @xmath74 ) , total semimajor axis ( @xmath75 ) , and secondary semimajor axis ( @xmath76 ) . accounting for covariances between these parameters , the derived primary mass is the most precise ( @xmath77 ) , followed by the total mass ( @xmath78 ) and secondary mass ( @xmath79 ) . the mass ratio ( @xmath80 ) is consistent with the radio - emitting secondary that is fainter in the nir being the less massive component , and 95.4% of our mcmc posterior values having @xmath81 . the derived primary and secondary masses are @xmath0(@xmath1 ) and @xmath2(@xmath3 ) , respectively . these individual masses are very nearly equal , within 2% of each other at 90% confidence . for additional discussion of the parallax and proper motion determined from this joint analysis see paper i. our new ao imaging allows us to compute resolved photometry for the two components of lspm j1314 + 1320ab . our @xmath22- and @xmath23-band flux ratios ( @xmath82mag , @xmath83mag ) are consistent with and more precise than the values of @xmath84mag and @xmath85mag from @xcite . using our @xmath86 and @xmath87 values and their @xmath44-band flux ratio , @xmath88mag , we computed resolved @xmath39 photometry for the components of lspm j1314 + 1320ab . for the integrated - light photometry we used the 2mass point source catalog @xcite along with 2mass - to - mko photometric system conversions calculated from our spex spectrum . we list the resulting resolved and integrated - light photometry on both the mko and 2mass photometric systems in table [ tbl : props ] . in order to compute the bolometric flux of lspm j1314 + 1320ab in integrated light , we combined and absolutely calibrated our optical and nir spectra following the method outlined in @xcite . briefly , we first collected published photometry from 2mass , sdss @xcite , the aavso all - sky photometric survey ( apass ; * ? ? ? * ) , and the _ wide - field infrared survey explorer _ ( _ wise _ ; * ? ? ? we converted this photometry to fluxes using the relevant zero - points , and we also calculated synthetic fluxes from our spectra using the corresponding filter profiles @xcite . at this point we noticed that the sdss @xmath49-band measurement was highly discrepant ( @xmath834mag ) with our spectrum and with the surrounding photometry , so we excluded it from our analysis . taking account of both random and correlated errors ( flux calibration and optical variability ) we scaled the optical and nir spectra to match the photometry and to match each other in the region overlapping between snifs and spex ( 0.800.95 ) . we replaced regions of high telluric contamination and those not covered by our spectra with a best - fit atmospheric model from the bt - settl grid @xcite , where the best - fit here was @xmath89k and @xmath90dex . the final calibrated and combined spectrum is shown in figure [ fig : lbol ] . we calculated the bolometric flux ( ) by integrating over the combined spectrum , accounting for errors in the flux calibration , optical variability of 1.5% @xcite , and errors introduced by replacing regions of spectra with a bt - settl model . this yielded @xmath91ergs@xmath26@xmath92 . we apportioned this integrated - light bolometric flux to the individual components by deriving a bolometric flux ratio from our @xmath23-band flux ratio . we examined bt - settl atmosphere models with @xmath932900k and @xmath905.0dex . relative to the best - fit @xmath89k and @xmath90dex model , the other five models had comparable or slightly larger amplitude @xmath23-band flux ratios spanning @xmath94mag to 0.12mag for hotter and cooler models , respectively . among these models , the relationship between their bolometric magnitude differences and @xmath23-band magnitude differences was consistently @xmath95mag . in comparison to our integrated - light bolometric magnitude error of 0.025mag , the correction itself is small and the uncertainty in the correction is essentially negligible ( though we do account for it in our analysis ) . combining this correction factor with our @xmath23-band photometry , we find @xmath96dex , with the resultant individual luminosities given in table [ tbl : props ] . we determined the integrated - light spectral type of lspm j1314 + 1320ab from our optical and nir data . we compared our optical snifs spectrum to the m dwarf spectral library of @xcite . to find a best fit spectral type and uncertainty we used the least - squares package mpfit in idl @xcite . in performing this fit we allowed the numerical spectral type to vary continuously , interpolating between standards to create the comparison spectra . we excluded wavelengths near h@xmath19 , allowed for a small wavelength shift to account for radial velocity , and neglect extinction given the small distance . we found a best fit spectral type for lspm j1314 + 1320ab of m@xmath97 , and figure [ fig : spt ] shows our data in comparison to the best - fit and adjacent spectral standards . we also used our spex sxd spectrum to determine a nir spectral type and gravity classification on the @xcite system . we find a spectral type of m@xmath98 ( figure [ fig : ir - spt ] ) , consistent with our optical type and the previous determinations of m@xmath99 in the optical from @xcite and m6.5 in the nir from @xcite . the gravity classification is determined from the amount of deviation in certain spectral features from field objects of similar spectral type . if most features are consistent with the field ( i.e. , having a score of zero ) , then the classification is fld - g . if most features are deviant from the field indicating low gravity , then the classification is vl - g . intermediate cases are given the classification int - g . for lspm j1314 + 1320 , only 3 scores are used to determine the gravity , since the strength of vo absorption is not applicable at this spectral type . based on the indices computed from our sxd spectrum ( table [ tbl : grav ] ) , we found that feh was consisent with the field , most individual alkali lines were not strongly deviant from the field , and the @xmath44-band continuum was marginally inconsistent with the field . this results in a score of 0n01 and a formal classification of fld - g . however , a closer examination of the alkali lines leads to a more nuanced interpretation . of the four features used in the alkali score , the line at 1.169 was consistently indicative of low gravity . figure [ fig : ir - grav ] shows all four alkali lines , and in fact all of them appear visually weaker than the field gravity m6 standard , consistent with the findings of @xcite from an independent spectrum . the other three are simply not weak enough to qualify as low gravity on the @xcite system . therefore , while the formal classification alone gives fld - g , we note that lspm j1314 + 1320 is more faithfully described as having a spectrum on the borderline between fld - g and int - g classifications . finally , we also used the @xmath23-band portion of our spex spectrum to determine the metallicity of lspm j1314 + 1320ab in integrated light . using the calibration of @xcite we found @xmath100 } = 0.04\pm0.08$]dex , i.e. , consistent with solar metallicity as assumed by theoretical models in the following analysis . we caution that this calibration was based on field objects and so might give a somewhat different value for [ fe / h ] than a relation based on low gravity dwarfs . given our precisely determined individual masses and luminosities for the components of lspm j1314 + 1320ab , we can uniquely infer other physical properties from evolutionary model tracks . we first interpolate the values of each physical parameter such as from the bhac15 evolutionary tracks onto uniform , 2-d grids in log(mass ) and log ( ) , using grid steps of 0.01dex in both axes . we draw random , normally distributed values of the individual component s bolometric fluxes for each step in our mcmc chains . we then bilinearly interpolate each resultant pair of ( mass , ) from our chain on a given 2-d grid of parameter values to compute the posterior distributions of that parameter . this approach preserves covariances between input parameters , e.g. , mass and both depend on distance , when deriving parameters like ( e.g. , * ? ? ? * ) . the resulting posterior distributions for model - derived values of age , , radius , , and fraction of lithium remaining ( li / li@xmath101 ) are summarized in table [ tbl : props ] . we find model - derived ages that are consistent with coevality at 0.7@xmath10 , giving a consensus age of @xmath5myr . ( this and other mean values given in table [ tbl : props ] represent the posterior distribution of the mean of primary and secondary values calculated from each step of the chain . ) other model - derived parameters are comparably consistent between the two components , as expected given the very similar component masses and luminosities , with mean values of @xmath102k , @xmath103dex , @xmath104 , and li / li@xmath105 . the very small formal uncertainties in our model - derived properties reflect the precision of the measured masses and luminosities projected onto the model grids ; we do not attempt to include any systematic errors that could be associated with the models . evolutionary models indicate that the components of lspm j1314 + 1320ab are well removed from the main sequence . for our measured masses of @xmath1 and @xmath3 , bhac15 models predict a main sequence luminosity of @xmath106dex attained by an age of @xmath8900myr . our measured luminosities , @xmath107dex and @xmath108dex , are a factor of @xmath84@xmath109 higher than the main sequence value . the detection of lithium by @xcite in the combined light spectrum of these @xmath80.09 objects further supports the pre main - sequence nature of this binary , as models predict lithium will be destroyed in objects of this mass hundreds of myr before they reach the main sequence . for very low - mass stars ( @xmath210.1 ) and high mass brown dwarfs , lithium is destroyed at a slow enough rate that it can be used to determine the ages of stellar associations up to at least @xmath12100myr . higher mass objects destroy their primordial lithium at a faster rate , so at older ages lithium disappears from the spectra of progressively lower mass objects . therefore , the brightest stars to display ( 6708 ) absorption in a given cluster define an empirical boundary that can be used as a relative age scale between different clusters , and evolutionary models can be used to infer absolute ages of individual clusters based on the location of this lithium depletion boundary ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? lspm j1314 + 1320ab allows us to test these model predictions with stars of known mass for the first time at ages comparable to that of nearby open clusters . @xcite measured a ( 6708 ) pseudo - equivalent width of @xmath110 from the integrated - light spectrum of lspm j1314 + 1320ab . they noted that this is consistent with measurements of comparable pleiades late - m dwarfs , e.g. , @xcite report three m7 dwarfs that all have @xmath1110.6 . given our mass and luminosity measurements for each component , bhac15 evolutionary models predict that the fraction of initial lithium remaining in the primary and secondary components is @xmath112 and @xmath113 , respectively . since the component masses are nearly equal , the predicted lithium depletion is correspondingly consistent within 1.2@xmath10 . the mean lithium fraction of the two components is predicted to be li / li@xmath114 . an equivalent width measurement can not be directly converted into lithium abundance . @xcite used a model dependent curve of growth approach to estimate the relationship between pseudo - equivalent widths and lithium abundance , defined by @xmath115 . if the lithium depletion level in the components of lspm j1314 + 1320ab is @xmath80.15 as predicted by evolutionary models , and the initial cosmic lithium abundance is @xmath116dex @xcite , then their present day abundance would be @xmath117dex . over a range of @xmath1183600k at @xmath90dex , @xcite found that @xmath119dex corresponds to ew = 0.410.51 , which is consistent with the lithium detection from @xcite . if this calibration from @xcite is accurate and applicable at the somewhat lower here ( @xmath83000k according to evolutionary models ; section [ sec : evol ] ) , then the detection of lithium absorption is fully consistent with the evolutionary model prediction that the components of lspm j1314 + 1320ab have depleted most of their initial lithium supply . as discussed in section [ sec : evol ] , models predict an age of @xmath5myr based on the mass and luminosity of the components of lspm j1314 + 1320ab . we can place this in the context of the relative age scale provided by nearby open clusters with well determined lithium depletion boundaries . for the pleiades , the boundary is at @xmath120mag @xcite using the vlbi parallax distance of @xcite . the components of lspm j1314 + 1320ab are 0.40.5mag brighter than this ( table [ tbl : props ] ) , implying that they must be significantly younger than the pleiades in order to still possess lithium . likewise for blanco 1 , @xcite found a lithium depletion boundary of @xmath121mag , @xmath80.7mag fainter than the components of lspm j1314 + 1320ab , so they must also be younger than blanco 1 . the younger cluster @xmath19 persei has a lithium depletion boundary of @xmath122mag @xcite , which is actually consistent within the errors for both components of lspm j1314 + 1320ab . therefore , we conclude that the age of lspm j1314 + 1320ab must be consistent with or younger than that of @xmath19 per . @xcite report an age of @xmath123myr for @xmath19 per , and recent age determinations for the pleiades and blanco 1 are , respectively , @xmath124myr @xcite and @xmath125myr @xcite . the age of lspm j1314 + 1320ab that we derived from models using mass and luminosity ( @xmath5myr ) is therefore consistent with the requirement from lithium for the system age to be equal to or younger than the age of @xmath19 per . finally , we compare the integrated - light color and resolved absolute magnitudes of lspm j1314 + 1320ab to the cluster sequences of @xmath19 per and the pleiades . using our observed spectrum , we compute integrated - light apparent magnitudes on the cousins system ( @xmath126mag ) and sdss system ( @xmath127mag ) . these give integrated - light colors of @xmath128mag and @xmath129mag . the @xmath19 per members from within 0.15mag of this color have apparent magnitudes of @xmath130mag , and assuming a distance of @xmath131pc gives an absolute magnitude of @xmath132mag . this is somewhat brighter than but consistent with the absolute magnitudes of the lspm j1314 + 1320ab components ( @xmath133mag and @xmath134mag ) . performing the same exercise for the dance sample of probable ( @xmath135 ) pleiades members from using our @xmath136 color gives @xmath137mag and thereby @xmath138mag assuming a pleiades distance of @xmath139pc @xcite . this is somewhat fainter than but consistent with the absolute magnitudes of the lspm j1314 + 1320ab components . the scatter in these cluster sequences on the color magnitude diagram ( @xmath80.3mag ) is relatively large compared to the change of @xmath80.6mag in absolute magnitude from @xmath19 per to the pleiades , which limits the discriminating power of this comparison . the components are somewhat fainter than expected for being as old or younger than @xmath19 per ( 0.40.5mag ) , but the effect is not significant . we therefore conclude that the location of the components of lspm j1314 + 1320ab on the color magnitude diagram is consistent with the more precise constraints on age from the lithium depletion boundary comparison , being comparable age to @xmath19 per and younger than the pleiades . combining our proper motion and parallax with the published radial velocity of the lspm j1314 + 1320ab system allows us to derive its space motion and thereby assess potential membership in known associations of young stars . @xcite reported spectrally resolved radial velocities of the two components of lspm j1314 + 1320ab from which they computed a system velocity of @xmath140 under the assumption that the two components are equal in mass . our astrometric mass ratio of @xmath141 now validates this assumption within their measurement uncertainty and thereby their reported system velocity . we derive a space motion of @xmath142 and plot this vector alongside various known young associations in figure [ fig : uvw ] . there are no clear visual associations , except perhaps with @xmath143 cha in @xmath144 , but lspm j1314 + 1320 is very far from @xmath143 cha in @xmath145 . using the banyan ii web tool ( v1.4 ; * ? ? ? * ; * ? ? ? * ) , we find a 99.98% membership probability in the young field population assuming the age of the system is @xmath1461gyr . therefore , according to banyan , lspm j1314 + 1320 is not likely a member of any of the seven young moving groups considered in their model ( ab dor , argus , @xmath147 pic , carina , columba , tuc - hor , and twa ) . we therefore conclude that the lspm j1314 + 1320ab does not belong to any known young association , despite being unambiguously pre main - sequence and located at only 17.25pc . it is possible that lspm j1314 + 1320ab belongs to an as yet unidentified @xmath146100 myr association in the solar neighborhood , and if so this should be testable with the upcoming release of _ gaia_astrometry . if no new associations are found that match lspm j1314 + 1320ab , then it will join the growing ranks of orphaned young objects in the solar neighborhood ( e.g. , liu et al.2016 , submitted ) . in principle , the rotation period ( @xmath148 ) and @xmath149 measurements from @xcite and @xcite provide an empirical constraint on the minimum stellar radius , @xmath150 . and @xmath151 here to indicate to the stellar rotation period and the inclination of the stellar rotation axis with respect to the plane of the sky , respectively , since we have already used @xmath47 and @xmath49 for the binary orbit s period and inclination . ] in practice , such a calculation is complicated by the fact that their measurements are made in integrated light , but for the sake of argument we will assume that they both correspond to one of the two components of lspm j1314 + 1320ab . @xcite reported a rotation period of @xmath152hr from multi - epoch vla observations and a @xmath153 from optical spectroscopy . in optical photometric monitoring , @xcite find two distinct rotation periods of @xmath154hr and @xmath155hr and conclude that the cause of the two very similar but distinct periods is not clear . assuming @xmath156 and 3.8hr gives @xmath157 . this is consistent with the model - derived average @xmath158 , which would correspond to @xmath159 . interestingly , this value is in good agreement with the measured orbital inclination @xmath160 , which would be consistent with a stellar spin axis aligned with the orbital plane . however , it is also possible that the true radius is smaller or larger than predicted by models , corresponding to @xmath161 or @xmath162 , respectively . the bhac15 evolutionary models that we have used to derive stellar parameters like radius and employ bt - settl model atmospheres as their boundary conditions for the surfaces of stars . therefore , we can test for consistency between the value of derived from bhac15 and an independent estimate based on the spectrum of lspm j1314 + 1320ab . direct fitting of our combined optical and nir spectrum with bt - settl models yields @xmath89k and @xmath90dex ( figure [ fig : btsettl ] ) , but these values necessarily lack in precision due to the somewhat coarse ( 100k , 0.5dex ) model grid steps . the spectral type scale of @xcite , which is based on bt - settl model atmospheres , gives @xmath163k for a spectral type of m7 , in good agreement with the direct fitting . therefore , we find the bt - settl models give @xmath8180k cooler values for than the average evolutionary model derived value of @xmath164k.k at m7 , in better agreement with the bhac15 . however , this could be due to the fact that the @xcite scale was designed to match evolutionary model isochrones of an earlier generation of the bhac15 models and thus does not provide a truly independent as needed for our consistency check . moreover , using a spectral type scale intended for higher gravity field dwarfs results in an even cooler , more inconsistent . ] in the absence of additional information , it is equally possible that this discrepancy could be caused by systematic errors in either or both of the evolutionary and atmosphere models . indeed , the bt - settl models do not fit the overall spectrum with high accuracy ( figure [ fig : btsettl ] ) , implying that any temperature based on these models will harbor some systematic error . observations of other pre main - sequence binaries at younger ages show discrepancies consistent with the 180k difference here . for example , @xcite found that for two unequal - mass m dwarf binary systems in the 10-myr - old upper scorpius subgroup , the evolutionary model - derived temperatures were 100300k higher than model atmospheres . if the discrepancy for lspm j1314 + 1320ab were due to evolutionary models , then at fixed luminosity this would imply model radii that are too small by 13% . interestingly , there are other cases of pre main - sequence m dwarfs for which such under - predicted model radii can explain observed discrepancies . @xcite found that for the 10myr old m5 eclipsing binary uscoctio 5 , multiple evolutionary models ( including bhac15 ) underpredicted the component radii by 10%15% while simultaneously overpredicting by @xmath8300k and thereby predicting consistent luminosities . in addition , evolutionary models of low - mass stars that include the effects of magnetic fields predict larger radii due a slowing down of their contraction ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? therefore , both observations and theory suggest that the dominant source of the discrepancy we observe for lspm j1314 + 1320ab is most likely due to evolutionary model radii , although we note that spectral type scale could still harbor systematic errors . lspm j1314 + 1320ab is much older than other pre main - sequence systems with dynamical mass measurements in nearby star - forming regions , implying that the same qualitative radius/ problem with evolutionary models extends to ages of at least @xmath1280myr . to our knowledge , this is the first mass measurement for an ultracool dwarf ( spectral type @xmath165m7 ) with spectral signatures of lower surface gravity than typical field objects . in section [ sec : spt ] , we classified the integrated - light spectrum as fld - g , but on the borderline of being int - g , and both @xcite and we note the presence of spectral features indicative of lower surface gravity relative to field dwarfs of similar spectral type . evolutionary models indicate that the surface gravities of the binary components are @xmath1664.87dex , which is 0.420.46dex lower than the predicted main sequence surface gravity of 5.29dex for a 0.09 star . if the model radii are too small by 13% as suggested above , then the model - derived gravities would be 0.1dex lower . overall , this implies that the borderline between fld - g and int - g designations for late - m dwarfs corresponds to a surface gravity @xmath80.5dex lower than field objects . while the lspm j1314 + 1320ab system itself does not belong to an association of independently determined age , numerous other ultracool dwarfs with gravity classifications do this allows us to check for consistency of spectral behavior between lspm j1314 + 1320ab and other young late - m dwarfs of known age . in the ab dor moving group ( 125myr ) , @xcite list 15 bona fide or strong candidate members . of these , all but one object have gravity classifications of int - g or vl - g , with only 2mass j03264225@xmath1672102057 ( l5 ) classified as fld - g ( but possessing some visual signs of youth , like lspm j1314 + 1320ab but at much later type ) . of 17 possible candidate members none are classified fld - g . moreover , @xcite report gravity classifications for eight ultracool dwarfs in the pleiades ; none were fld - g , and the two m7 dwarfs ppl 1 and teide 1 were int - g and vl - g , respectively . these comparisons imply that spectral signatures leading to int - gand vl - g classifications are typically still quite entrenched at ages significantly older than lspm j1314 + 1320ab . at even younger ages , e.g. , tuc - hor ( 50myr ) and @xmath147 pic ( 23myr ) , no known members are classified as fld - g either ( e.g. , see compilation of liu et al . 2016 , submitted ) . therefore , the fact that lspm j1314 + 1320ab has a gravity classification of fld - g is at odds with its model - derived age of @xmath5myr and that our lithium analysis that empirically places the system at a significantly younger age than the pleiades . we suggest a few possible explanations for the fact that lspm j1314 + 1320ab shows less distinct evidence for low gravity than older ultracool dwarfs . perhaps it is not that uncommon for late - m dwarfs to have more muted gravity signatures than l dwarfs , and the existing samples of ultracool dwarfs are too sparse to detect this yet . as noted by @xcite , the currently available infrared spectra of pleiades late - m dwarfs are of much lower s / n than was used to define the gravity classification system , so higher quality spectra may result in somewhat different , higher gravity , classifications . the classification of lspm j1314 + 1320ab is done in integrated light , so perhaps the gravity signatures are somehow obscured in combined light . @xcite tested such an idea , mostly using vl - gtemplates for components , and found that it is quite rare for two components to be classified as lower gravity in combined light . finally , maybe some third parameter is at work , e.g. , metallicity or rotation , causing the gravity classification to deviate slightly from other young ultracool dwarfs . our spectrum shows no signs of unusual metallicity , and it also seems unlikely that such a young object would different substantially in composition from other young associations in the solar neighborhood . pre main - sequence stellar models are commonly used to infer masses by placing objects on the h - r diagram ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . to test the accuracy of masses derived from models in this way , we used the effective temperatures and luminosities of lspm j1314 + 1320ab to derive mass and age . given that the masses and luminosities of the components are nearly equal , we simply consider the average integrated - light properties for this test . as mentioned above , the spectral type scale for young objects from @xcite gives @xmath163k , and they estimate a systematic error of 100k for their scale , which we adopt here . the mean measured luminosity of lspm j1314 + 1320ab is @xmath168dex . figure [ fig : hrd ] shows these values of and compared to bhac15 evolutionary model tracks . for this test , we interpolated model tracks on a uniform grid of @xmath169 and @xmath170 in the same fashion as described above in section [ sec : evol ] . we found an h - r diagram derived average component mass of @xmath171 , age of @xmath172myr , and of @xmath173dex . this would imply that lspm j1314 + 1320ab is actually a pair of young brown dwarfs , due to the h - r diagram derived mass being @xmath9% ( 2.0@xmath10 ) smaller than our directly measured component masses of @xmath892 . the h - r diagram age is also much smaller ( @xmath174dex ) than the age derived from the same models using mass and . as expected , the mass and age posteriors derived from the h - r diagram are highly correlated , where lower masses correspond to younger ages . the discrepancy between the h - r diagram derived mass and our dynamically measured masses suggests either large errors in the spectral type relations , which are calibrated using bt - settl model atmospheres , systematic errors in the evolutionary models , or some combination of both things . as we discuss in section [ sec : teff ] , we suggest the dominant source of this discrepancy is that evolutionary model radii are underpredicted and is thereby overpredicted at a given luminosity . regardless of the cause of the discrepancy , this test case shows that masses derived from the h - r diagram can harbor large systematic errors . unfortunately , this method is often the only practical option when attempting to infer masses of stars and brown dwarfs in young associations where age can be uncertain due to potential underlying age spreads . therefore , the systematic error we have identified here will have significant implications for efforts to determine the low - mass end of the initial mass function , suggesting that young low - mass stars may be mistakenly identified as young brown dwarfs . we present here individual dynamical masses for the components of lspm j1314 + 1320ab , a pre main - sequence binary located at a distance of only @xmath4pc . these masses and parallactic distance are made possible by a joint analysis of our resolved relative astrometry of the primary and secondary from keck ao imaging and masking along with absolute astrometry from vlba radio interferometry of the secondary ( i.e. , the component that is fainter in the optical and infrared ) . we also derive component luminosities using integrated - light spectroscopy and photometry and our resolved infrared photometry . the measured component masses of @xmath0 ( @xmath1 ) and @xmath2 ( @xmath3 ) and luminosities of @xmath175dex and @xmath176dex , respectively , are consistent with being coeval at an age of @xmath5myr according to bhac15 evolutionary models . our precise masses and luminosities are largely thanks to a remarkably precise vlba parallax ( @xmath177 ; paper i ) . we determine that lspm j1314 + 1320ab is unambiguously in the pre main - sequence phase of its evolution based on having lithium absorption and luminosities @xmath84@xmath109 higher than predicted for the main sequence at our measured masses . this combination of precise distance , masses , luminosities , and the detection of lithium by @xcite enables novel tests of pre main - sequence models distinct from previous work on objects in star forming regions with more uncertain distances . 1 . evolutionary models self - consistently predict luminosity and lithium depletion in this binary thereby passing the first test of the theoretical lithium depletion boundary using ultracool dwarfs of known mass . models predict the components have lost @xmath178% and @xmath179% of their initial lithium , leaving enough remaining that they are still expected to display lithium absorption . on the empirically defined relative cluster age scale , the presence of lithium and the component absolute magnitudes of lspm j1314 + 1320ab imply an age consistent with or younger than @xmath19 persei and significantly younger than the pleiades and blanco 1 . we compare the effective temperature derived from evolutionary models given our mass and luminosity ( component average @xmath180k ) to that derived from spectral type relations based on bt - settl models ( @xmath7k ) . the 180k discrepancy in these values would correspond to a 13% radius error at fixed luminosity , which is similar to a problem observed for low mass stars in the much younger upper scorpius star - forming region . the eclipsing binary usco ctio 5 shows that model radii are likely the dominant source of the problem , while the radius and errors cancel so that model luminosities are accurate . we suggest that the same physical cause is responsible for the 180k discrepancy for lspm j1314 + 1320ab , showing that this radius problem can extend to much older pre main - sequence ages than previously recognized . 3 . in a test that mimics the typical application of evolutionary models by observers , we used the derived from spectral type and the average component luminosity to infer mass and age from evolutionary model tracks on the h - r diagram . the derived average component mass of @xmath171 is much lower ( @xmath9% , 2.0@xmath10 ) than we measure dynamically . this highlights the large systematic errors possible when inferring masses of low - mass stars and brown dwarfs at young ages and implies that some stars may be mistakenly identified as brown dwarfs when using the h - r diagram . the integrated - light spectrum of lspm j1314 + 1320ab displays signatures of low surface gravity , although we formally classify it as fld - g ( on the borderline of int - g ) on the infrared @xcite system . this is the first time dynamical masses have been measured for ultracool dwarfs with low - gravity spectral features . however , contrary to expectations , lspm j1314 + 1320ab shows less distinct spectral signs of low gravity than ultracool dwarfs at older ages ( pleiades , ab dor ) , which we are unable to explain . lspm j1314 + 1320ab provides a high - precision benchmark for pre main - sequence models at a distance @xmath1210@xmath109 closer than even the nearest star - forming regions . one major unresolved question is why only the secondary component is radio emitting , as discussed in detail in paper i. given that we have shown that the component masses are within 2% of each other for this coeval , co - compositional binary system , one likely explanation for the divergent behavior is a difference in the angular momentum evolution of the two components . projected rotational velocities ( @xmath181 ) of other very low - mass binaries have hinted at such differences in angular momentum evolution ( e.g. , * ? ? ? * ) , but such measurements have yet to be obtained for lspm j1314 + 1320ab . the components of lspm j1314 + 1320ab are now the nearest , lowest mass pre main - sequence stars with direct mass measurements . in the future , _ gaia _ parallaxes combined with ongoing ground - based orbit monitoring efforts will make many more such tests of models possible for more distant binaries in star forming regions . _ gaia _ data could even help identify previously unrecognized associations of young stars in the solar neighborhood to which lspm j1314 + 1320ab , currently not associated with any known group , may belong . this work was supported by a nasa keck pi data award , administered by the nasa exoplanet science institute . we thank will best for assistance with some keck / nirc2 observations . it is a pleasure to thank joel aycock , carolyn jordan , jason mcilroy , luca rizzi , terry stickel , hien tran , and the keck observatory staff for assistance with our keck ao observations . we also thank p. k. g. williams , michael ireland , katelyn allers , joshua schlieder , and mark reid for useful discussions . the anonymous referee provided a remarkably rapid and thoughtful review that helped refine our discussion . james r. a. davenport for distributing his idl implementation of the cubehelix color scheme @xcite . our research has employed the 2mass data products ; nasa s astrophysical data system ; data 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 nasa and curated by the nasa / ipac infrared science archive ; and the simbad database operated at cds , strasbourg , france . finally , the authors wish to recognize and acknowledge the very significant cultural role and reverence that the summit of maunakea has always had within the indigenous hawaiian community . we are most fortunate to have the opportunity to conduct observations from this mountain . aldering , g. , adam , g. , antilogus , p. , et al . 2002 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 4836 , survey and other telescope technologies and discoveries , ed . j. a. tyson & s. wolff , 6172 allard , f. , homeier , d. , & freytag , b. 2011 , in astronomical society of the pacific conference series , vol . 448 , 16th cambridge workshop on cool stars , stellar systems , and the sun , ed . c. johns - krull , m. k. browning , & a. a. west , 91 cutri , r. m. , skrutskie , m. f. , van dyk , s. , et al . 2003 , 2mass all sky catalog of point sources . ( the irsa 2mass all - sky point source catalog , nasa / ipac infrared science archive . http://irsa.ipac.caltech.edu/applications/gator/ ) lantz , b. , aldering , g. , antilogus , p. , et al . 2004 , in society of photo - optical instrumentation engineers ( spie ) conference series , vol . 5249 , optical design and engineering , ed . l. mazuray , p. j. rogers , & r. wartmann , 146155 ) . contours are in logarithmic intervals from unity to 10% of the peak flux in each band . the image cutouts are all 05 across and interferogram cutouts are 1@xmath1822 across . in the interferograms , the binary can be seen by eye as an elongation or double peak in the center of the point - spread function . all cutouts have the same native pixel scale , and here we have rotated them so that north is up . [ fig : keck],title="fig:",height=144 ] 0.04 in ) . contours are in logarithmic intervals from unity to 10% of the peak flux in each band . the image cutouts are all 05 across and interferogram cutouts are 1@xmath1822 across . in the interferograms , the binary can be seen by eye as an elongation or double peak in the center of the point - spread function . all cutouts have the same native pixel scale , and here we have rotated them so that north is up . [ fig : keck],title="fig:",height=144 ] 0.04 in ) . contours are in logarithmic intervals from unity to 10% of the peak flux in each band . the image cutouts are all 05 across and interferogram cutouts are 1@xmath1822 across . in the interferograms , the binary can be seen by eye as an elongation or double peak in the center of the point - spread function . all cutouts have the same native pixel scale , and here we have rotated them so that north is up . [ fig : keck],title="fig:",height=144 ] 0.04 in ) . contours are in logarithmic intervals from unity to 10% of the peak flux in each band . the image cutouts are all 05 across and interferogram cutouts are 1@xmath1822 across . in the interferograms , the binary can be seen by eye as an elongation or double peak in the center of the point - spread function . all cutouts have the same native pixel scale , and here we have rotated them so that north is up . [ fig : keck],title="fig:",height=144 ] 0.04 in ) . contours are in logarithmic intervals from unity to 10% of the peak flux in each band . the image cutouts are all 05 across and interferogram cutouts are 1@xmath1822 across . in the interferograms , the binary can be seen by eye as an elongation or double peak in the center of the point - spread function . all cutouts have the same native pixel scale , and here we have rotated them so that north is up . [ fig : keck],title="fig:",height=144 ] 0.04 in ) . contours are in logarithmic intervals from unity to 10% of the peak flux in each band . the image cutouts are all 05 across and interferogram cutouts are 1@xmath1822 across . in the interferograms , the binary can be seen by eye as an elongation or double peak in the center of the point - spread function . all cutouts have the same native pixel scale , and here we have rotated them so that north is up . [ fig : keck],title="fig:",height=144 ] 0.04 in ) . contours are in logarithmic intervals from unity to 10% of the peak flux in each band . the image cutouts are all 05 across and interferogram cutouts are 1@xmath1822 across . in the interferograms , the binary can be seen by eye as an elongation or double peak in the center of the point - spread function . all cutouts have the same native pixel scale , and here we have rotated them so that north is up . [ fig : keck],title="fig:",height=144 ] 0.04 in band shown in comparison to m6 dwarfs with field gravity ( fld - g ) and very low gravity ( vl - g ) . although the alkali lines visually appear intermediate between field- and low - gravity m6 dwarfs , a full analysis of all gravity scores from @xcite yields a classification of fld - g . lspm j1314 + 1320ab is best described as having a spectrum on the borderline between fld - g and int - g classifications . _ bottom : _ the full spex spectrum with other indices ( feh , vo , and @xmath44-band continuum shape ) labeled . [ fig : ir - grav],height=336 ] band shown in comparison to m6 dwarfs with field gravity ( fld - g ) and very low gravity ( vl - g ) . although the alkali lines visually appear intermediate between field- and low - gravity m6 dwarfs , a full analysis of all gravity scores from @xcite yields a classification of fld - g . lspm j1314 + 1320ab is best described as having a spectrum on the borderline between fld - g and int - g classifications . _ bottom : _ the full spex spectrum with other indices ( feh , vo , and @xmath44-band continuum shape ) labeled . [ fig : ir - grav],height=336 ] 3% of both measured individual masses . isochrones from 1myr to 1gyr are indicated by dotted lines . the used for lspm j1314 + 1320ab here is calibrated off of the bt - settl model atmospheres , the same used for boundary conditions in the bhac15 evolutionary models . when we use this and luminosity to infer properties from bhac15 models , we find a mass @xmath9% ( 2.0@xmath10 ) lower than we measured dynamically . this discrepancy indicates either large errors in spectral type relations ( @xmath8180k ) or systematic errors in evolutionary models ( e.g. , 13% in radius ) . in either case , this result suggests that masses inferred for young stars from the h - r diagram will harbor large systematic errors , and young stars may be confused for young brown dwarfs . [ fig : hrd],width=480 ] lccccc 2013 jan 18 & @xmath23 & @xmath183 & @xmath184 & @xmath185 & masking + 2014 jan 22 & @xmath23 & @xmath186 & @xmath187 & @xmath188 & masking + 2014 may 9 & @xmath23 & @xmath189 & @xmath190 & @xmath191 & imaging + 2014 jun 15 & @xmath22 & @xmath192 & @xmath193 & @xmath194 & imaging + 2015 jan 14 & @xmath23 & @xmath195 & @xmath196 & @xmath197 & imaging + 2015 apr 10 & @xmath23 & @xmath198 & @xmath199 & @xmath200 & imaging + 2015 jun 28 & @xmath23 & @xmath201 & @xmath202 & @xmath203 & imaging + lcccc + + orbital period @xmath47 ( yr ) & @xmath204 & 9.58 & 9.45 , 9.74 & @xmath205 ( log - flat ) + semimajor axis @xmath56 ( mas ) & @xmath206 & 146.4 & 145.6 , 147.7 & @xmath207 ( log - flat ) + eccentricity @xmath48 & @xmath208 & 0.6014 & 0.5964 , 0.6060 & uniform , @xmath209 + inclination @xmath49 ( ) & @xmath210 & 49.19 & 48.77 , 49.82 & @xmath211 , @xmath212 + pa of the ascending node @xmath52 ( ) & @xmath213 & 60.2 & 59.6 , 61.3 & uniform + argument of periastron @xmath50 ( ) & @xmath214 & 205.8 & 204.1 , 207.1 & uniform + mean longitude at 2455197.5 jd @xmath51 ( ) & @xmath215 & 71.9 & 69.2 , 74.5 & uniform + @xmath216 ( mas ) & @xmath217 & 0.0 & @xmath1670.8 , 0.8 & uniform , @xmath218 + @xmath219 ( mas ) & @xmath220 & 0.0 & @xmath1671.7 , 1.7 & uniform , @xmath221 + proper motion in ra @xmath59 ( ) & @xmath222 & @xmath167248.01 & @xmath167248.19 , @xmath167247.78 & uniform + proper motion in dec @xmath60 ( ) & @xmath223 & @xmath167183.64 & @xmath167184.05 , @xmath167183.17 & uniform + parallax @xmath61 ( mas ) & @xmath224 & 57.988 & 57.856 , 58.082 & @xmath225 ( uniform volume density ) + semimajor axis of secondary @xmath55 ( mas ) & @xmath226 & 73.7 & 73.1 , 74.4 & uniform + ra vlbi error parameter @xmath227 [ deg@xmath228 & @xmath229 & @xmath16715.8 & @xmath16716.2 , @xmath16714.7 & uniform + dec vlbi error parameter @xmath230 [ deg@xmath228 & @xmath231 & @xmath16714.7 & @xmath16715.4 , @xmath16713.7 & uniform + + + + total mass @xmath232 ( ) & @xmath233 & 183.7 & 181.1 , 187.6 & + primary mass @xmath234 ( ) & @xmath0 & 92.5 & 91.5 , 94.0 & + secondary mass @xmath235 ( ) & @xmath2 & 91.2 & 89.6 , 93.8 & + mass ratio @xmath236 & @xmath237 & 0.986 & 0.975 , 1.002 & + distance @xmath238 ( pc ) & @xmath4 & 17.245 & 17.217 , 17.284 & + semimajor axis @xmath239 ( au ) & @xmath240 & 2.525 & 2.512 , 2.548 & + time of periastron @xmath241 ( jd ) & @xmath242 & 2456498.8 & 2456495.0 , 2456502.0 & + spectral type ( optical ) & & & m@xmath97 & + spectral type ( near - ir ) & & & m@xmath98 fld - g & + spectral type ( optical ) & & & m@xmath97 & + spectral type ( near - ir ) & & & m@xmath98 fld - g & + @xmath243 ( mag ) & @xmath244 & @xmath245 & @xmath246 & @xmath247 + @xmath248 ( mag ) & @xmath249 & @xmath250 & @xmath251 & @xmath252 + @xmath253 ( mag ) & @xmath254 & @xmath255 & @xmath256 & @xmath257 + @xmath258 ( mag ) & @xmath259 & @xmath260 & @xmath261 & @xmath247 + @xmath262 ( mag ) & @xmath263 & @xmath249 & @xmath264 & @xmath252 + @xmath265 ( mag ) & @xmath266 & @xmath267 & @xmath268 & @xmath257 + @xmath269 ( mag ) & @xmath270 & @xmath271 & @xmath246 & @xmath247 + @xmath272 ( mag ) & @xmath273 & @xmath274 & @xmath251 & @xmath252 + @xmath275 ( mag ) & @xmath276 & @xmath277 & @xmath256 & @xmath257 + @xmath278 ( mag ) & @xmath279 & @xmath280 & @xmath261 & @xmath247 + @xmath281 ( mag ) & @xmath282 & @xmath273 & @xmath264 & @xmath252 + @xmath283 ( mag ) & @xmath284 & @xmath285 & @xmath268 & @xmath257 + @xmath286 [ erg@xmath92s@xmath26 ] & @xmath288 & @xmath289 & @xmath290 & @xmath291 + @xmath292 ( mag ) & @xmath293 & @xmath294 & @xmath295 & @xmath296 + @xmath297 ( dex ) & @xmath298 & @xmath176 & @xmath299 & @xmath291 + + + age ( myr ) & @xmath300 & @xmath301 & @xmath5 & @xmath302 + ( k ) & @xmath303 & @xmath304 & @xmath164 & @xmath305 + radius ( ) & @xmath306 & @xmath307 & @xmath308 & @xmath309 + [ cms@xmath310 & @xmath311 & @xmath312 & @xmath313 & @xmath314 + li / li@xmath315 & @xmath112 & @xmath113 & @xmath316 & @xmath317 + lc feh@xmath318 & @xmath319 + feh@xmath320 & @xmath321 + vo@xmath318 & @xmath322 + @xmath320 & @xmath323 + @xmath44-cont & @xmath324 + & @xmath325 + ( 1.169 ) & @xmath326 + ( 1.177 ) & @xmath327 + ( 1.253 ) & @xmath328 + alkali score & 1000 + final score & 0n01 + gravity classification & fld - g +

we perform hartree calculations of symmetric and asymmetric semi - infinite nuclear matter in the framework of relativistic models based on effective hadronic field theories as recently proposed in the literature . in addition to the conventional cubic and quartic scalar self - interactions , the extended models incorporate a quartic vector self - interaction , scalar - vector non - linearities and tensor couplings of the vector mesons . we investigate the implications of these terms on nuclear surface properties such as the surface energy coefficient , surface thickness , surface stiffness coefficient , neutron skin thickness and the spin orbit force . = 22.5 cm = 16.4 cm = -0.4 cm = -0.5 cm addtoresetequationsection tempcntc citex[#1]#2@filesw auxout tempcnta@tempcntb@neciteaciteforciteb:=#2citeo#1 citeotempcnta > tempcntbciteacitea , tempcnta = tempcntbtempcnta * nuclear surface properties in + relativistic effective field theory * + m. del estal , m. centelles , x. vias + _ departament destructura i constituents de la matria , facultat de fsica , + universitat de barcelona , diagonal _ 647 _ , e-_08028 _ barcelona , spain _ _ pacs : _ 21.60.-n , 21.30.-x , 21.10.dr , 21.65.+f _ keywords : _ nuclear surface properties ; spin orbit potential ; semi - infinite nuclear matter ; non - linear self - interactions ; quantum hadrodynamics ; effective field theory .

nucl-th9810026

quantum hadrodynamics ( qhd ) and the relativistic treatment of nuclear systems has been a subject of growing interest during recent years @xcite . the model of walecka @xcite and its non - linear extensions with cubic and quartic self - interactions of the scalar - meson field @xcite have been widely used to this end . this model contains dirac nucleons together with neutral scalar and vector mesons as well as isovector - vector @xmath0 mesons . at the mean field ( hartree ) level , it already includes the spin orbit force , the finite range and the density dependence which are essential ingredients of the nuclear interaction . this simple model has become very popular in relativistic calculations and describes successfully many properties of the atomic nucleus . from a theoretical point of view , the non - linear model with cubic and quartic scalar self - interactions was classed within renormalizable field theories which can be characterized by a finite number of coupling constants . however , very recently , generalizations of this model that include other non - linear interactions among the meson fields and tensor couplings have been presented on the basis of effective field theories by serot et al.@xcite . the effective theory contains many couplings of non - renormalizable form that are consistent with the underlying symmetries of qcd . consequently , one must find some suitable expansion parameters and develop a systematic truncation scheme . for this purpose the concept of naturalness has been employed : it means that the unknown couplings of the theory should all be of the order of unity when written in appropriate dimensionless form using naive dimensional analysis @xcite . then , one can estimate the contributions coming from different terms by counting powers in the expansion parameters and truncating the lagrangian at a given level of accuracy . one important fact is the observation that at normal nuclear densities the scalar and vector meson fields , denoted by @xmath1 and @xmath2 , are small as compared with the nucleon mass @xmath3 and that they change slowly in finite nuclei . this implies that the ratios @xmath4 , @xmath5 , @xmath6 and @xmath7 are useful expansion parameters when the effective field theory is applied to the nuclear many - body problem . from this viewpoint , if all the terms involving scalar and meson self - interactions are retained in the lagrangian up to fourth order , one recovers the well - known non - linear model plus some additional terms @xcite . for the truncation to be consistent , the corresponding coupling constants should exhibit naturalness and can not be arbitrarily dropped out without an additional symmetry argument . the effective lagrangian truncated at fourth order contains thirteen free parameters that have been fitted to reproduce twenty - nine finite nuclei observables @xcite . remarkably , the fitted parameters turn out to be natural and the results are not dominated by the last terms retained . this evidence confirms the utility of the principles of naive dimensional analysis and naturalness and shows that truncating the effective lagrangian at the first lower orders is justified . the term with a vector - meson quartic self - interaction has been considered previously in relativistic mean field ( rmf ) calculations from a phenomenological point of view . bodmer @xcite considered this coupling to avoid the negative coefficient of the quartic scalar self - interaction that appears in many non - linear parametrizations that correctly describe the atomic nucleus @xcite . in some special situations this negative term can lead to a pathological behaviour of the scalar potential . on the other hand , the equation of state is softened at moderate high densities when the vector non - linearity is taken into account . the quartic vector self - interaction has also been phenomenologically used by gmuca @xcite in a non - linear model for parametrizing dirac brueckner fock calculations of nuclear matter . the same idea was developed by toki et al . and applied to study finite nuclei @xcite and neutron stars @xcite . recently , the properties of high - density nuclear and neutron matter have been analyzed in the rmf approach taking into account scalar and vector non - linearities @xcite . the tensor couplings of the vector @xmath8 and @xmath0 mesons to the nucleon were investigated by reinhard et al . @xcite as an extension of the rmf model , and more recently by furnstahl et al.@xcite from the point of view of relativistic effective field theory . in these works it was shown that the tensor coupling of the @xmath8 meson has an important bearing on the nuclear spin orbit splitting . the surface properties of nuclei play a crucial role in certain situations . this is the case , for instance , of saddle - point configurations in nuclear fission or fragment distributions in heavy - ion collisions . within a context related to the liquid droplet model ( ldm ) and the leptodermous expansion @xcite , the surface properties can be extracted from semi - infinite nuclear matter calculations either quantally or semiclassically ( though the total curvature energy coefficient can only be computed semiclassically @xcite ) . in the non - relativistic case most of the calculations of the surface properties have been carried out using skyrme forces , quantally @xcite or semiclassically with the help of the extended thomas fermi ( etf ) method @xcite . in the relativistic case the nuclear surface has been analyzed within the model since a long time ago . the calculations have been performed semiclassically using the relativistic thomas fermi ( tf ) method or its extensions ( retf ) , for symmetric @xcite and asymmetric @xcite matter , and also in the quantal hartree approach @xcite . in the framework of the relativistic model and effective field theory , the main purpose of the present work is to carefully analyze the influence on surface properties of the quartic vector non - linearity , of the newly proposed scalar - vector self - interactions and of the tensor coupling . we shall investigate quantities such as the surface energy coefficient , surface thickness , spin orbit strength , surface stiffness coefficient and neutron skin thickness obtained from hartree calculations of symmetric and asymmetric semi - infinite nuclear matter . the paper is organized as follows . section 2 is devoted to the basic theory . the results on the surface properties of symmetric matter are discussed in section 3 . section 4 addresses the case of asymmetric systems . the summary and conclusions are given in the last section . following ref . @xcite , to derive the mean field equations one starts from an energy functional containing dirac baryons and classical scalar and vector mesons . the energy functional can be obtained from the effective lagrangian in the hartree approach using many - body techniques @xcite . however , this energy functional can also be considered as an expansion in @xmath4 , @xmath5 , @xmath6 and @xmath7 of a general energy density functional that contains all the correlation effects . the theoretical basis of this functional lies on the extension of the hohenberg kohn theorem @xcite to qhd @xcite . using the kohn sham scheme @xcite with the mean fields playing the role of kohn sham potentials , one finds similar mean field equations to those obtained from the lagrangian @xcite , but including effects beyond the hartree approach through the non - linear couplings @xcite . a semi - infinite system of uncharged nucleons corresponds to a one - dimensional geometry where half the space is filled with nuclear matter at saturation and the other half is empty , so that a surface develops around the interface . the fields and densities change only along the direction perpendicular to the medium . specifying the energy density functional considered in refs . @xcite and @xcite to symmetric semi - infinite nuclear matter with the surface normal pointing into the @xmath9 direction one has ( z ) & = & _ _ { -i + + w(z ) - w(z ) } _ ( z ) + & & + ( 1 + _ 1 ) ( ( z))^2 + ( + + ) ^2(z ) + & & - ( 1 + _ 2 ) ( w(z ) ) ^2 - w^4 ( z ) + & & - ( 1 + _ 1 + ) w^2 ( z ) , [ eq1 ] where the index @xmath10 runs over all occupied states of the positive energy spectrum , @xmath11 and @xmath12 ( notation as in ref . @xcite ) . except for the terms with @xmath13 and @xmath14 , the functional ( [ eq1 ] ) is of fourth order in the expansion . we retain the fifth - order terms @xmath13 and @xmath14 because in refs . @xcite and @xcite they have been estimated to be numerically of the same magnitude as the quartic scalar term in the nuclear surface energy . the mean field equations are obtained by minimizing with respect to @xmath15 , @xmath1 and @xmath2 : \ { -i + + w(z ) - w(z ) } _ ( z ) = _ _ ( z ) , [ eq2 ] -(z ) + m_s^2 ( z ) & = & g_s^2 _ s(z ) -m_s^2m^2 ( z ) ( _ 32+_43!(z)m ) + & & + g_s^2 2 m ( _ 1+_2(z)m ) m_v^2g_v^2 w^2 ( z ) + & & + _ 1 2 m [ ( ( z))^2 + 2(z)(z ) ] + _ 2 2 m g_s^2g_v^2 ( w(z))^2 , [ eq3 ] + -w(z ) + m_v^2 w(z ) & = & g_v^2 ( ( z ) + _ t(z ) ) - ( _ 1+_22(z)m ) ( z)m m_v^2 w(z ) + & & -13!_0 w^3(z ) + _ 2 m [ ( z ) w(z ) + ( z)w(z ) ] . [ eq4 ] the baryon , scalar and tensor densities are respectively ( z ) & = & _ _ ^(z ) _ ( z ) , [ eq4a ] + _ s(z ) & = & _ _ ^(z ) _ ( z ) , [ eq4b ] + _ t(z ) & = & _ . [ eq4c ] the expression of the four - component spinors @xmath16 in the semi - infinite medium was given by hofer and stocker in ref.@xcite . in a semi - infinite nuclear matter calculation the sum over the single - particle states is replaced by an integration over momenta : _ 2 _ d*k * , [ eq5 ] where @xmath17 stands for the volume of the box , the factor 2 takes into account the isospin degree of freedom and @xmath18 describes the spin orientation of the nucleons . introducing the fermi momentum @xmath19 , the integration domain is restricted to @xmath20 , with @xmath21 if the bulk nuclear matter is located at @xmath22 . following the method outlined in ref . @xcite one finds two sets ( @xmath23 ) of first - order differential equations for the orbital part of the upper and lower components of the dirac spinors : - g_a ( z ) & = & f_a ( z ) , [ eq6a ] + - - f_a ( z ) & = & g_a ( z ) , [ eq6b]where @xmath24 and @xmath25 is the dirac effective mass of the nucleons . from the asymptotic behaviour at @xmath22 ( bulk nuclear matter ) , the condition on the energy eigenvalues is @xmath26 , with @xmath27 and @xmath28 being the nuclear matter values of @xmath29 and @xmath2 . the densities for each spin orientation @xmath30 read ^(z ) & = & _ 0^ k _ f dk_z _ 0^ dk _ k _ ( |g_a ( z ) |^2 + |f_a ( z ) |^2 ) , [ eq7 ] + _ s^(z ) & = & _ 0^ k _ f dk_z _ 0^ dk _ k _ ( |g_a ( z ) |^2 - |f_a ( z ) |^2 ) , [ eq8 ] + _ t^(z ) & = & _ 0^ k _ f dk_z _ 0^ dk _ k _ ( f_a(z ) g_a(z ) ) , [ eq9 ] and the total densities are given by ( z ) = _ ^(z ) , _ s(z ) = _ _ s^(z ) , _ t(z ) = _ _ t^(z ) . [ eq9b ] using the equations of motion the energy density of the semi - infinite nuclear matter system can be written as follows : ( z ) & = & _ _ 0^k _ f dk_z _ 0^ dk _ k _ ( + w _ ) ( | g_a ( z ) |^2 + |f_a ( z ) |^2 ) + & & + ( z)_s(z ) - w(z ) ( ( z ) + _ t(z ) ) - ( + ) ^2 ( z ) + & & + ( _ 1 + _ 2 ) w^2 ( z ) + w^4 ( z ) + & & - ( ( z ) ) ^2 + ( w(z ) ) ^2 . [ eq10 ] finally , the surface energy coefficient @xmath31 is obtained from the expression @xcite e_s = 4r_0 ^ 2 ^_- dz , [ eq11 ] where @xmath32 is the energy per particle in bulk nuclear matter and @xmath33 is the nuclear radius constant : @xmath34 , with @xmath35 the nuclear matter density . another important quantity in the study of the nuclear surface structure is the spin orbit interaction . by elimination of the lower spinor in terms of the upper spinor , one obtains a schrdinger - type equation with a term @xmath36 that has the structure of the single - particle spin orbit potential for the non - relativistic case @xcite . in our present model the orbital part of the spin orbit potential reads as v_so(z ) = . [ eq12]in the non - relativistic limit , by means of a foldy wouthuysen reduction , eq . ( [ eq12 ] ) becomes v_so^fw(z ) = [ eq13]and the nucleons are then moving in a central potential of the form v_c(z ) = w(z ) - ( z ) . although previous works @xcite have thoroughly investigated the properties of the nuclear surface in the standard non - linear model , for which @xmath37 and @xmath38 are the only non - linearities in ( [ eq1 ] ) , here we wish to enlarge this study by including the additional non - linear and tensor couplings considered in refs . @xcite . in concrete , we want to study the role of the quartic vector self - interaction @xmath39 that has been used after the work of bodmer @xcite , the role of the terms with @xmath40 and @xmath41 that couple the scalar and vector fields , of the terms with @xmath13 and @xmath14 that imply the gradients of the fields , and of the tensor coupling @xmath42 of the vector @xmath8 meson to the nucleon . while @xmath39 , @xmath40 and @xmath41 can be classified as volume contributions , the couplings @xmath13 , @xmath14 and @xmath42 are genuine surface terms . on the basis of the concept of naturalness there is no reason to omit any of these terms in the energy density functional ( [ eq1 ] ) , unless there exists a symmetry principle to forbid it . however , to clarify the impact on the surface properties of the aforementioned couplings we will analyze each one separately , as it has been similarly done in ref.@xcite . in this section we shall study symmetric systems , while in section 4 we shall address the case of asymmetric matter . in the conventional non - linear model the value of the coefficients @xmath43 , @xmath44 , @xmath37 and @xmath38 can be univocally obtained by imposing that for nuclear matter at saturation the density @xmath35 , energy per particle @xmath32 , effective mass @xmath45 and incompressibility modulus @xmath46 take given values . when the vector - meson quartic self - interaction is switched on , the dirac equation for the baryons and the klein gordon equation for the vector field in infinite nuclear matter become ( throughout this subsection we set @xmath47 ) : a_v & = & + w _ - m , [ eq3.1 ] + m_v^2 w _ & = & g_v^2 _ 0 - _ 0 w^3 _ . [ eq3.2 ] the saturation density @xmath35 and the fermi momentum @xmath19 are related as usual by @xmath48 . specifying @xmath35 , @xmath32 and @xmath45 , from the above equations one extracts the coupling constant @xmath49 as a function of @xmath39 ( the nucleon and @xmath8 masses take their empirical values : @xmath50 mev and @xmath51 mev ) . the steps to calculate @xmath43 , @xmath37 and @xmath38 are then the same as when @xmath52 , see e.g. refs . @xcite and @xcite , but now these coefficients become functions of @xmath39 . the reader will find a detailed study of the implications of the vector non - linearity @xmath39 in nuclear matter in refs . @xcite and @xcite . the assumption of naturalness requires that the couplings @xmath53 , @xmath54 , @xmath37 , @xmath38 and @xmath39 should all be roughly of the order of unity . figure 1 illustrates the variation of these couplings as a function of the non - dimensional parameter _ 0 = [ eq3.2.b ] used by bodmer was called @xmath9 . ] in ref . @xcite . with @xmath55 mev , the figure presents the results for four sets of nuclear matter properties : @xmath56 @xmath57 ( @xmath58 @xmath59 ) , @xmath60 mev , @xmath61 and 350 mev , and @xmath62 and 0.7 . when @xmath63 mev and @xmath62 the equilibrium properties of the interaction and the scalar mass @xmath64 are very close to those of the non - linear parametrization nl1 @xcite . we realize that @xmath65 and @xmath49 are only weakly affected by @xmath66 ( the change is not appreciable in the scale of figure 1 ) . however , @xmath66 has a direct effect on the scalar - meson quartic self - interaction @xmath38 , moving it from a negative value when the quartic vector term is absent ( @xmath67 ) to a desirable positive value when @xmath68 . the change of sign of @xmath38 takes place at larger values of @xmath66 for larger @xmath46 and @xmath45 ; in fact , @xmath38 is already positive at @xmath69 if @xmath70 mev and @xmath71 . both non - linearities of the scalar field @xmath37 and @xmath38 remain in the natural zone for @xmath72 approximately , excepting the case of @xmath70 mev and @xmath71 , but they start to depart appreciably from their natural values when @xmath73 . these trends fairly agree with the assumption of naturalness : @xmath74 corresponds to the region where @xmath39 can be considered as natural ( see the figure ) , and for this range of @xmath66 values also the rest of coupling constants are natural . the behaviour gleaned from figure 1 is rather independent of the saturation density @xmath35 and energy @xmath32 , and e.g. we have found similar trends with the specific nuclear matter properties used by bodmer in ref . @xcite . next we turn our attention to the surface properties . in figures 2 and 3 we have plotted , respectively , the surface energy coefficient @xmath31 and the surface thickness @xmath75 of the semi - infinite density profile ( the 90%10% fall - off distance ) against the @xmath66 parameter . the selected values of @xmath66 are those employed in table 1 of ref . @xcite . with @xmath76 @xmath57 and @xmath77 mev , we have performed the calculations for a few incompressibilities ( @xmath78 , 200 and 350 mev ) and effective masses ( @xmath79 and 0.7 ) . furthermore , we have considered two values of the mass of the scalar meson ( @xmath55 and 525 mev ) . this quantity governs the range of the attractive interaction and determines the surface fall - off : a larger @xmath64 results in a steeper surface and a reduction of @xmath31 and @xmath75 . from figures 2 and 3 we realize that the quartic vector self - interaction scarcely alters the values of the surface energy coefficient and of the surface thickness if @xmath39 remains in the natural domain , i.e. , if @xmath72 . at @xmath80 mev and @xmath81 , @xmath31 is raised by decreasing @xmath66 . this tendency may be inverted at smaller values of the incompressibility , depending also on the value of the nucleon effective mass , but the global trends are practically independent of the mass of the scalar meson . the thickness @xmath75 exhibits a more monotonous behaviour : in all cases it stays almost equal to its value at @xmath69 and goes down slightly for @xmath73 . noticeable departures of the surface energy coefficient and thickness from the @xmath82 values can be found only if @xmath66 is decreased beyond the natural limit , a situation where the interaction is mainly ruled by the vector - meson quartic self - interaction . even though the impact of the vector - meson quartic self - interaction is small , it can help to find parameter sets for which both the surface energy coefficient and the surface thickness lie in the empirical region . this fact is illustrated in figure 4 , where @xmath31 and @xmath75 are drawn versus @xmath45 for @xmath69 , 2 and @xmath83 and for @xmath84 , 500 and 550 mev , using the nuclear matter conditions of ref . @xcite : @xmath85 @xmath57 , @xmath86 mev and @xmath87 mev . ( we also performed the calculations for @xmath88 and @xmath89 to confirm the trends we discuss below . ) the horizontal dashed lines in figure 4 serve to indicate the empirical region for the surface energy and thickness . if at @xmath69 we concentrate , for instance , on the parametrizations with @xmath90 mev we see that the one with @xmath91 yields , simultaneously , @xmath31 and @xmath75 within the empirical region . the agreement between the calculated values with @xmath90 mev and the empirical region improves for @xmath92 , where both @xmath31 and @xmath75 are acceptable for practically all the values of the effective mass considered . for @xmath74 the dependence of @xmath31 and @xmath75 upon the nucleon effective mass is similar to that found in the usual model without a quartic vector self - interaction @xcite . the general tendencies start to change when @xmath66 is lowered and leaves the natural region . as @xmath66 becomes smaller figure 4 shows that the slope of the curves of @xmath31 and @xmath75 as a function of @xmath45 changes , and that the curves for the different @xmath64 come closer together . to get more insight about the influence of the vector - meson quartic self - interaction , we display in figure 5 the profiles of the baryon density @xmath93 and of the surface tension density ( swiatecki integrand ) @xmath94 , eq.([eq11 ] ) . in turn , we have represented in figure 6 the orbital part of the spin orbit potential @xmath36 , eq . ( [ eq12 ] ) , at the fermi surface ( i.e. , evaluated at @xmath95 ) and the central mean field @xmath96 defined in eq . ( [ eq14 ] ) . the properties of the interactions used in these figures are @xmath97 @xmath57 , @xmath98 mev , @xmath61 mev , @xmath79 and 0.7 , and @xmath55 mev . results are shown for @xmath69 , 2 and 0.5 . the corresponding values of the surface energy coefficient and surface thickness can be read from figures 2 and 3 . it can be seen that the local quantities depicted in figures 5 and 6 oscillate as functions of @xmath9 ( friedel oscillations ) . both the surface tension density and the single - particle spin orbit potential are confined to the surface and average to zero inwards as the bulk matter is approached . the local profiles for @xmath92 , which somehow marks the limit of naturalness as we have commented , are almost equal to those obtained in the absence of the quartic vector self - interaction . in agreement with figures 2 and 3 , only when @xmath66 is decreased to non - natural values one can notice some changes in the profiles , which are more visible for @xmath99 than for @xmath71 . decreasing @xmath66 makes the surface steeper and the thickness @xmath75 smaller , produces an enhancement in the surface region of the density @xmath93 and of the mean field @xmath96 , and builds up friedel oscillations in @xmath100 and in the spin orbit potential @xmath36 . it is well known that the experimental spin orbit splittings require within narrow bounds a dirac effective mass @xmath45 around 0.6 in the conventional relativistic model @xcite . figure 6 shows that introducing a quartic vector self - interaction makes the spin orbit well deeper . however , it is only a minor effect : with @xmath91 it is not possible to reproduce the spin orbit interaction of the case @xmath79 at @xmath69 , not even if one sets @xmath101 ( which in addition brings about an unreallistically small @xmath75 , see figure 3 ) . a similar conclusion was drawn in ref . @xcite from an analysis in nuclear matter . we also have computed the non - relativistic limit @xmath102 of the spin orbit potential given by the expression ( [ eq13 ] ) . in agreement with ref . @xcite we have found that while @xmath102 qualitatively reproduces the behaviour of @xmath36 , it strongly underestimates the quantitative depth of the fully relativistic spin orbit strength ( by @xmath103% for @xmath79 and by @xmath104% for @xmath91 ) . the next bulk terms in the energy density ( [ eq1 ] ) that contain non - linear meson interactions are - ( _ 1 + ) w^2 . [ eq3.7 ] in analogy to figure 1 for @xmath66 , in figure 7 we study the change of the couplings @xmath53 , @xmath54 , @xmath37 and @xmath38 with the parameters @xmath40 and @xmath41 ( introducing each one separately ) for some specific equilibrium properties . with @xmath105 @xmath57 , @xmath60 and @xmath55 mev , in part ( a ) of figure 7 it is @xmath61 mev and @xmath62 , while in part ( b ) it is @xmath80 mev and @xmath106 . for @xmath61 mev and @xmath62 both @xmath40 and @xmath41 have a considerable effect on the scalar non - linearities @xmath37 and @xmath38 . the coupling @xmath38 changes sign for @xmath107 , but it remains negative in the interval of @xmath40 values used . to keep all the coupling constants within natural values we see that the range for @xmath40 and @xmath41 ( when introduced separately ) is restricted to run roughly from @xmath108 to @xmath109 . consider now different values of the incompressibility and effective mass , as in part ( b ) of figure 7 . the dependence of the couplings on @xmath41 is not significantly altered . however , increasing either @xmath46 or @xmath110 results in a smoother slope of @xmath37 with @xmath40 , while it makes @xmath38 grow steadily with @xmath40 and become positive at some value of this parameter . we have checked for @xmath61 mev and @xmath111 , and for @xmath78 mev and @xmath62 , that @xmath38 remains negative at all values of @xmath40 and that @xmath37 grows with @xmath40 faster than in part ( a ) of figure 7 . the influence of the non - linear interactions @xmath40 and @xmath41 on the surface properties is analyzed in figure 8 . with @xmath112 , in this figure we have computed @xmath31 and @xmath75 taking into account the terms ( [ eq3.7 ] ) . the saturation conditions of the interaction and the scalar mass are the same as in part ( a ) of figure 7 . the results are displayed in the plane @xmath40@xmath41 in the form of contour plots of constant @xmath31 ( solid lines ) and of constant @xmath75 ( dashed lines ) . the range of variation of @xmath40 and @xmath41 lies in the region imposed by naturalness , and yields values of @xmath31 and @xmath75 within reasonable limits . as it can be inferred from the nearly vertical lines in the @xmath40@xmath41 plane , the surface energy coefficient and thickness depend mostly on @xmath40 and are rather independent of @xmath41 . the consequence of increasing @xmath40 is a reduction of the values of @xmath31 and @xmath75 . the lines of constant @xmath75 turn out to be , roughly speaking , parallel to the lines of constant @xmath31 . this means that from the interplay of the parameters @xmath40 and @xmath41 it is not possible to change the value of @xmath75 relative to that of @xmath31 ( for example , we see in figure 8 that @xmath113 fm if @xmath114 mev ) . we have calculated the spin orbit potential @xmath36 at the fermi surface for several values of @xmath40 and @xmath41 . we have found that these couplings have a marginal effect on the spin orbit strength , as it happened to be the case with the other bulk non - linearity @xmath66 . to get some information about the incidence on the surface energy and thickness of all the volume non - linear meson interactions together , we have repeated the calculations in the @xmath40@xmath41 plane setting @xmath92 for the quartic vector self - interaction . one finds similar features to those of figure 8 . the effect of @xmath92 is just shifting @xmath31 and @xmath75 towards smaller values as compared with the case @xmath69 ( @xmath115 ) , which is in accordance with what was found in figures 2 and 3 at @xmath61 mev and @xmath62 . now we discuss the non - linear interactions [ eq3.8 ] that vanish in infinite nuclear matter . we recall that these terms are actually of order 5 in the expansion of the effective lagrangian but , following refs . @xcite and @xcite , we include them because they can be relevant in the surface due to their gradient structure . using the same saturation properties and scalar mass of figure 8 , in figure 9 we have calculated @xmath31 and @xmath75 for several values of @xmath13 and @xmath14 with @xmath116 . one observes that the curves of constant @xmath31 are projected onto the plane @xmath13@xmath14 as almost parallel straight lines ( at least in the analyzed region , corresponding to natural values of @xmath13 and @xmath14 ) . the same happens to the curves of constant @xmath75 . but in contrast with the situation found in the plane @xmath40@xmath41 ( figure 8) , the slope of the lines of constant @xmath75 is different from that of the lines of constant @xmath31 . this means that by varying @xmath13 and @xmath14 one can achieve some modification on the surface thickness while keeping the same surface energy . for example , if we consider the contour line of @xmath117 mev we find that for @xmath118 it is @xmath119 fm , whereas for @xmath120 it is @xmath121 fm . from figure 9 we also see that increasing @xmath13 at constant @xmath14 brings about larger values of @xmath31 and @xmath75 , and that the opposite happens if one increases @xmath14 at constant @xmath13 . we have repeated the calculations of figure 9 ( @xmath61 mev , @xmath62 ) for @xmath80 mev and for @xmath122 , to verify to which extent the behaviour in the @xmath13@xmath14 plane is affected by the incompressibility and effective mass of the interaction . certainly , the contour lines of @xmath31 and @xmath75 are shifted with respect to figure 9 , but the trends with @xmath13 and @xmath14 turn out to be qualitatively the same . the range of variation of the surface energy and thickness in the @xmath13@xmath14 region we are considering is shorter when @xmath123 , while it is more or less the same when @xmath80 mev . to assess the importance of the bulk non - linear meson interactions on our study on @xmath13 and @xmath14 , we have performed calculations as in figure 9 but setting @xmath92 with @xmath124 , and setting @xmath125 with @xmath69 and @xmath126 ( as indicated , the effect of @xmath41 is much smaller than that of @xmath40 ) . the results show a completely similar behaviour to figure 9 . even the slope of the contour lines of @xmath31 and @xmath75 in the @xmath13@xmath14 plane changes only slightly . comparing with figure 9 , when @xmath92 one finds that @xmath31 is shifted by approximately @xmath127 mev , and that when @xmath125 then @xmath31 is shifted by around @xmath128 mev . the shifts of the surface thickness @xmath75 are less regular and their magnitude depends on the value of @xmath13 and @xmath14 . in order to investigate the impact of the gradient interactions @xmath13 and @xmath14 on the spin orbit potential , in figure 10 we have plotted @xmath36 at the fermi surface for a few selected values of @xmath13 and @xmath14 . the nuclear matter properties and the scalar mass are the same as in figure 6 , where we studied the dependence of @xmath36 on @xmath66 . one can see that the meson interaction with @xmath129 and @xmath130 reduces the strength of @xmath36 and shifts the position of the minimum slightly to the exterior . on the contrary , the interaction with @xmath131 and @xmath132 makes the potential well deeper . the combined effect is probed in the case @xmath133 . since in the relativistic model the spin orbit force is strongly correlated with the dirac effective mass , we compare in figure 10 the situation at @xmath99 and at @xmath71 . we realize that the incidence of @xmath13 on @xmath36 is weaker for @xmath71 . the small perturbations arising from the gradient interactions when @xmath71 are not sufficient to produce a spin orbit strength equivalent to that of the case @xmath99 . to conclude this section we investigate the influence of the @xmath8 tensor coupling _ _ ^(z ) _ ( z ) [ eq3.9 ] which adds some momentum and spin dependence to the interaction . the natural combination for this coupling is @xmath134 . well known from one - boson - exchange potentials ( where @xmath42 above is commonly written as @xmath135 ) , the tensor coupling was included in the fits to nuclear properties of refs.@xcite ( conventional qhd ) and @xcite ( effective field theory ) , and in the study of the nuclear spin orbit force in chiral effective field theories carried out in ref.@xcite . these works noticed the existence of a trade - off between the size of the @xmath8 tensor coupling and the size of the scalar field . in other words , the tensor coupling breaks the tight connection existing in relativistic models between the value of the nucleon effective mass at saturation and the empirical spin orbit splitting in finite nuclei ( which constrains @xmath45 to lie between 0.58 and 0.64 @xcite ) . including a tensor coupling the authors of refs . @xcite were able to obtain natural parameter sets that provide excellent fits to nuclear properties and spin orbit splittings with an equilibrium effective mass remarkably higher ( @xmath136 ) than in models that ignore such coupling . we want to analyze the nature of this effect in the simpler but more transparent framework of semi - infinite nuclear matter . in figure 11 we have drawn the surface energy coefficient and the surface thickness as functions of @xmath42 in the range @xmath137 $ ] for two values of the effective mass and of the incompressibility , having set @xmath55 mev , @xmath76 @xmath57 and @xmath77 mev . to exemplify the incidence of @xmath42 on the spin orbit potential , figure 12 displays @xmath36 at the fermi surface for a few of the cases of figure 11 . we also performed the calculations for @xmath138 mev : @xmath31 and @xmath75 are shifted downwards with respect to figure 11 and @xmath36 is deeper than in figure 12 , but the global trends with @xmath42 are the same . figure 11 shows the strong reduction of @xmath31 and @xmath75 as @xmath42 increases ( the slope of the curves is milder for @xmath139 than for @xmath79 ) . figure 12 reveals that this fact is associated with a deeper and wider spin orbit potential . this agrees with the results of hofer and stocker @xcite who showed in the standard rmf model that the spin orbit coupling reduces the surface energy and thickness . at variance with the individual values of @xmath31 and @xmath75 , the ratio @xmath140 stays to a certain extent constant with @xmath42 . figure 12 evinces the sensitivity of @xmath36 to @xmath42 . the lower the nucleon effective mass is , the larger the effect . for @xmath91 we realize that with positive values of @xmath42 ( @xmath141 in the present case ) one can get a spin orbit strength comparable , or even stronger , to that of the case @xmath79 and @xmath142 , something that could not be achieved with natural values of the couplings studied in the previous sections . since our parametrization with @xmath91 and @xmath61 mev at @xmath142 already has reasonable surface energy and thickness ( @xmath143 mev and @xmath144 fm ) , increasing @xmath42 results in smaller values of @xmath31 and @xmath75 . this should be compensated with the other couplings ( especially @xmath13 and @xmath14 ) that modify the spin orbit strength to a lesser degree than @xmath42 , or the starting point should have other values of the incompressibility @xmath46 and the scalar mass @xmath64 . orbit effect has to do with the explicit dependence of the nucleon orbital wave functions on the spin orientation @xmath18 . as described in ref . @xcite nucleons with @xmath145 feel an attractive spin orbit potential and are pushed to the exterior of the surface , whereas the spin orbit force is repulsive for nucleons with @xmath146 which are pushed to the interior . as a consequence of this a depletion of particles with @xmath146 occurs at the surface . this behaviour is contrasted in figure 13 for @xmath142 and @xmath147 in the case @xmath91 . the figure depicts the profiles of the total baryon and tensor densities as well as those of their spin components @xmath148 and @xmath149 for @xmath30 , eqs.([eq7])([eq9b ] ) . when the spin orbit strength is large , attraction dominates over repulsion and more particles accumulate at the surface than particles are removed from it . then the total baryon density is enhanced at the surface region and it falls down more steeply . we briefly recall some basic definitions concerning nuclear surface symmetry properties ( further details on the relativistic treatment of asymmetric infinite and semi - infinite nuclear matter can be found in refs . @xcite ) . for a bulk neutron excess @xmath150 ( i.e. , the asymptotic asymmetry far from the surface ) , a surface energy coefficient can be computed as e_s ( _ 0 ) = 4r_0 ^ 2 ^_- dz , [ eq4.3]where @xmath151 is the total energy density of the system of neutrons and protons , @xmath152 denotes the energy per particle in nuclear matter of asymmetry @xmath153 , and @xmath154 with @xmath155 and @xmath156 referring to the neutron and proton densities , respectively . according to the liquid droplet model ( ldm ) @xcite , for small values of the neutron excess @xmath157 can be expanded as follows : e_s(_0 ) = e_s + ( + ) _ 0 ^ 2 + . [ eq4.4]in this equation @xmath158 stands for the bulk symmetry energy coefficient , @xmath159 reads for the ldm coefficient that expresses the density dependence of the symmetry energy , and @xmath160 is the so - called surface stiffness coefficient that measures the resistance of the system against pulling the neutron and proton surfaces apart . all of these macroscopic coefficients are familiar from semi - empirical ldm mass formulae . another quantity of interest is the neutron skin thickness @xmath161 , namely the separation between the neutron and proton surface locations : = _ -^dz . [ eq4.5]in finite nuclei @xmath161 would correspond to the difference between the equivalent sharp radii of the neutron and proton distributions . in the small asymmetry limit the ldm predicts a linear behaviour of @xmath161 with @xmath153 : = _ 0 . [ eq4.6 ] for calculations of finite nuclei of small overall asymmetry @xmath162 , the ldm expansion of the energy can be written as e = ( a_v + j i^2 ) a + a^2/3 + a_c z^2 a^-1/3 + , [ eq4.7]where @xmath163 is the coulomb energy coefficient . notice that @xmath164 in finite nuclei . to describe asymmetric matter in the relativistic approach we need to generalize the energy density ( [ eq1 ] ) by including the isovector @xmath0 meson . in terms of the mean field @xmath165 , with @xmath166 the time - like neutral component of the @xmath0-meson field , the additional contributions to eq . ( [ eq1 ] ) read & & _ _ ^(z ) _ ( z ) + r(z ) + & & - ( r(z ) ) ^2 - ( 1 + _ ) r^2(z ) . [ eq4.1]the symmetry energy coefficient turns out to be j = + . [ eq4.2]in the conventional model one has @xmath167 . the isovector tensor coupling @xmath168 was included in the calculations of refs . the new non - linear coupling @xmath169 between the @xmath0- and @xmath170-meson fields is of order 3 in the expansion and it has been introduced in refs.@xcite . we will not consider higher - order non - linear couplings involving the @xmath0 meson since the expectation value of the @xmath0 field is typically an order of magnitude smaller than that of the @xmath8 field @xcite . for example , in calculations of the high - density nuclear equation of state , mller and serot @xcite found the effects of a quartic @xmath0 meson coupling ( @xmath171 ) to be only appreciable in stars made of pure neutron matter . on the other hand , in analogy to the couplings @xmath13 and @xmath14 for the @xmath170 and @xmath8 fields , we also tested a surface contribution @xmath172 and found that the impact it has on the properties we will study in this section is absolutely negligible . as we have seen , the quantity that governs the surface properties in the regime of low asymmetries is the surface stiffness @xmath160 . table 1 analyzes the effect on @xmath160 and @xmath159 of the couplings discussed in the preceding sections and of the @xmath168 and @xmath169 parameters . on the basis of eq . ( [ eq4.6 ] ) , we have extracted @xmath160 from a linear regression in @xmath153 to fit our results for @xmath161 up to @xmath173 . we have set the equilibrium properties to @xmath174 @xmath57 , @xmath98 mev , @xmath61 mev , @xmath79 and @xmath175 mev , and have used a scalar mass @xmath55 mev and a @xmath0-meson mass @xmath176 mev . though here we are interested in tendencies rather than in absolute values , for comparison we mention that nl1 has ( units in mev ) @xmath177 , @xmath178 and @xmath179 @xcite , the sophisticated droplet - model mass formula frdm @xcite implies @xmath180 , @xmath181 and @xmath182 , and the etfsi-1 mass formula @xcite based on microscopic forces predicts @xmath183 , @xmath184 and @xmath185 . table 1 shows that the influence on the surface stiffness of the volume self - interactions @xmath66 , @xmath40 and @xmath41 is not very large for natural values of these couplings . in the present case @xmath160 is slightly increased by decreasing @xmath66 ( i.e. , by increasing the quartic vector coupling @xmath39 ) . for @xmath125 we find a non - negligible increase of @xmath160 , which signals a larger rigidity of the nuclear system against the separation of the neutron and proton surfaces . the effect of @xmath41 is again moderate as compared to that of @xmath40 . @xmath160 is augmented by a positive @xmath169 coupling , while a negative @xmath169 induces a lower value of @xmath160 . some visible changes in @xmath160 take place when the @xmath13 and @xmath14 gradient interactions are taken into account . due to the opposite behaviour of @xmath160 with @xmath13 and @xmath14 , the tendencies compensate in a case like @xmath133 , but the net effect is reinforced e.g. if @xmath186 . as one could expect the isoscalar tensor coupling @xmath42 has a notable effect on @xmath160 , even for the relatively small value @xmath187 that we have used in table 1 . on the contrary , @xmath160 is virtually insensitive to the isovector tensor coupling @xmath168 . the reason is that the derivative of the @xmath188 field is much smaller than that of the @xmath189 field . in the least - square fits to ground - state properties of refs . @xcite nothing was gained by the @xmath0 tensor coupling . in any case , the best fits of refs.@xcite have @xmath190 . from table 1 we recognize that the main changes in the coefficient @xmath159 arise from the @xmath169 coupling . as a rule of thumb , increasing values of @xmath160 are associated with decreasing values of @xmath159 for the bulk couplings @xmath66 , @xmath40 , @xmath41 and @xmath169 . since @xmath159 is a bulk quantity , it is not modified by the surface interactions . figures 14 and 15 illustrate the dependence on asymmetry of the neutron skin thickness , surface energy and surface thickness for some of the cases considered in table 1 . the figures extend up to @xmath191 , which widely covers the range relevant for laboratory nuclei ( @xmath192 ) . as the system becomes neutron rich we can appreciate how a neutron skin develops and @xmath161 grows from its vanishing value at @xmath193 . for small asymmetries the growth is linear in @xmath153 , as predicted by the ldm . the surface energy coefficient grows quadratically with increasing neutron excess and the ldm equation ( [ eq4.4 ] ) is clearly a good approximation . in general , the interactions having thicker neutron skins ( smaller values of @xmath160 ) also have larger surface energies . the parameter @xmath169 can be used for the fine tuning of the symmetry properties of the interaction without spoiling the predictions for symmetric systems . if in the conventional ansatz @xmath194 is fixed by the value of the symmetry energy @xmath158 , in the extended model @xmath158 depends on a combination of @xmath194 and @xmath169 , eq . ( [ eq4.2 ] ) . therefore , @xmath169 provides in practice a mechanism that can help to simultaneously adjust @xmath160 ( to get the required neutron skin @xmath161 ) and @xmath158 ( to keep the fit to the masses ) preserving the symmetric surface properties . within relativistic mean field theory , we have investigated the influence on nuclear surface properties of the non - linear meson interactions and tensor couplings recently considered in the literature . these interactions , beyond standard qhd , are based on effective field theories . the effective field theory approach allows one to expand the non - renormalizable couplings , which are consistent with the underlying qcd symmetries , using naive dimensional analysis and the naturalness assumption @xcite . the quartic vector self - interaction @xmath39 makes it possible to obtain a desirable positive value of the coupling constant @xmath38 of the quartic scalar self - interaction , for realistic nuclear matter properties and within the bounds of naturalness . this @xmath39 coupling has only a slight impact on the surface properties . nevertheless , it helps to find parametrizations where both the surface energy coefficient @xmath31 ant the surface thickness @xmath75 lie in the empirical region . the @xmath39 vector non - linearity makes the spin orbit potential well deeper , although the effect is almost negligible . concerning the volume non - linear couplings @xmath40 and @xmath41 , they also allow one to obtain positive values of @xmath38 in the region of naturalness , depending somewhat on the saturation properties ( incompressibility and effective mass ) . the surface properties are not much affected by these bulk terms either , and it turns out that @xmath41 has a marginal effect as compared to that of @xmath40 . the equilibrium properties do not depend on the couplings @xmath13 and @xmath14 that involve the gradients of the fields . thus , these couplings serve to improve the quality of the surface properties without changing the bulk matter . in the conventional model the only parameter not fixed by the saturation conditions is the mass of the scalar meson . in the @xmath13@xmath14 plane the lines of constant @xmath31 have a different slope than those of constant @xmath75 . it is then possible to keep a fixed value of @xmath31 and to modify the value of @xmath75 by choosing @xmath13 and @xmath14 appropriately . the range of variation of @xmath31 and @xmath75 with @xmath13 and @xmath14 is wider than with the volume couplings . this justifies including these gradient terms in the energy functional in spite of being of order 5 in the expansion . the @xmath13 and @xmath14 surface meson interactions also influence the spin orbit potential , but the effect is not extremely significant . the effective model is augmented with a tensor coupling of the @xmath8 meson to the nucleon . an outstanding feature is the drastic consequences it has for the spin orbit force . we have emphasized how inclusion of @xmath42 permits to obtain a spin orbit strentgh similar to that of @xmath195 with larger values of the equilibrium nucleon effective mass , contrary to the phenomenology known from models without such a coupling . we have discussed the implications of the extra couplings of the extended model on various surface symmetry properties . we have restricted ourselves to the regime of low asymmetries , where the liquid droplet model can be applied and the surface stiffness coefficient @xmath160 is the key quantity . in particular we have pointed out the role that the non - linearity @xmath169 of the isovector @xmath0-meson field may play in the details of the symmetry properties . the authors would like to acknowledge support from the dgicyt ( spain ) under grant pb95 - 1249 and from the dgr ( catalonia ) under grant gr94 - 1022 . m. del estal acknowledges in addition financial support from the cirit ( catalonia ) . 99 = -1.0 mm b.d . serot and j.d . walecka , adv . phys . 16 ( 1986 ) 1 . celenza and c.m . shakin , relativistic nuclear physics : theories of structure and scattering ( world scientific , singapore , 1986 ) . reinhard , rep . 52 ( 1989 ) 439 . serot , rep . phys . 55 ( 1992 ) 1855 . serot and j.d . walecka , int . j. of mod . e 6 ( 1997 ) 515 . j. boguta and a.r . bodmer , nucl . phys . a 292 ( 1977 ) 413 . furnstahl , b.d . serot and h.b . tang , nucl . phys . a 598 ( 1996 ) 539 . h. mller and b.d . serot , nucl . a 606 ( 1996 ) 508 . furnstahl , b.d . serot and h.b . tang , nucl . phys . a 615 ( 1997 ) 441 . bodmer , nucl . phys . a 526 ( 1991 ) 703 . bodmer and c.e . price , nucl . phys . a 505 ( 1989 ) 123 , and references therein . s. gmuca , z. phys . a 342 ( 1992 ) 387 . s. gmuca , nucl . phys . a 547 ( 1992 ) 447 . y. sugahara and h. toki , nucl . phys . a 579 ( 1994 ) 557 . k. sumiyoshi , h. kuwabara and h. toki , nucl . phys . a 581 ( 1995 ) 725 . m. rufa , p .- reinhard , j.a . maruhn , w. greiner and m.r . strayer , phys . c 38 ( 1988 ) 390 . furnstahl , j.j . rusnak and b.d . serot , nucl . phys . a 632 ( 1998 ) 607 . myers and w.j . swiatecki , ann . 55 ( 1969 ) 395 ; ann . 84 ( 1974 ) 186 ; w.d . myers , droplet model of atomic nuclei ( plenum , new york , 1977 ) . m. centelles , x. vias and p. schuck , phys . c 53 ( 1996 ) 1018 . m. brack , c. guet and h .- b . hkansson , phys . rep . 123 ( 1985 ) 275 ; j. treiner and h. krivine , ann . 170 ( 1986 ) 406 ; k. kolehmainen , m. prakash , j.m . lattimer and j. treiner , nucl . phys . a 439 ( 1985 ) 537 . w. stocker , j. bartel , j.r . nix and a.j . sierk , nucl . phys . a 489 ( 1988 ) 252 . w. stocker and m.m . sharma , z. phys . a 339 ( 1991 ) 147 ; m.m . sharma , s.a . moszkowski and p. ring , phys . c 44 ( 1991 ) 2493 . m. centelles , x. vias , m. barranco and p. schuck , ann . ( n.y . ) 221 ( 1993 ) 165 ; m. centelles and x. vias , nucl . phys . a 563 ( 1993 ) 173 ; m. del estal , m. centelles and x. vias , phys . c 56 ( 1997 ) 1774 . c. speicher , r.m . dreizler and e. engel , nucl . phys . a 562 ( 1993 ) 569 . d. von - eiff , j.m . pearson , w. stocker and m.k . weigel , phys . b 324 ( 1994 ) 279 . m. centelles , m. del estal and x. vias , nucl . phys . a 635 ( 1998 ) 193 . d. hofer and w. stocker , nucl . phys . a 492 ( 1989 ) 637 . d. von - eiff , w. stocker and m.k . weigel , phys . c 50 ( 1994 ) 1436 . d. von - eiff , h. freyer , w. stocker and m.k . weigel , phys . b 344 ( 1995 ) 11 . p. hohenberg and w. kohn , phys . b 136 ( 1964 ) 864 . c. speicher , r.m . dreizler and e. engel , ann . phys . 213 ( 1992 ) 312 . w. kohn and l.j . sham , phys . rev . a 140 ( 1965 ) reinhard , m. rufa , j. maruhn , w. greiner and j. friedrich , z. phys . a 323 ( 1986 ) 13 . p. mller , j.r . nix , w.d . myers and w.j . swiatecki , at . data nucl . data tables 59 ( 1995 ) 185 . y. aboussir , j.m . pearson , a.k . dutta and f. tondeur , nucl . phys . a 549 ( 1992 ) 155 . table 1 . : : the surface stiffness coefficient @xmath160 and the coefficient @xmath159 for several values of the couplings analyzed in the text . we have set @xmath97 @xmath57 , @xmath98 mev , @xmath61 mev , @xmath79 , @xmath175 mev and @xmath55 mev . [ cols="^,>,>,>,>,>,>,>,^,^",options="header " , ] figure 1 . : : the couplings @xmath53 , @xmath54 , @xmath37 , @xmath38 and @xmath39 versus the parameter @xmath66 defined in eq.([eq3.2.b ] ) . the naturalness assumption requires all these couplings to be of order unity . we have taken @xmath76 @xmath57 ( @xmath58 @xmath59 ) , @xmath60 mev and @xmath55 mev . : : surface energy coefficient @xmath31 for several values of the parameter @xmath66 , @xmath46 , @xmath45 and @xmath64 , with @xmath76 @xmath57 and @xmath60 mev . figure 3 . : : surface thickness @xmath75 of the baryon density profile for several values of the parameter @xmath66 , @xmath46 , @xmath45 and @xmath64 , with @xmath76 @xmath57 and @xmath60 mev . : : surface energy coefficient @xmath31 and surface thickness @xmath75 for several values of the parameter @xmath66 , @xmath45 and @xmath64 , with @xmath85 @xmath57 , @xmath86 mev and @xmath87 mev ( ref . : : baryon density @xmath93 and surface tension density @xmath196 of semi - infinite nuclear matter for some values of the parameter @xmath66 . it is @xmath97 @xmath57 , @xmath98 mev , @xmath61 mev and @xmath55 mev . : : orbital part of the spin orbit potential @xmath36 at the fermi surface and central mean field @xmath96 , eqs . ( [ eq12 ] ) and ( [ eq14 ] ) respectively , for some values of the parameter @xmath66 . it is @xmath97 @xmath57 , @xmath98 mev , @xmath61 mev and @xmath55 mev . : : the couplings @xmath53 , @xmath54 , @xmath37 and @xmath38 against the parameters @xmath40 ( left ) and @xmath41 ( right ) . with @xmath76 @xmath57 , @xmath60 mev and @xmath55 mev , results are shown for @xmath61 mev and @xmath79 in part ( a ) , and for @xmath80 mev and @xmath91 in part ( b ) . : : level curves in the plane @xmath40@xmath41 of the surface energy coefficient @xmath31 ( in mev , solid lines ) and of the surface thickness @xmath75 ( in fm , dashed lines ) , with @xmath197 . the point @xmath124 is marked by a cross . it is @xmath76 @xmath57 , @xmath60 mev , @xmath198 mev , @xmath62 and @xmath55 mev . : : same as figure 8 in the plane @xmath13@xmath14 , with @xmath199 . the point @xmath200 is marked by a cross . figure 10 . : : orbital part of the spin orbit potential @xmath36 at the fermi surface for some values of the couplings @xmath13 and @xmath14 . the equilibrium properties of nuclear matter and the scalar mass are the same of figure 6 . figure 11 . : : surface energy coefficient @xmath31 and surface thickness @xmath75 as functions of the strength @xmath42 of the @xmath8-meson tensor coupling , with @xmath201 . we have set @xmath76 @xmath57 , @xmath60 mev and @xmath55 mev . figure 12 . : : orbital part of the spin orbit potential @xmath36 at the fermi surface for some values of the tensor coupling @xmath42 . the equilibrium properties of nuclear matter and the scalar mass are the same of figures 6 and 10 . figure 13 . : : total baryon density @xmath93 , total tensor density @xmath202 and their components @xmath148 and @xmath203 for the two spin orientations @xmath204 . they have been calculated for @xmath142 and @xmath147 , with @xmath91 , @xmath61 mev and @xmath55 mev . figure 14 . : : neutron skin thickness @xmath161 as a function of the bulk neutron excess @xmath153 . the solid line is the result of the conventional model ( @xmath205 ) . the other lines differ from the latter in the indicated parameter . figure 15 . : : same as figure 14 for the surface energy coefficient @xmath31 and the surface thickness @xmath75 as functions of the bulk neutron excess squared @xmath206 .

gravitational lensing directly measures mass density fluctuations along the lines of sight to very distant objects . no assumptions need to be made concerning bias , the ratio of fluctuations in galaxy density to mass density . hence , lensing is a very useful tool to study the universe at low to moderate redshifts . we describe in detail a new method to trace light rays from redshift zero through three dimensional mass distribution to high redshift . as an example , this method is applied here to a standard cold dark matter universe . we obtain a variety of results , some of them statistical in nature , others from rather detailed case studies of individual lines of sight " . among the former are the frequency of multiply imaged quasars , the distribution of separation of the multiple quasars , and the redshift distribution of lenses : all that as a function of quasar redshift . we find effects from very weak lensing to highly magnified multiple images of high redshift objects , which , for extended background sources , ( i.e. galaxies ) , range from slight deformations of the shapes through tangentially aligned arclets up to giant luminous arcs . different cosmological models differ , increasingly with redshift , in their predictions for the mass ( thus gravitational potential ) distributions . our ultimate goal is to apply this method to a number of cosmogonic models and to eliminate some models whose gravitational lensing properties are inconsistent with those observed .

astro-ph9610096

each specific model for the development of cosmogonic structure ( e.g. the hot dark matter [ hdm ] or cold dark matter [ cdm ] scenarios ) has one free parameter , the amplitude of the density power spectrum . in the light of the cobe observations ( smoot et al . 1992 ; gorski et al . 1994 ) , it occurs , for the first time , that this parameter is fixed ( @xmath0 ) by the determination on the @xmath1 scale in the linear regime . with its amplitude fixed , a secure determination of the potential fluctuation on any other scale provides a test for a particular cosmological model . any single conflict between the theory and reality can falsify the former . the most leverage is obtained for tests which are made on angular scales as far away as possible from the cobe measurements . but they should not be so small as to be greatly influenced by the difficulty inherent in modeling the physics of the gaseous , baryonic components ( @xmath2 ) . thus , critical tests are best made on scales @xmath3 . gravitational lensing measures _ directly _ the fluctuations in the gravitational potential along random lines of sight to distant objects . in contrast , the conventional tools for comparing cosmogonic theories with observations rely on either galaxy density or galaxy velocity information , both of which unavoidably suffer from the uncertain situations with regard to density or velocity bias of galaxies over the underlying mass distribution , hampering our attempts to understand the more fundamental " question concerning the mass evolution and distribution . hence gravitational lensing provides a powerful , independent test of cosmogonic models , because of its unique ability to directly measure the gravitational potential fluctuations . the sampled properties of cosmic structure on certain scales are also more fair " than other conventional measures owing to the fact that lines of sight to distant objects are _ random _ with regard to the foreground matter distribution . we present here a method to determine gravitational lensing effects for realistic mass distributions from the weak to the strong lensing regime . first , we fill the universe densely " with adjacent mass cubes , taken from cosmological simulations , along the line of sight , where inside each cube the matter is projected onto the middle plane . then , we study the lensing properties of the intervening matter by shooting " light rays through the lens ( mass ) planes representing matter distributions along redshifts ( cf . figure 1 for a schematic version of our actual scheme [ which uses of order 100 lens planes ] ) . the gravitational lens effects of many lens planes have been studied by several authors in the past ( e.g. , kochanek & apostolakis 1988 ; schneider & weiss 1988a , b ; jaroszyski 1989 , 1991 ; jaroszyski et al . 1990 ; lee & paczyski 1990 ; babul & lee 1991 ; bartelmann & schneider 1991 ) . these authors have typically used point lenses or isothermal spheres which were randomly distributed on individual lens planes . the number of lens planes used ranges from 2 to 32 . most of these studies involved the regime of weak lensing , i.e. slight effects on the magnification of sources , but not multiple imaging . rauch ( 1991 ) studied the microlensing effect of three - dimensionally distributed point lenses on high redshift supernovae . a complementary approach to the one described here , a semi - analytical method using the press - schechter formalism was first used by narayan and white ( 1987 ) , and more recently applied by kochanek ( 1995 ) to various cosmological models . a very preliminary attack on the problem of studying the lensing properties of different cosmological models has been presented by cen et al . ( 1994 , hereafter cgot ) . in that work no ray tracing was done . rather it was simply checked whether or not mass accumulations were greater than the critical level ( cf . turner , ostriker & gott 1984 , henceforth tog ) at which multiple imaging will occur . the mathematical and physical treatment here is far superior to that adopted in cgot ; the essential conclusions for the standard cdm scenario , however , are unchanged . some of these conclusions have already been presented in wambsganss et al . ( 1995 ) . here we provide the mathematical basis of that method as well as some tests of the method . we use cosmological simulations for the standard cobe - normalized @xmath4 , cdm universe as an example to illustrate the method . a first report on the results has been published elsewhere ( wambsganss et al . 1995 ) . in principle , mass distributions from any cosmological model can be treated with this method . the ultimate goal of this project is to provide a quantitative comparison of the lensing properties of different cosmological models ( e.g. , cdm with @xmath4 ; low density ( open ) cdm models ; mixed dark matter [ mdm ] models ; flat models with a cosmological constant ; isocurvature models etc ) with observations . for example , we anticipate that different cosmological models , which are typically tuned to give approximately the right properties when being compared with our local universe at redshift @xmath5 , will show different lensing properties because they reach the final state ( @xmath5 ) from quite different paths . thus , the typical lens will appear at different redshifts in different models . for example , in mdm models , where structure forms late , the lenses will have small redshift , but in isocurvature models , where the first structures form shortly after recombination , the average lens redshift will be much larger . for many questions ( e.g. the fraction of multiply imaged quasars as a function of angular separation , magnification ratio , redshift ) one needs a large number of simulations with different realizations of the matter distribution , in order to give quantitative answers on a good statistical basis . for other applications , e.g. detailed studies of lensing by clusters of galaxies and the effects of chance alignments with foreground or background matter , a specific realization must be studied in detail . while it is in principle possible to simulate , numerically in one simulation , the entire path from observer to source ( very large simulations with elongated geometry have been done ; e.g. , park & gott 1991 ) , technical limitations preclude this at the present time . for now we perform many independent simulations for different volumes along the line of sight ( since repeated structures are obviously to be avoided ) . we have developed an approximate technique for this which we term the convolution method " . physically it consists of assuming that separate pieces of an inhomogeneous universe act as if they were part of an homogeneous universe of the same mean density as themselves . details of the method , which is an improved version over that applied in cgot , are presented in the appendix , as well as detailed checks which indicate the accuracy of the approximations involved . although we think this convolution technique describes the properties accurately enough on the scales in which we are interested , we hope in the future to be able to avoid this step and be able to use simulations that have a large enough dynamic range to cover a representative part of the universe and have high enough resolution in order to represent the potential wells of galaxies , groups and clusters appropriately . the analytical techniques presented in this paper for the determination of the lens properties are independent of the details of the method for computing the evolution . we will use the same ray tracing method when later we utilize higher resolution numerical evolution techniques which do not depend on the convolution method . the rest of the paper is organized as follows . the ray - shooting technique is described and discussed in detail in section 2 . as a result of this ray - shooting , we obtain a mapping of the rays from the image plane ( sky plane ) to the source plane . with this mapping it is possible to determine the magnification distribution of background sources , the fraction of the source plane that is multiply imaged , the positions and shapes as well as the topology of the caustics and the corresponding critical lines ( section 3 ) . for a set of given source positions and shapes in the source plane at a given redshift , we determine and analyze the corresponding image configurations as they would appear in the sky , as is shown in sections 4 and 5 . finally in section 6 we give a short summary and describe the planned applications and the types of predictions we will be able to make for the different cosmological models . in schneider , ehlers and falco ( 1992 , subsequently sef ) a whole chapter deals with the theory of multiple light deflection , i.e. gravitational lensing by more than one plane . we refer the interested reader to this text book for details . here we just describe the multiplane lens equation as we use it : @xmath6 \left ( \begin{array}{c } x_{1 } \\ x_{2 } \end{array } \right ) = a({\vec x } ) { \vec x}. \eqno(1)\ ] ] here @xmath7 is the position vector of a light ray in the first lens plane ( or equivalently in the image / sky plane ) ; @xmath8 is its corresponding position in the source plane ; @xmath9 is the ( two dimensional ) deflection angle in lens plane # i ; @xmath10 and @xmath11 are the angular diameter distances between observer and source plane and between @xmath12th lens plane and source plane , respectively . the lens planes are labelled from @xmath13 through @xmath14 in order of increasing redshift . the part of the equation that we are essentially interested in is @xmath15 ; it describes just the mapping between the image / sky plane and the source plane . the relation between the positions @xmath16 in the image / sky plane and @xmath17 in the source plane is given by @xmath18 , the jacobian matrix of the mapping . in the case of multiple lens planes the jacobian @xmath19 does not have all the nice properties it has in the single lens plane case , in particular it is not symmetric any more in the general case : @xmath20 . furthermore it is not curl - free : @xmath21 . many other properties , however , are still valid . the magnification of a certain image at position @xmath16 is given as @xmath22 ^{-1 } = \left [ \ a_{11}({\vec x } ) a_{22}({\vec x } ) - a_{12}({\vec x } ) a_{21}({\vec x } ) \ \right ] ^{-1}. \eqno(3)\ ] ] we will see in later sections that other combinations of the components of @xmath23 represent useful quantities as well . the effective surface mass density @xmath24 ( which is basically the sum of the physical surface mass densities of the individual lens planes , weighted by the corresponding ratios between the angular diameter distances ) for any position @xmath7 in the sky plane is given as @xmath25 . \eqno(4)\ ] ] similarly , the components of the shear @xmath26 $ ] are @xmath27 ; \eqno(5a)\ ] ] @xmath28 . \eqno(5b)\ ] ] in order to shoot " a light ray @xmath12 through a single lens plane , what must be done is to determine the deflection angle @xmath29 of this ray due to all matter in this lens plane . for a number of @xmath30 point lenses this is just a summation of the deflection angle by each point lens @xmath31 : @xmath32 here @xmath33 and @xmath34 are the gravitational constant and the speed of light , respectively , @xmath35 is the mass of point lens @xmath31 , @xmath36 is the projected ( vector ) distance between the positions of light ray @xmath12 and point lens @xmath31 , and @xmath37 is its absolute value , @xmath38 here ( @xmath39,@xmath40 ) is the position of ray @xmath12 , and ( @xmath41,@xmath42 ) is the position of lens @xmath31 . the straightforward determination of one deflection angle takes about ten floating point operations . as a lens plane , we consider a two - dimensional distribution of matter that is given in a grid ; for the examples here we will always use a regular matter grid with 500@xmath43 positions ( in subsequent work we will increase the resolution to @xmath44 positions ; in principle this can be increased to any desired resolution , depending on the desired purpose ) . this is the projection of the three - dimensional matter distribution inside a cube with comoving side length @xmath45 ( where @xmath46 is the hubble constant divided by 100 ) onto one of the three faces of the cube . we treat this matter distribution as 250,000 point lenses , or more accurately , as smeared out " in matter or lens pixels , i.e. in a square region of comoving @xmath47 at a side . we must determine the deflection angle at least at as many positions as there are lens pixels , i.e. at least on a 500@xmath43 grid ( again , this is the resolution we use for the examples here ; this can be changed to whatever ray density is appropriate to the problem at hand ) . because we will use of order 100 lens planes , the total number of mathematical operations for a direct determination of the deflection angle amounts to about @xmath48 floating point operations . this is non - trivial , even for fast computers . in order to speed up the calculation we determine the deflection angle in each lens plane with a hierarchical tree code in two dimensions , as described in wambsganss ( 1990 ) . the idea is to group together lenses that are far away from the light ray , and treat these cells " ( whose sizes increase with the distance to the ray ) as pseudo - lenses . nearby lenses are treated individually , lenses at intermediate distances to the light ray are grouped into small cells of a few lenses , and those lenses very far away from the light ray considered , are in cells containing up to a few thousand lenses . this tree code treatment speeds up the determination of the deflection angle by a large factor . the number of lenses and pseudo - lenses necessary for an accurate determination of the deflection angle in one plane is of order 100 ( rather than 250,000 for the direct determination ) . there is some overhead , namely the determination of the tree structure for each ray , i.e. the configuration of which lenses have to be treated directly and which ones are to be put in cells of different sizes . however , this is done just once , and used again in all subsequent planes , since the relative positions of the lenses are exactly the same . the cost of this additional calculation is negligible compared to that of a direct determination of the deflection angles . in its simplest application the tree code assumes that all matter inside a cell is located at its center of mass . we do , however , use higher order multipoles of the mass distribution in order to increase the accuracy ( cf . wambsganss 1990 ) . for the actual calculations we use multipoles up to order 6 . in figure 2 we show the deviation of the deflection angle determined with this tree - code method from the directly determined , i.e. accurate deflection angle : @xmath49 as a function of @xmath50 in arbitrary units ( the top row indicates the x - component , the bottom row the y - component ) . the panels correspond to determinations of @xmath51 including multipoles of different order ; they represent : all matter in center of mass , i.e. only monopole term ( leftmost column ) , monopole term plus quadrupole moments ( 2nd column ) , monopole term plus all multipoles up to order 4 ( 3rd column ) , and monopole term plus all multipoles up to order 6 ( rightmost column ) . in the actual calculation of the deflection angle we use multipoles up to order 6 , corresponding to the rightmost panel . it can be seen that for this case the largest deviation between the two methods of determining the deflection angle is of order @xmath52 , with the rms deviation quite a bit smaller . this seems to be a good enough approximation for our purposes . for point lenses the deflection angle formally diverges when the distance between light ray and lens position becomes zero , @xmath53 . however , in our case this is an artificial divergence , because the underlying mass distribution is smooth , we only approximate it by a large number of point lenses . in order to avoid this artificial divergence , we always determine the deflection angle at the points directly in the center between four lens positions so that nearest neighbor effects cancel . that means , for each lens plane we determine the deflection angles for a regular grid of test - rays " . but , since the real positions of the light rays are offset from the positions of these test - rays " , for which we determined the deflection angles , we calculate the actual deflection angles of the rays that are followed through the planes using a bicubic interpolation between the four test - rays surrounding the real ray position . in this way we obtain a smooth two - dimensional field of deflection angles for each lens plane , as is to be expected for a continuous matter distribution . the angular size of the ( square ) bundle of light rays that we consider is determined by the angular size of the highest redshift lens plane that we use . as an example , the angular size of a cube with comoving size @xmath45 at a redshift @xmath54 is about @xmath55 arcsec for a standard cdm universe . the ( average ) angular size of the field of rays is fixed , it is the same for all lens planes ; however , as the physical size of the underlying cosmological cubes is expanding with decreasing redshift , the angular size of the lens planes increases rapidly with decreasing redshift . this means that the field of real - rays intercepts only a small part of the lens plane for small redshifts ( cf . figure 1 ) . this effect is reflected in the matter distribution inside the field of view : for low redshift lens planes , one lens - pixel ( which has fixed comoving size ) covers quite a large angular region . the matter distribution of each lens plane inside the field of rays is shown in figure 3a for a particular line of sight : dark ( light ) gray means high ( low ) surface mass density ( in grams per cm@xmath43 ) . from top left to bottom right you see the matter inside the beam for 60 different lens redshifts , starting with @xmath57 and increasing to @xmath58 . for the low redshift planes one can easily identify the pixel size of the matter distribution , because there one matter pixel ( physical size 5 h@xmath59 mpc @xmath60kpc ) corresponds to many tens of arcsecs . in figure 3b the integrated matter distribution up to a redshift @xmath58 is shown . this is the physical " sum , i.e. in units of mass per physical area . this map could be converted into a map of light , if one assumes , e.g. , a constant mass - to - light ratio and weights the individual lens planes by the inverse of the ( luminosity ) distance . if one compares figures 3a and 3b , it is quite obvious that the total surface mass density is dominated by a few planes along the line of sight . in general , lensing is a very convergent process : it is not due to the sum of a large number of weak lensing events ; rather it is due to the occasional passage of a light ray past a large mass concentration . this fact makes the requirements for statistical validation of our results especially stringent . many independent realizations of the universe must be made and many lines of sight examined before one can have confidence ( by examination of convergence ) that the derived statistical results are robust ( i.e. the statistically averaged results become independent of the number of realizations studied ) . for each lens plane the critical surface mass density @xmath61 necessary for multiple imaging can be calculated ( cf . sef ) : @xmath62 comparing the surface mass density in each matter pixel @xmath63 with this number identifies the regions that should produce multiple images of a background source : @xmath64 where @xmath65 is the physical surface mass density in pixel @xmath63 , and @xmath66 is its normalized surface mass density . when @xmath67 , then multiple lensing occurs . we note , however , that overdensity in a single plane is a sufficient but not necessary condition of multiple imaging ; chance alignments of two or more individually subcritical but collective supercritical regions can produce multiple images as well as slightly underdense regions in combination with a sufficiently large value of shear . a comparison of such regions with the actual location of multiple images serves as a good check of the whole method . for each complete run of the light rays through all the lens planes , the positions of the light rays in the image / sky plane and the corresponding positions in the source plane are stored . in subsequent sections we will describe how the matter distribution is obtained in each lens plane , and how we analyze the mapping of light rays through essentially a realistic 3-dimensional matter distribution . but first we like to add a few paragraphs discussing the issue of numerical resolution , and describing an additional test of our method . many authors have studied magnification cross sections and caustic structures of various lens models . to name just a few : blandford & kochanek ( 1987 ) and kochanek & blandford ( 1987 ) studied magnification cross sections and probability distributions for isolated elliptical potentials ; hinshaw & krauss ( 1987 ) determined the lensing probabilities for isothermal spheres with finite cores ; wallington & narayan ( 1993 ) investigated the influence of the core radius on the imaging properties of elliptical lenses ; kovner ( 1987a , b ) looked into marginal " lenses ; kassiola & kovner ( 1993 ) compared the lensing properties of elliptical mass distributions with elliptical potentials . most of these authors studied idealized lens models in great detail . in order to get a good representation of , e.g. , the magnification probability of a certain elliptical lens , one needs a grid of rays with good coverage of the source plane , in particuar of the multiply imaged region ( of order 1000@xmath43 rays ) . we have a very different objective here , we want to study large scale statistical lensing properties of the matter rather than to model and understand individual lenses in great detail . therefore our resolution is very different from those examples mentioned above , i.e. we have many fewer rays covering the caustics region of our multiply imaging lenses than they have . this is due to the different questions we try to answer . nevertheless , we like to know how well the lensing properties are reproduced in the high magnification regions . therefore we compare the lensing properties of ideal elliptical lenses as obtained with our method to the high resolution results obtained by others . first we describe here briefly how we determine the magnification : we evaluate the magnification in a regular grid in the source plane , as is explained in more detail in chapter 4 . this is done by using the differential deflection properties of the four rays closest to the grid position to be evaluated . subsequently we interpolate two - dimensionally , to determine the value at the desired position . this way we do get a moderate resolution though certainly not a perfect coverage of the multiply imaged region : a region with size about ( 350 arcsec)@xmath43 is covered by about 500@xmath43 resolution elements in the examples here . however , we would like to emphasize that the resolution is not hardwired in this method . it can and will be increased with the availability of faster computers with more memory or for the investigation of specific problems . below we present a test of the accuracy of the magnification distribution obtained by our method . for that purpose we use elliptical lenses with different parameters ( ellipticity , core radius , velocity dispersion ) . we test the accuracy of our method by using elliptical lenses as described by wallington & narayan ( 1993 , henceforth wn93 ) . we run tests with a single lens plane populated by an isolated elliptical lens with various values of ellipticity , velocity dispersion and core radius . the results of these tests show that the qualitative lensing properties ( the caustic geometry , the evolution of the caustic / critical line structure with increasing core radius , as shown in fig.1 of wn93 ) are well reproduced . we also did a detailed quantitative comparison of the magnification distributions produced by the various elliptical lenses with the cross sections presented by wn93 ( see their fig.3 ) . we determined the cross sections for lensing of elliptical lenses with nine different velocity dispersions ranging from @xmath68 km / sec to 1000 km / sec . the ellipticities were chosen between 0.0 and 0.2 and we varied the core radii from 10% to 100% of the einstein radius of a corresponding circular symmetric isothermal sphere . we compared the cross sections for magnifications larger than 2 , 4 , 8 , 16 , 32 , ... with the values obtained by wn93 . for each lens " as specified by the parameter set of velocity dispersion , ellipticity , and core radius , we run 10 different realisations ( in terms of angle relative to the coordinate axes and position of center of lens ) . here is the summary of the comparison of our results with those of wn93 , for three values of the velocity dispersion @xmath69km / sec , @xmath70km / sec , @xmath71km / sec , all with ellipticity @xmath72 , and with various core radii . the distributions @xmath73 for magnifications @xmath74 , 4 and 8 agreed within about 15% or better for all three values of @xmath75 . for @xmath76 the deviations were still moderate for the two higher @xmath75-values ( 23% and 11% ) , whereas it was worse than that for the low velocity dispersion @xmath77 . even for @xmath78 , the deviation was only 30% for @xmath79 , but quite large for the two lower values . this indicates that the cross sections of gravitational lenses with velocity dispersions at or below 1000 km / sec are relatively well represented in the low and intermediate magnification regime . at high magnifications they are not very accurately represented in our simulations with the given resolution . however , the fluctuations in the cross sections are mostly of statistical nature ( the high magnification regions are covered only by a small number of resolution elements ) , as we could infer from comparisons among the 10 realisations we had used for each parameter set . in the applications of our method , there are always _ many _ lenses in each of our lines - of sight " , and since we will use 100 or more different lines of sight for our statistical analyses , the deviations relative to the wn93 values will get ( much ) smaller , again , due to the decreasing statistical noise . hence we do get a fair enough representation for the low and moderately high magnification regime . as we are interested mainly in the global statistical properties rather than the local ones , this test shows that our approach is justified . the matter evolution used in this paper is simulated by a particle - mesh ( pm ) n - body code ( cen 1992 ) . but a direct , brute force calculation with such a code is unable to cover the dynamic range needed : a ratio of cluster - cluster separation ( @xmath80mpc ) to the lensing scale of arcseconds ( @xmath81kpc ) translates to @xmath82 mesh points , which is far beyond the current generation of supercomputers . in order to overcome this problem a novel scheme involving convolution of a small scale simulation box ( which provides good resolution ) with a large scale simulation box ( which provides good statistical information concerning clusters ) was developed ( cgot ) . we basically use their scheme of convolving two box sizes , but we add a few essential improvements . our scheme is described in detail in the appendix . the three dimensional matter distribution from these calculations is given for any cosmological time as three two - dimensional projections onto the x - y , y - z , and x - z - planes , with an effective resolution of comoving 10@xmath83kpc , which is necessary in order to study the lens effects on small angular scales of approximately 5 arcsec . were we to wish to study still smaller scales , hydrodynamic simulations ( cf . e.g. cen 1992 ; katz , hernquist , & weinberg 1993 ; navarro , frenk & white 1994 ; summers , davis & evrard 1995 ; steinmetz & mller 1994 ) would be required since the baryonic component begins to dominate over the dark matter component within 10 kpc . the essence of the convolution method is to note that small regions of given over ( or under ) density act for most processes as if they were in homogeneous universes of larger ( or smaller ) mean density . for details and checks of the method see the appendix . for a typical lensing simulation a few hundred lens planes with size @xmath84 are required . ideally one would like to have as many independent cosmological sequences as lens planes used . this is not achievable for reasons of limited cpu time . but neither is it really necessary . the reason is that for most of the planes only a small fraction of the matter distribution is illuminated " by the bundle of light rays ( see figure 1 ) . in practice , we use ten independent evolutionary runs . for a given lens redshift the particular run to be used for this particular lens plane is chosen randomly , as is the projection to be used out of the three possible ones , as is the location and orientation of the square region that is illuminated " by the rays . the computing time for such a lensing simulation is directly proportional to the number of lens planes used . most of the lens planes have densities that are below the average density of the universe ( because a few have a high overdensity ) , some are almost completely empty . these lens planes contribute very little to the deflection angle . we did some experiments in bunching together a couple of lens planes at adjacent redshifts , that means adding up the matter distributions , and treating the combined matter distribution as a single pseudo "- plane . we found that for a grouping of about 10 real lens planes into one pseudo lens plane , the results ( see next sections ) are indistinguishable from treating all the lens planes individually . so we concluded by running the simulations with only 60 pseudo - planes ( up to a redshift of @xmath85 ) , where each one represents a grouping of 10 real planes . that means , in fact that one of our pseudo - lens planes represents a rectangular parallelepiped of size @xmath86mpc@xmath87 . in figure 4a ( color plate ) a two - dimensional magnification pattern is shown , representing the magnification as a function of source position at a redshift of @xmath85 for one particular lensing run of our example cosmology , a standard cdm model with @xmath88 and @xmath89 . the magnification is shown as a function of position in the source plane . by definition , the average magnification is @xmath90 , or @xmath91 , i.e. equal to the case were all the matter is smoothly distributed . whenever we use magnification " ( or demagnification ) subsequently , we mean magnification ( or demagnification ) relative to a situation in which all matter is smoothly distributed . the magnification as a function of position in the source plane is shown color - coded in figure 4a : it increases from blue through green and red to yellow . the border line between green and red corresponds roughly to the average magnification @xmath92 . green regions have magnifications slightly less than one ( i.e. , they are slightly de - magnified , compared to a universe in which all matter were smoothly distributed ) . red color indicates magnifications of up to one magnitude , yellow magnification by more than one magnitude . note that there is quite a large region that gets demagnified by a small amount , and a few relatively small spots that get quite high magnification . such a magnification pattern in the source plane is similar to the magnification patterns used in microlensing simulations at much smaller angular scales ( e.g. , wambsganss , paczyski & schneider 1990 ) . one difference between the magnification patterns used here and those used in microlensing is that there one is interested in the change of magnification as a function of time ( due to the relative motion of observer , lens and source ) , whereas here we only study the changes in the magnification as a function of position . another difference is that here we find caustics ( and multiple images ) only in rare cases , because we basically study the lens properties of a more or less smooth mass distribution ; in such a situation , the surface mass density must be above the critical value , in order to produce caustics . in ( quasar ) microlensing , ensembles of point lenses are studied , which always produce many ( micro-)images and , correspondingly , many caustics . an example of a two dimensional distribution @xmath93 of magnification in the _ image / sky plane _ eq(3 ) ) is shown in figure 4b ( color plate ) . it shows the magnification that an image seen at this particular position in the sky would have . the color assignment is the same as for figure 4a , except that here there are locations with black color , that means formally very high negative values of the magnification . the boundaries between regions with high positive and those with high negative magnifications ( transitions from yellow to black ) are easy to identify . these are the critical lines . this image plane magnification distribution is quite similar , but not identical , to the magnification distribution in the source plane shown in figure 4a . the differences are most obvious for highly magnified positions that correspond to multiple images . whereas in figure 4a ( magnification pattern in the _ source plane _ ) , the _ total _ magnification of all images corresponding to this source positions is displayed , in figure 4b the _ sky plane _ magnification of the different images at the corresponding positions are shown . from the distribution in figure 4b alone , however , it is not obvious , which images belong to the same source position . in figure 5 the magnification probability distribution for the underlying cosmological model is shown . the top panel shows the differential distribution , the bottom panel the integrated one . it shows the relative frequency @xmath94 of magnification @xmath95 , sampled in ten equally spaced logarithmic intervals per decade . ( in contrast to most other figures , this one shows the result of all 100 different realizations , rather than that of an example " line - of - sight only ) . in the differential distribution ( top panel ) there is a strong peak just left of magnification @xmath92 , and a power law tail ( @xmath96 ) to high magnifications ( the same power law dependence as found in tog for point lenses and in subsequent work for more realistic matter distributions ) until the probability drops to zero due to the finite resolution . the thin solid line in figure 5 indicates the magnification distribution for single image cases only , whereas the dotted line is the one for multiple images only ( the bold line is the sum of the two ) . because of the rareness of very high magnifications , the high magnification end naturally is quite noisy . the magnification distribution in figure 5 can be interpreted as the transfer function " of the matter in the universe ( i.e. , for this particular cosmological model ) . in other words : any intrinsic luminosity function of , say , quasars , will be folded with this magnification distribution , and what we see as the observed quasar luminosity function is the convolution of the intrinsic luminosity function with this transfer function " of the universe ( vietri & ostriker 1983 ) . in principle one can determine the intrinsic quasar luminosity function for a given cosmological model , once the observed luminosity function and the transfer function of this cosmological model are known ( ostriker & vietri 1986 ) . in practice this will turn out to be non - trivial , mainly because the observed luminosity function of quasars is not well known ( boyle , shanks & peterson 1988 ) . the effect of this transfer function is even more obvious when applied to a population of perfect standard candles " at high redshift : for a given source redshift , the luminosity function ( originally a delta function in such a case ) will be broadened to exactly the shape @xmath94 indicated by the curve in the top part of figure 5 . the slope of this magnification probability function for gravitational lensing is known analytically in the limit of very high magnifications ( tog ; schneider 1987 ; blandford & narayan 1986 ) and should therefore be independent of the cosmological model . however , in the intermediate range from small to moderate magnifications , different cosmological models produce different magnification probability distributions . as described in section 2 , the positions of all rays in the image plane and in the source plane are stored for each lensing run . therefore it is straightforward to determine characteristics of the mapping @xmath97 between the positions in the image plane @xmath16 and those in the source plane @xmath17 ( see sef , chapter 5 ) . of particular interest are the points in the image plane @xmath7 at which the jacobian determinant @xmath98 changes its sign ( in the previous section it was pointed out that these points can be seen in figure 4b as the locations where the color changes from yellow to black ) . by definition these are the _ critical lines_. in figure 6a the critical lines are plotted for this particular line of sight . mapping these critical lines onto the source plane results in the _ caustics_. these are shown in figure 6b . comparing figure 6a and 6b , it is not too difficult to find out which caustic belongs to which critical line . the shapes of the critical lines and of the caustics found in these lensing simulations are quite irregular , they are certainly different from the symmetric shapes that are obtained for simple models of lenses , e.g. singular isothermal sphere plus external shear or elliptic potentials ( e.g. , blandford & kochanek 1987 ; kormann , schneider & bartelmann 1994 ) . the reason is simply that the mass distributions we use are not at all symmetric , but rather representing the irregular real " distribution of matter in the universe . whenever the source position is inside such a closed caustics line , a pair of new ( usually highly magnified ) images emerges . when the source is outside , but near a caustic , one highly magnified image is resulting . in figure 7 various image configurations are shown that are produced by the caustic structure in the top right part of figure 4a/6b . they indicate the variety of image shapes and morphologies that can be produced by this particular caustic structure . it ranges from a very stretched ( and highly magnified ) single image at the top left through elongated and aligned double and triple images of similar magnification through cases with faint central ( radial ) images . for comparison , the shape and size of a unlensed source is indicated at the bottom right corner of the bottom right panel . one can look at these image configurations from two different viewpoints : if the source is extended ( e.g. , a distant galaxy ) , then figure 7 shows the highly extended images that could be observed : giant arcs , arclets , tangentially aligned images . on the other hand , if the source is an unresolved quasar , then the area of the images reflects the magnification of the two or more quasars that are seen . the separation between the centers of light " would be the distance between the multiple images . for comparison , in the lower right corner an unlensed source is shown , which represents magnification @xmath92 . in this section we show the effects of gravitational lensing due to our three dimensional ray tracing on various ensembles of sources . in the left column of figure 8 we show four different fields of sources : a regular grid of circular sources , randomly distributed circular sources , a regular grid of elliptical sources ( with fixed ellipticity and random position angle ) and randomly distributed elliptical sources with fixed ellipticity . the diameter of the circular sources shown here is about 3 arcsec . at the corresponding right - hand side the resulting image configurations are shown : this is what a telescope would see if a galaxy distribution at a redshift of @xmath99 like that at the left side were seen through the matter distribution in front of it . the size of an image reflects the magnification , since surface brightness is conserved by gravitational lensing : the area of the images , divided by the area of the corresponding source , is the magnification of this particular image . since all source sizes and shapes are identical in one panel , the relative areas of the images reflect the relative magnifications . in the panels with a regular grid of source positions , it is very obvious , from the deviations of the chains of galaxies from straight lines , where the largest matter concentrations exist along the line of sight ( cf . with the map of the matter in figure 3b ) . since all the sources have identical intrinsic sizes and shapes in each panel ( circular or constant ellipticity ) , each deviation from that shape on the right hand side is lens - induced . one can see a large variety of lens - induced image shapes : slight ellipses from circular sources , small arcs , straight arcs " , giant arcs , big blobs , and more complicated image structures . perhaps most obvious is the tendency of tangentially aligned " ( roughly ) elliptical images , whose major axes are about perpendicular to the radius vector from the center of the mass concentration . in figure 9 the lensing effect of our example line of sight is shown for three different source redshifts : the left columns from top to bottom show three sets of circular sources with random positions at source redshifts @xmath100 . the second column shows the respective image configurations that one would see in the sky , if seen through the deflecting matter between source plane and observer . in the lowest row , the three source populations are superimposed , with the different redshifts indicated by three different gray scales . in section 2 we described how one can obtain the surface mass density @xmath24 ( cf . ( 4 ) ) along the line of sight , as measured by the light rays . in figure 4c ( color plate ) we show the distribution of the effective surface mass density @xmath24 for our example field of view . the highest density clumps are easily identifiable with the highest magnification regions in figures 4a and 4b . it is also quite obvious , how the magnification @xmath93 ( figure 4b ) follows the surface mass density distribution very closely ( figure 4c ) in the low magnification regime . this indicates that it is the convergence , the matter inside the beam , that is mainly responsible for the magnification , and not so much the shear field from outside the beam . for each of our artificial lines of sight , we can obtain the distribution of the shear fields as shown in eq . ( 5 ) in section 2 . for our example line of sight , this shear distribution @xmath101 is shown for a source redshift of @xmath102 in figure 4d ( color plate ) . the length of the straight lines is proportional to the absolute value of the shear , and the direction represents the direction of the shear . it is quite obvious that the big lump of matter in the top right part produces the strongest shear , the lines clearly indicate the tangential alignment expected for galaxies that are behind such a cluster of galaxies . similar shear distributions have recently been obtained observationally by various groups who study the distribution of shear in fields around rich clusters of galaxies ( e.g. , bonnet et al . 1993 ) or in empty " fields ( mould et al . observationally , the shear fields are obtained by averaging the ellipticities of galaxies in small fields of a few tens of arcseconds . the idea is that this averaging should cancel the intrinsic distribution of position angles , and as a net effect just show the effective shear produced by the gravitational lens effect of the foreground matter . we have described a method to determine the lensing properties of a three dimensional , fully nonlinear distribution of matter , obtained from cosmological simulations . we explain the techniques we use to calculate the light deflection by a large number of lenses in many lens planes . the simulations produce maps of the magnification in the source and image plane . we can study the caustics and critical lines for each matter realization individually , or in a statistical way for a large number of realizations . here one particular line of sight " with the matter distribution from a standard cdm model with @xmath88 and @xmath89 is used , as an example , to illustrate diagnostic power of the applications . it is straightforward to apply this method to other cosmological models , and we plan to do so with significantly higher resolutions in both cosmological simulations and ray - tracing method . there are numerous possible applications of this ray shooting technique . just to mention a few , which we plan to address in the future : \1 ) study of frequency , geometry , magnification ratio , and separation of multiple images . what fraction of quasars is multiply imaged ? how many double , triple , quadruple images do we obtain ( for given observational limits of spatial resolution and dynamic range ) ? \2 ) study of the distribution of magnification of certain classes of objects ( galaxies , quasars ) ; this can be done separately for single images and multiple image cases , or for both combined , addressing question like are there many highly magnified quasars that are single ( rather than multiply imaged ) ? " . another straightforward application is a quantitative study of the dispersion effect of lensing on standard candles at moderate redshifts ( i.e. on supernovae of type 1 ) . \3 ) study of correlations between single / multiple quasars and strong ( visible " ) mass concentrations at or near the line of sight ( which may affect the issues of dark lenses and/or quasar - galaxy - associations ) \4 ) study of effects of chance alignments of dense matter clumps along the line of sight : how accurate are the mass determinations which assume that all matter reponsible for the lensing is in a single plane ? \5 ) study of the shear distribution / tangentially aligned galaxies around ( rich ) clusters and in empty " fields . what is the average or rms - shear value in random lines of sight ? \6 ) use techniques of reconstructing mass profiles of galaxy clusters from weak lensing ( kaiser & squires 1993 ) , and compare the reconstructed mass profile with the actual one . are there any systematic effects due to foreground and background matter ? \7 ) compare the frequency and properties of arcs and straight arcs " produced by extreme lensing in clusters with that seen in the real world . \8 ) study of the effects of three dimensional lensing on the cosmic microwave background . most of the points above can be investigated for sources at different redshifts . we know that the expected and computable gravitational lensing properties of the different models , that have been proposed for the growth of structure in the universe , will vary greatly from model to model . thus a quantitative comparison between observed and expected lensing properties should offer great power in selecting among the competing models , and ideally allow us to eliminate models which fail to correspond to reality . even at present the strongest constraints on the size of the cosmological constant are derived from lensing statistics ( fukugita & turner 1991 ; kochanek 1992 ; maoz & rix 1993 , kochanek 1995 ) in a recent application of the method described here , we ( wambsganss et al . 1995 ) found that the standard cdm model predicts far more large splitting multiple quasars than are observed . it will be very interesting to see if other models such as mixed dark matter , cdm@xmath103 or open models can survive this gravitational lensing test . it is a pleasure to thank michal jaroszyski , avi loeb , jordi miralda - escud , peter schneider , and ed turner for many useful discussions at various stages of this project . we also want to thank an anonymous referee for his constructive criticism of the first version of the manuscript , which significantly improved the quality of this paper . here we describe the convolution method " , which is used to compensate the lack of sufficient spatial dynamic range in the n - body simulations . as stated earlier in the main body of this paper , for gravitational lensing applications one must have enough scale - resolving power to reach scales which are interesting and accessible to observations ( @xmath104 arcsecond ) . but it is not possible at the same time , in a single simulation , to reach the other end of the spectrum , the large scale waves of the universe , to have a fair sample of the relevant cosmic systems , which are clusters of galaxies in this case . a minimal desired dynamic range will be a few times the ratio of cluster separation ( @xmath105mpc ) to cluster core size ( @xmath106kpc ) , i.e. , a few thousand , which was not achievable at the time when the simulations used in this paper were computed . as an alternative to the brute - force approach , we made two kinds of simulations : one with box size @xmath107mpc ( box - s hereafter ) and resolution of @xmath81kpc , the other with box size @xmath108mpc ( box - l hereafter ) and resolution of @xmath109mpc ( both simulations use @xmath110 cells with a pm n - body code ) . in computations of both boxes ( s , l ) we assume periodic boundary conditions . that is we assume that both boxes have the same mean density as the universe at any given redshift . for the large box that is a good approximation since the amplitudes of waves on scales larger than @xmath111mpc are quite small at @xmath5 and even smaller at the redshifts relevant for lensing . but for the small box , this assumption is poor . the nonlinear scale at @xmath5 is approximately @xmath112mpc so , in the l box we will find @xmath113mpc subboxes at a variety of over and under densities . then using this distribution we convolve box - s ( which has small scale resolution ) with box - l ( which has large scale power ) , details of which are described below . we also note that , at the time this paper is being written , we are using new n - body techniques to make simulations which can cover this needed range directly . we note that the gravitational ray - tracing method presented in the paper is independent of the convolution method described here . before a step - by - step description of the convolution method , it is useful to define a few quantities . the actual over- or under - density of a particular plane , @xmath114 , drawn from the density distribution @xmath115 on the scale of box - s ( @xmath116mpc@xmath87 ) , which is computed from box - l ( containing @xmath117 subboxes each of which is a @xmath116mpc@xmath87 cube ) . @xmath118 is the ratio of the density in the box - s to the mean cosmic density . the projection of the matter in a box - s onto one of the three orthogonal planes makes a surface mass density screen called @xmath119 . we now detail steps to do the convolution . \1 ) we first choose @xmath118 randomly from @xmath115 . \2 ) then a mosaic square region , @xmath120 , consisting of @xmath121 @xmath122 screens , where @xmath123 , is set up . the number of repetition screens , @xmath14 , in each dimension on the projection plane is determined by @xmath118 such that the shrunk @xmath124 ( when @xmath125 , see step 4 below ) still has a comoving length larger than @xmath126mpc ) to prevent situations of no defined densities from happening , when rays shoot outside the original region . \3 ) from @xmath127 we obtain a new screen called , @xmath128 equal to @xmath129 , where the power operation operates on each individual cell of screen @xmath127 . this step is to mimic the evolution of regions with different densities and the value @xmath130 is empirically determined using simulations . this step does not , to first order , change the overall density but it increases the contrast ( for @xmath125 ) . since @xmath131 is always very close to unity , this step affect the results very mildly . \4 ) next , we rescale each of the two dimensions of the screen @xmath128 by @xmath132 . this involves resampling the surface mass onto a smaller ( or larger ) grid for @xmath125 ( @xmath133 ) , increasing ( decreasing ) the surface density by @xmath134 , resulting in a new screen , @xmath135 . this step is to take account of the fact that high density regions should have been compressed to smaller comoving regions during their evolution . \5 ) the density in each of the new pixels of @xmath136 is multiplied by @xmath137 to take into account the effect that the line of sight dimension is also reduced , yielding @xmath138 . \6 ) randomly choose an origin and randomly rotate screen @xmath139 to get , by bilinear interpolation , a new screen @xmath140 with the original comoving pixel size ( @xmath81kpc ) . \7 ) finally , renormalize @xmath140 such that @xmath141 , which will be the end product of this scheme . the renormalization factor @xmath118 is typically very close to unity . in figure 10 we illustrate this convolution . the four panels correspond to the original surface mass density distribution @xmath142 ( top left ) , a part of the shrunk or expanded distribution @xmath143 ( top right ) , the full @xmath143 ( bottom left ) , and the final rotated and normalized distribution @xmath144 ( bottom right ) . in figure 10a this is shown for a factor @xmath145 ( over dense ) , in figure 10b for @xmath146 ( under dense ) this convolution method is tested against a high resolution p@xmath87 m simulation ( bertschinger 1995 ) where the boxsize is @xmath80mpc and therefore it contains large scale power . of cource , this p@xmath87 m simulation does not reach the resolution of @xmath81kpc on the small scale end , so we can only make comparisons in the overlapping region . figure 11 shows the surface density distributions on a cell of area @xmath147kpc@xmath43 with depth of @xmath80mpc using our convolution method ( solid histogram ) and from direct calculation using bertschinger s simulation ( dotted histogram ) . the agreement is satisfactory . = 0.0pt babul , a. & lee , m.h . 1991 , mnras 250,407 bartelmann , m. & schneider , p. 1991 , a&a 248,353 . bertschinger , e. 1995 , private communication blandford , r.d . & kochanek , c.s . 1987 apj 321 , 658 blandford , r.d . , narayan , r. 1986 , apj 310 , 568 bonnet , h. , fort , b. , kneib , j .- p . , mellier , y. , soucail , g. , 1993 , a & a 280 , l7 boyle , b.j . , shanks , t , & peterson , b.a . 1988 , _ asp conf . : `` the evolution of optically selected qsos '' _ * 2 * , 1 cen r. 1992 , apjs , 78 , 341 cen , r. , gott iii , j.r . , ostriker , j.p . , turner , e.l . 1994 , apj 423 , 1 ( cgot ) fukugita , m. & turner , e.l . 1991 , mnras 253,99 gorski et al . 1994 , apj 430 , 89 hinshaw , g. , krauss , l.m . 1987 , apj 320 , 468 jaroszyski , m. 1989 , acta astr . , 39 , 301 jaroszyski , m. 1991 , mnras 249,430 jaroszyski , m. , park , c. , paczynski , b. & gott , j.r . 1990 , apj 365 , 22 kaiser , n. , & squires , g. 1993 , apj 404 , 441 kassiola , a. , kovner , i. 1993 , apj 417 , 450 katz , n. , hernquist , l. , & weinberg , d.h 1993 , apj , 399 , l109 kochanek , c.s . 1992 apj 384 , 1 kochanek , c.s . 1995 apj 453 , 545 kochanek , c.s . & apostolakis , j. 1988 mnras 235 , 1073 kochanek , c.s . & blandford , r.d . 1987 , apj 321 , 676 kormann , r. , schneider , p. , bartelmann , m. 1994 , a & a 284 , 285 kovner , i. 1987a , apj 321 , 686 kovner , i. 1987b , nature 325 , 507 lee , m.h . & paczyski , b. 1990 , apj 357,32 maoz , d. & rix h .- w . 1993 apj 416 , 425 mould , j. , , blandford , r. , villumsen , j. , brainerd , t. , smail , i . 1994 , apj 271 , 31 navarro , j .. f , frenk , c. , and white , s.d.m . 1994 , mnras , 267 , l1 narayan , r. , & white , s.d.m . 1987 , mnras , 231 , 97p ostriker , j.p . & vietri , m. 1986 , apj 300 , 68 park , c.b . , & gott , j.r . , iii 1991 , mnras , 249 , 288 rauch , k.p . 1991 , apj 374 , 83 schneider , p. 1987 , a & a 183 , 189 schneider , p. & weiss , a. 1988a , apj 327,526 schneider , p. & weiss , a. 1988b , apj 330,1 schneider , p. , ehlers , j. , & falco , e.e . , _ gravitational lensing _ ( springer verlag , berlin , 1992 ) ( sef ) smoot , g.f . , et al . 1992 , apj(letters ) , 396 , l1 steinmetz , m , and mller , e. 1994 , a & a 281 , 97 summers , f. , davis , m. , and evrard , a. 1995 , apj , 454 , 1 turner , e.l . , ostriker , j.p . , and gott , j.r . iii 1984 apj 284 , 1 ( tog ) vietri , m. & ostriker , j.p . 1983 , apj 267 , 488 wallington , s. , & narayan , r. 1993 , apj 403 , 517 ( wn93 ) wambsganss , j. 1990 ph.d . thesis , munich university ; available as mpa report # 550 wambsganss , j. , paczynski , b. , & schneider , p. 1990 apj 358 , l33 wambsganss , j. , cen , r. , ostriker , j. p. , & turner , e. l. 1995 , science 268 , 274 schematic depiction of the bundle of light rays with fixed opening angle traversing many planes with constant comoving ( but increasing physical ) side length . for simplicity we show only three lens planes , in fact we use of order 100 individual planes . the observer is at the left hand side , the source plane is at the right hand side . redshifts of the lens planes are increasing from left to right , ( i.e. , @xmath148 ) . the bundle of light rays illuminates " a small fraction of the lens planes at low redshift . this fraction increases with redshift to almost one at the highest lens redshift . displayed is the effect of increasing the accuracy of the hierarchical tree code for the determination of the deflection angle by using multipoles of higher order . the top ( bottom ) row deals with the x-(y-)component of the deflection angle . the difference between direct determination and tree - code determination is shown versus the direct determination ( in arbitrary units ) . in the leftmost column the tree code uses just the monopole term , i.e. it assumes all the matter is at the location of the center of mass of each cell . the accuracy of the angle determination in this case is about 1 % . in the next columns the same is shown for a tree code including quadrupole moment ( column 2 ) and tree code with all moments up to order 4 ( column 3 ) . in the right most column all multipole moments up to order 6 are included in the determination of the deflection angle . the maximum deviation in the last panel is clearly better than 0.1 % . all our lensing calculations are done with this accuracy . the matter distribution in physical units ( grams per cm@xmath43 ) inside the beam of rays is shown . dark color indicates high values of the surface mass density . figure 3a : 60 panels contain the matter distribution of individually treated lens planes . the first " plane is top left ( lowest redshift ) , with redshift increasing to the right . the highest redshift plane is the last one in the bottom row at @xmath54 . in figure 3b this matter integrated along the line of sight up to a redshift z = 3 is displayed . magnification pattern of one particular line of sight " through the universe ( the same as portrayed in figure 3 ) . this is a two dimensional map of the magnification as a function of position in the _ source _ plane at @xmath85 . the color indicates the _ total _ magnification of all images for a particular source position : it increases from blue through green and red to yellow . the transition between green and red marks roughly the average magnification . for the strong lens in the top right part the caustics are identifiable as sharp yellow lines , indicating very high magnification . the sidelength of this and all the other maps is 343 arcsec , comparable to the hst wfpc side length of @xmath149 arcsec . magnification map in the _ image / sky _ plane . here the magnification of each image appearing at a certain position in the sky is indicated . the low and intermediate magnification regions are very similar to those in the map shown in figure 4a . at the high magnification end there are obvious differences : here each image of a multiply imaged source would appear at a different position . note the transitions from bright yellow to black : here the magnification changes from very high positive values to ( formally ) very high negative values . that means this boundary marks the critical lines " which separate multiple images with positive and negative parity . map of effective surface mass density @xmath24 or convergence along the line of sight . this distribution is quite similar to the matter integrated along the line of sight , as shown in figure 3b . however , here the matter is weighted " by the angular diameter distances . shear map for a source plane at @xmath99 , due to the foreground mass distribution in the field of view . the length of the lines indicates the relative strength of the shear , the angles of the lines show the direction of the shear . the regions of strong shear are easily recognized as those with the highest mass concentrations ( cf . figures 3b , figures 4a / b / c ) . magnification distribution from 100 different realizations ( i.e. , about @xmath150 different source positions at a redshift of @xmath151 : bold line : all source positions ; thin solid line : singly imaged source positions only ; dashed line : multiply imaged source positions only . the top panel displays the differential distribution f(@xmath152 , the bottom panel the integrated one f(@xmath153 . the typical ( unlensed ) source becomes slightly fainter , but there is a powerlaw tail of highly magnified but rare sources . critical lines in the image plane ( figure 6a ) and caustics in the source plane ( figure 6b ) for our example line of sight are displayed , for a source redshift of @xmath151 . the critical lines are the locations at which the determinant of the jacobian matrix disappears , the caustics are these points , mapped onto the source plane ( cf . with figures 4b and 4a , respectively ) . various image configurations with large magnifications , showing the range of possible geometrical arrangements of images . in the bottom right part of the bottom right panel the size and shape of the unlensed source is indicated ( the relative scale is in arcsec ) . various distributions of sources at redshift @xmath102 and corresponding images as seen through the three dimensional matter distribution , i.e. weakly and strongly lensed sources . the source configurations are always shown left , the image configurations right . from top to bottom the source ensembles are : circular sources in a regular grid and randomly distributed , and elliptical sources ( random orientation ) in a regular grid and randomly placed . the source sizes are about 3 arcsec . the source density is equivalent to about 40 000 galaxies per square degree . similar to figure 8 , but now source planes ( left ) and image planes ( right ) are shown for three different source redshifts . starting at the top : @xmath99 , @xmath154 , @xmath85 . in the bottom row the superposition of the three source redshifts is shown . the panels in this figure illustrate the convolution technique as described in the appendix . in figure 10a the top left panel represents the matter distribution @xmath142 in one of the @xmath155 cubes with average density at redshift @xmath156 . here , this particular plane should represent a randomly determined overdensity of 1.6 ( a larger than typical case chosen for purpose of illustration ) ; therefore , the matter is contracted in x- and y - direction each by @xmath157 , and multiplied by @xmath157 ( corresponding to a compression in z - direction ) . the resulting matter distribution is displayed in the top right panel ( part of @xmath143 , in the notation of the appendix ) . now , the matter distribution is extended in the empty " parts of the square , taking advantage of the periodic boundary conditions ( lower left panel , full @xmath143 ) . this distribution , now , is shifted in x- and y - direction and rotated by arbitrary amounts , so that finally the lower right panel is obtained ( @xmath144 ) . the apparent double appearance of the two dense matter clumps is of no worry , since the bundle of light rays with fixed angular size cuts out only a small square ( as indicated in the final panel ) , so that no repeated structures appear in the illuminated field " . in figure 10b the same is shown for a ( under- ) density of @xmath158 . comparison of surface mass density distributions in @xmath159kpc@xmath160kpc@xmath161mpc cylinders using the convolution technique ( see appendix a ; solid histogram ) and a direct simulation ( bertschinger 1996 ; dotted histogram ) . it indicates that the convolution method gives results which are in satisfactory agreement with direct simulations , where comparisons can be made .

let be the teichmller space of an oriented surface of finite type . we discuss the action of subgroups of the mapping class group of on the-boundary of the completion of with respect to the weil - petersson metric . we show that the set of invariant borel probability measures for the weil - petersson flow on moduli space which are supported on a closed orbit is dense in the space of all ergodic invariant probability measures . partially supported by sonderforschungsbereich 611 .

0901.4301

for an oriented surface @xmath1 of genus @xmath3 with @xmath4 punctures and _ complexity _ @xmath5 let @xmath0 be the _ teichmller space _ of all isotopy classes of complete hyperbolic metrics on @xmath1 of finite volume . then @xmath0 is a contractible manifold which can be equipped with the _ weil - petersson metric _ , an incomplete khler metric of negative sectional curvature . in spite of the lack of completeness , any two points in @xmath0 can be connected by a unique weil - petersson geodesic which depends smoothly on its endpoints @xcite . as a consequence , @xmath0 can be completed to a _ hadamard space _ @xmath6 , i.e. a complete simply connected @xmath2-space which however is not locally compact . a hadamard space @xmath7 admits a _ visual boundary _ @xmath8 , and the action of the isometry group of @xmath7 extends to an action on @xmath8 . for surfaces @xmath1 of complexity at most three , the visual boundary @xmath9 of @xmath6 was identified by brock and masur @xcite , but for higher complexity it is not known . however , it follows from the work of brock @xcite that the boundary is not locally compact . the _ extended mapping class group _ @xmath10 of all isotopy classes of diffeomorphisms of @xmath1 acts on @xmath0 as a group of isometries . since every isometry of @xmath0 extends to an isometry of the completion @xmath6 , the extended mapping class group also acts isometrically on @xmath6 . this fact was used by masur and wolf @xcite to show that in fact every isometry of @xmath6 is an extended mapping class . an isometry @xmath11 of @xmath6 is called _ axial _ if @xmath11 admits an axis , i.e. if there is a geodesic @xmath12 and a number @xmath13 such that @xmath14 for all @xmath15 . the endpoints @xmath16 are then fixed points for the action of @xmath11 on @xmath9 . every _ pseudo - anosov _ mapping class @xmath17 is axial @xcite . limit set _ @xmath18 of a subgroup @xmath19 of @xmath10 is the set of accumulation points in @xmath20 of an orbit of the action of @xmath19 on @xmath6 . the group @xmath19 is called _ non - elementary _ if its limit set contains at least three points . we show . [ thm1 ] let @xmath21 be a non - elementary subgroup with limit set @xmath18 which contains a pseudo - anosov element . 1 . @xmath18 does not have isolated points , and the @xmath19-action on @xmath18 is minimal . pairs of fixed points of pseudo - anosov elements are dense in @xmath22 . 3 . there is a dense orbit for the action of @xmath19 on @xmath22 . there is a compactification of @xmath0 whose boundary consists of the sphere @xmath23 of projective measured geodesic laminations on @xmath1 . the mapping class group acts on @xmath24 as a group of homeomorphisms . limit sets for subgroups of @xmath10 in @xmath23 were investigated by mccarthy and papadopoulos @xcite . we also look at properties of the _ weil - petersson geodesic flow _ @xmath25 on the quotient of the unit tangent bundle @xmath26 of @xmath0 under the action of the mapping class group . even though this quotient space @xmath27 is non - compact and this flow is not everywhere defined , it admits many invariant borel probability measures . particular such measures are measures supported on periodic orbits . each of these measures is ergodic . the space of all @xmath25-invariant borel probability measures on @xmath27 can be equipped with the weak@xmath28-topology . our second result is a version of theorem 1 for the weil - petersson geodesic flow . a @xmath25-invariant borel probability measure on @xmath27 can be approximated in the weak@xmath28-topology by measures supported on periodic orbits . the organization of this note is as follows . in section 2 we review some geometric properties of hadamard spaces . section 3 explains some geometric properties of pseudo - anosov mapping classes . in section 4 we look at groups of isometries and establish the first and the second part of theorem 1 . in section 5 , we complete the proof of theorem 1 and show theorem 2 . the purpose of this section is to collect some general geometric properties of @xmath2-spaces which are needed for the investigation of weil - petersson space . a @xmath2-space is defined as follows . a _ triangle _ @xmath29 in a geodesic metric space consists of three vertices connected by three ( minimal ) geodesic arcs @xmath30 . a _ comparison triangle _ @xmath31 for @xmath29 in the euclidean plane is a triangle in @xmath32 with the same side - lengths as @xmath29 . by the triangle inequality , such a comparison triangle exists always , and it is unique up to isometry . for a point @xmath33 the comparison point of @xmath34 in the comparison triangle @xmath35 is the point on the side @xmath36 of @xmath35 corresponding to @xmath37 whose distance to the endpoints of @xmath38 coincides with the distance of @xmath34 to the corresponding endpoints of @xmath37 . a geodesic metric space @xmath39 is called a _ @xmath2-space _ if for every geodesic triangle @xmath29 in @xmath7 with sides @xmath30 and every comparison triangle @xmath35 in the euclidean plane with sides @xmath40 and for all @xmath41 and all comparison points @xmath42 we have @xmath43 a complete @xmath2-space is called a _ hadamard space_. in a hadamard space @xmath7 , the distance function is convex : if @xmath44 are two geodesics in @xmath7 parametrized on the same interval then the function @xmath45 is convex . for two geodesics @xmath44 issuing from the same point @xmath46 , the _ alexandrov angle _ between @xmath44 is defined . if @xmath7 is a riemannian manifold of non - positive curvature , then this angle coincides with the angle between the tangents of @xmath44 at @xmath47 ( see @xcite ) . for a fixed point @xmath48 , the _ visual boundary _ @xmath8 of @xmath7 is defined to be the space of all geodesic rays issuing from @xmath34 equipped with the topology of uniform convergence on bounded sets . this definition is independent of the choice of @xmath34 . we denote the point in @xmath8 defined by a geodesic ray @xmath49 by @xmath50 . we also say that @xmath51 _ connects _ @xmath34 to @xmath50 . the union @xmath52 has a natural topology which restricts to the usual topology on @xmath7 and such that @xmath7 is dense in @xmath52 . the isometry group of @xmath7 acts as a group of homeomorphisms on @xmath52 . a subset @xmath53 is _ convex _ if for @xmath54 the geodesic connecting @xmath34 to @xmath55 is contained in @xmath56 as well . for every complete convex set @xmath53 and every @xmath48 there is a unique point @xmath57 of smallest distance to @xmath34 ( proposition ii.2.4 of @xcite ) . now let @xmath58 be a closed connected set and let @xmath59 be a geodesic arc . then @xmath60 is complete and convex and hence there is a shortest distance projection @xmath61 . the projection @xmath61 is distance non - increasing . the following definition is due to bestvina and fujiwara ( definition 3.1 of @xcite ) . [ contracting ] a geodesic arc @xmath59 is _ @xmath62-contracting _ for some @xmath63 if for every closed metric ball @xmath64 in @xmath7 which is disjoint from @xmath65 the diameter of the projection @xmath66 does not exceed @xmath62 . we call a geodesic _ contracting _ if it is @xmath62-contracting for some @xmath63 . as an example , every geodesic in a @xmath67-space for some @xmath68 is @xmath62-contracting for a number @xmath69 only depending on @xmath70 . the next lemma ( lemma 3.2 and 3.5 of @xcite ) shows that a triangle containing a @xmath62-contracting geodesic as one of its sides is uniformly thin . [ thintriangle ] let @xmath71\to x$ ] be a @xmath62-contracting geodesic . if @xmath48 and if @xmath72}(x)$ ] then for every @xmath73 $ ] the geodesic connecting @xmath34 to @xmath74 passes through the @xmath75-neighborhood of @xmath76 . on the other hand , thinness of triangles with a fixed geodesic @xmath51 as one of the three sides guarantees that @xmath51 is contracting this is formulated in the following useful criterion to detect contracting geodesics . for its formuation , note that in a @xmath2-space , the _ alexandrov angle _ between two geodesics issuing from the same point is defined ( see @xcite ) . [ quadrangle ] let @xmath59 be a geodesic such that there is a number @xmath63 with the following property . assume that for all @xmath77\subset j$ ] with @xmath78 and every geodesic quadrangle @xmath79 in @xmath7 with one side @xmath80 $ ] and an angle at least @xmath81 at @xmath82 the geodesic arc connecting the two vertices of @xmath79 which are distinct from @xmath82 passes through the @xmath83-neighborhood of @xmath80 $ ] . then @xmath65 is @xmath62-contracting . let @xmath59 be a geodesic which satisfies the assumption in the lemma . we have to show that @xmath84 for all @xmath48 with @xmath85 and every @xmath86 with @xmath87 . for this let @xmath48 with @xmath88 . assume that @xmath89 . let @xmath86 and write @xmath90 where @xmath91 without loss of generality ( otherwise reverse the orientation of @xmath51 ) . the angle at @xmath92 between the geodesic connecting @xmath92 to @xmath55 and the subarc of @xmath51 ( with reversed orientation ) connecting @xmath92 to @xmath76 is not smaller than @xmath81 . since the angle sum of a triangle in a @xmath2-space does not exceed @xmath93 , this implies that if @xmath94 then the angle at @xmath95 of the quadrangle with vertices @xmath96 is not smaller than @xmath81 . thus by the assumption in the lemma , the geodesic connecting @xmath34 to @xmath55 passes through a point @xmath97 in the @xmath83-neighborhood of @xmath98 $ ] . then @xmath99 $ ] , moreover also @xmath100 . now assume that @xmath101 . then we have @xmath102 . since the projection @xmath103 is distance non - increasing we conclude that @xmath104 in other words , @xmath65 is @xmath62-contracting . let @xmath1 be an oriented surface of genus @xmath3 with @xmath4 punctures and @xmath5 . the metric completion @xmath6 of the teichmller space @xmath0 of @xmath1 with respect to the weil - petersson metric @xmath105 is a hadamard space . the completion locus @xmath106 of @xmath0 can be described as follows @xcite . a _ surface with nodes _ is defined by a degenerate hyperbolic metric on @xmath1 where at least one essential simple closed curve on @xmath1 ( i.e. a curve which is homotopically nontrivial and not freely homotopic into a puncture ) has been pinched to a pair of punctures . for the free homotopy class of an essential simple closed curve @xmath107 on @xmath1 , the degenerate surfaces with a single node at @xmath107 define a _ @xmath108 in the completion locus @xmath109 of teichmller space . this stratum equipped with the induced metric is isometric to the teichmller space equipped with the weil - petersson metric of the ( possibly disconnected ) surface obtained from @xmath110 by replacing each of the two ends corresponding to @xmath107 by a cusp . if @xmath111 is disconnected then a point in the teichmller space of @xmath110 is given by a pair of points , one for each of the two components of @xmath110 . the stratum @xmath108 is a convex subset of @xmath6 . the completion locus @xmath106 then is the union of the completions of the strata @xmath108 where @xmath107 runs through all free homotopy classes of simple closed curves and with the obvious identifications . the extended mapping class group @xmath10 of all isotopy classes of diffeomorphisms of @xmath1 acts on @xmath112 properly discontinuously as a group of isometries . this action extends to an action on the completion @xmath6 preserving the completion locus @xmath106 . masur and wolf @xcite used the action of the extended mapping class group on the completion locus of teichmller space to show . [ isometrygroup ] the isometry group of @xmath112 coincides with the extended mapping class group . for @xmath113 let @xmath114 be the subset of @xmath0 of all hyperbolic metrics whose _ systole _ , i.e. the length of a shortest closed geodesic , is at least @xmath115 . the mapping class group preserves @xmath114 and acts on it properly discontinuously and cocompactly . in particular , the weil - petersson distance between @xmath114 and the completion locus @xmath106 of teichmller space is positive . moreover , the sectional curvature of the restriction of the weil - petersson metric to @xmath114 is bounded from above by a negative constant . this fact together with lemma [ quadrangle ] is used to show that weil - petersson geodesic segments which are entirely contained in the thick part of teichmller space are contracting . an analogous result for the _ teichmller metric _ on teichmller space ( which is however much more difficult ) was established by minsky @xcite . as a convention , in the sequel a weil - petersson geodesic is always parametrized on a closed connected subset of @xmath116 . [ wpcontr ] for every @xmath113 there is a number @xmath117 such that every geodesic @xmath118 is @xmath62-contracting . it was shown in lemma 3.1 of @xcite ( see also @xcite for an earlier argument along the same line ) that for every @xmath113 there is a constant @xmath117 only depending on @xmath115 such that every geodesic @xmath118 satisfies the hypothesis in lemma [ quadrangle ] for @xmath62 . for an isometry @xmath11 of @xmath0 define the _ displacement function _ @xmath119 of @xmath11 to be the function @xmath120 . [ semisimple ] an isometry @xmath11 of @xmath6 is called _ semisimple _ if @xmath119 achieves its minimum in @xmath6 . if @xmath11 is semisimple and @xmath121 then @xmath11 is called _ elliptic_. a semisimple isometry @xmath11 with @xmath122 is called _ axial_. by the above definition , an isometry is elliptic if and only if it fixes at least one point in @xmath6 . by proposition 3.3 of @xcite , an isometry @xmath11 of @xmath6 is axial if and only if there is a geodesic @xmath12 such that @xmath14 for every @xmath123 where @xmath124 . such a geodesic is called an _ oriented axis _ for @xmath11 . note that the geodesic @xmath125 is an oriented axis for @xmath126 . the endpoint @xmath50 of @xmath51 is a fixed point for the action of @xmath11 on @xmath9 which is called the _ attracting fixed point_. the closed convex set @xmath127 of all points for which the displacement function of @xmath11 is minimal is isometric to @xmath128 where @xmath129 is closed and convex ( theorem ii.2.14 of @xcite ) . for each @xmath130 the set @xmath131 is an axis of @xmath11 . by the nielsen - thurston classification , a mapping class @xmath17 either is _ pseudo - anosov _ or it is of finite order or it is _ reducible_. an example of a reducible mapping class is a _ multi - twist _ which can be represented in the form @xmath132 where each @xmath133 is a dehn - twist about a simple closed curve @xmath134 in @xmath1 and where the curves @xmath134 are pairwise disjoint . we allow the multi - twist to be trivial . we have . [ semisimple2 ] every isometry @xmath135 of @xmath6 is semi - simple , and @xmath135 is elliptic if and only if there is some @xmath136 such that @xmath137 is a multi - twist . in a hadamard space @xmath7 , an isometry @xmath11 with a finite orbit on @xmath7 has a fixed point which is the center of the orbit . this means the following . for a fixed orbit @xmath138 for @xmath11 there is a unique point @xmath86 such that the radius of the smallest closed metric ball centered at @xmath55 which contains the set @xmath139 is minimal ( proposition ii.2.7 of @xcite ) . since this point is defined by purely metric properties , it is a fixed point for @xmath11 . as a consequence , an element @xmath17 is elliptic if and only if this holds true for @xmath140 for every @xmath141 , and every element of finite order is elliptic . now assume that @xmath17 is a multi - twist about a multi - curve @xmath142 . let @xmath143 be the completion of the stratum in @xmath6 of all surfaces with nodes at the curves @xmath144 . then @xmath11 fixes each point in @xmath143 and hence @xmath11 is elliptic . if @xmath11 is pseudo - anosov then it was shown in @xcite that @xmath11 has an axis in @xmath0 and hence it is axial . now assume that @xmath11 is reducible . then up to replacing @xmath11 by @xmath140 for some @xmath141 , @xmath11 preserves a non - trivial multi - curve @xmath107 component - wise , and it preserves each connected component of @xmath110 . moreover , the multi - curve @xmath107 can be chosen in such a way that for every component @xmath145 of @xmath110 , either @xmath145 is a three - holed sphere or the restriction of @xmath11 to @xmath145 is pseudo - anosov . if @xmath11 is not a multi - twist then there is at least one component @xmath145 of @xmath110 such that the restriction of @xmath11 to @xmath145 is pseudo - anosov . then the restriction of @xmath11 to @xmath145 viewed as an element of the mapping class group of @xmath145 has an axis in @xmath146 . the weil - petersson metric on the stratum @xmath108 induced from the weil - petersson metric on @xmath1 is the product of the weil - petersson metrics on the teichmller spaces of the connected components of @xmath110 . this implies that the restriction of @xmath11 to @xmath108 has an axis . now the completion @xmath147 of @xmath108 is a closed convex subset of @xmath6 . the shortest distance projection @xmath148 is distance non - increasing and equivariant with respect to the action of @xmath11 . therefore the infimum of the displacement function @xmath119 of @xmath11 equals the infimum of @xmath119 on @xmath143 . thus this infimum is a minimum and once again , @xmath11 is axial . * remark : * 1 ) the flat strip theorem ( theorem ii.2.14 of @xcite ) states that two geodesic lines in a hadamard space @xmath7 whose endpoints in the boundary @xmath8 coincide bound a flat strip . since the sectional curvature of the weil - petersson metric is negative , the proof of lemma [ semisimple2 ] shows that an axial isometry of @xmath6 which admits an axis intersecting @xmath0 is pseudo - anosov . \2 ) the celebrated solution of the nielsen realization problem states that each finite subgroup of @xmath10 has a fixed point in @xmath0 @xcite . the discussion in the proof of lemma [ semisimple2 ] immediately implies that such a group has a fixed point in @xmath6 . it is not difficult to establish that there is also a fixed point in @xmath0 , however we omit this discussion here . the following definition is due to bestvina and fujiwara ( definition 5.1 of @xcite ) . [ rankone ] an isometry @xmath11 of a @xmath2-space @xmath7 is called _ @xmath62-rank - one _ for some @xmath63 if @xmath11 is axial and admits a @xmath62-contracting axis . we call an isometry @xmath11 _ rank - one _ if @xmath11 is @xmath62-rank - one for some @xmath63 . since a pseudo - anosov element has an axis @xmath51 in @xmath0 , by invariance and cocompactness of the action of @xmath11 on @xmath51 , the geodesic @xmath51 entirely remains in @xmath114 for some @xmath113 . thus the following result ( proposition 8.1 of @xcite ) is an immediate consequence of lemma [ wpcontr ] . [ flatrank ] a pseudo - anosov element in @xmath10 is rank - one . * example : * an axial isometry @xmath11 of @xmath6 which admits an axis @xmath51 bounding a flat half - plane is not rank - one . an example of such an axial isometry of @xmath6 can be obtained as follows . let @xmath107 be a simple closed separating curve on @xmath1 such that none of the two components @xmath149 of @xmath110 is a three - holed sphere . let @xmath150 be a pseudo - anosov mapping class . then @xmath11 defines a reducible element in @xmath10 . if @xmath151 is the axis for the action of @xmath11 on @xmath152 then for each point @xmath153 the curve @xmath154 is an axis for the action of @xmath11 on the stratum @xmath155 and hence @xmath154 is an axis for the action of @xmath11 on @xmath6 . in particular , for every infinite geodesic @xmath156 the set @xmath157 is an isometrically embedded euclidean plane in @xmath6 containing an axis for @xmath11 . thus @xmath11 is axial but not rank - one . a homeomorphism @xmath11 of a topological space @xmath64 is said to act with _ north - south dynamics _ if there are two fixed points @xmath158 for the action of @xmath11 such that for every neighborhood @xmath159 of @xmath37 , @xmath160 of @xmath161 there is some @xmath141 such that @xmath162 and @xmath163 . the point @xmath37 is called the _ attracting fixed point _ for @xmath11 , and @xmath161 is the _ repelling fixed point_. teichmller space equipped with the _ teichmller metric _ can be compactified by adding the _ thurston boundary _ @xmath23 of projective measured geodesic laminations which is a topological sphere . this compactification however is different from the @xmath2-boundary @xmath9 of @xmath6 . the action of the extended mapping class group on @xmath0 naturally extends to an action on @xmath23 . an element @xmath17 acts on @xmath23 with north - south - dynamics if and only if @xmath11 is pseudo - anosov . lemma 3.3.3 of @xcite shows that a rank - one isometry of a proper hadamard space @xmath7 acts on the boundary @xmath8 with north - south dynamics . the proof of this fact given in @xcite ( proof of lemma 4.4 ) does not use the assumption of properness of @xmath7 . thus we obtain . [ northsouth ] a rank - one isometry @xmath11 of @xmath6 acts with north - south dynamics on @xmath9 . in this section we investigate the action on @xmath9 of non - elementary subgroups of @xmath10 which contain a pseudo - anosov element . we begin with recalling some standard terminology used for groups of isometries on hadamard spaces . let @xmath21 be any subgroup . the _ limit set _ @xmath18 of @xmath19 is the set of accumulation points in @xmath20 of one ( and hence every ) orbit of the action of @xmath19 on @xmath6 . if @xmath164 is axial with axis @xmath51 , then @xmath165 . in other words , the two fixed points for the action of a pseudo - anosov element on @xmath20 are contained in @xmath18 . [ limitsetmod ] the limit set of @xmath10 is the entire boundary @xmath20 of @xmath6 . for suficiently small @xmath113 , the set @xmath166 of all hyperbolic metrics whose systole is at least @xmath115 is connected , and the mapping class group @xmath10 acts cocompactly on @xmath114 . there is a number @xmath167 such that the weil - petersson distance between any point in @xmath6 and @xmath114 is at most @xmath168 @xcite . thus there is a number @xmath169 such that for all @xmath170 and all @xmath171 there is some @xmath17 with @xmath172 . this just means that @xmath20 is the limit set of @xmath10 . [ minimal ] let @xmath21 be a subgroup which contains a pseudo - anosov element @xmath11 . then the limit set @xmath18 of @xmath19 is the closure in @xmath20 of the set of fixed points of conjugates of @xmath11 in @xmath19 . if @xmath19 is non - elementary then @xmath18 does not have isolated points . let @xmath21 be a subgroup which contains a pseudo - anosov element @xmath164 . let @xmath18 be the limit set of @xmath19 . we claim that @xmath18 is contained in the closure of the @xmath19-orbit of the two fixed points of @xmath11 . for this let @xmath51 be the axis of @xmath11 . by proposition [ flatrank ] , @xmath51 is @xmath62-contracting for some @xmath63 . let @xmath173 and let @xmath174 be a sequence such that @xmath175 converges to @xmath176 . there are two cases possible . in the first case , up to passing to a subsequence , the geodesics @xmath177 eventually leave every bounded set . let @xmath178 and for @xmath179 let @xmath180 . then @xmath181 @xmath182 . on the other hand , @xmath177 is @xmath62-contracting and hence by lemma [ thintriangle ] a geodesic @xmath183 connecting @xmath184 to @xmath185 passes through the @xmath75-neighborhood of @xmath186 , and the same is true for a geodesic @xmath187 connecting @xmath184 to @xmath188 . by @xmath2-comparison , the angles at @xmath184 between the geodesics @xmath189 converge to zero as @xmath190 . since @xmath191 , the sequence @xmath192 converges to @xmath176 as well . but @xmath188 is a fixed point of the conjugate @xmath193 of @xmath11 . thus @xmath176 is contained in the closure of the fixed points of all conjugates of @xmath11 . in the second case there is a bounded neighborhood @xmath64 of @xmath184 in @xmath6 such that @xmath194 for all @xmath195 . for @xmath196 let @xmath183 be the geodesic connecting @xmath184 to @xmath185 . since @xmath191 , the geodesics @xmath183 converge as @xmath190 locally uniformly to the geodesic ray connecting @xmath184 to @xmath176 . after passing to a subsequence and perhaps a change of orientation of @xmath51 we may assume that for large @xmath195 the point @xmath185 lies between a point @xmath197 and @xmath188 . this means that @xmath185 is contained in the geodesic connecting @xmath198 to @xmath188 . since the distance between @xmath198 and @xmath184 is uniformly bounded , by @xmath2-comparison the alexandrov angle at @xmath185 between the inverse of the geodesic @xmath183 ( which connects @xmath185 to @xmath184 ) and the inverse of the geodesic @xmath177 ( which connects @xmath185 to @xmath198 ) tends to zero as @xmath190 . this implies that the angle at @xmath185 of the ideal triangle in @xmath6 with vertices @xmath199 tends to @xmath93 as @xmath190 . since in a @xmath2-space the sum of the alexandrov angles of a geodesic triangle ( with possibly one vertex at infinity ) does not exceed @xmath93 , the angle at @xmath184 between the geodesic @xmath183 and the geodesic @xmath200 connecting @xmath184 to @xmath188 tends to zero as @xmath190 . but @xmath191 and therefore the points @xmath188 converge to @xmath176 @xmath182 in @xmath9 . thus @xmath176 is indeed contained in the closure of the fixed points of conjugates of @xmath11 . now assume that the limit set @xmath18 of @xmath19 contains at least 3 points . let @xmath11 be any pseudo - anosov element of @xmath19 . since by lemma [ northsouth ] @xmath11 acts with north - south dynamics on @xmath20 , the set @xmath18 contains at least one point @xmath176 which is not a fixed point of @xmath11 . the sequence @xmath201 consists of pairwise distinct points which converge as @xmath202 to the attracting fixed point of @xmath11 . similarly , the sequence @xmath203 consists of pairwise distinct points which converge as @xmath202 to the repelling fixed point of @xmath11 . moreover , by the above , a point @xmath173 which is not a fixed point of a pseudo - anosov element of @xmath19 is a limit of fixed points of pseudo - anosov elements . this shows that @xmath18 does not have isolated points and completes the proof of the lemma . we need the following simple ( and well known to the experts ) observation which parallels the properties of the action of @xmath10 on the space of projective measured geodesic laminations . this observation follows immediately from the work of brock , masur and minsky @xcite . [ pseudofix ] let @xmath204 be pseudo - anosov elements . if there is a common fixed point for the action of @xmath205 on @xmath9 then the fixed point sets of @xmath205 coincide . let @xmath204 be pseudo - anosov elements and assume that there is a common fixed point for the action of @xmath205 on @xmath20 . since @xmath205 act with north - south dynamics on @xmath9 , this implies that the axis @xmath51 for @xmath11 and the axis @xmath206 for @xmath207 have a common endpoint , say @xmath208 . we may also assume that @xmath50 is the attracting fixed point for both @xmath205 . by theorem 1.5 of @xcite , up to a reparametrization we have @xmath209 @xmath210 . after another reparametrization , there is a number @xmath211 such that the semi - group @xmath212 acts cocompactly on the closed @xmath213-neighborhood @xmath214 of @xmath215 , and that the @xmath216-neighborhood of @xmath215 contains @xmath217 . let @xmath218 be the translation length of @xmath11 on @xmath51 . if there are no integers @xmath219 such that @xmath220 then there are infinitely many distinct elements of @xmath10 of the form @xmath221 which map @xmath222 into the @xmath216-neighborhood of @xmath223 $ ] . namely , let @xmath224 be arbitrary . then there is a unique number @xmath225 such that @xmath226 . since the point @xmath227 is contained in the @xmath216-neighborhood of @xmath228 , the point @xmath229 is contained in the @xmath216-neighborhood of @xmath223 $ ] . however , this violates the fact that @xmath10 acts properly discontinuously on @xmath0 . thus there are number @xmath230 with @xmath220 and hence the fixed point sets for the action of @xmath205 on @xmath9 coincide . the action of a group @xmath19 on a topological space @xmath231 is called _ minimal _ if every @xmath19-orbit is dense . [ moveaway ] let @xmath21 be a non - elementary group with limit set @xmath18 which contains a pseudo - anosov element @xmath164 with fixed points @xmath232 . then for every non - empty open set @xmath233 there is some @xmath234 with @xmath235 . moreover , the action of @xmath19 on @xmath18 is minimal . let @xmath21 be a non - elementary subgroup with limit set @xmath18 which contains a pseudo - anosov element @xmath164 . let @xmath236 be the attracting and repelling fixed points of @xmath11 , respectively , and let @xmath233 be a non - empty open set . by lemma [ minimal ] , the limit set @xmath18 does not have isolated points and up to replacing @xmath11 by @xmath126 ( and exchanging @xmath37 and @xmath161 ) there is an element @xmath237 which maps @xmath37 to @xmath238 . then @xmath239 is a pseudo - anosov element with fixed points @xmath240 . by lemma [ pseudofix ] , we have @xmath241 . by lemma [ northsouth ] , @xmath207 acts with north - south dynamics on @xmath9 and hence @xmath242 for all sufficiently large @xmath243 . every closed @xmath19-invariant subset @xmath244 of @xmath20 contains every fixed point of every pseudo - anosov element . namely , if @xmath245 are the two fixed points of a pseudo - anosov element @xmath164 and if there is some @xmath246 then also @xmath247 since @xmath244 is closed and @xmath11 acts with north - south dynamics on @xmath9 . on the other hand , if @xmath248 then by the above consideration there is some @xmath249 with @xmath250 and once again , we conclude by invariance that @xmath251 as well . now the set of all fixed points of pseudo - anosov elements of @xmath19 is @xmath19-invariant and hence the smallest non - empty closed @xmath19-invariant subset of @xmath20 is the closure of the set of fixed points of pseudo - anosov elements . this set contains the limit set @xmath18 of @xmath19 by lemma [ minimal ] and hence it coincides with @xmath18 . in other words , the action of @xmath19 on @xmath18 is minimal . the lemma is proven . note the following immediate corollary of lemma [ pseudofix ] . [ nofixinf ] let @xmath21 be a non - elementary subgroup which contains a pseudo - anosov element . then @xmath19 does not fix a point in @xmath9 . * example : * there are non - elementary groups @xmath21 which fix a point in @xmath9 . namely , let @xmath107 be a simple closed separating curve on @xmath1 so that @xmath252 where neither @xmath253 nor @xmath254 is a three - holed sphere . let @xmath255 be a pseudo - anosov element @xmath256 . then @xmath257 generate a free abelian subgroup @xmath19 of @xmath10 whose limit set is a circle which is fixed pointwise by @xmath19 . we are now ready to show . [ pairfixdense ] let @xmath21 be a non - elementary subgroup with limit set @xmath18 which contains a pseudo - anosov element . 1 . the pairs of fixed points of pseudo - anosov elements of @xmath19 are dense in @xmath22 . 2 . for any two non - empty open subsets @xmath258 of @xmath22 there is some @xmath164 with @xmath259 . let @xmath21 be a non - elementary subgroup with limit set @xmath18 . assume that @xmath19 contains a pseudo - anosov element @xmath11 with attracting fixed point @xmath260 and repelling fixed point @xmath261 . let @xmath262 be a non - empty open set . our goal is to show that @xmath159 contains a pair of fixed points of a pseudo - anosov element . since @xmath18 does not contain isolated points , for this we may assume that there are small open sets @xmath263 with disjoint closure @xmath264 @xmath256 and such that @xmath265 . choose some @xmath234 which maps @xmath266 into @xmath267 . such an element exists by lemma [ moveaway ] . then @xmath268 is a pseudo - anosov element with fixed points @xmath269 . similarly , there is a pseudo - anosov element @xmath270 with both fixed points in @xmath271 . via replacing @xmath272 by sufficiently high powers we may assume that @xmath273 and that @xmath274 then we have @xmath275 and @xmath276 . by a result of mccarthy @xcite , up to possibly replacing @xmath277 and @xmath278 by even higher powers we may assume that @xmath279 is pseudo - anosov . then @xmath279 acts on @xmath9 with north - south - dynamics . since @xmath280 and @xmath281 , the pair of fixed points of @xmath279 is necessarily contained in @xmath282 and hence in @xmath159 . the first part of the proposition is proven . to show the second part of the proposition , let @xmath283 be non - empty open sets . we have to show that there is some @xmath164 such that @xmath259 . for this we may assume without loss of generality that @xmath284 where @xmath285 and @xmath286 are non - empty open subsets of @xmath18 with disjoint closure . since @xmath18 does not have isolated points , by possibly replacing @xmath287 by proper non - empty open subsets we may assume that the sets @xmath287 are pairwise disjoint . by the first part of the proposition , there is a pseudo - anosov element @xmath234 with attracting fixed point in @xmath288 and repelling fixed point in @xmath289 . since @xmath290 acts on @xmath20 with north - south dynamics , there is some @xmath141 and a small open neighborhood @xmath291 of the attracting fixed point of @xmath290 such that @xmath292 . the same argument produces an element @xmath270 , a number @xmath293 and an open subset @xmath294 of @xmath295 such that @xmath296 . this completes the proof of the proposition . as noted in the example after corollary [ nofixinf ] , in general the second part of proposition [ pairfixdense ] does not hold true for non - elementary subgroups of @xmath10 which do not contain a pseudo - anosov element . in this section we discuss some implications of the results in the previous section to the dynamics of the weil - petersson geodesic flow on moduli space . let @xmath26 be the unit tangent bundle of @xmath0 for the weil - petersson metric . the _ weil - petersson geodesic flow _ @xmath25 acts on @xmath26 by associating to a direction @xmath277 and a number @xmath297 the unit tangent @xmath298 at @xmath15 of the geodesic with initial velocity @xmath277 . note that this flow is not everywhere defined due to the existence of finite length geodesics which end in a point in @xmath106 . define @xmath299 to be the space of all directions of biinfinite geodesics , i.e. such that the flow line of @xmath25 through a point @xmath300 is defined for all times . note that the set @xmath301 is invariant under the action of the extended mapping class group . a direction at a point @xmath303 either defines a geodesic ray ( i.e. a geodesic defined on the half - line @xmath304 ) or a geodesic which ends at a point in @xmath106 . the set @xmath106 is a countable union of closed convex strata , each of real codimension two . since by the @xmath305-property , any two points @xmath306 can be connected by a unique geodesic depending continuously on @xmath307 , the set of directions of geodesics issuing from a point in @xmath0 and which terminate in the closure of a fixed stratum is a closed subset of @xmath26 of real codimension one . thus @xmath301 is the complement in @xmath26 of a countable union of closed subsets of codimension one , i.e. it is a dense @xmath302-set ( we refer to @xcite for details ) . wolpert @xcite also observed that for every @xmath303 the set of directions of geodesic rays issuing from @xmath34 has full lebesgue measure in the unit sphere at @xmath34 . then @xmath301 has full lebesgue measure . to each @xmath300 we can associate the ordered pair @xmath308 of endpoints of the biinfinite geodesic @xmath51 with initial velocity @xmath277 ( here ordered means that @xmath309 ) . the map @xmath93 clearly is invariant under the action of the geodesic flow on @xmath26 and hence it factors through a map of the quotient space @xmath310 . since the weil - petersson metric is negatively curved , by the flat strip theorem ( theorem ii.2.13 of @xcite ) the induced map @xmath311 is injective . this means that the set @xmath312 can be equipped with two natural topologies : the topology as a quotient of @xmath301 , and the induced topology as a subset of @xmath313 . we next observe that these two topologies coincide . we have to show that @xmath93 is open for the topology of @xmath301 as a subset of @xmath26 and for the topology of @xmath315 as a subset of @xmath316 . for this let @xmath300 and let @xmath159 be a neighborhood of @xmath277 in @xmath301 . we have to find a neighborhood @xmath160 of @xmath317 in @xmath318 such that the unit tangent line of every biinfinite geodesic @xmath51 whose pair of endpoints is contained in @xmath160 passes through @xmath159 . for this let @xmath303 be the footpoint of @xmath277 . since a geodesic depends smoothly on its initial velocity , if no such neighborhood @xmath160 of @xmath317 exists then there is a sequence of points @xmath319 which can be connected by a geodesic line @xmath320 in @xmath6 and there is a number @xmath113 such that @xmath321 for all @xmath195 . let @xmath322 be the ideal triangle in @xmath0 with vertices @xmath323 . since @xmath324 , the angle at @xmath34 of the triangle @xmath322 converges to @xmath93 as @xmath190 . connect each point on the geodesic ray from @xmath34 to @xmath325 to the point @xmath326 by a geodesic ray . this defines a _ ruled surface _ in @xmath0 with smooth interior which we denote again by @xmath322 . the intrinsic gau curvature of this surface at a point @xmath55 is bounded from above by an upper bound for the curvature of the weil - petersson metric at @xmath55 . since the weil - petersson metric is negatively curved , there is a number @xmath327 such that the gau curvature of the intersection of @xmath322 with the ball of radius @xmath216 about @xmath34 is bounded from above by @xmath328 . ( such a argument has been used in the literature many times . we refer to @xcite for a more detailed explanation and for additional references ) . let @xmath329 be the side of @xmath322 connecting @xmath34 to @xmath330 . the intrinsic angle at @xmath34 of the triangle @xmath322 coincides with the weil - petersson angle at @xmath34 . by assumption , the distance between @xmath34 and @xmath320 is at least @xmath331 . if @xmath196 is sufficiently large that the angle of @xmath322 at @xmath34 exceeds @xmath81 then by convexity of the distance function , a geodesic in @xmath322 for the intrinsic metric which issues from a point in @xmath332 $ ] and is perpendicular to @xmath183 does not intersect the side of @xmath322 connecting @xmath34 to @xmath325 . therefore the maximal length of such a geodesic is not smaller than @xmath333 . the union of these geodesic segments is an embedded strip in @xmath322 which is contained in the ball of radius @xmath216 about @xmath34 in @xmath0 . hence the gau curvature of @xmath322 at each point in the strip is at most @xmath328 . moreover , comparison with the euclidean plane shows that the area of the strip is at least @xmath334 . since the gau curvature of @xmath322 is negative , this implies that the integral of the gau curvature over @xmath322 does not exceed @xmath335 . on the other hand , the angle of @xmath322 at @xmath34 tends to @xmath93 as @xmath190 . since the gau curvature of @xmath322 is negative , the gau - bonnet theorem shows that the integral of the gau curvature of @xmath322 tends to zero as @xmath190 which is a contradiction to the estimate in the previous paragraph . as a consequence , the image of the open set @xmath159 under the projection @xmath93 contains indeed an open subset of @xmath312 equipped with the topology as a subspace of @xmath336 which shows the lemma . let @xmath27 be the quotient of the unit tangent bundle @xmath26 of @xmath0 under the action of the ( extended ) mapping class group . by equivariance , the weil - petersson geodesic flow projects to a flow on @xmath27 . every pseudo - anosov element @xmath17 defines a periodic orbit for the weil - petersson flow . this periodic orbit is the projection of the unit tangent line of an axis of @xmath11 . these are the only periodic orbits . namely , if @xmath337 is a periodic orbit in @xmath27 then there is a biinfinite weil - petersson geodesic @xmath51 in @xmath0 whose unit tangent line projects to the orbit . this geodesic is invariant under an element @xmath17 . then @xmath11 is axial , with axis @xmath338 , and hence @xmath11 is pseudo - anosov . since the subset @xmath301 of @xmath26 is dense and @xmath10-invariant , it suffices to show that the unit tangents of all axes of all pseudo - anosov elements are dense in @xmath301 . now by lemma [ limitsetmod ] and proposition [ pairfixdense ] the set of pairs of endpoints of all axes of pseudo - anosov elements in @xmath10 is dense in @xmath339 and hence in @xmath315 and therefore the proposition follows from lemma [ fiber ] . let @xmath21 be a non - elementary subgroup which contains a pseudo - anosov element . let as before @xmath299 be the space of all directions of biinfinite weil - petersson geodesics in @xmath0 and let @xmath340 be the space of all directions of geodesics with both endpoints in @xmath18 . then @xmath341 is a closed @xmath19-invariant subset of the ( non - locally compact ) space @xmath301 . by lemma [ fiber ] , the restriction to @xmath341 of the map @xmath93 factors through a homeomorphism @xmath342 . since a pair of fixed points of a pseudo - anosov element @xmath164 is contained in @xmath312 , by proposition [ pairfixdense ] the set @xmath343 is dense in @xmath22 . let @xmath344 be the canonical projection . we claim that for all nonempty open sets @xmath345 with @xmath346 and @xmath347 and every @xmath297 there is some @xmath348 and some @xmath349 such that @xmath350 . for this let @xmath351 be the preimages of @xmath352 in @xmath26 . then @xmath351 are open @xmath19-invariant subsets of @xmath26 . by lemma [ fiber ] the projection @xmath353 is open , there are open subsets @xmath258 of @xmath354 with @xmath355 , @xmath356 and such that the intersections of @xmath258 with @xmath22 are non - empty . since @xmath357 , by the second part of proposition [ pairfixdense ] there is some @xmath249 such that @xmath358 and that @xmath359 . let @xmath360 be the pair of endpoints of an axis of a pseudo - anosov element @xmath164 . such an element exists by the first part of proposition [ pairfixdense ] . then the axis of the conjugate @xmath361 of @xmath11 in @xmath19 has a pair of endpoints @xmath362 in @xmath363 . since the unit tangent lines of axes of pseudo - anosov elements which are conjugate in @xmath19 project to the same periodic orbit in @xmath364 for the weil - petersson flow , this implies that the projection of the unit tangent line of the axis of @xmath11 passes through both @xmath159 and @xmath160 . in particular , for @xmath297 and for a point @xmath365 contained in this periodic orbit , there is some @xmath349 such that @xmath366 this shows our claim . we use this observation to complete the proof of the corollary . namely , the closure @xmath367 of @xmath368 in @xmath364 is locally compact and separable since this is the case for @xmath364 . moreover , @xmath369 is invariant under the weil - petersson flow @xmath25 . hence we can choose a countable basis @xmath287 of open sets for @xmath369 . let @xmath370 and for each @xmath371 define inductively a nonempty open set @xmath372 in @xmath369 with @xmath373 and a number @xmath374 such that @xmath375 . this is possible by the above consideration and by continuity of the weil - petersson flow . then @xmath376 , and the forward @xmath25-orbit of any point in @xmath377 is infinite and dense in @xmath369 . with the same argument we can also guarantee that the backward @xmath25-orbit of a point @xmath378 is infinite and dense in @xmath369 . but this just means that for a lift @xmath379 of @xmath277 the @xmath19-orbit of @xmath380 is dense . this completes the proof of the corollary . a flow @xmath25 is called _ topologically transitive _ if it admits a dense orbit . as an immediate consequence of lemma [ fiber ] and proposition [ thm1finish ] we obtain the following result of brock , masur and minsky @xcite . * remark : * for two points @xmath381 it is in general difficult to decide whether @xmath382 , i.e. whether there is a geodesic line connecting @xmath176 to @xmath206 . however , brock , masur and minsky @xcite showed the following . let @xmath383 be a geodesic ray such that there is a number @xmath113 and a sequence @xmath384 with @xmath385 . then @xmath50 can be connected to every @xmath386 by a geodesic . even though the space @xmath27 is non - compact and the weil - petersson geodesic flow @xmath25 on @xmath27 is not everywhere defined , it admits uncountably many invariant borel - probability measures . indeed , it was shown in @xcite that there is a continuous injection from the space of invariant probability measures for the teichmller flow into the space of invariant probability measures for the weil - petersson geodesic flow . specific such measures are measures which are supported on periodic orbits . each such measure is _ ergodic_. the space of @xmath25-invariant borel probability measures for the weil - petersson flow can naturally be equipped with the weak@xmath28-topology . with respect to this topology , it is a closed convex set in the topological vector space of all finite signed borel measures on @xmath387 . the extreme points of this convex set are just the ergodic measures . thus theorem 2 from the introduction is an immediate consequence of the following let @xmath388 be a borel probability measure on @xmath27 which is invariant and ergodic for the weil - petersson geodesic flow @xmath25 . we have to find a sequence of periodic orbits for @xmath25 so that the normalized lebesgue measures @xmath389 supported on these orbits converge weakly as @xmath190 to @xmath388 . this means that for every continuous function @xmath390 with compact support we have @xmath391 for this let @xmath392 be a ( typical ) density point for the measure @xmath388 . by the birkhoff ergodic theorem , we have @xmath393 for every continuous function @xmath394 on @xmath27 with compact support . thus it suffices to find a sequence of numbers @xmath384 and a sequence of periodic orbits for @xmath25 which are the supports of normalized @xmath25-invariant measures @xmath389 such that @xmath395 for every continuous function @xmath390 with compact support . the weil - petersson metric induces a riemannian metric and hence a distance function @xmath396 on ( the orbifold ) @xmath27 ( the so - called sasaki metric ) . since a continuous function @xmath397 with compact support is bounded and uniformly continuous , for every @xmath113 there is a number @xmath398 depending on @xmath394 such that @xmath399 whenever @xmath400 are such that @xmath401 for all @xmath402 $ ] . since the sectional curvature of the weil - petersson metric is negative , comparison with the euclidean plane shows that the sasaki distance in the covering space @xmath26 can geometrically be estimated as follows . let @xmath403 be the canonical projection . then for every @xmath398 there is a number @xmath404 with the following property . let @xmath405 be two unit tangent vectors such that the flow - lines @xmath406 of @xmath407 are defined on the interval @xmath408 $ ] . if @xmath409 for all @xmath410 $ ] then @xmath411 . let @xmath412 be a geodesic whose initial tangent @xmath413 is a preimage of @xmath277 . by the above discussion and convexity of the distance function , it suffices to find a sequence of numbers @xmath384 and a sequence @xmath414 of pseudo - anosov elements with the following properties . 1 . there is a number @xmath415 such that the translation length of @xmath416 is contained in the interval @xmath417 $ ] for all @xmath195 . 2 . for every number @xmath398 there is a number @xmath418 not depending on @xmath195 such that the distance between the points @xmath419 and the axis @xmath320 of @xmath416 is at most @xmath420 for all sufficiently large @xmath195 . for the construction of such a sequence of pseudo - anosov elements , note that by the poincar recurrence theorem we may assume that there is a sequence @xmath384 such that @xmath421 @xmath182 . let as before @xmath412 be a geodesic whose initial velocity is a preimage of @xmath277 and write @xmath178 . then there is a sequence @xmath414 such that @xmath422 in particular , @xmath423 @xmath182 . let @xmath424\to { { \mathcal{t}}}(s)$ ] be the geodesic connecting @xmath184 to @xmath425 . it follows from the above discussion that it suffices to show that for sufficiently large @xmath195 the element @xmath426 is pseudo - anosov and its axis has property 2 ) above where the geodesic arc @xmath427 $ ] in the statement is replaced by the geodesic arc @xmath183 . for this recall from lemma [ semisimple2 ] that for each @xmath195 , the element @xmath428 either is axial or it is elliptic . if @xmath416 is axial then let @xmath429 be an oriented axis of @xmath416 . let @xmath430 . then we have @xmath431 where @xmath432 is the minimum of the displacement function of @xmath416 . note that @xmath433 for all large @xmath195 . if @xmath416 is not axial then let @xmath186 be a fixed point of @xmath416 . consider the ( possibly degenerate ) geodesic quadrangle @xmath434 in @xmath6 with vertices @xmath435 . the @xmath2-angle of @xmath434 at the vertices @xmath436 is not smaller than @xmath81 . since @xmath437 , by convexity of the distance function the angles of @xmath434 at @xmath438 do not exceed @xmath81 . on the other hand , by equivariance under the action of @xmath416 , the sum of the angles of @xmath434 at @xmath184 and at @xmath185 is not smaller than the angle at @xmath184 between the tangent of the geodesic arc @xmath183 and the negative @xmath439 of the tangent of the geodesic arc @xmath440 . now @xmath441 , @xmath442 and therefore this angle converges to @xmath93 as @xmath190 . thus we may assume that the minimum of the two angles of @xmath434 at @xmath438 is bigger than @xmath443 for all @xmath195 . let @xmath444 be the ruled triangle with vertices @xmath445 which we obtain by connecting the vertex @xmath185 to each point on the opposite side by a geodesic ( see the figure ) . a geodesic for the weil - petersson metric connecting a point in @xmath0 to a point in @xmath6 is entirely contained in @xmath0 except possibly for its endpoint @xcite . thus @xmath446 is a smooth embedded surface in @xmath0 whose intrinsic gau curvature is defined . the gau curvature at a point @xmath447 does not exceed the maximum of the sectional curvatures of the weil - petersson metric at @xmath34 ( compare the proof of lemma [ fiber ] and see @xcite for more and references ) . let @xmath448 be such that @xmath449 and let @xmath450 be the projection to @xmath27 of the set of all unit tangent vectors in @xmath26 with foot - point in @xmath114 . then @xmath244 is compact . let @xmath451 be the characteristic function of @xmath244 . since @xmath277 is a density point for @xmath388 by assumption and since there is a neighborhood of @xmath277 in @xmath27 which is entirely contained in @xmath244 , we have @xmath452 . let @xmath211 be sufficiently small that the sectional curvature of the restriction of the weil - petersson metric to the @xmath216-neighborhood of @xmath453 is bounded from above by a negative constant @xmath454 . such a number exists by invariance under the action of the mapping class group and cocompactness . by @xmath2-comparison , there is a number @xmath455 with the following property . let @xmath456 and let @xmath457\to \overline{{{\mathcal{t}}}(s)}$ ] , @xmath458\to \overline{{{\mathcal{t}}}(s)}$ ] be any two geodesics issuing from @xmath459 which enclose an angle at least @xmath443 . then @xmath206 does not pass through the @xmath216-neighborhood of @xmath460 $ ] . for this number @xmath461 and for any number @xmath462 let @xmath463 be such that @xmath464 and @xmath465 . such a number exists since by the birkhoff ergodic theorem we have @xmath466 let @xmath196 be sufficiently large that @xmath467 , and that @xmath468 . by convexity , we then have @xmath469 for all @xmath470 $ ] . let @xmath187 be the side of @xmath322 connecting @xmath186 to @xmath185 . if @xmath107 is a geodesic arc in @xmath322 of length at most @xmath216 issuing from a point in @xmath471 $ ] then @xmath107 does not intersect the side of @xmath322 connecting @xmath186 to @xmath184 and hence either it ends on a point in @xmath187 or it can be extended . thus if the distance between @xmath472 and @xmath187 is bigger than @xmath420 then by convexity the ruled triangle @xmath322 contains an embedded strip of width @xmath473 with the arc @xmath474 $ ] as one of its sides . this strip is the union of all geodesic arcs of length @xmath420 in @xmath322 with respect to the intrinsic metric which issue from a point in @xmath474 $ ] and which are perpendicular to @xmath183 . if @xmath107 is such a geodesic arc issuing from a point @xmath475 where @xmath476 is such that @xmath477 , then the gau curvature of @xmath322 on each point of @xmath107 does not exceed @xmath478 . by the choice of @xmath479 and by volume comparison , the lebesgue measure of the set of all points on such geodesic arcs is not smaller than @xmath480 . since the gauss curvature of @xmath322 is negative , this implies that the integral of the gauss curvature of @xmath322 over this strip is smaller than @xmath481 . however , this violates the gauss - bonnet theorem ( compare @xcite for more details for this argument , and see also the proof of lemma [ fiber ] ) . as a consequence , the geodesic @xmath187 passes through the @xmath420-neighborhood of the point @xmath472 . this implies that for sufficiently large @xmath195 the isometry @xmath416 is not elliptic . namely , we have @xmath482 and @xmath483 and hence if @xmath484 then @xmath485 on the other hand , if @xmath416 is elliptic then we have @xmath486 which contradicts inequality ( [ elliptic ] ) . thus @xmath416 is axial for all sufficiently large @xmath195 . let @xmath320 be an oriented axis for @xmath416 . we observe next that @xmath416 is pseudo - anosov for all @xmath195 . namely , using the above notation , let @xmath487 be such that @xmath488 . if the distance between the axis @xmath320 of @xmath416 and @xmath489 is smaller than @xmath420 then the axis @xmath320 of @xmath416 passes through the @xmath490-neighborhood of a point in @xmath114 and hence it passes through a point in @xmath0 . as a consequence , @xmath416 is pseudo - anosov ( see the remark after lemma [ semisimple2 ] ) . on the other hand , if the distance between @xmath320 and @xmath489 is bigger than @xmath420 then let @xmath491 be such that @xmath492 and @xmath493 . apply the above consideration to the ruled triangle @xmath494 with vertices @xmath495 which we obtain by connecting @xmath186 to each point on the opposite side by a geodesic segment and to the subarc @xmath496 $ ] of @xmath187 . we conclude that @xmath497 and once again , @xmath320 passes through a point in @xmath0 and @xmath416 is pseudo - anosov . by the above consideration , the axis @xmath320 of @xmath416 passes through the @xmath420-neighborhood of @xmath498 where @xmath499 only depends on @xmath420 . the same argument shows that this axis also passes through the @xmath420-neigbhorhood of @xmath500 where once more @xmath501 only depends on @xmath420 ( assume without loss of generality that @xmath502 is a density point for the image of @xmath388 under the flip @xmath503 and use the fact that two orbit segments of the same finite length are uniformly close if their initial points are close enough ) . in particular , the translation length of @xmath416 is contained in the interval @xmath504 $ ] . as a consequence , @xmath414 is a sequence of pseudo - anosov elements which satisfies the conditions 1),2 ) above . therefore the normalized lebesgue measures on the projections to @xmath27 of the unit tangent lines of the axes of the elements @xmath416 converge weekly to @xmath388 . this completes the proof of the proposition .

ill - posed inverse problems are ubiquitous in applications . understanding of algorithms for their solution has been greatly enhanced by a deep understanding of the linear inverse problem . in the applied communities ensemble - based filtering methods have recently been used to solve inverse problems by introducing an artificial dynamical system . this opens up the possibility of using a range of other filtering methods , such as 3dvar and kalman based methods , to solve inverse problems , again by introducing an artificial dynamical system . the aim of this paper is to analyze such methods in the context of the ill - posed linear inverse problem . statistical linear inverse problems are studied in the sense that the observational noise is assumed to be derived via realization of a gaussian random variable . we investigate the asymptotic behavior of filter based methods for these inverse problems . rigorous convergence rates are established for 3dvar and for the kalman filters , including minimax rates in some instances . blowup of 3dvar and a variant of its basic form is also presented , and optimality of the kalman filter is discussed . these analyses reveal a close connection between ( iterative ) regularization schemes in deterministic inverse problems and filter based methods in data assimilation . numerical experiments are presented to illustrate the theory .

1512.01955

in many geophysical applications , in particular in the petroleum industry and in hydrology , distributed parameter estimation problems are often solved by means of iterative ensemble kalman filters @xcite . the basic methodology is to introduce an artificial dynamical system , to supplement this with observations , and to apply the ensemble kalman filter . the methodology is described in a basic , abstract form , applicable to a general , possibly nonlinear , inverse problem in @xcite . in this basic form of the algorithm regularization is present due to dynamical preservation of a subspace spanned by the ensemble during the iteration . the paper @xcite gives further insight into the development of regularization for these ensemble kalman inversion methods , drawing on links with the levenberg - marquardt scheme @xcite . in this paper our aim is to further the study of filters for the solution of inverse problems , going beyond the ensemble kalman filter to encompass the study of other filters such as 3dvar and the kalman filter itself see @xcite for an overview of these filtering methods . a key issue will be the implementation of regularization with the aim of deriving optimal error estimates . we focus on the linear inverse problem @xmath0 where @xmath1 is a compact operator acting between hilbert spaces @xmath2 and @xmath3 . the exact solution is denoted by @xmath4 and @xmath5 is a noise polluting the observations . we will consider two situations : * data model @xmath6 * where multiple observations are made in the form [ eq : base ] ; and * data model @xmath7 * where a single observation is made . for modelling purposes we will assume that the noise @xmath5 is generated by the gaussian @xmath8 , independently in the case of multiple observations . in each case we create a sequence @xmath9 ; for data model @xmath6 the elements of this sequence are i.i.d . @xmath10 whilst for data model @xmath7 they are @xmath11 , with @xmath12 a single draw from @xmath10 . the case where multiple independent observations are made is not uncommon in applications ( for example in electrical impedance tomography ( eit , @xcite ) and , although we do not pursue it here , our methodology also opens up the possibility of considering multiple instances with correlated observational noise , by means of similar filtering - based techniques . the artificial , partially observed linear dynamical system that underlies our methodology is as follows : @xmath13 in _ deriving _ the filters we apply to this dynamical system , it is assumed that the @xmath14 are i.i.d . from @xmath8 . note , however , that whilst the data sequence @xmath9 we use in data model @xmath6 is of this form , the assumption is not compatible with data model @xmath7 ; thus for data model @xmath7 we have a form of _ model error _ or _ model mis - specification _ @xcite . by studying the application of filtering methods to the solution of the linear inverse problem our aim is to open up the possibility of employing the filtering methodology to ( static ) inverse problems of the form [ eq : base ] , and nonlinear generalizations . we confine our analysis to the linear setting as experience has shown that a deep understanding of this case is helpful both because there are many linear inverse problems which arise in applications , and because knowledge of the linear case guide methodologies for the more general nonlinear problem @xcite . the last few decades have seen a comprehensive development of the theory of linear inverse problems , both classically and statistically see @xcite and the references therein . consider the tikhonov - phillips regularization method @xmath15 this can be reformulated from a probabilistic perspective as the map estimator for bayesian inversion given a gaussian smoothness prior , with mean @xmath16 and cameron - martin space @xmath17 compactly embedded into @xmath2 , and a gaussian noise model as defined above ; this connection is eludicated in @xcite . we note that from the point of view of tikhonov - phillips regularization only the parameter @xmath18 is relevant , but that each of @xmath19 and @xmath20 have separate interpretations in the overarching bayesian picture , the first as a scaling of the prior precision and the second as observational noise variance . in this paper we deepen the connection between the bayesian methodology and classical methods . the recent paper @xcite opens up the prospect for a statistical explanation of iterative regularization methods in the form of @xmath21 with a general kalman gain operator @xmath22 . in this paper , we establish the connection between iterative regularization methods ( c.f . @xcite ) and filter based methods @xcite with respect to an artificial dynamic system . more precisely , for a linear inverse problem , we verify that the iterative tikhonov regularization method @xmath23 is closely related to filtering methods such as 3dvar and the kalman filter when applied to the partially observed linear dynamical system [ eq : dynamic ] . the similarity between both schemes provides a probabilistic interpretation of iterative regularization methods , and allows the possibility of quantifying uncertainty via the variance . on the other hand , we will employ techniques from the convergence analysis arising in regularization theories @xcite to shed light on the convergence of filter based methods , especially when the linear observation operator is ill - posed . the paper is organized as follows . we first introduce filter based methods for the artificial dynamics [ eq : dynamic ] in section [ se_artidyna ] . section [ se_assumptions ] describes some general useful formulae which are relevant to all the filters we study , and lists our main assumptions on the inverse problem of interest . in sections [ se_kalman ] and [ se_3dvar ] respectively , detailed asymptotic analyses are given for the kalman filter method and 3dvar , for both data models . the final section [ se_num ] presents numerical illustrations confirming the theoretical predictions . recall the artificial dynamics , where the observation operator @xmath1 also defines the inverse problem , and @xmath14 is an i.i.d . sequence with @xmath24 . the aim of filters is to estimate @xmath25 given the data @xmath26 in particular , probabilistic filtering aims to estimate the probability distribution of the conditional random variable @xmath27 if we assume that @xmath28 then the desired conditional random variable is gaussian , because of the linearity inherent in , together with the assumed gaussian structure of the noise sequence @xmath14 . furthermore the independence of the elements of the noise sequence means that the gaussian can be updated sequentially in a markovian fashion . if we denote by @xmath29 the mean , and by @xmath30 the covariance , then we obtain the kalman filter updates for these two quantities : @xmath31 the operator @xmath32 is known as the _ kalman gain _ and the inverse of the covariance , the _ precision operator _ @xmath33 , may be shown to satisfy @xmath34 all of these facts concerning the kalman filter may be found in chapter 4 of @xcite . expression requires careful justification in infinite dimensions , and this is provided in @xcite in certain settings . however we will only use as a quick method for deriving useful formulae , not expressed in terms of precision operators , which can be justified directly under the assumptions we make . a simplification of the kalman filter method is the 3dvar algorithm @xcite which is not , strictly speaking , a probabilistic filter because it does not attempt to accurately track covariance information . instead the covariance is fixed in time at @xmath35 for some fixed positive and self - adjoint operator @xmath36 the parameter @xmath19 is a scaling constant the inverse of which measures the relative size of the fixed covariance of the filter relative to that of the data . imposing this simplification on equations , gives [ eq:3dvarn ] @xmath37 it is also helpful to define , from , @xmath38 notice @xcite that the iteration looks like a stationary iterative tikhonov method with @xmath1 replaced by @xmath39 throughout the paper @xmath40 stands for kalman gain , updated mean and updated covariance for the kalman filter method and @xmath41 is the related sequence of quantities for 3dvar . we will view the filters as methods for reconstructing the truth @xmath42 ; in particular we will study the proximity of @xmath29 ( for the kalman filter ) and @xmath43 ( for 3dvar ) to @xmath42 for various large @xmath44 asymptotics . although the assumption in the _ derivation _ of the filters is that @xmath45 is an i.i.d . sequence of the form @xmath46 , we will not always assume that the data available is of this form ; to be precise data model @xmath6 is compatible with this assumption whilst data model @xmath7 is not . recall that data model @xmath6 refers to the situation where the data used in the kalman and 3dvar filters has the form @xmath47 , where the @xmath48 are i.i.d . @xmath49 given such a data sequence we can generate an auxiliary element @xmath50 with @xmath51 and @xmath52 . the law of large numbers and central limit theorem thus allows us to consider an inverse problem of the form with noise level reduced by a factor of @xmath53 we will study , in the sequel , whether the kalman or 3dvar filters are able to automatically exploit the decreased uncertainty inherent in an i.i.d . data set of this form . for data model @xmath7 we simply assume that the data used in the filters is of the form @xmath11 where @xmath12 is given by with @xmath54 from the discussion in the preceding paragraph , we clearly expect less accurate reconstruction in this case . for this data model we may view 3dvar as a stationary iterative tikhonov regularization , whilst the kalman filter is an alternative iterative non - stationary regularization scheme , since @xmath32 is updated in each step . in addition , the statistical perspective not only allows us to obtain an estimator ( the mean ( [ eq : mean ] ) or ( [ eq:3dvarmean ] ) ) , but also in the case of the kalman filter method , to quantify the uncertainty ( via the covariance ( [ eq : covariance ] ) ) . this uncertainty quantification perspective provides additional motivation for the filtering approaches considered herein . in this paper our primary focus is the asymptotic large @xmath44 behavior of the kalman filter method and 3dvar . more precisely , we are interested in the accurate recovery of the true state @xmath55 when the noise variance vanishes , i.e. @xmath56 for data models @xmath6 and @xmath7 , or as @xmath57 for data model @xmath6 ( by the law of large numbers / central limit theorem discussion above ) . to highlight the difficulties inherent in ill - posed inverse problems in this regard , we note the following which is a straightforward consequence of theorem 4.10 in @xcite when specialized to linear problems . here , and in what follows , @xmath58 denotes both the norm on @xmath2 and the operator norm from @xmath2 into itself . consider the 3dvar algorithm with @xmath59 . assume that there exists a constant @xmath60 such that @xmath61 and that @xmath62 then , for data model @xmath7 , it yields @xmath63 note however , that if the observation operator @xmath1 is compact or the inversion is ill - posed , the assumption @xmath64 in the preceding proposition can not hold . more precisely , the operator @xmath65 is no longer contractive since the spectrum of the operator @xmath66 clusters at the origin . our focus in the remainder of the paper will be on such ill - posed inverse problems . recall that @xmath58 denotes both the norm on @xmath2 and the operator norm from @xmath2 into itself . throughout @xmath67 will denote a generic constant , independent of the key limiting quantities @xmath20 and @xmath44 . our main assumption is : [ assp_main ] for both the kalman filter and the 3dvar filter , we assume ( i ) : : the variance @xmath68 and @xmath69 , where @xmath19 is a positive constant and @xmath70 is positive self - adjoint , and @xmath71 is a densely defined unbounded self - adjoint strictly positive operator ; ( ii ) : : the forward operator @xmath1 satisfies @xmath72 on @xmath2 for some constants @xmath73 and @xmath74 ; ( iii ) : : the initial mean satisfies @xmath75 ( or @xmath76 ) with @xmath77 ; ( iv ) : : the operator @xmath70 in item ( i ) is trace class on @xmath2 . we briefly comment on these items . item @xmath78 allows a well defined operator @xmath79 which is essential in carrying out our analysis . item @xmath80 is often called the _ link condition _ and it connects both operators @xmath1 and @xmath70 ( or @xmath81 ) . the third item @xmath82 is the _ source condition _ ( regularity ) of the true solution @xcite . the final item @xmath83 makes @xmath81 a well - defined covariance operator on @xmath2 @xcite . item @xmath80 in the preceding assumption is automatically satisfied if @xmath84 and @xmath70 have the same eigenfunctions and certain decaying singular values . item @xmath82 can then be expressed in this eigenbasis . when studying the kalman filter we will , in some instances , employ the following specific form of items @xmath80 , @xmath82 . comparison of assumptions [ assp_main ] and [ assp_main2 ] we see that they are identical if @xmath85 and @xmath86 [ assp_main2 ] let the variance @xmath87 . the operators @xmath70 and @xmath84 have the same eigenfunctions @xmath88 with their eigenvalues @xmath89 and @xmath90 satisfying @xmath91 for some @xmath92 , @xmath93 and @xmath94 . furthermore , by choosing the initial mean @xmath95 , the true solution @xmath42 with its coordinates @xmath96 in the basis @xmath88 obeys @xmath97 . we start by deriving properties of the kalman filter method under data model @xmath6 . other cases can be simply derived from minor variants of this setting . recall from ( [ eq : mean ] ) @xmath98 and note that @xmath99 under data model @xmath6 we have @xmath100 and hence the total error @xmath101 satisfies @xmath102 here @xmath103 to establish a rigorous convergence analysis , the mean squared error ( mse ) @xmath104 is of particular interest . since @xmath55 is deterministic and each @xmath48 is i.i.d gaussian we obtain a bias - variance decomposition of the mse : @xmath105 to proceed further , both terms @xmath106 and @xmath107 need to be calibrated more carefully . we consider the operator @xmath108 which appears in both terms . by ( [ eq : covariance ] ) , we obtain @xmath109 which is equivalent to @xmath110 notice that ( [ eq : preci ] ) yields @xmath111 by ( [ eq : prod ] ) and ( [ eq : recur_preci ] ) we obtain @xmath112 we will use this expression ( which is well - defined in view of assumption [ assp_main ] @xmath78 ) and the labelled equations preceding it in this subsection , frequently in what follows . in this section we investigate the asymptotic behaviour of the kalman filter ( [ eq : kgain])-([eq : covariance ] ) , under assumption [ assp_main ] . in particular , we are interested in whether we can reproduce the minimax convergence rate . this minimax rate is achieved by adopting assumption [ assp_main ] in the diagonal form of assumption [ assp_main2 ] . we present the main results in current subsection . [ thm_kfdm1 ] let assumption [ assp_main ] hold . then the kalman filter method ( [ eq : kgain])-([eq : covariance ] ) yields a bias - variance decomposition of the mse @xmath113 for the data model @xmath6 . setting @xmath114 and stopping the iteration when @xmath115 then gives @xmath116 [ thm_kfdm1ass2 ] let assumption [ assp_main2 ] hold . then the kalman filter method ( [ eq : kgain])-([eq : covariance ] ) yields a bias - variance decomposition of the mse @xmath117 for the data model @xmath6 . setting @xmath118 and stopping the iteration when @xmath115 then gives the following minimax convergence rate : @xmath119 * \(i ) under the assumption [ assp_main2 ] we prove unconditional convergence of the kalman filter method for any fixed @xmath120 and @xmath121 , noticing that both the bias and variance vanish when @xmath44 goes to infinity . the key ingredient which leads to this unconditional convergence , in comparison with assumption [ assp_main ] , is that the rate of decay of the eigenvalues of the variance operator @xmath70 is made explicit under assumption [ assp_main2 ] ; this is to be contrasted with the weaker assumption @xmath122 made in assumption [ assp_main ] @xmath83 . * \(ii ) by choosing @xmath19 depending on the update step @xmath123 , again with fixed @xmath20 , both theorems [ thm_kfdm1 ] and [ thm_kfdm1ass2 ] yield convergence rates in the mse sense . indeed in the second case , where we use assumption [ assp_main2 ] , the minimax rate of @xmath124 is achieved . this minimax rate may also be achieved from the bayesian posterior distribution with appropriate tuning of the prior in terms of the ( effective ) noise size @xmath125 @xcite ; the tuning of the prior is identical to the tuning of the initial condition for the covariance @xmath81 , via choice of @xmath126 @xmath127 proof of theorems [ thm_kfdm1 ] and [ thm_kfdm1ass2 ] is straightforward by means of a bias - variance decomposition . let assumption [ assp_main ] @xmath78 hold , noting that then @xmath128 is well - defined . we thus obtain , by ( [ eq : kgexplicit ] ) , @xmath129 where @xmath130 the operator - valued function @xmath131 ( [ eq_rn ] ) has been verified to be powerful in the convergence analysis of deterministic regularization schemes see ( * ? ? ? * ch.2 ) . in that context the following inequality is useful : @xmath132 , \quad 0\leq t \leq 1.\end{aligned}\ ] ] following these ideas we obtain the next two lemmas describing the bias and variance error bounds . we leave the proofs of both lemmas to the appendix . theorems [ thm_kfdm1 ] and [ thm_kfdm1ass2 ] are consequences , by choosing the parameter @xmath19 appropriately . [ lemma_kalmanbias ] let assumption [ assp_main ] @xmath78-@xmath82 hold . then the kalman filter method ( [ eq : kgain])-([eq : covariance ] ) yields @xmath133 furthermore , if assumption [ assp_main2 ] is valid , the bias obeys @xmath134 [ lemma_kalmanvariance ] let assumption [ assp_main ] @xmath78 , @xmath83 hold and @xmath135 in ( [ eq : dynamic ] ) be an i.i.d sequence with @xmath136 . then the kalman filter method ( [ eq : kgain])-([eq : covariance ] ) yields @xmath137 furthermore , if assumption [ assp_main2 ] is valid , the variance obeys @xmath138 the key difference between data model @xmath7 and data model @xmath6 is that , in the case @xmath7 , the noises @xmath48 appearing in the expression for the term @xmath107 are identical , rather than i.i.d . mean zero as in case @xmath6 . this results in a reduced rate of convergence in case @xmath7 over case @xmath6 , as seen in the following two theorems : [ thm_kfdm2 ] let assumption [ assp_main ] hold . then the kalman filter method ( [ eq : kgain])-([eq : covariance ] ) yields a bias - variance decomposition of the mse @xmath139 for the data model @xmath7 . fix @xmath140 and assume that the noise variance @xmath141 . if the iteration is stopped at @xmath115 then the following convergence rate is valid : @xmath142 [ thm_kfdm2ass2 ] let assumption [ assp_main2 ] hold . then the kalman filter method ( [ eq : kgain])-([eq : covariance ] ) yields a bias - variance decomposition of the mse @xmath143 for the data model @xmath7 . fix @xmath140 and assume that the noise variance @xmath144 if the iteration is stopped at @xmath115 then the following convergence rate is valid : @xmath145 both convergence rates in theorems [ thm_kfdm2 ] and [ thm_kfdm2ass2 ] are of the same order since the variance has been tuned to scale in the same way as the bias by choosing to stop the iteration at @xmath123 , depending on @xmath146 , appropriately . in comparison with the convergence rates in theorems [ thm_kfdm1 ] and [ thm_kfdm1ass2 ] , the ones in this section under data model @xmath7 require small noise @xmath20 ; those in the preceding subsection do not because multiple observations , with additive independent noise , are made of @xmath147 proof of the two preceding theorems is straightforward : the @xmath106 terms is analyzed as in the preceding subsection and the @xmath107 term must be carefully analyzed under the assumptions of data model @xmath7 . the key new result is stated in the following lemma , whose proof may be found in the appendix . [ lemma_kalmanvariancedm2 ] let assumption [ assp_main ] hold and each observation @xmath148 be fixed . then the kalman filter method ( [ eq : kgain])-([eq : covariance ] ) yields @xmath149 furthermore , if assumption [ assp_main2 ] is valid , the variance obeys @xmath150 the mean of the 3dvar algorithm is given by ( [ eq:3dvarmean ] ) and has the form @xmath151 furthermore @xmath152 if we define @xmath153 then we obtain , since @xmath100 , @xmath154 with @xmath155 . we further derive @xmath156 by inserting the definition of the kalman gain ( [ eq:3dvarkgain ] ) , and by assuming assumption [ assp_main ] @xmath78 . the operator - valued function @xmath157 is defined @xmath158 similarly to the analysis of the kalman filter , we derive @xmath159 thus , the mse takes the bias - variance decomposition form @xmath160 this leads to the following two theorems : [ thm_3dvardm1 ] let assumption [ assp_main ] hold . then 3dvar filter ( [ eq:3dvarkgain])-([eq:3dvarmean ] ) yields a bias - variance decomposition of the mse @xmath161 for the data model @xmath6 . setting @xmath114 and stopping the iteration when @xmath115 then gives @xmath162 [ thm_3dvardm1ass2 ] let assumption [ assp_main2 ] hold . then 3dvar filter ( [ eq:3dvarkgain])-([eq:3dvarmean ] ) yields a bias - variance decomposition of the mse @xmath163 for the data model @xmath6 . setting @xmath164 and stopping the iteration when @xmath115 then gives @xmath165 the decay rate at the end of the preceding theorem is the same as that in theorem [ thm_3dvardm1 ] if @xmath85 and @xmath166 . this is the setting in which assumptions [ assp_main ] and [ assp_main2 ] are identical . the preceding two theorems show that , under data model @xmath6 and for any fixed @xmath19 , the ( bound on the ) mse of the 3dvar filter blows up logarithmically as @xmath57 under assumption [ assp_main ] , and is asymptotically bounded for assumption [ assp_main2 ] . in contrast , for the kalman filter method the mse is asymptotically bounded or unconditionally converges in @xmath44 under the same assumptions see theorems [ thm_kfdm1 ] and [ thm_kfdm1ass2 ] . with optimal choice of @xmath19 in terms of the stopping time of the iteration at @xmath115 , comparison of the convergence rates in theorems [ thm_kfdm1 ] and [ thm_3dvardm1 ] ( or theorems [ thm_kfdm1ass2 ] and [ thm_3dvardm1ass2 ] ) shows that the kalman filter outperforms 3dvar , but only by a logarithmic factor ( or a hlder factor ) . for simplicity we only analyze and discuss data model @xmath6 for 3dvar filter under the additional assumption [ assp_main2 ] ; as for data model @xmath7 one can derive consequences similar to those in the preceding section in an analogous manner . @xmath127 we now study data model @xmath7 and 3dvar . we consider only assumption [ assp_main ] ; however the reader may readily extend the analysis to include assumption [ assp_main2 ] . in the case of data model @xmath7 , both the kalman and 3dvar filters have the same error bounds : [ thm_3dvardm2 ] let assumption [ assp_main ] hold . then the 3dvar algorithm ( [ eq:3dvarkgain])-([eq:3dvarmean ] ) yields a bias - variance decomposition of the mse @xmath167 for the data model @xmath7 . fix @xmath140 and assume that the noise variance @xmath141 . if the iteration is stopped at @xmath115 then the following convergence rate is valid : @xmath168 the preceding three theorems can be proved by the bias - variance decomposition and application of the following three lemmas , whose proofs are left to the appendix . however the proof of theorem [ thm_3dvardm1ass2 ] is not as straightforward as the others and we present details in the appendix . [ lemma_3dvarbias ] let assumption [ assp_main ] @xmath78-@xmath82 hold . then 3dvar ( [ eq:3dvarn ] ) yields @xmath169 furthermore , if assumption [ assp_main2 ] is valid , the bias obeys @xmath170 [ lemma_3dvarvariancedm1 ] let assumption [ assp_main ] @xmath78 , @xmath83 hold and each noise @xmath48 in ( [ eq : dynamic ] ) i.i.d . generated by @xmath171 . then 3dvar ( [ eq:3dvarn ] ) yields @xmath172 for data model @xmath6 . furthermore , if assumption [ assp_main2 ] is valid , the variance obeys @xmath173 and simultaneously @xmath174 [ lemma_3dvarvariancedm2 ] let assumption [ assp_main ] hold and each observation @xmath148 be fixed . then 3dvar ( [ eq:3dvarn ] ) yields @xmath175 for data model @xmath7 . recall that the 3dvar iteration looks like a stationary iterative tikhonov method @xcite , with @xmath1 replaced by @xmath176 and with a fixed parameter @xmath19 . the non - stationary iterative tikhonov regularization , with varying @xmath19 , has been proven to be powerful in deterministic inverse problems @xcite . we generalize this method to the 3dvar setting . furthermore we demonstrate that , for data model @xmath6 , it is possible to see blow - up phenomena with this algorithm . starting from the classical form of the 3dvar filter as given in , , we propose a variant method in which @xmath19 varies with @xmath44 . we obtain @xmath177 if we define @xmath178 then we obtain , analogously to the derivation of ( [ eq : generaldecomposition ] ) and ( [ eq:3dvarerror ] ) , the bias - variance decomposition as follows : @xmath179 with @xmath180 by calibrating both terms @xmath181 , @xmath182 carefully , we obtain the following blow - up result , proved in the appendix . although the result only provides an upper bound , numerical evidence does indeed show blow - up in this regime . [ thm_3dvarvariant ] let assumption [ assp_main ] hold and let @xmath183 be a sequence satisfying @xmath184 , with constant @xmath185 , for @xmath186 then the variant enkf method ( [ eq : variantkgain])-([eq : variantrcovariance ] ) yields a bias - variance decomposition of the mse @xmath187 for data model @xmath188 in particular , the geometric sequence @xmath189 with @xmath120 and @xmath190 yields @xmath191 in this section we provide numerical results which display the capabilities of 3dvar and kalman filter for solving linear inverse problems of the type described in section [ se_intro ] . in addition , we verify numerically some of the theoretical results from section [ se_kalman ] and section [ se_3dvar ] . we consider the two - dimensional domain @xmath192 ^ 2 $ ] and define the operators @xmath193 with @xmath194 with this domain @xmath195 is positive , self - adjoint and invertible.note that ( [ eq : ca ] ) from assumption [ assp_main ] is satisfied with @xmath196 and @xmath197 . in the following subsections we produce synthetic data from a true function @xmath198 that we generate as a two - dimensional random field drawn from the gaussian measure @xmath199 with covariance @xmath200 with domain of definition @xmath201 and where @xmath202 is selected as described below . the shift of @xmath195 by a constant introduces a correlation length into the true function . furthermore , for simplicity we consider @xmath203 and note @xcite that the given selection of @xmath198 yields @xmath204 for all @xmath205 . therefore , for discussion of the present experiments we simply assume that @xmath206 . consequently , in order to satisfy assumption [ assp_main ] ( iii ) we need @xmath207 such that @xmath208 . note that operator @xmath70 in ( [ num1 ] ) satisfies assumption [ assp_main ] ( iv ) . the numerical generation of @xmath198 is carried out by means of the karhunen - loeve decomposition of @xmath198 in terms of the eigenfunctions of @xmath209 which , from the definition of @xmath210 , are cosine functions . we recall that for the application of the kalman filter and 3dvar with data model 1 we need to generate @xmath123 instances of synthetic data @xmath211 where @xmath123 is the maximum number of iterations of the scheme . below we discuss the selection of such @xmath123 . the aforementioned synthetic data are generated by means of @xmath212 with @xmath213 and @xmath20 specified below . for data model 2 we produce synthetic data simply by setting @xmath214 with @xmath215 . in order to avoid inverse crimes @xcite , all synthetic data used in our experiments are generated on a finer grid ( @xmath216 cells ) than the one ( of @xmath217 cells ) used for the inversion . we use splines to interpolate synthetic data on the coarser grid that we use for the application of the filters . for both 3dvar and kalman filter with data model 1 , we fix the number of iterations @xmath123 for the scheme and consider the selection @xmath218 stated in theorem [ thm_kfdm1 ] and theorem [ thm_3dvardm1 ] . provided that the filters are stopped according to @xmath115 , these theorems ensure the convergence rates in ( [ rate1_kf ] ) and ( [ rate1_3dvar ] ) that we verify numerically in the following subsection . for data model 2 , theorem [ thm_kfdm2 ] and theorem [ thm_3dvardm2 ] suggest that convergence rates ( [ rate2_kf ] ) and ( [ rate2_3dvar ] ) are satisfied for @xmath140 and provided that @xmath141 . the latter equality provides an expression for the number of iterations @xmath219 that we may use as stopping criteria for these filtering algorithms applied to data model 2 . in algorithm [ al1 ] we summarize the kalman filter and 3dvar schemes applied to both data models . kalman filter/3dvar ( data model @xmath220data model @xmath7)[al1 ] + let @xmath221 and @xmath222 for @xmath223 , update @xmath224 and @xmath225 as follows @xmath226 @xmath227 where @xmath228 and where we recall that @xmath229 note that for data model @xmath6 we need at least @xmath123 independent instances of data . in this subsection we demonstrate how the filters under consideration in the iterative framework described in algorithm [ al1 ] can be used , with data model 1 and data model 2 , to solve the linear inverse problem presented in section [ se_intro ] . let us consider the truth @xmath198 displayed in figure [ fig1 ] ( top ) generated , as described in subsection [ se_num : setup ] , from a gaussian measure with covariance ( [ num3 ] ) and @xmath230 . synthetic data are generated as described above with three different choices of @xmath20 that yield noise levels of approximately @xmath231 , @xmath232 , and @xmath233 of the norm of the noise free data ( i.e. @xmath234 ) . we apply algorithm [ al1 ] to this synthetic data generated both for the application of data model @xmath6 and data model @xmath7 . for data model @xmath6 we consider a selection of @xmath235 . algorithm [ al1 ] states that the these schemes should be stopped at iteration level @xmath115 . however , in order to observe the performance of these schemes , in these experiments we allowed for a few more iterations . in the left column of figure [ fig2 ] we display the plots of the error w.r.t the truth of the estimator @xmath236 as a function of the iterations , i.e. @xmath237 where @xmath238 denotes the interpolation of @xmath198 on the coarse grid used for the inversion . note that the error w.r.t . the truth of the estimates produced by both schemes decreases monotonically . interestingly , the value at the final iteration displayed in these figures is approximately the same for all these experiments independently of the noise level . moreover , the stability of the scheme does not seem to depend on the early termination of the scheme . in addition , we note that both kalman filter and 3dvar exhibit very similar performance in terms of reducing the error w.r.t the truth . however , for larger noise levels we observe small fluctuations in the error obtained with 3dvar . in figure [ fig1 ] we display the estimates @xmath29 obtained with 3dvar with data model @xmath6 at iterations @xmath239 for noise levels ( determined by the choice of @xmath20 ) of @xmath231 ( top - middle ) , @xmath232 ( middle - bottom ) and @xmath233 ( bottom ) . we can visually appreciate from figure [ fig1 ] that the estimates obtained at the final iteration @xmath240 are indeed similar one to another even though the one in the bottom row was computed by inverting data five times noisier than the one in the first row . similar estimates ( not shown ) were obtained with the kalman filter for data model @xmath6 . for data model @xmath7 , the selection of @xmath20 corresponding to noise levels of @xmath231 , @xmath232 , and @xmath233 yields a maximum number of iterations @xmath241 respectively . clearly , for data model @xmath7 , smaller observational noise results in schemes that can be iterated longer , presumably achieving more accurate estimates . similarly to data model @xmath6 , we are required to stop the algorithm at the iteration @xmath115 . however , for the purpose of this study we iterate until @xmath240 . in the right column of figure [ fig2 ] we display the plots of the error w.r.t the truth of @xmath29 . in contrast to data model @xmath6 , we observe that the error w.r.t the truth increases for @xmath242 . in other words , the kalman filter and 3dvar , when applied to data model @xmath7 , need to be stopped at @xmath115 in order to stabilize the scheme and obtain accurate estimates of the truth . moreover , stopping the scheme at @xmath115 results in estimates whose accuracy increases with smaller noise levels . clearly , in the small noise limit , both data models tend to exhibit similar behaviour . in figure [ fig3 ] we display @xmath29 obtained from 3dvar applied to data model @xmath7 at iterations @xmath239 for data with noise levels of @xmath231 ( top - middle ) , @xmath232 ( middle - bottom ) and @xmath233 ( bottom ) . similar estimates ( not shown ) were obtained with the kalman filter for data model @xmath7 . it is clear indeed , that for data model @xmath6 , the application of kalman filter and 3dvar results in more accurate and stable estimates when the noise level in the data is not sufficiently small . the results from this subsection show that the reduction in the variance of the noise due to the law of large numbers and central limit theorem effect in data model @xmath6 produces more accurate algorithms . although data model @xmath6 requires multiple instances of the data , in some applications such as in eit @xcite , the data collection can be repeated multiple times in order to obtain these data . . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 1 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) . , title="fig : " ] + . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 1 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) . , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 1 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) . , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 1 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) . , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 1 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) . , title="fig : " ] + . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 1 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) . , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 1 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) . , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 1 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) . , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 1 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) . , title="fig : " ] + . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 1 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) . , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 1 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) . , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 1 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) . , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 1 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) . , title="fig : " ] + + . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 2 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) , title="fig : " ] + . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 2 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 2 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 2 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 2 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) , title="fig : " ] + . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 2 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 2 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 2 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 2 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) , title="fig : " ] + . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 2 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 2 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 2 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) , title="fig : " ] . top - middle , bottom - middle and bottom : estimates obtained with 3dvar and data model 2 at iterations ( from left to right ) 1 , 10 , 20 , 30 ) for noise levels of @xmath231 ( top - middle ) , @xmath232 ( bottom - middle ) and @xmath233 ( bottom ) , title="fig : " ] . [ fig3 ] in this subsection we test the convergence rates from theorems [ thm_kfdm1 ] , [ thm_3dvardm1 ] , [ thm_kfdm2 ] , and [ thm_3dvardm2 ] . for the verification of each of these rates we let @xmath243 denote the covariance from , and we consider @xmath244 experiments corresponding to different truths @xmath198 generated from @xmath245 with the four selections of @xmath246 . note that assumption [ assp_main ] ( iii ) is satisfied only for @xmath247 . again , for data model 1 we generate synthetic data for each of the truths and for as many iterations used for the application of both schemes . inverse crimes are avoided as described in subsection [ se_num : setup ] . the verification of theorem [ thm_kfdm1 ] and theorem [ thm_3dvardm1 ] by means of algorithm [ al1 ] in the case of data model 1 is straightforward . for each of the set of synthetic data associated to each of the 20 truths @xmath198 previously mentioned , we fix @xmath248 . for each @xmath123 ( with @xmath249 ) we run algorithm [ al1 ] , stop the schemes at @xmath115 and record the value of @xmath250 . in the right ( resp . left ) figure [ fig4 ] we display a plot of @xmath250 vs @xmath251 for the kalman filter ( resp . 3dvar ) for each of the set of 20 experiments associated to different truths ( red solid lines ) generated as described above with ( from top to bottom ) @xmath246 . from theorem [ thm_kfdm1 ] we note that the corresponding slopes of the convergence rates should be approximately given by @xmath252 . for theorem [ thm_3dvardm1 ] there is an additional term of @xmath253 , but this is of course negligible compared to the algebraic decay and we ignore it for the purposes of this discussion . for comparison , a line ( black dotted ) with slope @xmath252 is displayed in figure [ fig4 ] . we now verify the convergence rates of theorem [ thm_kfdm2 ] and theorem [ thm_3dvardm2 ] . note first that in algorithm [ al1 ] for data model @xmath7 we define @xmath123 in terms of the given small noise @xmath20 , in order to obtain convergence . however , for the purpose of the verification of the aforementioned convergence rates we define @xmath20 in terms of @xmath123 by means of the same expressions . in other words , for each @xmath123 ( @xmath249 ) we produce synthetic data ( or each of the 20 truths ) with @xmath254 and @xmath255 . we then run algorithm [ al1 ] and stop the schemes at @xmath115 . in the right ( resp . left ) figure [ fig5 ] we display a plot of @xmath250 vs @xmath251 for the kalman filter ( resp . 3dvar ) for each of the set of 20 experiments associated to different truths ( red solid lines ) generated as before with ( from top to bottom ) @xmath246 . we again include a line ( black dotted ) with slope of @xmath256 which is the asymptotic behavior predicted by theorems [ thm_kfdm2 ] and [ thm_3dvardm2 ] . we can clearly appreciate that , for @xmath207 satisfying assumption [ assp_main ] ( iii ) ( i.e. @xmath257 ) , the numerical convergence rates fit very well the ones predicted by the theory . note that the higher the regularity of the truth ( i.e. the larger the @xmath207 ) , the smaller the error . w.r.t the truth in the estimates . we note that for @xmath258 , the aforementioned assumption is violated and , in the case of data model 1 , the slopes of the numerical convergence rates are slightly smaller than the theoretical ones . in this case ( @xmath258 ) there are also fluctuations of the error w.r.t . the truth obtained with 3dvar . these fluctuations may be associated with the fact that since for the error w.r.t . the truth is very small for sufficiently large iterations and for data model @xmath6 the noise level is fixed a priori ( recall @xmath248 ) . however , for the kalman filter these fluctuations are not so evident ; presumably updating the covariance has a stabilizing effect . for data model @xmath7 , as @xmath123 increases , the corresponding @xmath20 decreases and so these fluctuations in the error are non existent . * we have presented filter based algorithms for the linear inverse problem , based on introduction of an artificial dynamic . this results in methods which are closely related to iterative tikhonov - type regularization . two data scenarios are considered , one ( data model @xmath6 ) involving multiple realizations of the data , with independent noise ; the other ( data model @xmath6 ) involving a single realization of the data ; both are relevant in applications . * we present theoretical results demonstrating convergence of the algorithms in the two data scenarios . for multiple realizations of the noisy the convergence is induced by the inherent averaging present in the iterative method , and the link to the law of large numbers and central limit theorem . for the single instance of data the small observational noise limit must be considered . * for both data model @xmath6 and data model @xmath7 the kalman filter and 3dvar produced very similar results for relatively small @xmath123 ( @xmath259 ) . in practice it is clear that 3dvar is preferable as the kalman filter requires covariance updates which may be impractical for large scale models . however , updating the covariance in the kalman filter seems to have an stabilizing effect in the error w.r.t the truth . * for data model @xmath6 the level of accuracy of the estimator is independent of the noise level . moreover , the stability of the scheme is not conditioned to the early termination of the scheme . in contrast , for data model @xmath7 we need to stop at @xmath115 to avoid an increase in the error w.r.t the truth . again , this illustrates that , whenever multiple instances of the data are available , data model @xmath6 offers a more stable and accurate framework for solving the inverse problems under consideration . * the theoretical results from this work are verified numerically whenever the assumptions of the theory are satisfied . with @xmath207 ( from top to bottom ) 1,2,3 and 4 . , title="fig : " ] with @xmath207 ( from top to bottom ) 1,2,3 and 4 . , title="fig : " ] + with @xmath207 ( from top to bottom ) 1,2,3 and 4 . , title="fig : " ] with @xmath207 ( from top to bottom ) 1,2,3 and 4 . , title="fig : " ] + with @xmath207 ( from top to bottom ) 1,2,3 and 4 . , title="fig : " ] with @xmath207 ( from top to bottom ) 1,2,3 and 4 . , title="fig : " ] + with @xmath207 ( from top to bottom ) 1,2,3 and 4 . , title="fig : " ] with @xmath207 ( from top to bottom ) 1,2,3 and 4 . , title="fig : " ] with @xmath207 ( from top to bottom ) 1,2,3 and 4 . , title="fig : " ] with @xmath207 ( from top to bottom ) 1,2,3 and 4 . , title="fig : " ] + with @xmath207 ( from top to bottom ) 1,2,3 and 4 . , title="fig : " ] with @xmath207 ( from top to bottom ) 1,2,3 and 4 . , title="fig : " ] + with @xmath207 ( from top to bottom ) 1,2,3 and 4 . , title="fig : " ] with @xmath207 ( from top to bottom ) 1,2,3 and 4 . , title="fig : " ] + with @xmath207 ( from top to bottom ) 1,2,3 and 4 . , title="fig : " ] with @xmath207 ( from top to bottom ) 1,2,3 and 4 . , title="fig : " ] andrew m. stuart is funded by epsrc ( under the programme grant equip ) , darpa ( under equips ) and onr . shuai lu is supported by special funds for major state basic research projects of china ( 2015cb856003 ) , nsfc ( 91130004 , 11522108 ) , shanghai science and technology commission grant ( 14qa1400400 ) and the programme of introducing talents of discipline to universities ( number b08018 ) , china . 1 agapiou s. ; larsson s. and stuart a. m. : _ posterior contraction rates for the bayesian approach to linear ill - posed inverse problems _ , stochastic process . ( 2013 ) , no . 10 , 38283860 . bissantz n. ; hohage t. ; munk a. and ruymgaart f. : _ convergence rates of general regularization methods for statistical inverse problems and applications _ , siam j. numer . , 45 ( 2007 ) , 2610,c2636 . bogahcev v. i. : _ gaussian measures _ , vol . 62 of _ mathematical surveys and monographs_. american mathematical society , ( 1998 ) . borcea l , _ electrical impedance tomography . _ , inverse problems series 18(6 ) , 2002 dashti m ; and stuart a. m. : _ the bayesian approach to inverse problems_. to appear in the handbook of uncertainty quantification , editors r. ghanem , d. higdon and h. owhadi , springer , 2016 . arxiv:1302.6989 engl h. w. ; hanke m. and a. neubauer : _ regularization of inverse problems _ , mathematics and its applications , vol . 375 , kluwer academic publishers group , dordrecht , 1996 . hanke m. : _ a regularizing levenberg - marquardt scheme , with applications to inverse groundwater filtration problems . _ inverse problems , 13 ( 1997 ) , 7995 . hanke m. and groetsch c. w. : _ nonstationary iterated tikhonov regularization _ , journal of optimization theory and applications 98 ( 1998 ) , no . 1 , 3753 . iglesias m. a. ; law k. j. and stuart a. m. : _ ensemble kalman methods for inverse problems _ , inverse problems 29 ( 2013 ) , no . 4 , 045001 . iglesias m. a. : . to apper in inverse problems . arxiv:1505.03876 . kaipio j. ; somersalo e. : _ statistical and computational inverse problems_. applied mathematical sciences , 160 . springer - verlag , new york , 2005 . knapik b. t. ; van der vaart a. w. and van zanten j. h. : _ bayesian inverse problems with gaussian priors _ , the annals of statistics 39 ( 2011 ) , no . 5 , 26262657 . law k. j. ; stuart a. m. and zygalakis k.c . : _ data assimilation : a mathematical introduction _ , springer , 2015 . lu s. and pereverzev s. v. , _ regularization theory for ill - posed problems . selected topics . _ , inverse and ill - posed problems series , vol . 58 , walter de gruyter , berlin , 2013 . oliver d. ; reynolds a. , and liu n. : . cambridge univ pr , 2008 . _ proof of lemma [ lemma_kalmanbias ] . _ the proof follows the classic arguments on tikhonov regularization with hilbert scales , c.f . * ch8.4 ) . we recall ( [ eq : ca ] ) in assumption [ assp_main ] @xmath80 and rewrite it in the form @xmath260 notice that the definition of @xmath261 in gives , since @xmath2 is a hilbert space , and using assumption [ assp_main ] @xmath78 to ensure that @xmath261 and @xmath262 are well - defined bounded linear operators , @xmath263 combining the two preceding displays we obtain @xmath264 and a duality argument yield @xmath265 for any @xmath266 . let @xmath267 $ ] . then the inequality of heinz ( * ? ? ? * ch.8.4 , pp . 213 ) and an additional duality argument gives @xmath268 which yields @xmath269 . let assumption [ assp_main ] @xmath82 be valid with @xmath270 , and define @xmath271 . since @xmath272 $ ] , and consequently @xmath273,$ ] we obtain from ( [ eq : cb ] ) that @xmath274 furthermore there exists a @xmath275 such that @xmath276 noting that @xmath277 , and employing ( [ eq : cb ] ) with @xmath278 , together with ( [ eq : rapproxi_km2 ] ) and ( [ eq : pos_rr ] ) , we have @xmath279 in case of assumption [ assp_main2 ] we insert @xmath280 and @xmath281 . _ proof of lemma [ lemma_kalmanvariance ] . _ notice that @xmath283 and @xmath284 thus we obtain @xmath285 by virtue of ( [ eq : prod ] ) and ( [ eq : cnkn ] ) we derive @xmath286 we denote @xmath287 and obtain @xmath288 by the definition of @xmath289 and assumption [ assp_main ] @xmath78 , @xmath83 we obtain and consequently derive @xmath291 with the operator - valued function @xmath292 . such an observation then yields @xmath293 concerning assumption [ assp_main2 ] , we further estimate , by exploiting ( * ? ? ? * lemma 8.2 ) , @xmath294 and @xmath295 _ proof of theorem [ thm_3dvardm1ass2 ] . _ as for the other theorems , the proof rests , of course , on the bias variance decomposition , and then use of lemmas [ lemma_3dvarbias ] and [ lemma_3dvarvariancedm1 ] . this yields @xmath163 and simultaneously @xmath298 choosing @xmath299 for the former inequality and @xmath300 for the latter inequality , we conclude that , by stopping the iteration when @xmath115 , @xmath301 @xmath282 _ proof of lemma [ lemma_3dvarbias ] . _ analogously to the proof of lemma [ lemma_kalmanbias ] , it may be shown that @xmath302 the final inequality follows from the asymptotic behavior of @xmath303 , established , for example . in ( * ch.2 , pp . in the case of assumption [ assp_main2 ] we insert @xmath280 and @xmath281 . _ proof of lemma [ lemma_3dvarvariancedm1 ] . _ we denote @xmath304 and obtain @xmath305 furthermore , we derive @xmath306 and @xmath307 thus , @xmath308 for assumption [ assp_main2 ] we need to estimate @xmath309 carefully . we substitute the given decay rate of the different eigenvalues , to obtain @xmath310 by arguments similar to those used in the proof of lemma [ lemma_kalmanvariance ] ( or ( * ? ? ? * lemma 8.2 ) ) , we further estimate @xmath311 and @xmath312 where the summation term in the right - hand side is bounded . _ proof of lemma [ lemma_3dvarvariancedm2 ] . _ since @xmath315 in this case , we derive @xmath316 and by operator - valued calculation we obtain @xmath317 where @xmath318 . thus we obtain , by the asymptotic behavior of @xmath319 derived in ( * ? ? ? * ch.2 , pp . 64 ) , @xmath320 @xmath282 _ proof of theorem [ thm_3dvarvariant ] . _ similar to the kalman filter method and 3dvar , by assumption [ assp_main ] @xmath78-@xmath82 , we obtain the bias error estimate @xmath321 now we need upper bounds of the following operator - valued function @xmath322 define @xmath323 and assume the sequence @xmath324 satisfying @xmath325 with a constant @xmath185 . then the results in @xcite yield @xmath326 notice that for any @xmath330 @xmath331 we derive , @xmath332 and @xmath333 a rough variance estimate for the variant method is @xmath334 the first term vanishes but the second term blows up when @xmath57 and @xmath335 . to further investigate the blow up , we consider a special geometric sequence @xmath336 with @xmath190 . thus , we have @xmath337 and ( [ eq : sequence ] ) is satisfied with @xmath338 . actually , we derive @xmath339 thus the results in @xcite refine , by the asymptotic behavior of ( [ eq : f_nv ] ) , @xmath340 and we derive @xmath341 summing up , for the geometric sequence , we obtain @xmath342 the second term blows up exponentially when @xmath44 goes to infinity .

more than two thirds of disk galaxies are barred to some degree . many today harbor massive concentrations of gas in their centers , and some are known to possess supermassive black holes ( smbhs ) and their associated stellar cusps . previous theoretical work has suggested that a bar in a galaxy could be dissolved by the formation of a mass concentration in the center , although the precise mass and degree of central concentration required is not well - established . we report an extensive study of the effects of central masses on bars in high - quality-body simulations of galaxies . we have varied the growth rate of the central mass , its final mass and degree of concentration to examine how these factors affect the evolution of the bar . our main conclusions are : ( 1 ) bars are more robust than previously thought . the central mass has to be as large as several percent of the disk mass to completely destroy the bar on a short timescale . ( 2 ) for a given mass , dense objects cause the greatest reduction in bar amplitude , while significantly more diffuse objects have a lesser effect . ( 3 ) the bar amplitude always decreases as the central mass is grown , and continues to decay thereafter on a cosmological time - scale . ( 4 ) the first phase of bar - weakening is due to the destruction by the cmc of lower - energy , bar - supporting orbits , while the second phase is a consequence of secular changes to the global potential which further diminish the number of bar - supporting orbits . we provide detailed phase - space and orbit analysis to support this suggestion . thus current masses of smbhs are probably too small , even when dressed with a stellar cusp , to affect the bar in their host galaxies . the molecular gas concentrations found in some barred galaxies are also too diffuse to affect the amplitude of the bar significantly . these findings reconcile the apparent high percentage of barred galaxies with the presence of central massive concentrations , and have important implications for the formation and survival of bars in such galaxies .

astro-ph0310194

bars are one of the most prominent morphological features presented by disk galaxies ; eskridge ( 2000 ) find that more than two thirds of disk galaxies are either strongly or weakly barred in the near - infrared band . central mass concentrations ( cmcs ) are also frequently found in galaxies of all types , including barred galaxies . a few examples of cmcs are : * large condensations of molecular gas with scales of @xmath1 kpc and masses of @xmath2 are found in the central regions ( e.g. sakamoto 1999 ; regan 2001 ) . they are particularly evident in barred galaxies , where they are believed to be created by bar - driven inflow ( athanassoula 1992 ; heller & shlosman 1994 ) . * central supermassive black holes ( smbhs ) seem to be a ubiquitous component in spiral galaxies as well as in ellipticals . typical masses are in the range of @xmath3 , and there is a loose correlation with bulge luminosity ( magorrian 1998 ) , suggesting the mass of the smbh is roughly @xmath4 , where @xmath5 is the mass of the bulge . recent measurements with improved accuracy constrained the typical mass of a smbh to be about @xmath6 ( ferrarese & merritt 2000 ; gebhardt 2000 ; tremaine 2002 ) . furthermore , a black hole is thought to be surrounded by a stellar cusp ( young 1980 ) , which augments the effective mass of a smbh for the purpose of large - scale morphological changes . * dense star clusters are found near the centers of many spiral galaxies ( carollo 2003 ; walcher 2003 ) , which are young , very compact ( a few to up to 20 pc ) and relatively massive ( @xmath7 ) . for the purposes of this paper , a cmc is any sufficiently large mass at the center of galaxy that is likely to have a dynamical effect on the evolution of its host galaxy . in other words we focus on the dynamical consequences of such objects regardless of their nature . studies of single - particle dynamics in rotating bar potentials with a cmc ( hasan & norman 1990 ; hasan , pfenniger & norman 1993 ) , and some limited @xmath0-body simulations ( see below ) have given rise to a general belief that a cmc can weaken or dissolve the bar . yet cmcs appear to coexist with bars in many spiral galaxies . in order to determine whether this presents a genuine paradox , we need to know how massive a cmc is needed to destroy the bar completely and on what timescale . the best way to address these questions is with fully self - consistent @xmath0-body simulations . unfortunately , different papers in the literature give apparently discrepant results . the simulations by norman , sellwood & hasan ( 1996 ) showed that a 5% mass can cause the bar to dissolve on a dynamical timescale . that by friedli ( 1994 ) , which included both stars and gas , indicated that objects with 2% of the disk mass could dissolve the bar on the time scale of about 1 gyr . hozumi & hernquist ( 1998 ) , who employed a 2-d self - consistent field ( scf ) method , found that black holes with 0.5 to 1% of the total disk mass are sufficient to weaken the bar substantially within a few gyrs . berentzen ( 1998 ) found , from a single experiment , that the gas inflow driven by a stellar bar concentrates a gas mass some 1.6% of that of the galaxy into the center , which causes the bar to decay on a timescale of 2 gyrs . bournaud & combes ( 2002 ) also found that bars are very fragile . there are a number of reasons to regard these results and their implications as tentative . first , the high frequency of bars in galaxies suggests they can not be destroyed with any great efficiency ( miller 1996 ) . indeed , the massive gas content of the central regions of some barred galaxies ( sakamoto 1999 ; regan 2001 ) , as large as a few @xmath8 in some cases , suggest that the destruction of bars by cmcs is less efficient than claimed . second , large discrepancies between the different experimental results suggest possible numerical problems with the simulations or misunderstood implications . the simulations to study these processes are highly challenging numerically : there is a wide range of timescales on which the particles need to be integrated , and a large number of particles is also very important in order to make the mass of an individual particle much smaller than the central mass and to minimize poisson noise . moreover , none of the previous studies presented a _ systematic _ exploration of the parameter space relevant to the evolution of bars . in particular , most restricted themselves to varying the mass of central objects , and unfortunately did not pay close attention to the `` compactness '' of the cmc , which we show here to be almost as important . studies of bar - forming mechanisms ( sellwood 2000 ) and secular evolution of barred galaxies are seriously hampered by our inadequate understanding of the influence of cmcs on the bar . the existence of cmcs in early stages of galaxies makes this problem even more urgent . there is mounting evidence to indicate that massive black holes grow very rapidly in their early stages , some reaching @xmath9 at @xmath10 ( fan 2001 ; vestergaard 2004 ) . our main motivation of this work is to verify ( or otherwise ) previous results , and to determine the timescale on which the bar is weakened by a cmc , the critical mass of the cmc which causes rapid bar dissolution , and other parameters vital to the bar - weakening process by cmcs . we have carried out a systematic investigation with extensive high - quality simulations . our aim in this study , is not to attempt perfect realism , but rather to study the dynamics of bar destruction in isolation from other possible evolutionary behavior . in outline , our overall strategy is to first construct a model galaxy which is bar unstable . we follow its evolution until the bar rotates at nearly constant pattern speed ( rotation rate ) and with a roughly constant amplitude , in the absence of a cmc . long after the bar has settled , we grow a cmc from zero to its final mass within a certain growth time , and study how the bar is weakened in response to the cmc growth . we learn how the bar is affected by some parameter , such as the growth time of the cmc , from a series of simulations as we change that parameter while keeping the rest fixed . in this work we mainly study two initial bars formed in different ways , which we name weak and strong initial bars because of their relative strengths . we create the weak initial bar simply by evolving a kuzmin - toomre ( k - t ) disk , which has the surface density distribution @xmath11 here @xmath12 is the cylindrical radius and @xmath13 is a length scale . we select disk particles from a distribution function designed to yield the toomre stability parameter @xmath14 ( kalnajs 1976 ; athanassoula & sellwood 1986 ) . we truncate the k - t disk at @xmath15 , spread the particles vertically as the isothermal sheet with a locally - defined equilibrium vertical velocity dispersion , and embed it in the halo ( see [ sec : halo ] ) . the values of the parameters used in the initial setup are listed in table 1 . we allow the model , which is globally unstable , to form a bar . some time after the bar develops , the well - known buckling instability ( toomre 1966 ; raha 1991 ; sellwood & wilkinson 1993 ) thickens the bar . after these events , we obtain a disk containing a long - lived bar having moderate strength , which we designate our weak initial bar . in order to create a strong initial bar , we add fresh particles on circular orbits in the disk mid - plane to a similar weak bar model , which started from a slightly thicker initial disk , so that the buckling instability was milder . we adopt one of the addition rules described in sellwood & moore ( 1999 ) , selecting the angular momentum @xmath16 of each new particle from a gaussian distribution with mean @xmath17 ( in the units described in the next paragraph ) and dispersion @xmath18 ( also see table 1 ) ; empirically , we find these parameters cause the bar to strengthen . this particle - adding process mimics the gradual growth of the disk in a simplified way . we adopt @xmath13 and @xmath19 as our units of length and mass , respectively , and our time units are therefore dynamical times @xmath20 . from here on all quantities are expressed in units such that @xmath21 unless otherwise noted . these units can be scaled to physical values as desired ; we adopt one possible scaling , choosing @xmath22 and @xmath23kpc , which implies a unit of time of roughly @xmath24yr . we mimic a cmc as a plummer sphere for simplicity , its potential is given by : @xmath25 where @xmath26 is the cmc mass . we regard the scale - length of the cmc , @xmath27 ( or softening length ) , as a physically interesting parameter , since it controls the compactness of the cmc . for example , we use a very small @xmath27 ( @xmath28 a few pc , much less than the influence radius of a smbh with about one percent of galactic mass ) for a `` hard '' cmc to mimic a smbh and relatively large @xmath27 ( @xmath28 a few hundred pc ) for a `` soft '' cmc to represent a massive gas concentration . we grow the cmc by increasing its mass according to the relation @xmath29 with @xmath30 , for a cmc introduced at @xmath31 . this sinusoidal time dependence is almost exactly the same as the cubic - type form adopted by merritt & quinlan ( 1998 ) , hozumi & hernquist ( 1998 ) , and others . note that in this work we hold @xmath27 fixed as the cmc mass is grown , in contrast to norman ( 1996 ) who decreased @xmath27 at fixed mass . we adopt @xmath32 for most of our simulations . even though @xmath33 may not be comparable to the rotational period of our initial bars ( @xmath34 or 50 time units ) , we regard this as adiabatic growth because @xmath33 is much longer than the orbital periods of particles near the disk center . we include a rigid halo in most of our simulations , having a potential of the form @xmath35 this choice yields an asymptotically flat circular velocity of @xmath36 for @xmath37 , with @xmath38 being the `` core radius . '' we choose a large core radius , @xmath39 , and @xmath40 . we include this large - core halo as a rigid component instead of a `` live '' one represented by real particles , mainly for computational economy . in [ sec : livehalotest ] we demonstrate that a live halo , with very similar potential to the rigid one , gives nearly same evolution in the bar - destruction process by cmcs . so we are assured that our results are not compromised by this simplification . in most of our simulations , we employ the 3-d cylindrical polar grid - based @xmath0-body code described in detail by sellwood & valluri ( 1997 ) . we solve separately for the three components of the gravitational acceleration and for the potential of the mass distribution using fast fourier transforms ( ffts ) in the vertical and azimuthal directions and by direct convolution in the radial direction . the gravitational field at a distance @xmath41 from a unit mass follows the standard ( plummer sphere ) softened potential @xmath42 where the softening length , @xmath43 , is a constant . we generally employ more than one million particles in our runs ; most numerical parameters adopted are listed in table 1 . we advance the motion of each particle using simple time - centered leap - frog in cartesian coordinates . a fixed time step for all particles is impractical because of the wide range of time - scales , and we vary the time step length in two separate ways . our fundamental time step , @xmath44 , defines the interval between successive determinations of the gravitational field of all the particles . we divide the computation volume into 4 zones and advance particles in the outer zones using longer time steps ; the spherical zone boundaries are at @xmath45 , @xmath46 , and @xmath47 , and time steps are doubled in each successive zone . contributions to the gravitational field from the outer zones to accelerations near the center are computed from time - interpolated estimates of the slowly - changing outer mass distribution ( sellwood 1985 ) . we also implement `` guard shells '' ( see appendix [ app : guard ] ) around a cmc in which the most rapidly moving particles are integrated on shorter time steps ; the shortest step can be as small as @xmath48 for hard cmcs . since the gravitational field inside the tiny guard shells is dominated by the cmc , we do not update the field of the mobile particles during these sub - steps . for test runs with a live halo , we use a hybrid pm scheme ( appendix b of sellwood 2003 ) , in which the self - gravity of the disk is computed on a high - resolution cylindrical polar grid while that of the halo is computed using a surface harmonic expansion on a spherical grid . the code conserves total energy and angular momentum to a satisfactory degree in these challenging experiments . in our fiducial run ( [ sec : fiducial ] ) , @xmath49 and @xmath50 over 100 time units , once cmc growth has ceased . our estimator of bar strength is @xmath51\right|\ ] ] which is the relative amplitude of the bisymmetric ( @xmath52 ) fourier component of the mass density averaged over certain inner radial range where the bar dominates . here @xmath53 is the coordinate of each of the @xmath0 particles in the radial range @xmath54 in the simulation at time @xmath55 . our strong initial bar has a greater radial extent , and therefore @xmath56 underestimates its true amplitude , but we have held this radial range fixed for both bars in order to minimize differences between the quantities measured ; we are most interested in changes to @xmath56 and not its absolute value . we also measure the _ ellipticity _ , @xmath57 , where @xmath13 and @xmath58 are respectively the semimajor and semiminor axes of the bar . we estimate @xmath59 by fitting iso - density contours using the image reduction and analysis facility ( iraf ) to smooth projected density profiles constructed from the particle distribution using an adaptive kernel scheme ( silverman 1986 ) . observers generally use the maximum value of ellipticity as the measure of bar amplitude ( laurikainen , salo & rautiainen 2002 ) . however , the maximum ellipticity in our simulations sometimes fluctuates due to complications such as spiral waves . we therefore use the ellipticity measured at certain reasonably - chosen fixed semi - major axis , which yields a measure with much smaller fluctuations . laurikainen , salo & rautiainen ( 2002 ) have shown that the ellipticity of a bar correlates quite well with the maximum relative tangential force @xmath60 ( buta & block 2001 ) , another indicator of bar strength , in the bar region . we find that the two estimators of bar strength , @xmath61 and @xmath59 , behave in similar ways as the cmc grows ( see [ sec : fiducial ] ) . we therefore generally rely on @xmath61 only , since it is much easier to calculate from an @xmath0-body simulation . we first present the result of a fiducial run in [ sec : fiducial ] , then describe results from our systematic exploration of parameter space ; checks of our simulations are discussed in [ sec : tests ] . figure [ fig : snapshots ] illustrates how a typical barred model reacts as a cmc with mass @xmath62 and scale - length @xmath63 is grown according to eq . ( [ eqn : bhmass ] ) with @xmath64 . figure [ fig : snapshots](a ) shows snapshots of particle positions and figure [ fig : snapshots](b ) the corresponding iso - density contours , obtained by smoothing every particle with an adaptive kernel . this experiment starts with the strong initial bar , which rotates steadily with period @xmath65 at constant amplitude . the cmc starts to grow at the first time shown ( @xmath66 ) in figure [ fig : snapshots ] , and its mass rises to @xmath67 by @xmath68 . the bar is weakened by the cmc , but not destroyed , and retains a moderate strength to the last time shown . the heavy solid line in figure [ fig : ell_t ] represents @xmath69 , the evolution of the amplitude defined in ( [ eqn : adef ] ) , for our fiducial run . the bar weakens rapidly as the cmc mass rises , but the bar amplitude decreases much more slowly after @xmath68 . figure [ fig : ell_t ] also shows the time evolution of the ellipticity @xmath59 . the striking similarity of the overall evolutionary trends of these two estimators assures us that @xmath61 is indeed a good bar - strength indicator . the scale - length of this cmc is roughly two orders of magnitude smaller than the influence radius ( @xmath70 ; here @xmath71 is the flat central value of the velocity dispersion before the cmc is grown ) of a smbh with the same mass , so this cmc can be regarded as a smbh - type object . however , its mass is of course much greater than typical smbh masses in spiral galaxies , which are about @xmath6 ( ferrarese & merritt 2000 ; gebhardt 2000 ; tremaine 2002 ) . our intention here is to demonstrate that a bar can survive even with such an extraordinarily massive hard cmc . figure [ fig : a - tgrowth ] shows how the bar amplitude evolves for different @xmath33 ; the uppermost line indicates that the bar amplitude stays roughly constant in a comparison run with no central mass . the oscillations visible around @xmath72 and @xmath73 are due to interference by spiral features just outside the bar region that have different pattern speeds from that of the bar . they do not cause any lasting change to the bar amplitude . the other lines show that bar is rapidly weakened as the central mass grows . shortly after the cmc mass reaches its maximum value , the bar amplitude @xmath61 decays on a much longer timescale for this fixed central mass . the transition between the two trends becomes less sharp for large @xmath33 ( @xmath74 ) . thus , we confirm , as found in previous work ( e.g. norman 1996 ; hozumi & hernquist 1998 ) , that that @xmath33 is not very important in determining the final bar amplitude long after cmc growth , even when the mass is introduced impulsively and not adiabatically . we use @xmath64 in all subsequent work . the importance of the compactness ( the scale - length ) of cmcs is explicitly shown in figure [ fig : a - ecmc ] . the bar amplitude in figure [ fig : a - ecmc ] is measured at a fixed time , long ( @xmath75 time units ) after the cmc starts to grow . the final @xmath76 for all runs in figure [ fig : a - ecmc ] . the solid and dashed curves represent experiments with weak and strong initial bars , respectively . both types of initial bar show a similar trend : _ hard cmcs are much more destructive to bars than are softer ones_. note that the trend of bar amplitude with @xmath27 is flattening out as the cmc becomes very hard . the typical sizes of molecular gas concentrations in the centers of galaxies range from a few hundred parsecs to 2 kpc or so ( e.g. sakamoto 1999 , regan 2001 ) , which corresponds to @xmath77 in our simulation units . thus diffuse gaseous cmcs cause much less damage to bars than would equally massive but significantly more compact counterparts such as smbhs . figure [ fig : a - mcmc ] shows the final bar amplitude at the end of simulations ( @xmath78 gyrs after cmc growth ) as a function of @xmath79 , for both types of initial bar . this figure indicates that the final bar amplitude decreases continuously as @xmath79 is increased , and again we see that hard cmcs cause much more damage to bars than do soft ones ( also see figure [ fig : a - ecmc ] ) . for both initial bars , we find that the bar is effectively destroyed when the mass of a hard ( @xmath63 ) cmc is greater than 4% or 5% of @xmath80 , whereas soft ( @xmath81 ) cmcs as massive as @xmath82 do not destroy the bar within @xmath78 gyr ( 500 time units ) . these simulations are technically very challenging . here we report a few of our many tests to check that our main results and conclusions are reliable . some of our findings might account for the discrepant results obtained by other groups . figure [ fig : basictests ] presents tests ( for the weak initial bar ) of the number of particles @xmath0 , grid size and particle softening , respectively . we have conducted a similar suite of tests for the strong initial bar , with similar results . the amplitude difference shown in the top panel for the two experiments with different @xmath0 starts and remains very small ( it is hard to achieve precisely the same bar parameters ) . there is a mild dependence on particle softening , slightly more pronounced for the strong initial bar , which we discuss below . these figures confirm that variations of particle number and grid size by a factor of two or more around our adopted values make negligible difference to our main results . we do , however , find that the evolution is sensitive to the adopted time step @xmath44 , and one obtains more rapid , but erroneous , bar destruction by cmcs if the orbit integration is not handled with sufficient care . if a single time step is used over the whole simulation domain , @xmath44 needs to be as small as @xmath83 to ensure that all particles are integrated accurately ( see appendix [ app : guard ] ) , for a typical simulation containing a hard central mass with @xmath62 and @xmath63 . time steps this short are , of course , too expensive for practical use , which is the reason for the multiple zones and guard shell schemes described above . figure [ fig : timesteptest ] shows the effect of changing the time step while all other parameters are held fixed . when no guard shells are employed , the bar amplitude at a fixed time becomes weaker as the time step is increased ; there is no indication of convergence even for very small @xmath44 . the bar amplitude is always greater when guard shells are employed , and is less sensitive to the particular choice of the basic @xmath44 , since we divide it even more finely as @xmath44 is increased . even with guard shells , the final amplitude does increase mildly as the time step is reduced . since the severity of the time - step problem is greater for harder and more massive cmcs , it requires great care to determine the threshold for bar destruction . figure [ fig : livehalotest ] shows , for one case , that the bar amplitude evolution is hardly affected when the rigid halo is replaced by a similar live pseudo - isothermal halo with a large core resembling our rigid form ( eq . [ eqn : halopotential ] ) . thus , our approximation of a rigid halo appears to be adequate . while not a parameter test , we report here that a _ dense _ or _ cuspy _ live halo stimulates the growth of a bar , as noted by debattista & sellwood ( 2000 ) and discussed in more detail by athanassoula & misiriotis ( 2002 ) and athanassoula ( 2003 ) . thus , a bar in a denser or cuspy halo is _ even harder _ to destroy , which underscores one of our main conclusions that bars are very robust . we are unable to understand all discrepancies with other experiments in the literature because many authors give insufficient information . we believe the parameters adopted by norman ( 1996 ) were adequate , and their results are largely consistent with ours . we suspect inadequately short time steps could be a factor in accounting for the rapid bar decay claimed by some groups . for example , the shortest time step ( 0.4 myr ) used in bournaud & combes ( 2002 ) is equivalent to @xmath84 in our units , which our figure [ fig : timesteptest ] suggests is dangerously long . but an inadequate time step may not be able to account for all the discrepancies . for example , hozumi & hernquist ( 1998 ) used a timestep @xmath85 which should be adequate for @xmath86 , but they found the bar to be quite fragile , inconsistent with our findings here . we speculate that the difference might be due to the scf ( self - consistent field ) simulation method they used , in which the gravitational field is computed from a rather low - order expansion in some set of basis functions . the method therefore raises a severe restriction on the complexity of shapes which their models could support . in particular such simulations may be unable to sustain a shape which is axisymmetric near the center while remaining non - axisymmetric in the outer parts , which might spuriously hasten the decay of bars . a nave explanation for the robustness of a bar against the central mass is that the orbits of most bar - supporting particles are little affected by the introduction of a light , dense central mass . this is because most bar - supporting orbits are loops that naturally tend to avoid the center of the bar due to the coriolis force in a rotating frame ( sparke & sellwood 1987 ; hasan & norman 1990 ; sellwood & wilkinson 1993 ; etc . ) . contopoulos has designated this main family of bar - supporting orbits as @xmath87 orbits ( see contopoulos & grosbl 1989 for a review ) , but they are also sometimes described as b - orbits . a more careful examination of figure [ fig : a - tgrowth ] reveals an intriguing phenomenon : the bar is generally weakened in two phases , divided by the time at which the cmc mass growth stops , which are easily distinguishable unless @xmath33 is extremely long ( @xmath88 ) . the first phase , which coincides with the cmc mass - growing period , is clearly controlled by @xmath33 , except when it is very abrupt . thus the response is effectively instantaneous because typical dynamical times of particles near the cmc are short in comparison with @xmath33 . the second phase of bar weakening is on a much longer time scale ( @xmath89 hubble time ) and begins right after the period of cmc mass growth . why is the bar weakened by the cmc in two distinct rates ? we suggest that the first phase of bar weakening is due mainly to the rapid scattering of @xmath87 particles with lower energies . , energy is not conserved , but the jacobi constant @xmath90 is constant along on an orbit ( binney & tremaine 1987 ) . in this paper we loosely refer to the jacobi constant @xmath91 as the energy . ] to support this suggestion , we present a study of orbits in the mid - plane ( @xmath92 ) of the bar . one of the more powerful tools of orbit analysis is the surface of section ( sos ) which can be constructed at any accessible energy . we adopt cartesian coordinates in the rotating frame of the bar , with the major axis of the bar aligned with the @xmath93-axis , and integrate the orbits of test particles in a steadily rotating bar potential from some moment in the simulation . a sos records the values of @xmath94 for a number of orbits all having the same energy , each time they cross the minor ( @xmath95 ) axis with @xmath96 . particles in prograde motion in the frame of the bar have @xmath97 when @xmath96 , and particles in retrograde motion have @xmath98 when @xmath96 . an invariant curve , a series of points all lying on a closed curve in the sos , is produced by a regular orbit that conserves a second integral in addition to the energy . irregular , or chaotic , orbits conserve only one integral ( @xmath91 ) and fill an area in this plane . the adopted energy of the sos determines the maximum possible excursion , @xmath99 , on the bar minor - axis , which is the parameter we prefer to specify . figure [ fig : sos_examples ] shows the sos at two energies for the potential of the rotating bar at @xmath66 before the cmc is grown . the series of nested invariant curves on the left arises from the retrograde ( @xmath100 ) orbit family , which is not very eccentric even in this strongly non - axisymmetric potential . because of their round shape , @xmath100 orbits do not support the bar and a self - consistent bar can contain rather little mass moving on orbits of this family . the other series of nested invariant curves on the right of figure [ fig : sos_examples ] is of much greater importance . these regular prograde orbits are members of the @xmath87 family , and are widely believed to form the backbone of the bar . figure [ fig : x1_examples](a ) shows three examples of @xmath87 orbits selected from figure [ fig : sos_examples](a ) , showing clearly that many particles on lower - energy @xmath87 orbits pass very close to the center ; such orbits will be strongly affected by the introduction of even a low - mass , but dense , cmc . regular @xmath87 orbits selected from the higher - energy sos ( figure [ fig : sos_examples]b ) , on the other hand , are larger ( figure [ fig : x1_examples]b ) and generally avoid the central region . figure [ fig : sos_evo ] shows how the sos at lower energies changes as the cmc in our fiducial run grows , corresponding to the first phase of bar weakening . the breakdown of the large regular areas once occupied by the @xmath87 family is very striking . more and more bar - supporting particles become chaotic , as the growing central mass affects orbits that pass at ever increasing distances . since chaotic orbits tend to have more nearly round long - time average shapes , it is clear that the bar is weakened in this phase because many regular bar - supporting orbits become chaotic . since the orbital periods of particles near the center are much shorter than @xmath33 , low - energy orbits adjust essentially instantly to the new potential and the rate of change of bar amplitude tracks the cmc growth . as the orbits of particles change , the global potential of the system evolves secularly , an effect that continues even after the cmc has reached its final mass . this structural adjustment is illustrated in figure [ fig : ymax1.4sos_evo ] for the fiducial run ( @xmath63 , @xmath76 ) , which shows the slow evolution after @xmath33 . the sos at @xmath101 keeps evolving even after the cmc has reached full mass at @xmath68 . the figure clearly shows that the area occupied by regular @xmath87 orbits at this lower - energy gradually diminishes and the @xmath102 orbit family , which does not support bar , becomes increasingly important among the prograde orbits at this energy level . the potential of the system also tends to favor @xmath87 orbits that are somewhat fatter in shape ( figure [ fig : fatter_x1 ] ) at higher energies , since the once - strong bar has now been weakened . based on figure [ fig : a - tgrowth ] , this structural adjustment happens on a very long timescale , @xmath89 hubble time . however , there is no reason to expect that the rate of this second - phase bar - weakening should remain constant over long time , in fact it becomes slightly slower during the course of our simulation ( figure [ fig : a - tgrowth ] ) . we also notice that the second phase bar decay rate is somewhat slower for the weak initial bar than for the strong initial bar , suggesting that the rate of secular structural change differs from case to case . note that this second phase of bar weakening could not be observed in studies of single particle dynamics in a rigid potential . it happens because the global potential evolves _ self - consistently _ , which requires an @xmath0-body simulation . furthermore , we find that the area occupied by regular @xmath87 orbits shrinks rapidly for @xmath103 when the potential is allowed to evolve , which could be achieved only by further increases in @xmath79 if the bar strength is fixed ( also see hasan & norman 1990 ) . we find that two - phase bar destruction occurs only for compact cmcs of modest mass ( @xmath104 ) . a more massive cmc causes a larger fraction of @xmath87 orbits to become chaotic in the first phase , causing the second - phase structural adjustment to also happen very rapidly . thus the self - consistent bar can no longer withstand the effect of the massive cmc and starts to dissolve this idea may explain why the critical mass is about the same for both strong and weak initial bars . the fact that harder ( more compact ) cmcs cause more damage to bars than do softer ( more diffuse ) ones can also be explained qualitatively in our picture . the weaker forces from a soft cmc might be expected to introduce less chaos . this expectation is suported by the evidence in figure [ fig : sos_cmc_softening ] , which compares the sos for orbits of similar size in the presence of hard and soft cmcs . figure [ fig : sos_cmc_softening](a ) shows the sos at @xmath105 at @xmath72 for the fiducial run with a hard ( @xmath106 ) cmc , while figure [ fig : sos_cmc_softening](b ) shows the same for a very diffuse ( @xmath107 ) cmc . the area of @xmath87 orbit family in figure [ fig : sos_cmc_softening](b ) is much larger than the corresponding region in figure [ fig : sos_cmc_softening](a ) ; similar sos plots at other energies also show the same result . this is in agreement with the finding by gerhard & binney ( 1985 ) that the chaotic region in the sos increases as the central mass is hardened . our picture may also be able to account for the erroneously faster bar decay resulting from inaccurate orbit integration ( [ sec : timesteptest ] ) . if a bar - supporting @xmath87 orbit is not followed very accurately as it approaches a cmc , values of integrals may change for numerical reasons , and a particle on a regular @xmath87 orbit could become chaotic . the enhanced loss of @xmath87 particles will lead to structural adjustment , making the inverse of this change , a chaotic orbit converting to a regular one , less likely . thus such errors can only accelerate bar decay . figure [ fig : a - tgrowth ] also shows that the bar amplitude at the end of cmc growth does not have the same value for the different @xmath33 , implying that secular structural change may have started the moment the cmc starts to grow . figure [ fig : basictests](c ) showed a mild dependence of the final bar amplitude on particle softening , indicating that the decay rate in the second - phase is very slightly accelerated as softening is decreased . a slight increase in the rate of collisional relaxation is the most natural explanation but we stress that the effect is very minor even though the softening length changes by a factor of five . such a weak dependence indicates that the models are indeed almost collisionless . since 3-d orbit dynamics is complicated and very difficult to visualize , we do not wish to give a detailed discussion here . the preceding discussion concerned only those orbits confined to the @xmath92 plane of the 3-d potential . this could be misleading , since many cases are known ( pfenniger 1984 ) where 2-d orbits are irregular while 3-d orbits , with projected shapes very similar to those in 2-d , may be regular . we therefore need to show that allowing 3-d motion does not change our arguments , especially in regard to the importance of chaos . the sos in figure [ fig : sos_evo](a ) , with no cmc , is a nice example of a strong bar that is supported by a large contribution from regular 2-d @xmath87 orbits , yet it was constructed from a 3-d bar . it therefore seems likely that the main bar - supporting orbits are not drastically altered when 3-d motion is allowed . pfenniger & friedli ( 1991 ) show that @xmath87-like orbits in 3-d are generally 2:2:1 resonant , that is they bob up and down about the mid - plane twice for each turn around the center in the rotating frame of the bar . skokos , patsis & athanassoula ( 2002a , b ) give a more comprehensive discussion . since these orbits are simple 3-d generalizations of the @xmath87 orbit family that sustains 2-d bars , we expect the essentials of the phase space structure we find to carry over to 3-d bars . we also need to show , however , that the evidence for chaos in our 2-d sos is in fact accompanied by chaos in the full 3-d potential . in order to address this issue , and to complement our 2-d sos study , we have carried out a floquet analysis ( binney & tremaine 1987 ) . the existence of chaos is indicated by a positive lyapunov exponent , which is defined as the @xmath59-folding divergence rate of two orbits starting with an infinitesimal separation in phase space . valluri & merritt ( 2000 ) show that lyapunov exponents in stellar systems are not good predictors of regular or chaotic orbits , and suggest that a more practical criterion is whether the exponential divergence _ continues _ until the phase - space separations reach the system size . as an example , figure [ fig : lya_ex ] shows the evolution of phase - space separation from a very small value for three orbits ; the left - hand panel ( a ) shows the case when motion continues to be confined to the @xmath92 plane , while the right - hand panel ( b ) shows cases in which full 3-d motion is allowed . the latter orbits start with the same @xmath108 position , but move away from the mid - plane because the potential is not perfectly reflection symmetric . the two orbits in each panel for which the separation quickly reaches the system size are clearly chaotic while the orbit for which the separation remains at some value much smaller than the system size for a long time ( @xmath109gyr ) is regular , according to the criterion proposed by valluri & merritt ( 2000 ) . thus for these three examples , at least , the ( ir)regular nature is unchanged when 3-d motion is permitted . in order to know whether a region that appears to be chaotic in 2-d remains fully , or mostly , chaotic when 3-d motion is allowed , we examine the phase - space separation of many pairs of orbits started at random in the region in question . we focus on two 3-d potentials at @xmath110 and @xmath73 , since both show large areas of chaos at some energy in the 2-d sos plots ( figure [ fig : sos_evo](e ) for @xmath111 at @xmath110 and figure [ fig : ymax1.4sos_evo](f ) for @xmath112 at @xmath73 ) . for 500 orbits , we choose the initial @xmath113 at random from the range @xmath114 for @xmath110 and @xmath115 for @xmath73 , the initial @xmath116 , and the initial @xmath117 is one of five values between @xmath118 . for velocities , we select @xmath119 , @xmath120 and @xmath121 randomly to satisfy the selected @xmath91 . from the criterion for a chaotic orbit described above ( figure [ fig : lya_ex ] ) , we find that about 90% of our randomly - chosen orbits are chaotic for @xmath111 at @xmath110 and about 80% are chaotic for @xmath112 at @xmath73 . thus this 3-d test confirms that most of the realistic 3-d orbits are chaotic , consistent with expectation of the 2-d sos results . we therefore contend that a 2-d sos indeed provides a simplified , yet useful , way to picture the underlying 3-d orbital structure in our study . the experiments described here are rather artificial . we simply grow central masses in a pre - existing bar in order to learn how the bar is affected by cmcs having a range of properties . our intention was to isolate and understand a single aspect of the dynamics , with as few complications as possible . despite this , and other limitations , it may be reasonable to apply our main results to real galaxies . we have restricted our simulations to collisionless stellar systems only , and have not attempted to include gas . we justify this mainly because simulations with gas are much more complicated and we have shown that it requires great care to get even the simpler collisionless results right . the forced inflow of gas is of considerable importance , and star formation , feedback , etc . , will also affect the accumulation of gas near the center . however , gas mass fractions in galaxies today are thought to be modest , and the mass accumulations in the centers of bars appear not to be large enough to destroy the bars ( as we discuss next ) ; thus our approximation of treating all mobile material as if it were collisionless may capture most of the essential physics relevant to bar survival . we have not included a bulge component in our models , largely because it would increase the number of parameters ; we leave a study that includes realistic bulges for future work . our results therefore apply most directly to small - bulge galaxies ( late - type spirals ) . while smbhs are associated directly with bulges , and not with disks , other types of cmcs are found in late - type galaxies . however , we do not expect the bar dissolution behavior to be substantially affected by the introduction of a modest bulge component because bulges are relatively diffuse objects . even if the bulge density profile has an inner cusp of slope @xmath122 , the enclosed mass rises as @xmath123 from the center , and therefore only a tiny fraction of the bulge mass will behave as a hard cmc . we have studied two bars with different strengths and found consistent results . but the exact properties of a cmc to dissolve a bar may also depend on other details of the initial bar in host galaxies . for example , one might expect long slender bars to be more fragile than short stubby bars ( also see hasan & norman 1990 ) . the pattern speed might also affect the bar dissolution process , since teuben & sanders ( 1985 ) showed that dynamically slow bars are supported by different orbit families than are fast bars . however , it is hard to vary the bar pattern speed only while keeping other properties of the bar unchanged in a self - consistent study . in fact , the pattern speed and length of the bar appear to be closely related ( aguerri , debattista & corsini 2003 ) . it may be more appropriate to express the critical cmc mass as a fraction of the bar mass ( @xmath124 ) , since the outer disk presumably has little effect on the dynamics near the center of a bar . unfortunately we found this impractical , as an accurate measure of @xmath124 alone requires us to separate the bar from the other components , which is hard to do both for simulations and for real galaxies . we have shown that the timescale for bar destruction depends on the mass of the cmc and that smbh - like hard cmcs cause much more damage to bars than do soft ones . the critical smbh mass to destroy bars on short timescale is about 4% or 5% of the disk mass for the bars used in our study . ferrarese & merritt ( 2000 ) , gebhardt ( 2000 ) , and tremaine ( 2002 ) estimated the mass of smbhs found in many galaxies to be @xmath125 of the _ bulge _ mass , and bulges are generally less luminous ( palunas & williams 2000 ) , and probably less massive , than the disk , except for rare cases . thus smbhs in galaxies are much less than this critical value and are therefore too small to pose a significant threat to the survival of bars today . for the same reason , central dense star clusters in some galaxies ( carollo 2003 ; walcher 2003 ) are also incapable of dissolving bars . our finding that bars are more robust than previously thought eliminates any possible paradox presented by the coexistence of smbhs and gas concentrations in the centers of barred galaxies . it is widely believed that bars drive gas into the central regions ( athanassoula 1992 ) , yet our findings indicate they should be able to survive such massive gas accumulations . sakamoto ( 1999 ) and regan ( 2001 ) detected much greater concentrations of molecular gas in barred galaxies than in unbarred ones . this is the opposite of what would be expected from rapid bar destruction by central gas concentrations , since greater accumulations would be observed in unbarred ( _ but previously barred _ ) galaxies , provided the gas concentration survives for some time . das ( 2003 ) find an inverse correlation between the deprojected bar ellipticity ( @xmath59 ) and the ratio of the dynamical mass within the bulge to that within the bar radius ( @xmath126 ) in a small sample of 13 nearby barred galaxies taken from the bima song survey . they argue that @xmath126 is a good indicator , although not a direct measure , of the central mass concentration , and they further suggest that the correlation with @xmath59 may be the consequence of the bar dissolution by cmcs . their correlation needs to be confirmed in a larger sample . in particular , 11 of the 13 sample members are in the range of @xmath127 , while only two have small ellipticities and essentially set the `` tightness '' of their inverse correlation . if we regard the two small-@xmath59 galaxies as unreliable , because of uncertainties in deprojection ( barnes & sellwood 2003 ) , the inverse correlation between @xmath59 and @xmath126 is rather loose . if indeed real , the correlation would provide an important constraint on the bar weakening process by cmcs . for example , the modest diffuse cmcs they report can not cause _ rapid _ bar destruction ( @xmath128 a couple of gyrs ) , otherwise most bars should have been destroyed already . to avoid this conclusion , one would require bars to be rejuvenated in a special way , strong bars must reform in galaxies with smaller cmcs and vice versa . on the other hand , the correlation does not rule out our finding of slow bar dissolution ; bars are weakened to differing extents , more by a large cmc and less by a small counterpart , and then survive at their weakened amplitude for a long time . detailed predictions are impossible , however , because we currently lack a coherent picture of when , and in what sequence , bars and cmcs formed . our results suggest it is unlikely that large fractions of bulges are formed by bar dissolution , because the cmcs observed in galaxies today do not destroy bars . however , the bar weakening process does add some disk material to bulges , since those orbits which become chaotic because of the presence of the cmc fill a more nearly spherical volume near the center of the bar . chaotic orbits which were quite elongated before the introduction of the cmc will contribute to the bulge density to a radius several times that of the usual sphere of influence of a smbh . furthermore , such a process might be more important in young galaxies when disks / bars were smaller and cmcs could have been proportionately more massive . cmcs do not , of course , offer the only means to destroy or weaken bars . in fact , we have shown that neither realistic smbhs nor diffuse molecular gas clouds can dissolve bars with any great efficiency . there are other ways to weaken and destroy bars more efficiently than by growing a cmc . bars can be destroyed in minor mergers ( berentzen 2003 and references therein ) . also sellwood & moore ( 1999 ) found that ongoing spiral activity in the outer disk triggered by newly formed `` cold '' material accreted onto the disk , which is not included in our present simulations , could either complete the destruction of the bar or cause it to grow again ; buckling also weakens bars ( raha 1991 ) . our problem is similar to the question of whether a smbh destroys the large - scale triaxiality of a typical elliptical galaxy . as these objects , which are believed to lack significant figure rotation , are supported mostly by box orbits and not by tubes , as in our case , the dynamics is quite different , yet it is interesting that the survival of triaxiality in this case is also not yet settled . holley - bockelmann ( 2002 ) found that triaxiality may be long - lived even with a smbh as massive as @xmath129 , while merritt & quinlan ( 1998 ) found that smbhs with @xmath130 are able to destroy large - scale triaxiality within a hubble time . further investigations in this area would also seem to be warranted . we have conducted a systematic study of the effects of central massive concentrations ( cmcs ) on bars using high - quality @xmath0-body simulations . we have experimented with both strong and weak initial bars and a wide range of physical parameters of cmcs , such as the final mass , scale - length and mass growth time @xmath33 etc . we have demonstrated that our main findings are insensitive to most numerical parameters in our simulation , and shown that the time step requires special care particularly for the cases with compact cmcs . we find that , for a given mass , compact cmcs ( such as supermassive black holes ) are more destructive to bars than are more diffuse ones ( such as molecular gas clouds in many galactic centers ) . we have shown that the former are more efficient scatterers of bar - supporting @xmath87 orbits that pass close to the center , therefore decrease the number of regular @xmath87 orbits and increase the size of the chaotic region in phase space . a bar is generally weakened by a cmc in two connected phases . the bar strength decreases rapidly as the central mass grows due to the rapid scattering of stars on lower - energy bar - supporting @xmath87 orbits as they pass close to the cmc . the time scale for this first phase is therefore the orbital period of the stars in the bar or the growth rate of the central mass , whichever is the longer . the bar continues to decay thereafter on a cosmological timescale ( @xmath89 hubble time for a compact cmc with @xmath131 ) . the second phase reflects slow evolution of the gravitational potential causing a gradual loss of bar - supporting @xmath87 orbits . bars are more robust against cmcs than previously thought : the central object , even for the most destructive smbh - like cmcs , has to be as massive as a few percent of the disk mass to destroy a bar completely within a hubble time . on the other hand , diffuse cmcs need a tremendous amount of mass ( @xmath132 ) to achieve the same effect . our findings clearly show that neither typical smbhs in spirals ( @xmath133 ) nor typical central molecular gas concentrations ( mass @xmath134 a few percent of @xmath80 , scale @xmath135 a few hundred pcs ) can have any significant weakening effect on the bar within a hubble time the former are generally not massive enough , whereas the latter are too diffuse . thus , our results can naturally account for the coexistence of cmcs and bars in many spiral galaxies . we thank the anonymous referee for a thoughtful report and for suggesting further examination of the 3-d `` chaos '' question . this work was supported by nsf grant ast-0098282 . aguerri , j. a. l. , debattista , v. p. , & corsini , e. m. 2003 , , 338 , 465 athanassoula , e. 1992 , , 259 , 345 . 2003 , , 341 , 1179 athanassoula , e. , & misiriotis , a. 2002 , , 330 , 35 athanassoula , e. & sellwood , j. a. 1986 , , 221 , 213 barnes , e. i. , & sellwood , j. a. 2003 , , 125 , 1164 berentzen , i. , athanassoula , e. , heller , c. h. , & fricke , k. j. 2003 , , 341 , 343 berentzen , i. , heller , c. h. , shlosman , i. , & fricke , k. j. 1998 , , 300 , 49 binney , j. , & tremaine , s. 1987 , galactic dynamics ( princeton : princeton univ . press ) bournaud , f. , & combes , f. 2002 , , 392 , 83 buta , r. , & block , d. l. 2001 , , 550 , 243 carollo , c. m. 2003 , carnegie observatories astrophysics series , vol . 1 : coevolution of black holes and galaxies , ed . l. c. ho ( cambridge : cambridge univ . press ) contopoulos , g. & grosbl , p. 1989 , , 1 , 261 das , m. , teuben , p. j. , vogel , s. n. , regan , m. w. , sheth , k. , harris , a. i. , & jefferys , w. h. 2003 , , 582 , 190 debattista , v. p. , & sellwood , j. a. 2000 , , 543 , 704 eskridge , p. b. , 2000 , , 119 , 536 fan , x. 2001 , , 122 , 2833 ferrarese , l. , & merritt , d. 2000 , , 539 , l9 friedli , d. 1994 , in mass - transfer induced activity in galaxies , ed . i. shlosman ( cambridge : cambridge univ . press ) , p. 268 gebhardt , k. 2000 , , 539 , l13 gerhard , o. e. , & binney , j. 1985 , , 216 , 467 hasan , h. , & norman , c. 1990 , , 361 , 69 hasan , h. , pfenniger , d. , & norman , c. a. 1993 , , 409 , 91 heller , c. h. & shlosman , i. 1994 , , 424 , 84 holley - bockelmann , k. , mihos , c. j. , sigurdsson , s. , hernquist , l. , & norman , c. 2002 , , 567 , 817 hozumi , s. , & hernquist , l. 1998 , astro - ph/9806002 kalnajs , a. j. 1976 , , 205 , 751 laurikainen , e. , salo , h. , & rautiainen , p. 2002 , , 331 , 880 magorrian , j. 1998 , , 115 , 2285 merritt , d. , & quinlan , g. 1998 , , 498 , 625 miller , r. h. 1996 , in `` barred galaxies '' asp conference series , volume 91 ( san francisco : astronomical society of the pacific ) ed . by r. buta , d. a. crocker and b. g. elmegreen , p. 569 norman , c. a. , sellwood , j. a. , & hasan , h. 1996 , , 462 , 114 palunas , p. , & williams , t. b. , 2000 , , 120 , 2884 pfenniger , d. 1984 , , 134 , 373 pfenniger , d. , & friedli , d. 1991 , , 252 , 75 raha , n. , sellwood , j. a. , james , r. a. , & kahn , f. d. 1991 , nature , 352 , 411 regan , m. w. , , , 2001 , 561 , 218 sakamoto , k. , okamura , s. k. , ishizuki , s. , & scoville , n. z. 1999 , , 525 , 691 skokos , c. , patis , p. a. , & athanassoula , e. 2002a , , 333 , 847 , 2002b , , 333 , 861 sellwood , j. a. 1985 , , 217 , 127 sellwood , j. a. 2000 , in `` dynamics of galaxies : from the early universe to the present '' , 15th iap meeting held in paris , france , july 9 - 13 , 1999 , eds . : francoise combes , gary a. mamon , and vassilis charmandaris , asp conference series , vol . 197 , p. 3 . 2003 , , 587 , 638 sellwood , j. a. & evans , n. w. 2001 , , 546 , 176 sellwood , j. a. , & moore , e. m. 1999 , , 510 , 125 sellwood , j. a. , & valluri , m. 1997 , , 287 , 124 sellwood , j. a. , & wilkinson , a. 1993 , rep . , 56 , 173 silverman b. w. 1986 , density estimation for statistics and data analysis ( london : chapman and hall ) sparke , l. s. , & sellwood , j. a. 1987 , , 225 , 653 teuben , p. j. , & sanders , r. h. 1985 , , 212 , 257 toomre , a. 1966 , in geophysical fluid dynamics , notes on the 1966 summer study program at the woods hole oceanographic institution , ref . 66 - 46 , p. 111 toomre , a. 1981 , in the structure and evolution of normal galaxies , ed . s. m. fall & d. lynden - bell ( cambridge : cambridge university press ) , 111 tremaine , s. 2002 , , 574 , 740 valluri , m. , & merritt , d. 2000 , in `` the chaotic universe '' , proceedings of the second icra network workshop , advanced series in astrophysics and cosmology , vol.10 , edited by v. g. gurzadyan and r. ruffini , world scientific , p.229 . also see astro - ph/9909403 vestergaard , m. 2004 , , in press , astro - ph/0309521 walcher , c. j. 2003 , carnegie observatories astrophysics series , vol . 1 : coevolution of black holes and galaxies , ed . l. c. ho ( pasadena : carnegie observatories , http://www.ociw.edu/ociw/symposia/series/symposium1/proceedings.html young , p. 1980 , , 242 , 1232 we have devised the following `` guard shells '' scheme in order to improve the accuracy of orbit integration when particles experience strong acceleration near a compact central mass . we divide the volume around the cmc into many concentric shells and successively halve the time step for particles in each shell as they approach the cmc ( figure [ fig : ga ] ) . since accelerations in this small region are dominated by the field of the cmc , we integrate the orbit for these sub - timesteps in a fixed field , and update the self - consistent part from the bar and the disk at the basic time step interval @xmath44 . we choose the outermost guard radius @xmath139 so that @xmath140 is @xmath141 . particles inside @xmath139 , which have shorter periods , require finer time steps . we adopt time steps in adjacent shells that differ by a factor of two , making it appropriate to require @xmath142 . since @xmath143 ( if @xmath144 ) , the ratio of adjacent guard radii is @xmath145^{2/3}= \left ( \frac{1}{2 } \right)^{2/3 } \approx 0.63.\ ] ] we keep this ratio constant for simplicity , although this is somewhat over - conservative for @xmath146 . the shortest possible orbital period is @xmath147 the number of guard shells ( @xmath148 ) , or the boundary of the innermost zone ( @xmath149 , is determined so that @xmath150 say , which we find to be adequate . + initial number of particles & @xmath151 & @xmath152 + final number of particles & @xmath153 + grid size @xmath154 & @xmath155 & @xmath155 + vertical plane spacing & @xmath156 & @xmath156 + grid boundaries @xmath157 & @xmath158 & @xmath158 + particle softening length & @xmath156 & @xmath159 + time step @xmath160 w / o cmc & @xmath161 & @xmath161 + time step @xmath44 w/ cmc & @xmath162 & @xmath162 + number of guard shells + with the fiducial cmc & @xmath163 & @xmath163 + outermost guard radius @xmath164 & @xmath165 & @xmath165 + innermost guard radius @xmath166 & @xmath167 & @xmath167 + + , with other cmc parameters fixed . the solid and dashed curves represent the runs with a hard ( @xmath63 ) and soft cmc ( @xmath81 ) , respectively . ( b ) as for ( a ) , but for the strong initial bar . the final bar amplitude decreases continuously as @xmath79 is increased , and hard cmcs always cause significantly more damage to the bar than do soft ones . the bar is destroyed on short time scales by a hard cmc with a mass a few percent @xmath80 , but the mass of a soft cmc needs to be @xmath175 . ] , @xmath176 and @xmath64 ) . ( a ) two runs with different numbers of particles . note that the initial bars in these two runs start with slightly different @xmath61 . ( b ) two with different grid sizes . the test run with a much finer grid yields almost exactly same result as the run with our standard grid . ( c ) tests with different particle softening from 0.05 to 0.01 , with other physical parameters unchanged . ] the dashed curve shows the zero - velocity curve ( the boundary of the region energetically accessible ) for the energy specified by the limiting distance , @xmath99 , that a particle may reach on the bar minor axis . ( a ) sos for @xmath177 and ( b ) for @xmath178 . ] = 1.0 , 1.2 and 1.4 ( from top to bottom ) after a hard cmc ( @xmath63 , @xmath179 ) is grown from @xmath66 to @xmath180 . column ( a ) shows the soss at @xmath66 , just before the cmc is added . columns ( b ) to ( f ) show the sos at the same values of @xmath181 at times 710 , 720 , 730 , 740 & 750 respectively.,title="fig:"]= 1.0 , 1.2 and 1.4 ( from top to bottom ) after a hard cmc ( @xmath63 , @xmath179 ) is grown from @xmath66 to @xmath180 . column ( a ) shows the soss at @xmath66 , just before the cmc is added . columns ( b ) to ( f ) show the sos at the same values of @xmath181 at times 710 , 720 , 730 , 740 & 750 respectively.,title="fig : " ] = 1.2 ) at @xmath72 for the fiducial run with a hard cmc ( @xmath63 ) on the left ( a ) , and for a similar case with a soft cmc ( @xmath81 ) on the right ( b).,title="fig:"]= 1.2 ) at @xmath72 for the fiducial run with a hard cmc ( @xmath63 ) on the left ( a ) , and for a similar case with a soft cmc ( @xmath81 ) on the right ( b).,title="fig : " ] gyr ) of phase space separation due to a infinitesimal perturbation for the potential at @xmath110 , @xmath111 . ( a ) 2-d case : orbits are confined to the @xmath92 midplane . the corresponding 2-d sos plot is the top panel of figure [ fig : sos_evo](e ) . from top to bottom for both panels , the three curves correspond to initial @xmath113=0.2 , 0.5 and 0.8 , respectively . as for ( a ) , but in this case full 3-d motion is allowed.,title="fig:"]gyr ) of phase space separation due to a infinitesimal perturbation for the potential at @xmath110 , @xmath111 . ( a ) 2-d case : orbits are confined to the @xmath92 midplane . the corresponding 2-d sos plot is the top panel of figure [ fig : sos_evo](e ) . from top to bottom for both panels , the three curves correspond to initial @xmath113=0.2 , 0.5 and 0.8 , respectively . ( b ) . as for ( a ) , but in this case full 3-d motion is allowed.,title="fig : " ]

we study multiple orthogonal polynomials of meixner - pollaczek type with respect to a symmetric system of two orthogonality measures . our main result is that the limiting distribution of the zeros of these polynomials is one component of the solution to a constrained vector equilibrium problem . we also provide a rodrigues formula and closed expressions for the recurrence coefficients . the proof of the main result follows from a connection with the eigenvalues of block toeplitz matrices , for which we provide some general results of independent interest . the motivation for this paper is the study of a model in statistical mechanics , the so - called six - vertex model with domain wall boundary conditions , in a particular regime known as the free fermion line . we show how the multiple meixner - pollaczek polynomials arise in an inhomogeneous version of this model . * keywords * : multiple orthogonal polynomial , meixner - pollaczek polynomial , recurrence relation , block toeplitz matrix , potential theory , six - vertex model .

1101.2982

in this paper we study a system of polynomials , orthogonal with respect to two different weight functions of meixner - pollaczek type . our work is motivated by the analysis of the six - vertex model in statistical mechanics introduced in @xcite and studied in many papers since then , see e.g. @xcite . colomo and pronko @xcite studied the role of orthogonal polynomials in the six - vertex model , and in particular the meixner - pollaczek polynomials . an object of major physical interest is the _ partition function _ of this model . a rather complete analysis of the large @xmath0 asymptotics in the homogeneous case is based on the analysis of a riemann - hilbert problem for orthogonal polynomials @xcite . in the inhomogeneous case , the corresponding reasoning leads to questions of asymptotics of _ multiple orthogonal _ polynomials , as we will discuss in section [ section : sixvertex ] . here we will focus on a particular regime of the six - vertex model , known as the `` free fermion line '' . in the homogeneous case , this case is trivial , the partition function being identically @xmath1 . also in the inhomogeneous case , a closed expression for the partition function can be calculated explicitly , so the present results do not give any new insights into the original model , but should rather be considered as providing exact and asymptotic information in its own right on the associated polynomials , the _ multiple meixner - pollaczek polynomials_. an important tool in the study of ( usual ) orthogonal polynomials @xmath2 on the real line , appropriately scaled in order to have @xmath0-dependent weights , is that their limiting zero distribution satisfies an equilibrium problem . this equilibrium problem is an important ingredient for the steepest descent analysis of the riemann - hilbert problem for orthogonal polynomials , thereby allowing to obtain strong and uniform asymptotics of the polynomials , see e.g.@xcite . for the case of multiple orthogonal polynomials , however , no general result is known about the existence of an equilibrium problem . the aim of this paper is to obtain such an equilibrium problem for multiple meixner - pollaczek polynomials . the equilibrium problem will be posed in terms of a couple of measures @xmath3 and it involves both an external source acting on @xmath4 and a constraint acting on @xmath5 . this structure is very similar to the equilibrium problem for the gue with external source model , @xcite . our method for obtaining the equilibrium problem is similar to that applied to other systems of multiple orthogonal polynomials @xcite , but a distinguishing feature is the characterization in terms of the eigenvalue distributions of block toeplitz matrices , rather than usual ( scalar ) toeplitz matrices . we will provide some general results on this topic that are of independent interest . along the way we also obtain rodrigues formulae and closed expressions for the recurrence coefficients for multiple meixner - pollaczek polynomials . inspired by the scalar case @xcite , one might hope that the equilibrium problem presented in this paper can be used in the steepest descent analysis of the riemann - hilbert problem for multiple meixner - pollaczek polynomials , thereby obtaining strong and uniform asymptotics for these polynomials . this approach , although interesting , will not be carried out in this paper . we also hope that our equilibrium problem might serve as an inspiration to obtain similar results for the general inhomogeneous six - vertex model , not necessarily on the free fermion line , which in turn could serve as the first step in finding the large @xmath0 asymptotic analysis for this model . let @xmath6 and @xmath7 denote two distinct weight functions of meixner - pollaczek type ; @xmath8 with @xmath9 where @xmath10 , @xmath11 and @xmath12 are fixed parameters and @xmath13 denotes euler s gamma function @xcite . note that in the gamma function is evaluated in a complex argument and that @xmath14 furthermore , for @xmath15 fixed , @xmath16 as @xmath17 , so the restrictions on @xmath18 guarantee that @xmath19 is exponentially decaying for @xmath20 . [ lemma : interlacing ] ( existence , uniqueness , real and interlacing zeros ) for any nonnegative integers @xmath21 and @xmath22 , there is a unique monic polynomial @xmath23 of degree @xmath24 satisfying the orthogonality conditions @xmath25 the zeros of these polynomials are real and interlacing , in the sense that each @xmath26 has @xmath27 distinct real zeros @xmath28 such that @xmath29 and @xmath30 whenever @xmath31 . lemma [ lemma : interlacing ] will be proved in section [ subsection : proofinterlacing ] . in analogy with the case of meixner - pollaczek orthogonal polynomials @xcite we refer to the @xmath26 as _ multiple meixner - pollaczek polynomials _ ; see also ( * ? ? ? these polynomials are related to the six - vertex model ; see section [ section : sixvertex ] . for information on other systems of multiple orthogonal polynomials in the literature , see e.g. @xcite . in this paper we will derive a rodrigues type formula for the polynomials @xmath26 , enabling us to compute explicitly the following four term recurrence relations , to be proved in section [ section : recurrence ] . [ theorem : recurrencemmpp ] ( recurrence relations ) let @xmath32 with @xmath33 . then , for nonnegative integers @xmath21 and @xmath22 , the multiple meixner - polleczek polynomials satisfy the recurrence relations @xmath34 and @xmath35 where @xmath36 @xmath37 @xmath38 @xmath24 , and where we set @xmath39 , @xmath40 for any @xmath41 . the above theorem is the basis for the main purpose of the paper , namely to study the asymptotic zero distribution of the appropriately rescaled multiple meixner - pollaczek polynomials . fix @xmath42 and @xmath43 . for simplicity we choose a particular sequence of indices @xmath44 along which we analyze the large @xmath27 asymptotics of @xmath45 , and form the single sequence @xmath46 of polynomials . we will show that the zero - distribution of the rescaled polynomials @xmath47 has a weak limit as @xmath27 goes to infinity , and in order to study this we let @xmath48 and introduce the doubly indexed sequence of monic polynomials @xmath49 defined by @xmath50 from ( [ recurrencemp ] ) and we obtain an explicit recurrence relation for @xmath51 , @xmath52 with initial conditions @xmath53 , where @xmath54 for @xmath27 even and @xmath55 for @xmath27 odd . using the recurrence relation , standard considerations show that the zeros of @xmath56 are the eigenvalues of the @xmath57-diagonal matrix @xmath58 so the problem of the asymptotic zero - distribution of @xmath56 reduces to finding the eigenvalue asymptotics of ( [ jacobi ] ) . if @xmath27 and @xmath0 both tend to infinity in such a way that @xmath59 for some constant @xmath60 , then the coefficients @xmath61 have two subsequential limits @xmath62 and @xmath63 for even and odd @xmath27 respectively . similarly , @xmath64 and @xmath65 have subsequential limits @xmath66 and @xmath67 along subsequences consisting of even and odd @xmath27 respectively . using the identity @xmath68 , these limits become @xmath69 @xmath70 if we consider ( [ jacobi ] ) in @xmath71 blocks , it is tri - diagonal with blocks @xmath72 @xmath73 and @xmath74 in view of the limits and , the blocks are slowly varying along the diagonals if @xmath0 is large . in other words , for large @xmath0 , the matrix ( [ jacobi ] ) has a _ locally block toeplitz _ structure with square blocks of size @xmath75 . this allows its limiting eigenvalue distribution to be obtained from a general machinery to which we turn now . in this subsection we will work in the following general setting . let @xmath0 be a fixed parameter and let @xmath76 be a sequence of monic polynomials , where @xmath51 has degree @xmath27 and depends parametrically on @xmath0 . assume that the @xmath51 are generated by the recurrence relation @xmath77 where @xmath78 is a semi - infinite matrix with unit lower hessenberg structure , i.e. , the strictly upper triangular part of @xmath78 is equal to zero , except for the first superdiagonal , on which all entries are @xmath1 . we also assume that the lower triangular part of @xmath78 has a finite bandwidth , which is independent of @xmath0 . the entries of @xmath78 are assumed to have asymptotically periodic behavior with period @xmath79 ( @xmath80 ) . more precisely , suppose @xmath78 is partitioned into blocks of size @xmath81 , with one superdiagonal and a fixed finite number @xmath82 of subdiagonal non - zero blocks , @xmath83 where @xmath84 where the @xmath85 s denote arbitrary constants . we will assume that the block entries @xmath86 in are slowly varying with @xmath0 , in the sense that the limits @xmath87 exist for any @xmath88 and @xmath89 . here the notation @xmath90 means that we let both @xmath27 and @xmath0 tend to infinity , in such a way that the ratio @xmath91 tends to a limit @xmath89 . the relation is on the level of @xmath81 matrices , with the limit taken entry - wise . if holds then the matrix in is said to have _ locally block toeplitz _ structure , in the spirit of @xcite . for fixed @xmath60 , we collect the limiting matrices in into the following matrix - valued laurent polynomial @xmath92 : @xmath93 we will sometimes refer to @xmath92 as the _ symbol_. from it follows that @xmath94 where @xmath95 denotes all the terms in that tend to zero as @xmath96 . define for each @xmath97 , the algebraic equation @xmath98 where @xmath99 denotes the identity matrix of size @xmath79 . note that @xmath100 depends on two complex variables @xmath101 and @xmath102 . by expanding the determinant and using , we can write @xmath100 as a ( scalar ) laurent polynomial in @xmath101 : @xmath103 where @xmath104 , @xmath105 , are polynomials in @xmath102 , with @xmath106 . we define @xmath107 in as the largest positive integer for which @xmath108 . let us solve for @xmath101 ; this yields @xmath109 roots @xmath110 we assume that for each @xmath111 these roots are ordered such that @xmath112 in the cases we are interested in , we will have that @xmath113 supposing this to hold , we define a measure @xmath114 on @xmath115 with density @xmath116 here the prime denotes the derivative with respect to @xmath102 , and @xmath117 are the boundary values of @xmath118 obtained from the @xmath119-side ( upper side ) and @xmath120-side ( lower side ) respectively of @xmath115 . these boundary values exist for all but a finite number of points . as discussed in @xcite , the measure @xmath114 can be interpreted as the weak limit as @xmath121 of the normalized eigenvalue counting measures for the block toeplitz matrices @xmath122 associated to the symbol . [ lemma : mu0prob ] with the above notation , we have * @xmath123 as @xmath124 . * @xmath114 in is a borel probability measure on @xmath125 . see @xcite . with this notation in place , let us return to the sequence of polynomials @xmath126 in . this sequence is said to have _ real and interlacing _ zeros if each @xmath51 has @xmath27 distinct real zeros @xmath127 such that @xmath128 whenever @xmath31 . the next result states that , under certain conditions , the normalized zero - counting measures of the polynomials @xmath56 have a ( weak ) limit for @xmath121 , which is precisely the average of the measures . here the average is with respect to the parameter @xmath60 . [ theorem : locallytoepeig ] ( limiting zero distribution of @xmath56 ) let the sequence of polynomials @xmath126 be such that hold . assume that @xmath126 has real and interlacing zeros for each @xmath0 , as described above . also assume that holds for every @xmath89 . then as @xmath129 , the normalized zero - counting measure @xmath130 of the polynomial @xmath56 , @xmath131 where @xmath132 denotes the dirac point mass at @xmath101 , has the limit @xmath133 in the sense of weak convergence of measures , where @xmath114 is defined in . theorem [ theorem : locallytoepeig ] will be proven in section [ section : zeroasymptotics ] . this theorem generalizes a result for the scalar case @xmath134 by kuijlaars - romn ( * ? ? ? * theorem 1.2 ) ; see also @xcite . we apply theorem [ theorem : locallytoepeig ] to the polynomials @xmath51 in section [ subsection : mmpp ] . in what follows , we will assume the condition of symmetric weights , @xmath135 for this case we can characterize the limiting zero distribution in terms of a constrained vector equilibrium problem , theorem [ theorem : equilibriumproblem ] , which is the main result of the paper . the symmetry condition in proposition [ theorem : zeroseigenvalues ] is needed to prove that @xmath136 ; in the general case this may fail . the recurrence coefficients in and then become @xmath137 we partition the matrix @xmath78 from into blocks as in with @xmath75 and @xmath138 . then the limiting values in exist and are given by @xmath139 @xmath140 and @xmath141 with @xmath62 , @xmath142 and @xmath143 as in . the symbol @xmath92 in now becomes @xmath144 so reduces to @xmath145 this equation has three roots , @xmath146 , @xmath147 which we label in order of increasing modulus : @xmath148 [ theorem : zeroseigenvalues ] ( limiting zero distribution ) suppose @xmath149 . as @xmath129 , the normalized zero - counting measure @xmath130 of the multiple meixner - pollaczek polynomial @xmath56 in section [ subsection : mmpp ] , see , converges weakly to the average @xmath150 of the measures @xmath114 in . this is a consequence of theorem [ theorem : locallytoepeig ] . the assumptions in the latter theorem are indeed satisfied : the interlacing condition follows from lemma [ lemma : interlacing ] , and the fact that @xmath115 follows from proposition [ monotonoussupports ] . in addition to the set @xmath125 in , we define @xmath151 we will now characterize the limiting zero distribution @xmath4 in terms of a vector equilibrium problem from logarithmic potential theory @xcite . recall that for a pair of borel measures @xmath152 supported in the complex plane , the _ mixed logarithmic energy _ of @xmath153 and @xmath154 is defined as @xcite @xmath155 [ theorem : equilibriumproblem ] ( equilibrium problem ) suppose @xmath149 . then the asymptotic zero distribution , @xmath4 , of @xmath56 is the first component of the unique minimizer @xmath3 of the energy functional @xmath156 among all vectors @xmath152 of positive measures such that @xmath157 , @xmath158 , @xmath159 , @xmath160 and @xmath154 is absolutely continuous with density satisfying @xmath161 the measures have the properties @xmath162\ ] ] and @xmath163 where @xmath164 is the positive , absolutely continuous measure on @xmath165 with constant density @xmath166 , and where @xmath167 and @xmath168 are positive constants given by @xmath169 and @xmath170 with @xmath171 furthermore , @xmath4 is absolutely continuous with respect to lebesgue measure and has density @xmath172,\ ] ] where @xmath173 and @xmath174 is the complex solution to the algebraic equation @xmath175 such that @xmath176 . for ( a ) @xmath177 and ( b ) @xmath178.,title="fig : " ] for ( a ) @xmath177 and ( b ) @xmath178.,title="fig : " ] in general , there is no closed expression for the density of @xmath4 , but in the limiting case @xmath179 , corresponding to the ordinary meixner - pollaczek orthogonal polynomials , ( [ densityrel3 ] ) can be explicitly solved , @xmath180 , giving @xmath181 . thus ( [ densityrel2 ] ) becomes @xmath182 and the density ( [ densityrel3 ] ) takes the form @xmath183 for non - zero values of @xmath184 , the density of @xmath4 turns out to have the same qualitative features as ; see figure [ fig : mp ] for some illustrations . theorem [ theorem : equilibriumproblem ] will be proven in section [ section : equilibrium ] . the proof uses a general result for block toeplitz matrices @xcite , see also @xcite , together with some explicit calculations using . multiple meixner - pollaczek polynomials appear in the study of the six - vertex model in statistical mechanics , as we explain now . consider an @xmath185 square lattice in the plane . a configuration of the six - vertex model is an assignment of an orientation to the edges of the lattice in such a way that each vertex is surrounded by precisely two incoming and two outgoing edges . see fig . [ fig : grid5x5 ] for a configuration with @xmath186 . the name _ six - vertex model _ refers to the fact that the local behavior near each vertex is given by six possible edge configurations ( see fig . [ fig:6vertices ] ) . we consider the six - vertex model with domain wall boundary conditions ( dwbc ) . this means that the edges at the top and bottom of the lattice must be directed outwards and those at the left and right of the lattice must be directed inwards . ( 80,50)(-20,20 ) ( 10,10 ) ( 10,20 ) ( 10,30 ) ( 10,40 ) ( 10,50 ) ( 20,10 ) ( 20,20 ) ( 20,30 ) ( 20,40 ) ( 20,50 ) ( 30,10 ) ( 30,20 ) ( 30,30 ) ( 30,40 ) ( 30,50 ) ( 40,10 ) ( 40,20 ) ( 40,30 ) ( 40,40 ) ( 40,50 ) ( 50,10 ) ( 50,20 ) ( 50,30 ) ( 50,40 ) ( 50,50 ) ( 1,10)(1,0)58 ( 1,20)(1,0)58 ( 1,30)(1,0)58 ( 1,40)(1,0)58 ( , 50)(1,0)58 ( 10,1)(0,1)58 ( 20,1)(0,1)58 ( 30,1)(0,1)58 ( 40,1)(0,1)58 ( 50,1)(0,1)58 ( 5,10 ) ( 5,20 ) ( 5,30 ) ( 5,40 ) ( 5,50 ) ( 55,10 ) ( 55,20 ) ( 55,30 ) ( 55,40 ) ( 55,50 ) ( 10,5 ) ( 20,5 ) ( 30,5 ) ( 40,5 ) ( 50,5 ) ( 10,55 ) ( 20,55 ) ( 30,55 ) ( 40,55 ) ( 50,55 ) ( 15,10)(1,0)1 ( 15,20)(1,0)1 ( 15,30)(1,0)1 ( 15,40)(-1,0)1 ( 15,50)(1,0)1 ( 25,10)(1,0)1 ( 25,20)(1,0)1 ( 25,30)(-1,0)1 ( 25,40)(1,0)1 ( 25,50)(-1,0)1 ( 35,10)(-1,0)1 ( 35,20)(1,0)1 ( 35,30)(-1,0)1 ( 35,40)(1,0)1 ( 35,50)(-1,0)1 ( 45,10)(-1,0)1 ( 45,20)(-1,0)1 ( 45,30)(1,0)1 ( 45,40)(-1,0)1 ( 45,50)(-1,0)1 ( 10,15)(0,-1)1 ( 20,15)(0,-1)1 ( 30,15)(0,1)1 ( 40,15)(0,-1)1 ( 50,15)(0,-1)1 ( 10,25)(0,-1)1 ( 20,25)(0,-1)1 ( 30,25)(0,1)1 ( 40,25)(0,1)1 ( 50,25)(0,-1)1 ( 10,35)(0,-1)1 ( 20,35)(0,1)1 ( 30,35)(0,1)1 ( 40,35)(0,-1)1 ( 50,35)(0,1)1 ( 10,45)(0,1)1 ( 20,45)(0,-1)1 ( 30,45)(0,1)1 ( 40,45)(0,1)1 ( 50,45)(0,1)1 ( 60,49)@xmath187 ( 60,39)@xmath188 ( 60,29)@xmath189 ( 60,19)@xmath190 ( 60,9)@xmath191 ( 48,-1)@xmath192 ( 38,-1)@xmath193 ( 28,-1)@xmath194 ( 18,-1)@xmath195 ( 8,-1)@xmath196 ( 20,20)(0,0 ) ( 10,10 ) ( 10,10)(1,0)8 ( 10,10)(-1,0)8 ( 10,10)(0,1)8 ( 10,10)(0,-1)8 ( 5,10)(1,0)1 ( 15,10)(-1,0)1 ( 10,5)(0,-1)1 ( 10,15)(0,1)1 ( 9,-5)1 ( 20,20)(0,0 ) ( 10,10 ) ( 10,10)(1,0)8 ( 10,10)(-1,0)8 ( 10,10)(0,1)8 ( 10,10)(0,-1)8 ( 5,10)(-1,0)1 ( 15,10)(1,0)1 ( 10,5)(0,1)1 ( 10,15)(0,-1)1 ( 9,-5)2 ( 20,20)(0,0 ) ( 10,10 ) ( 10,10)(1,0)8 ( 10,10)(-1,0)8 ( 10,10)(0,1)8 ( 10,10)(0,-1)8 ( 5,10)(1,0)1 ( 15,10)(1,0)1 ( 10,5)(0,1)1 ( 10,15)(0,1)1 ( 9,-5)3 ( 20,20)(0,0 ) ( 10,10 ) ( 10,10)(1,0)8 ( 10,10)(-1,0)8 ( 10,10)(0,1)8 ( 10,10)(0,-1)8 ( 5,10)(-1,0)1 ( 15,10)(-1,0)1 ( 10,5)(0,-1)1 ( 10,15)(0,-1)1 ( 9,-5)4 ( 20,20)(0,0 ) ( 10,10 ) ( 10,10)(1,0)8 ( 10,10)(-1,0)8 ( 10,10)(0,1)8 ( 10,10)(0,-1)8 ( 5,10)(-1,0)1 ( 15,10)(-1,0)1 ( 10,5)(0,1)1 ( 10,15)(0,1)1 ( 9,-5)5 ( 10,10)(0,0 ) ( 10,10 ) ( 10,10)(1,0)8 ( 10,10)(-1,0)8 ( 10,10)(0,1)8 ( 10,10)(0,-1)8 ( 5,10)(1,0)1 ( 15,10)(1,0)1 ( 10,5)(0,-1)1 ( 10,15)(0,-1)1 ( 9,-5)6 to each of the @xmath197 rows of the lattice we associate a parameter @xmath198 and similarly to each column a parameter @xmath199 , @xmath200 . we also fix a positive parameter @xmath201 and we assume that @xmath202 for all @xmath203 and @xmath204 . we define the weight of the vertex in row @xmath203 and column @xmath204 according to its type as @xmath205 ( type 1 or 2 ) , @xmath206 ( type 3 or 4 ) , or @xmath207 ( type 5 or 6 ) . note that the weights are parameterized according to the so - called disordered phase convention . the weight of a configuration is defined as the product of the weights of all the @xmath185 vertices in the configuration . the partition function @xmath208 is defined as the sum of the weights of all the consistent configurations of the @xmath185 six - vertex model with dwbc . an explicit expression for the partition function in terms of an @xmath185 determinant was found by izergin and korepin @xcite , see also @xcite . in the case where @xmath209 for all @xmath200 , for some fixed parameter @xmath210 , the izergin - korepin formula reduces to @xmath211^{n^2}}{\left ( \displaystyle \prod_{n=0}^{n-1 } n ! \right)^2}\det m,\ ] ] where the matrix @xmath212 has entries @xmath213 with @xmath214 see e.g. @xcite . the matrix @xmath215 is then precisely the _ moment matrix _ corresponding to the weight function @xmath216 on the real line . ( e.g. @xcite ) show that @xmath217 can be expressed in terms of the monic orthogonal polynomials @xmath218 defined by @xmath219 for all @xmath0 and @xmath220 for all @xmath221 with @xmath222 . in fact , @xmath217 is expressed in terms of the numbers @xmath223 in through the formula @xmath224 special choices of parameters lead to known families of orthogonal polynomials . indeed , colomo and pronko @xcite showed that the continuous hahn , meixner - pollaczek and continuous dual hahn polynomials appear in this way . in more general cases , the expressions were used to compute the asymptotics of the partition function @xmath225 for large @xmath197 in great detail by means of the riemann - hilbert method @xcite , see also @xcite . the situation in this paper corresponds to the case where @xmath226 for some @xmath227 and all @xmath228 . following the reasoning in @xcite , see also the appendix in @xcite , one sees that the izergin - korepin formula reduces to @xmath229^{n_1n}[\sin(\gamma+t_2)\sin(\gamma - t_2)]^{n_2n}}{\left(\displaystyle \prod_{n=0}^{n_1 - 1 } n ! \right ) \left ( \displaystyle \prod_{n=0}^{n_2 - 1 } n!\right ) \left(\displaystyle \prod_{n=0}^{n-1 } n!\right ) } \det m\ ] ] where the matrix @xmath230 now has entries @xmath231 with @xmath232 still given by ( [ def : mx ] ) . thus @xmath215 is a moment matrix with respect to the system of weight functions @xmath233 and @xmath234 on the real line . the inhomogeneous model was studied in @xcite in connection with the calculation of the arctic curve . leading order asymptotics of the partition function for the case @xmath235 in was computed in @xcite for the disordered regime , and in @xcite for the anti - ferroelectric regime . the analysis is valid in fact for any @xmath236 as long as @xmath237 as @xmath238 . it turns out that @xmath217 can be expressed in terms of monic _ multiple _ orthogonal polynomials @xmath239 with respect to the system of weight functions @xmath233 and @xmath234 . the polynomial @xmath239 is defined for any nonnegative integers @xmath41 by @xmath240 and @xmath241 we now have the following generalization of the formula . ( partition function ) let @xmath242 , @xmath243 , @xmath227 and recall the notation @xmath244 in and @xmath245 in . then the moment matrix @xmath215 in has determinant @xmath246 where @xmath247 is any sequence of @xmath1 s and @xmath248 s such that @xmath1 appears @xmath236 times and @xmath248 appears @xmath249 times . and where the associated sequences @xmath250 , @xmath251 , are defined recursively by @xmath252 and @xmath253 for any @xmath254 and @xmath251 . we will give the proof for the particular sequence @xmath255 ; it will be straightforward to extend the proof to the more general sequences @xmath247 in the statement of the proposition . from the definition it follows that @xmath256 where we define the inner product @xmath257 and where we use the functions @xmath258 and @xmath259 define functions @xmath260 and @xmath261 , @xmath262 , by bi - orthogonalizing the functions @xmath263 and @xmath264 in the following way : @xmath265 @xmath266 for appropriate coefficients @xmath267 and @xmath268 , subject to the orthogonality relations @xmath269 it is not hard to see that we can identify @xmath270 and @xmath271 . then we have that @xmath272 where the last step follows in a trivial way from . in view of the identifications mentioned in the previous paragraph , we then obtain . it is straightforward to generalize the above reasoning to the case of multiple values of the differences in . in general , one could even allow _ both _ @xmath273 and @xmath274 to take multiple values and then one should deal with _ multiple orthogonal polynomials of mixed type_. we leave the details to the interested reader . the value @xmath275 corresponds to the so - called free fermion line . as first observed in @xcite , in this case the above weight functions are related to the meixner - pollaczek weight . indeed , we then have @xmath276 we may compare this with the classical meixner - pollaczek weight function @xcite , @xmath277 by invoking the identity @xmath278 we see that for @xmath279 and @xmath280 reduces to @xmath281 thus the weight functions and are the same up to a scaling of the variable @xmath102 by a factor @xmath57 . it is an easy job to evaluate the izergin - korepin formula for the partition function explicitly on the free fermion line , so our results will not lead to new insights in that perspective . they should rather be considered as giving exact and asymptotic information on the multiple meixner - pollaczek polynomials in its own right . the rest of this paper is organized as follows : in section [ section : recurrence ] we consider multiple meixner - pollaczek polynomials with respect to two general weights and prove theorem [ theorem : recurrencemmpp ] . in section [ section : zeroasymptotics ] we establish theorem [ theorem : locallytoepeig ] and lemma [ lemma : interlacing ] , which together lead to proposition [ theorem : zeroseigenvalues ] . in section [ section : equilibrium ] we prove theorem [ theorem : equilibriumproblem ] using the theory of eigenvalue asymptotics for banded block toeplitz matrices . in this section we prove lemma [ lemma : interlacing ] on the fact that the multiple meixner - pollaczek polynomials @xmath26 exist and are unique , and have real and interlacing zeros . it is well known that the corresponding result about orthogonal polynomials with respect to one weight function holds ; we will rely on a generalization of this fact due to kershaw , @xcite . in our context it amounts to the statement that a sufficient condition is that for any non - negative integer @xmath24 and any polynomials @xmath282 and @xmath283 ( not both identically zero ) of degrees at most @xmath21 and @xmath284 ( or @xmath285 and @xmath22 ) respectively , the function @xmath286 has at most @xmath27 zeros . since @xmath287 the conclusion will certainly follow if we can show that , for any real @xmath288 , @xmath289 has at most @xmath27 zeros whenever @xmath282 and @xmath283 are polynomials such that @xmath290 . ( by convention , the zero polynomial has degree @xmath291 . ) this can easily be shown by induction , see e.g. @xcite . @xmath292 for @xmath293 , let @xmath19 be integrable real functions on the real line such that the measures @xmath294 have moments of all orders . suppose that for any non - negative integers @xmath21 and @xmath22 there exists a unique monic multiple orthogonal polynomial @xmath26 with respect to the weights @xmath295 , that is , a polynomial of degree @xmath24 satisfying the orthogonality conditions @xmath296 let @xmath297 denote the sub - leading coefficient of @xmath23 , so that @xmath298 . for @xmath293 , put @xmath299 the first non - vanishing moments . we begin by stating a general four term recurrence formula for multiple orthogonal polynomials on the real line in terms of their sub - leading coefficients and first non - vanishing moments . this standard fact is shown in @xcite and can also be derived from the riemann - hilbert problem for multiple orthogonal polynomials , see @xcite . [ generalrecurrence ] for any positive integers @xmath21 and @xmath22 , the multiple orthogonal polynomials satisfy the following four term recurrence relation : @xmath300 by symmetry between the two indices , a corresponding recurrence relation for @xmath301 is obtained by interchanging @xmath21 and @xmath22 and superindices . from standard results on the ordinary meixner - pollaczek polynomials , ( see e.g. equations ( 1.7.2 ) ( 1.7.4 ) in @xcite with @xmath302 ) , we have the orthogonality relation @xmath303 and the recurrence relation @xmath304 for any real parameter @xmath288 , define the finite difference operator @xmath305 , acting on functions @xmath306 , by the equation @xmath307 [ commutation ] for any real @xmath308 , @xmath43 , the operators @xmath309 and @xmath310 commute . straightforward calculation . [ pi ] let @xmath288 be a real number and @xmath311 , @xmath312 analytic functions in a domain containing the strip @xmath313 , and assume that there are positive numbers @xmath314 and @xmath315 such that @xmath316 for all @xmath317 . then the following integration by parts formula holds : @xmath318 using the definition of @xmath305 , we can split the integral of the left hand side into two terms and shift the contours of integration from the real line to @xmath319 for the first and second terms , respectively , by cauchy s theorem . this gives @xmath320 we can now derive a rodrigues type formula for the multiple meixner - pollaczek polynomials , which will be the tool to calculate explicit recurrence coefficients . [ representation ] let @xmath21 and @xmath22 be non - negative integers and put @xmath24 . let @xmath321 denote the @xmath322th iterate of an operator @xmath323 . then , for any @xmath32 with @xmath33 , and any @xmath12 , the multiple meixner - pollaczek polynomial @xmath26 satisfies the rodrigues formula @xmath324 where @xmath325 define the function @xmath326 for any non - negative integer @xmath322 . first of all , we note that by the properties @xmath327 and @xmath328 of the gamma function , it follows immediately from the definitions that if @xmath329 is a polynomial of degree @xmath322 with leading coefficient @xmath330 , then @xmath331 where @xmath332 is a polynomial of degree @xmath333 with leading coefficient @xmath334 equivalently , by simply replacing the parameter @xmath15 by @xmath335 , @xmath336 by induction over @xmath27 , it follows from ( [ recursionlambda ] ) and ( [ lofpolynomial ] ) that @xmath337 where @xmath338 is a polynomial of degree @xmath27 in @xmath102 with leading coefficient @xmath339 . therefore , @xmath340 for some monic @xmath27th degree polynomial @xmath341 . using this representation we can check the orthogonality conditions . let @xmath342 be a non - negative integer . note that the choices @xmath343 for any real @xmath344 and @xmath312 a polynomial satisfy condition ( [ decay ] ) of lemma [ pi ] ; this can be seen from the asymptotics of the gamma function valid as @xmath345 with @xmath346 < @xmath347 , @xmath348 ( stirling s formula ) . by definition of the weight function and applying lemma [ pi ] @xmath21 times , we get @xmath349 since @xmath350 acting on non - zero polynomials decreases their degree by one . by lemma [ commutation ] , the same argument applies in checking the orthogonality relations with respect to @xmath7 . we will need explicit expressions for the first non - vanishing moments , defined by ( [ def : h ] ) . these are readily calculated using proposition [ representation ] . [ moment ] the first non - vanishing moments @xmath245 of the multiple meixner - pollaczek polynomial @xmath26 are given by @xmath351 and @xmath352 consider the case @xmath353 ; by lemma [ commutation ] the @xmath354 case is completely analogous . reasoning as in the proof of the orthogonality relations , and noting that @xmath355 is a polynomial of degree @xmath356 with leading coefficient @xmath357 , and that @xmath358 , we find that @xmath359 here we made use of the orthogonality relation ( with @xmath360 ) to compute the integral . [ leadingcoeffprop ] the sub - leading coefficient @xmath297 of @xmath239 is given by @xmath361 let @xmath24 be fixed . the polynomial @xmath362 is clearly of degree @xmath27 and satisfies @xmath363 orthogonality conditions with respect to @xmath19 , for @xmath293 . it is thus a multiple of @xmath26 , and reading off the leading coefficient gives @xmath364 . multiplying this relation by @xmath365 and integrating with respect to @xmath6 gives the equation @xmath366 which by proposition [ moment ] can be written @xmath367 identifying coefficients in the recurrence relation ( [ recurrenceorthogonal ] ) for the ordinary monic orthogonal meixner - pollaczek polynomials @xmath368 with respect to @xmath6 , gives @xmath369 and so @xmath370 then repeated application of ( [ gammadifference ] ) leads to the claim , for any @xmath371 . finally we are ready for the proof of theorem [ theorem : recurrencemmpp ] . with the explicit expressions for @xmath297 and @xmath372 given by propositions [ moment ] and [ leadingcoeffprop ] , this follows from the general recurrence relation ( [ recurrence1 ] ) . in this section we will prove the general theorem [ theorem : locallytoepeig ] on the asymptotic zero distribution of a sequence of polynomials @xmath51 generated by a recurrence relation . the main idea of the proof follows the scalar case @xmath134 by kuijlaars - romn ( * ? ? ? * theorem 1.2 ) , see also @xcite . but we will need some nontrivial modifications due to the fact that @xmath79 may be greater than @xmath1 . the main tool in the proof is the following result on ratio asymptotics for the @xmath51 , compare with ( * ? ? ? * lemma 2.2 ) . [ lemma : roman ] ( ratio asymptotics ) under the assumptions of theorem [ theorem : locallytoepeig ] , we have that for each @xmath373 there exists @xmath374 so that all zeros of @xmath51 belong to @xmath375 $ ] whenever @xmath376 . moreover , @xmath377 uniformly on compact subsets of @xmath378 $ ] , where @xmath118 is the solution to the algebraic equation ( [ fzlambda0 ] ) with smallest modulus . the claim about the boundedness of the zeros of @xmath51 follows in a rather standard way from the assumptions , see e.g. ( * ? ? ? * proof of lemma 2.2 ) . now we turn to the claim . we consider the family of functions @xmath379 from the assumption that the zeros of @xmath380 are real and interlacing , it follows that @xmath381 is a normal family ( in the sense of montel ) on @xmath382 $ ] , see e.g. ( * ? ? ? * proof of lemma 2.2 ) . using induction on @xmath383 , we will show the following . * claim * : for any @xmath384 and for each @xmath385 , the following holds . if @xmath386 , @xmath387 are sequences of non - negative integers with @xmath388 , @xmath389 as @xmath390 , so that @xmath391 exists for @xmath392 , then @xmath393 as @xmath124 . let us prove this claim . we have @xmath394 as @xmath124 ( lemma [ lemma : mu0prob](a ) ) , and so it is clear that the claim holds for @xmath395 . now assume that the claim holds for @xmath385 . we will prove that it also holds for @xmath396 . we will prove this when @xmath397 ; the proof for the other values of @xmath322 can be given in a similar way . letting @xmath386 , @xmath387 be as in the claim , our goal will be to prove that the function @xmath398 defined by @xmath399 satisfies @xmath400 for @xmath124 . let us prove this . since @xmath389 as @xmath390 , we may assume that @xmath401 for every @xmath203 . for @xmath402 , we then have that @xmath403 belongs to the family @xmath381 . since @xmath381 is a normal family , we may assume , by passing to a subsequence if necessary , that @xmath404 exists for @xmath405 $ ] and @xmath406 . taking the @xmath407th block row in with @xmath408 and using , we obtain the matrix - vector relation @xmath409 where we denote with @xmath410 the @xmath411 column vector @xmath412 dividing both sides of by the scalar function @xmath413 , and taking the limit @xmath390 , we find @xmath414 where we used and where we set @xmath415 which exists entrywise due to our assumptions . it will be convenient to rewrite @xmath416 , @xmath417 , as a telescoping product : @xmath418 where by abuse of notation we write @xmath419 and @xmath420 for two vectors @xmath421 of length @xmath79 to denote their entrywise quotient and product respectively . each of the limits in exists again entrywise due to our assumptions . applying the induction hypothesis to each of the limits in the telescoping product , we find that @xmath422 for any @xmath417 . next we turn to the term @xmath423 in the expansion of the right hand side of . denoting @xmath424 , we can write this as @xmath425 where the first step follows by definition , the second step follows since the matrix @xmath426 is zero except for its bottom left entry , cf . , the third step is a consequence of , and the last step uses that @xmath427 in has its first entry equal to @xmath1 . inserting in yields the matrix - vector relation @xmath428 where the matrix @xmath429 satisfies @xmath430 we can rewrite as @xmath431 by using that @xmath432 as @xmath124 ( lemma [ lemma : mu0prob](a ) ) . relation clearly implies that @xmath433 expanding the determinant for @xmath434 large , with the help of , we obtain @xmath435 here the terms in the second line of can be justified by using the special structure of @xmath436 and @xmath426 in , and using again the fact that @xmath437 as @xmath124 . now the determinant in the first line of vanishes identically , since by definition @xmath438 is a root of , cf . . so reduces to @xmath439 which implies in turn that @xmath440 as @xmath124 . this proves the induction step , thereby establishing the claim . having proved the claim , the proof of lemma [ lemma : roman ] can now be finished from a standard normal family argument as in @xcite . with lemma [ lemma : roman ] in place , the proof of theorem [ theorem : locallytoepeig ] can be finished as in ( * ? ? ? * proof of theorem 1.2 ) . @xmath292 the above proof also shows that @xmath441 satisfies @xmath442,\ ] ] by virtue of and the fact that hold with @xmath383 arbitrarily large . so @xmath427 is a vector with first component equal to @xmath1 which lies in the null space of the matrix @xmath443 . in fact , it can be shown that there is a unique vector @xmath444 satisfying this condition , for all @xmath445 $ ] , and hence we have @xmath446.\ ] ] recall the functions @xmath448 , defined as the solutions to the algebraic equation ( [ algebraicequation : p ] ) ordered by increasing modulus , ( [ roots ] ) , and the definitions ( [ def : gamma0 ] ) and ( [ def : gamma2 ] ) of the sets @xmath125 and @xmath447 , whose structure we now describe . [ monotonoussupports ] @xmath449 $ ] and @xmath450 , where @xmath167 and @xmath168 are explicit constants given by ( [ supportnu1 ] ) and ( [ constraintactive ] ) , respectively . by ( [ algebraicequation : p ] ) , specializing ( [ fzlambda0 ] ) to the present setting gives @xmath451 where @xmath452 it is clear that @xmath453 if and only if @xmath454 , so it will be convenient to consider for a moment the sets @xmath455 we begin by establishing that @xmath456 @xmath457 ; the proof of this fact will be very similar to the proof of lemma 4.1 in @xcite . suppose @xmath458 . we can assume without loss of generality that @xmath459 is not a branch point , since the number of branch points is finite and @xmath125 and @xmath447 in our case have no isolated points @xcite . thus there exist distinct @xmath460 such that @xmath461 and @xmath462 . by the factorization of @xmath463 , we see that @xmath464 has only negative real zeros and therefore the even function @xmath465\ni\theta \mapsto which implies in turn that @xmath466 . but @xmath467 so @xmath459 is real , and hence @xmath456 @xmath468 . this argument also shows that @xmath469 may contain only branch points , since otherwise there would be three distinct values @xmath470 with the same modulus and @xmath471 which is clearly impossible . studying the function @xmath472 for fixed @xmath373 , we see that it has two local minima at the points @xmath473 and @xmath474 with @xmath475 and @xmath476 and that @xmath477 is a local maximum attained at the point @xmath478 ( see fig . [ plot_symbol ] ) . since @xmath479 is a polynomial equation in @xmath101 with real coefficients , it has two complex conjugate solutions if @xmath480 . if @xmath481\cup [ 0,y_1(s)]\right)$ ] , then all three solutions to are real and it is easy to see that there can be at most a finite number of such @xmath459 for which two of these roots have the same modulus . but as already mentioned , @xmath125 and @xmath447 can not have isolated points , hence @xmath482\cup [ 0,y_1(s)]$ ] . now consider the interval @xmath483 $ ] . for the branch point @xmath484 , we have that @xmath485 is a double root to , and there is also a negative root @xmath486 which is smaller than @xmath487 ( see fig . [ plot_symbol ] ) . therefore @xmath488 i.e. , the real negative root @xmath486 has larger modulus than the double positive root @xmath489 , so @xmath490 . in a similar way it follows that @xmath491 and @xmath492 . thus the branch points are not in @xmath469 , and therefore @xmath493 since we already proved that non - branch points are not in the intersection . we have now established that @xmath494 \cup ( -\infty , -(c_2s)^2]\ ] ] with @xmath495 and @xmath496 . we also established that @xmath497 , and then it follows by continuity that @xmath498 , \qquad \gamma_2 ^ 2(s ) = ( -\infty , -(c_2s)^2].\ ] ] this completes the proof . given a measure @xmath153 in the complex plane , define the _ logarithmic potential _ of @xmath153 , @xmath499 for @xmath293 , define the measures @xmath500 where @xmath501 is the ( complex ) line element on @xmath502 . [ equilibriumdelvaux ] fix @xmath89 . the pair @xmath503 is the unique minimizer of the energy functional @xmath504 among all pairs @xmath152 of positive measures such that @xmath505 , @xmath506 ; @xmath507 and @xmath508 . for all @xmath509 , @xmath503 satisfies the euler - lagrange variational conditions @xmath510 for some constant @xmath511 . see @xcite . we will now integrate ( [ explicitequilibrium ] ) to get an equilibrium problem for @xmath3 , where @xmath512 in analogy with the definition of @xmath4 in ( [ def : nu_1 ] ) . [ integratedproposition ] for all complex @xmath102 , the vector @xmath3 of measures satisfies the following conditions . firstly , @xmath513 where @xmath514 and @xmath383 is some constant , with equality in if and only if @xmath515 $ ] . secondly , @xmath516 with equality if and only if @xmath517 . furthermore , @xmath518 where @xmath519 by proposition [ monotonoussupports ] , the supports of the measures @xmath114 and @xmath520 are subsets of the real and imaginary lines respectively , which increase / decrease linearly in @xmath60 . this means that we can integrate the logarithmic potentials of @xmath114 and @xmath520 with respect to @xmath60 and change the order of integration to obtain @xmath521 and similarly for @xmath5 . the integrated variational conditions ( [ explicitequilibrium ] ) thus become @xmath522 @xmath523 furthermore , since by definition @xmath524 , @xmath525 with equality if and only if @xmath526 $ ] . clearly , @xmath5 must satisfy @xmath527 since @xmath528 for @xmath529 , with equality in if and only if @xmath530 . inserting into ( [ integratedequilibrium1a ] ) and ( [ integratedequilibrium2a ] ) and putting @xmath531 gives the stated inequalities . equations ( [ integratedequilibrium1 ] ) and ( [ integratedequilibrium2 ] ) are the euler - lagrange variational conditions for the equilibrium problem in theorem [ theorem : equilibriumproblem ] . the external field @xmath532 , density of the measure @xmath4 and upper constraint measure @xmath164 can be calculated explicitly , and the following subsections are devoted to these computations . in the calculations that follow we make use of the function @xmath533 which is such that @xmath534 for every @xmath535 . so in particular @xmath536 . we also define @xmath537 and these are the solutions of @xmath538 . because of we have @xmath539 and it follows that @xmath540 [ extfieldv ] the external field @xmath532 is given by @xmath541 because of symmetry , we may assume @xmath542 . it follows from proposition [ monotonoussupports ] that @xmath125 is increasing with @xmath60 and @xmath543 if and only if @xmath544 . thus @xmath545 for @xmath544 , and the integral that defines @xmath546 can be restricted to an integral over @xmath547 $ ] . using we obtain @xmath548 where in the last step we made the change of variables @xmath549 ( recall that @xmath542 ) . note that does not depend on @xmath102 . to evaluate we note that both @xmath550 and @xmath551 are one - to - one for @xmath552 and they map the interval @xmath553 onto @xmath554 $ ] and @xmath555 , respectively . we split the integral into two integrals , and apply a change of variables @xmath556 to each of them . then combining the two integrals again , and noting that @xmath557 and that @xmath558 if @xmath556 with @xmath293 , we obtain @xmath559 the integral in can be calculated explicitly , since @xmath560 = \frac{4}{(4 + ( 1+b^2)z ) \sqrt{z(4+b^2z ) } } = - \frac{1}{2 z \sqrt{q(z)}}.\ ] ] therefore @xmath561 using that @xmath562 and applying trigonometric identities , we finally obtain @xmath563 noting that @xmath564 since @xmath565 for every @xmath535 , we find the claimed expression for the external field by integrating with respect to @xmath102 . note that the external field @xmath532 has the form to be expected by analyzing directly the asymptotics of the weight functions @xmath566 : for the rescaled polynomials @xmath51 the orthogonality conditions read @xmath567 thus we have new effective weights @xmath568 . using stirling s formula , @xmath569 asymptotically , the dominant weight determining the potential associated with the distribution of zeros , will be @xmath570 , giving the external field @xmath571 . next we turn to the density of @xmath4 . [ propdensity ] the measure @xmath4 is absolutely continuous with density given by . by and ( [ explicitdensity ] ) , @xmath572.\ ] ] we now make essentially the same calculations as in the proof of proposition [ extfieldv ] . assuming @xmath573 and using we obtain as in @xmath574 a change of variables @xmath575 leads to @xmath576 which is an integral in the complex @xmath101-plane . since @xmath577 , we have that @xmath578 and @xmath579 are each others complex conjugate , and it can be shown that @xmath580 . the integral in is from @xmath578 in the lower half plane to its complex conjugate in the upper half plane along a path in @xmath581 $ ] . the square root @xmath582 is defined and analytic in @xmath583 $ ] and it is positive for @xmath101 real and positive . let @xmath584 then by and @xmath585 the arctangent is understood here as an analytic function @xmath586 thus @xmath587 using the defining relation @xmath588 in we can see that is equal to and then follows . note also that @xmath589 since @xmath4 has a positive density . [ propconstraint ] the upper constraint measure @xmath164 in is a multiple of the lebesgue measure on the imaginary axis with density @xmath590 . because of symmetry it is enough to consider the density on the positive imaginary axis . let @xmath591 . then @xmath592 and by a calculation as in the proof of proposition [ extfieldv ] this is equal to @xmath593 now we are going to make the change of variables @xmath594 . since @xmath595 , @xmath596 , and @xmath597 , we obtain @xmath598 with integration along a contour from @xmath599 to @xmath600 that intersects the real line in @xmath601 see and figure [ plot_symbol ] for @xmath487 . it is easy to check ( and this can also be seen from figure [ plot_symbol ] ) that @xmath602 and so the contour intersects the real line in a point lying in between the simple root and the double root of @xmath603 . the branch of @xmath604 that is used in is the one that is defined and analytic in @xmath605 \cup [ 0 , \infty)$ ] and that is in @xmath606 for @xmath607 . thus @xmath608 with principal branches of the fractional powers . we now deform the contour in to the positive real axis . in doing so , we pick up a residue contribution at @xmath609 , which is @xmath610 since the residue turns out to be @xmath611 , and so is independent of @xmath612 . the deformed contour goes along the positive real axis , starting at @xmath613 on the lower side , and ending at @xmath613 on the upper side of the real axis . it gives the contribution @xmath614 with our choice of square root we have @xmath615 where @xmath616 is the positive square root of a real and positive number . so the contribution from the positive real line is the integral @xmath617 and this integral we already calculated in and . its value is @xmath618 adding and we find @xmath619 and this proves the proposition . the equilibrium problem has a unique minimizer which will satisfy the euler - lagrange variational conditions , and proposition [ integratedproposition ] shows that @xmath3 has this property . the explicit expressions for the external field , upper constraint measure and density of @xmath4 are calculated in propositions [ extfieldv ] , [ propconstraint ] and [ propdensity ] . we thank pavel bleher , filippo colomo and karl liechty for useful discussions and pointers to the literature . m. abramowitz and i.a . stegun , _ handbook of mathematical functions _ , new york : dover publications , 1968 . aptekarev , multiple orthogonal polynomials , j. comp . appl . math . 99 ( 1998 ) , 423447 . r. baxter , _ exactly solved models in statistical mechanics _ , academic press , san diego , ca . b. beckermann , j. coussement and w. van assche , multiple wilson and jacobi - pieiro polynomials , j. approx . theory 132 ( 2005 ) , 155181 . p.m. bleher , s. delvaux and a.b.j . kuijlaars , random matrix model with external source and a constrained vector equilibrium problem , comm . pure appl . math . 64 ( 2011 ) , 116160 . p.m. bleher and v.v . fokin , exact solution of the six - vertex model with domain wall boundary conditions . disordered phase , commun . 268 , 223284 ( 2006 ) . p.m. bleher and k. liechty , exact solution of the six - vertex model with domain wall boundary conditions . ferroelectric phase , commun . phys . 286 ( 2009 ) , 777801 . p.m. bleher and k. liechty , exact solution of the six - vertex model with domain wall boundary condition . critical line between ferroelectric and disordered phases , j. statist . 134 ( 2009 ) , 463485 . p.m. bleher and k. liechty , exact solution of the six - vertex model with domain wall boundary condition : antiferroelectric phase , comm . pure appl . math . 63 ( 2010 ) , 779829 . f. colomo and a. pronko , on the partition function of the six - vertex model with domain wall boundary conditions , j. phys . a. 37 ( 2004 ) , 19872002 . f. colomo and a.g . pronko , square ice , alternating sign matrices , and classical orthogonal polynomials , j. stat . theory exp . ( 2005 ) , 33 pp . ( electronic ) . f. colomo and a.g . pronko , the arctic curve of the domain - wall six - vertex model , j. stat . ( 2010 ) , 662770 . f. colomo , a.g . pronko , and p. zinn - justin , the arctic curve of the domain wall six - vertex model in its antiferroelectric regime , j. stat . theory exp . ( 2010 ) , l03002 , 11pp . e. coussement , j. coussement and w. van assche , asymptotic zero distribution for a class of multiple orthogonal polynomials , trans . ( 2008 ) , 55715588 p. deift , t. kriecherbauer , k.t - r mclaughlin , s. venakides , and x. zhou , uniform asymptotics for polynomials orthogonal with respect to varying exponential weights and applications to universality questions in random matrix theory , comm . pure appl . 52 ( 1999 ) , 13351425 . p. deift , t. kriecherbauer , k.t - r mclaughlin , s. venakides , and x. zhou , strong asymptotics of orthogonal polynomials with respect to exponential weights , comm . pure appl . 52 ( 1999 ) , 14911552 . s. delvaux , equilibrium problem for the eigenvalues of banded block toeplitz matrices , preprint arxiv:1101.2644 . m. duits , d. geudens and a.b.j . kuijlaars , equilibrium problem for the two matrix model with an even quartic and a quadratic potential , nonlinearity 24 ( 2011 ) , 951993 . m. duits and a.b.j . kuijlaars , an equilibrium problem for the limiting eigenvalue distribution of banded toeplitz matrices , siam j. matrix anal . appl . 30 ( 2008 ) , 173196 . izergin , partition function of the six - vertex model in a finite volume , sov . 32 ( 1987 ) , 878 . izergin , d.a . coker and v.e . korepin , determinant formula for the six - vertex model , j. phys . a 25 ( 1992 ) , 4315 . d. kershaw , a note on orthogonal polynomials , proc . edinburgh math . 17 ( 1970 ) , 8394 . r. koekoek and r.f . the askey - scheme of hypergeometric orthogonal polynomials and its @xmath620-analogue , arxiv : math/9602214 . v. korepin , calculation of norms of bethe wave functions , comm . 86 ( 1982 ) , 391418 . v.e . korepin and p. zinn - justin , thermodynamic limit of the six - vertex model with domain wall boundary conditions , j. phys . a 33 ( 2000 ) , 7053 . kuijlaars and p. romn , recurrence relations and vector equilibrium problems arising from a model of non - intersecting squared bessel paths , j. approx . theory 162 ( 2010 ) , 20482077 . a.b.j . kuijlaars and w. van assche , the asymptotic zero distribution of orthogonal polynomials with varying recurrence coefficients , j. approx . theory 99 ( 1999 ) , 167197 . g. kuperberg , another proof of the alternating sign matrix conjecture , int . res . not . ( 1996 ) , 139150 . k. mahler , perfect systems , compositio math . 19 ( 1968 ) , 95166 . e. m. nikishin and v. n. sorokin , _ rational approximations and orthogonality _ , amer . math . soc . , providence , ri , 1991 . saff and v. totik , logarithmic potentials with external field , springer - verlag , berlin , 1997 . g. szeg , _ orthogonal polynomials _ , vol 23 , amer . soc . , providence , ri , 1975 . p. tilli , locally toeplitz sequences : spectral properties and applications , linear algebra appl . 278 ( 1998 ) , 91120 . w. van assche , multiple orthogonal polynomials , irrationality and transcendence , in `` continued fractions : from analytic number theory to constructive approximation '' , contemporary mathematics 236 ( 1999 ) , 325342 . w. van assche , j.s . geronimo and a.b.j . kuijlaars , riemann - hilbert problems for multiple orthogonal polynomials , in : j. bustoz , m. ismail and s. k. suslov ( eds . ) , _ special functions 2000 : current perspective and future directions _ ( tempe , az ) , 2359 , nato sci . ser . , 30 , kluwer acad . , dordrecht , 2001 . h. widom , asymptotic behavior of block toeplitz matrices and determinants , advances in math . 13 ( 1974 ) , 284322 . l. zhang and p. romn , asymptotic zero distribution of multiple orthogonal polynomials associated with macdonald functions , j. approx . theory 163 , ( 2011 ) , 143162 . p. zinn - justin , six - vertex model with domain wall boundary conditions and one - matrix model , phys . e 62 ( 2000 ) , 34113418 .

we present two confirmed wide separation l - dwarf common proper motion companions to nearby stars and one candidate identified from the two micron all sky survey . spectral types from optical spectroscopy are l0 v , l2.5 v , and l8 v. near - infrared low resolution spectra of the companions are provided as well as a grid of known objects spanning m6 v t dwarfs to support spectral type assignment for these and future l - dwarfs in the _ zjhk _ bands . using published measurements , we estimate ages of the companions from physical properties of the primaries . these crude ages allow us to estimate companion masses using theoretical low - mass star and brown dwarf evolutionary models . the new l - dwarfs in this paper bring the number of known wide - binary ( au ) l - dwarf companions of nearby stars to nine . one of the l - dwarfs is a wide separation companion to the f7 iv - v extrasolar planet system hd89744ab .

astro-ph0108424

the study of binary and multiple systems in the local neighborhood is important for several reasons . it is desirable to have an accurate accounting of stellar systems in the local neighborhood . improving the statistics of binary primary masses , system mass fractions , orbital separations , and eccentricities are crucial for refining stellar and planetary formation theories . also , studying wide separation systems in particular gives insight into their dynamic stability with time as a function of orbital separation and mass fraction ( and hence binding energy ) against disruptive events such as close approaches to giant molecular clouds , stars , and the gravitational potential of the inner galaxy . in this paper we present three wide separation l - dwarf companions to nearby stars . one companion has an apparent separation @xmath2 of 100 au @xmath3 au , and two have @xmath4 au . these discoveries , along with similar companions previously reported , highlight the success in searching new portions of binary system phase space ( i.e. wide separations ) by surveys such as two micron all sky survey ( 2mass ; @xcite ) , the deep near - infrared survey ( denis ; @xcite ) , and the sloan digital sky survey ( sdss ; @xcite ) . the surveys sensitivities and large areal coverages allow for very low - mass ( vlm ) stars and brown dwarf companions to be found at separations @xmath5 au . ) , to mean ` apparent separation . ' ] the new detections will in turn improve the statistics of systems with mass ratios ( @xmath6 ) far from unity . , and use the term ` low mass ratio ' to mean @xmath6 near zero . ] for instance , the low - mass companions already discovered show that the ` brown dwarf desert ' at close separation ( @xmath7 % of fgkm main sequence primaries have brown dwarf companions at @xmath8 au ; @xcite ) may not extend to wide separations @xcite . the added statistics can also be used to test the mass ratio distribution results of @xcite : their distribution peaks at @xmath9 and is then either flat or decreases for @xmath10 . @xcite , however , used the radial velocity technique for searching for unknown companions , a technique poorly suited for discovering @xmath11 systems at wide separation . the study of vlm stars and brown dwarfs as singular objects also benefits from the discovery of these objects as companions in binary systems . because brown dwarfs continually cool with time since they can not sustain hydrogen fusion , and vlm stars near the main sequence limit can take a hubble time to come to equilibrium , degeneracies occur in the hr diagram for these objects . for example , a field l1 v can either be a very old and vlm star or a less massive , young brown dwarf that is slowly cooling to later spectral types . binary systems can be used to break this degeneracy and provide examples of l - dwarfs with estimated masses as follows : assuming that the primary and low - mass companion in a system are coeval , age estimates from the primary can be adopted for the secondary . the ages along with inferred effective temperatures can then be used to determine crude mass estimates from theoretical evolutionary curves . metallicity and distance information is also acquired for l - dwarfs in binary systems from known properties of the primaries . there are six previously discovered l - dwarfs that are wide separation ( @xmath0 au ) companions to nearby stars ( see @xcite , table 7a , for a summary ) , including three from 2mass . the latter were l - dwarf candidates found to be serendipitously located close on the sky to known nearby stars . in this paper we present the first results from a _ dedicated 2mass search for l - dwarf candidates in close proximity on the sky to known nearby stars . the search area around any given candidate primary is an annulus with inner radius defined by the object brightness and the capability of the 2mass point source extraction algorithm to identify faint objects adjacent to bright sources . the outer radius is defined by the distance to the primary and the maximum companion separation plausible from dynamical binding considerations . the candidate primaries are mainly drawn from the catalogue of nearby stars @xcite and other nearby objects discovered more recently . a future paper will discuss in more detail the search criteria and success rate . in this paper we present two confirmed ( spectroscopic and common proper - motion ) companions discovered while searching the vicinity of two ` gliese ' objects : 2massw j0912145 + 145940 ( l8 v companion to a g8 v@xmath1k1 v spectroscopic binary ) and 2massw j1620261 - 041631 ( l2.5 v companion to an m0 v ) . _ we also present the spectroscopically ( but not common proper - motion ) confirmed candidate companion 2massi j1022148 + 411426 ( l0 v ) . this object was an l dwarf candidate discovered serendipitously to be nearby on the sky to hd 89744ab , an f7 iv - v + giant planet system . motivated by this discovery , we will add all stars with known planets to the list of candidate primaries for the dedicated search . 2mass survey observations and evidence of common proper motion are discussed in 2 . red - optical ( 0.651.00 @xmath12 ) and near - infrared ( nir ) spectroscopic follow - up observations are presented in 3 , along with a grid of known late - m and l - dwarf nir spectra for classification in @xmath13 . and @xmath14 . the long wave cut - off is determined by ccd sensitivity @xcite . ] age estimates for the candidates are assigned in 4 based upon known physical characteristics of their primaries . we use these age assignments to determine masses for the candidates using theoretical evolutionary curves . a discussion follows in 5 . the 2mass working database was used to search for color - selected low - mass candidates in close proximity to nearby stars . the color - selection of late - m and l - dwarfs from nir surveys is now well established ( @xcite ; hereafter k99 and @xcite ; hereafter k00 ) . because of their low effective temperatures , they possess extreme visual - ir colors , e.g. @xmath15 . with the exception of the nearest examples , late - m and l - dwarf infrared candidates will have no optical counterparts on visual sky - survey plates . @xmath16 for these objects reddens monotonically ( allowing for cosmic scatter ) from @xmath17 for m8 v to a maximum of @xmath18 for late l - dwarfs using classifications from the red - optical ( k99 , k00 ) . the three wide separation companion candidates followed - up in this paper met the color - selection criteria above , were in close proximity on the sky to known nearby stars , and were bright enough @xmath19 for nir spectral follow - up with the cornell massachusetts slit spectrograph ( cormass ; @xcite ) on the palomar 60-inch telescope . the objects coordinates , photometric magnitudes and @xmath20 colors are listed in table [ objects ] . finding charts are presented in figure [ finders ] . the companion candidate 2massw j0912145 + 145940 ( hereafter 2m0912 + 14 ) has @xmath21 and @xmath22 . it has an angular separation of @xmath23 ( 881 au ) from the binary gl 337ab ( fin 347 , hip 45170 ) , from gl 337ab at position angle ( @xmath24 ) and aa - d was located @xmath25 away with @xmath26 @xcite . ] which has an hipparcos measured distance of @xmath27 @xcite . at this distance the companion would have @xmath28 . using the absolute magnitude spectral class relation of k00 , @xmath29 where subclass is @xmath30 for m9 v , @xmath31 for m9.5 v , 0 for l0 v , 0.5 for l0.5 v , etc . , we infer a candidate companion spectral type of l7 v. k00 found @xmath32 with @xmath33 for the l7 subclass , so the color of 2m0912 + 14 is about 1 @xmath34 from the mean . 2massw j1620261 - 041631 ( hereafter 2m1620 - 04 ) has @xmath35 and @xmath36 . it has an angular separation of @xmath37 ( 1090 au ) from the m0 v gl 618.1 ( g 17 - 11 , hip 80053 ) , which has an hipparcos measured distance of @xmath38 @xcite . at this distance the companion would have @xmath39 . using equation ( [ subclass ] ) we infer l4 v. k00 found @xmath40 with @xmath41 for the l4 subclass , so the color of 2m1620 - 04 is also about 1 @xmath34 from the mean . 2massi j1022148 + 411426 ( hereafter 2m1022 + 41 ) has @xmath42 and @xmath43 . it has an angular separation of @xmath44 ( 2460 au or 0.012 pc ) from the f7 iv - v star hd 89744 ( hr 4067 , hip 50786 ) , which has an hipparcos measured distance of @xmath45 @xcite . at this distance the companion would have @xmath46 . using equation ( [ subclass ] ) we infer l1 v. k00 found @xmath47 with @xmath48 for the l1 subclass , so the color of 2m1620 - 04 is within 1 @xmath34 of the mean . the proper motions of the primaries gl 337ab and gl 618.1 are @xmath49 and @xmath50 @xcite . the 2mass survey astrometric uncertainty is @xmath51 ( 1 @xmath34 ; @xcite ) so these motions allow confirmation that the candidates are indeed proper motion companions by comparing two epoch 2mass survey images separated by at least a few months . the field containing 2m0912 + 14 was scanned twice with 2mass , once on 1997 november 18 and again on 2000 april 26 . the scans showed the position of 2m0912 + 14 to move @xmath52 towards position angle @xmath53 relative to other stars in the field . this relative motion compares well with the gl 337ab hipparcos absolute proper motion of @xmath54 with @xmath55 , thus confirming common proper motion ( cpm ) companionship . hereafter we refer to 2m0912 + 14 as gl 337c . the field containing 2m1620 - 04 was scanned twice with 2mass , once on 1999 july 29 and again on 2000 may 30 . the scans showed the position of 2m1620 - 04 to move @xmath56 towards position angle @xmath57 relative to other stars in the field . this relative motion compares well with the gl 618.1 hipparcos absolute proper motion of @xmath58 with @xmath59 , thus confirming cpm companionship . hereafter we refer to the primary as gl 618.1a and 2m1620 - 04 as gl 618.1b . the proper motion of hd 89744 is only @xmath60 @xcite , so a longer baseline is needed to confirm cpm companionship . this field was scanned once with 2mass on 1998 april 05 . the palomar observatory sky survey ii ( poss - ii ) also weakly detected 2m1022 + 41 in the photographic nir passband ( ivn plate @xmath1 rg9 filter @xmath61 ; @xcite ) on 1998 december 30 . unfortunately , the nine month baseline between the 2mass and poss - ii images is not sufficient to establish common proper motion . we can , however , assess the probability that 2m1022 + 41 is an unassociated low - mass object based upon the likelihood that an l - dwarf such as this object would be found in the same line of sight as hd 89744 . the surface density on the sky of 2mass l - dwarfs to @xmath62 is 1 per 20 square degrees ( k99 ) . there are @xmath63 stars within 50 pc listed in the hipparcos catalogue ; the simbad database contains @xmath64 . using the larger of the two estimates , less than one 2mass l - dwarf will be randomly found within @xmath65 of _ any of the @xmath64 cataloged stars . in addition , the correlation between the estimated spectral type of 2m1022 + 41 from @xmath66 and the distance of the suspected primary , l1 v ( [ cand ] above ) , and its actual spectral type of l0 v from red - optical spectroscopy ( [ os ] below ) strongly supports companionship . based on these arguments we consider it safe to assume companionship and hereafter refer to the primary as hd 89744a and 2m1022 + 41 as hd 89744b . @xcite have discovered a planet circling the f star , and we shall refer to this planet as hd 89744ab . _ red - optical spectra of the candidates were obtained with the low resolution imaging spectrograph ( lris ; @xcite ) at the 10 m w. m. keck observatory ( keck i ) . a 400 groove mm@xmath67 grating blazed at 8500 with a @xmath68 slit and a tektronics @xmath69 ccd were used to yield @xmath70 pixel@xmath67 ( @xmath71 resolution ) spectra spanning the wavelength range 6300 @xmath72 . an og570 blocking filter was utilized to exclude second order . further details regarding data acquisition and reduction are given in k99 and k00 . table [ observations ] summarizes observations . figure [ optical spectra ] displays the reduced optical spectra . the spectra have not been corrected for telluric absorption . they confirm that the objects are l - dwarfs . using the l - dwarf classification scheme of k99 we assign the following spectral types : gl 337c ( l8 v ) , gl 618.1b ( l2.5 v ) , and hd 89744b ( l0 v ) . these spectral types correspond to effective temperatures of l8 v ( @xmath73 k ) , l2.5 v ( @xmath74 ) , and l0 v ( @xmath75 ) , where the lower and upper limits are from the classifications schemes of k00 and @xcite , respectively . the ( @xmath76 ) feature is an important indirect gauge of mass @xcite . when seen in absorption in all but the youngest fully convective very low - mass objects , this feature indicates insufficient mass ( @xmath77 , i.e. sub - stellar ) to produce central temperatures high enough ( @xmath78 ) to fully fuse the primordial lithium through @xmath79 reactions ( see e.g. @xcite ) . lithium is not detected in any of the objects to the following equivalent width ( ew ) upper limits : @xmath80 for gl 337c and gl 618.1b , and @xmath81 for hd 89744b . this indicates that gl 618.1b and hd 89744b have @xmath82 , an important mass constraint on the secondaries that we will utilize in [ primasses ] . gl 337c is too cool to use the lithium test . as temperatures fall in the atmospheres of l - dwarfs , lithium is expected to give way to licl as the dominant lithium bearing gas for @xmath83 , thereby depleting existing atomic lithium @xcite . since our assigned @xmath84 upper limit for gl 337c ( 1600 k ) is only slightly warmer than this transition temperature , we consider the use of this test misleading for this object . k00 found diminishing li absorption line strengths for l7 and l8 dwarfs when present , and interpreted this as evidence of increasing lithium bearing molecular formation . licl spectral band heads occur in the mid - infrared at @xmath85 ( @xmath86 ) , @xmath87 ( @xmath88 ) , and @xmath89 ( @xmath90 ) @xcite . unfortunately , these wavelengths are not accessible from the ground due to atmospheric absorption just beyond the @xmath91 passband , but they will be observable with the infrared spectrograph ( irs ) on the upcoming space infrared telescope facility ( sirtf ) . @xmath92 seen in emission is an important measure of chromospheric activity ( and youth ) in early - mid m - dwarfs , just as h and k lines are used for fgk stars . @xmath92 emission is not detected with upper limits @xmath7 for gl 337c and @xmath81 for hd 89744b . gl 618.1b is a possible detection with @xmath93 . these results are compatible with the l - dwarf activity statistics from the k99 and k00 samples , in which @xmath92 emission declined from 60% ( 80% if marginal detections are included ) for l0 v to 8% ( 25% including two marginal detections ) for l8 v @xcite . statistically we might have expected emission in hd 89744b , but it is not surprising that gl 618.1b and gl 337c are not active . the wavelength of peak energy emission shifts redward into the nir with decreasing effective temperature , so l - dwarfs become very faint in the red - optical . as noted in k99 , a large aperture telescope is needed to use red - optical spectral typing schemes effectively . because a more practical means of assigning spectral types to color - selected late - m and l - dwarfs was sought , the cornell massachusetts slit spectrograph ( cormass ; @xcite ) was built for use on the palomar 60-inch telescope . over 100 new field late - m and l - dwarf candidates have been observed with cormass as part of a magnitude - limited spectroscopic survey to improve the statistics of the luminosity function across the stellar sub - stellar boundary in the solar neighborhood . the instrument has also been used for the spectral confirmation of a bright t - dwarf @xcite and now for the follow - up of some candidate wide separation companions in @xcite and this paper . cormass is a prism cross - dispersed low - resolution ( @xmath94 ) near - infrared ( nir ) spectrograph . its 2-dimensional spectral format provides simultaneous coverage from @xmath95 to @xmath96 ( _ zjhk _ bands ) . a 40 lines mm@xmath67 grating , blazed at @xmath97 , is used with a fixed @xmath98 slit to image the raw spectra across 6 orders on a nicmos3 detector . figure [ nir grid ] presents a grid of cormass nir spectra for known low - mass objects spanning spectral classes m6t . an observing log for the grid is presented in table [ grid log ] . the m - dwarf spectral types are on the @xcite system , and l - dwarf classifications are on the k99 system . the data reduction technique follows that described in @xcite . our grid qualitatively agrees with similar nir grids of l - dwarfs already published @xcite . as seen in figure [ nir grid ] , the nir spectra are dominated by molecular absorption features that smoothly change with spectral type through the end of the l - class . the @xmath99 absorption features bracketing h - band change most obviously . in particular , the slope of the blue side of the h - band spectra appears to increase monotonically with later spectral type as @xmath99 absorption increases in the l - dwarf atmospheres . this monotonic relationship has been quantitatively confirmed for a grid of l - dwarfs by @xcite and @xcite and for a grid of m - dwarfs by @xcite . @xcite also use this feature to help classify denis candidate low - mass objects . other spectral changes in the grid include varying strengths of the feh features at @xmath100 and 1.19 - 1.24 @xmath12 , as well as changes in the pair of doublets in j - band . the evolution of the j - band features were investigated by @xcite using higher resolution spectra of l - dwarfs . k - band evolution seen in figure [ nir grid ] includes increasing @xmath101 absorption on the blue end , increasing curvature longward of @xmath102 due to collision induced absorption of @xmath103 , and a deepening of the co bandhead feature at @xmath104 ( see e.g. @xcite ) . the slow spectral evolution for late - m and l - dwarfs is in marked contrast to the rapid change in features observed from l8 v to t dwarfs as the primary carbon - bearing molecule changes from co to @xmath105 . cormass observations of the companions are listed in table [ observations ] . the spectra have been inserted into figure [ nir grid ] at positions bracketed by known objects with similar spectra based upon comparisons by eye . the nir spectra of gl 337c supports a spectral type assignment of l8 v. the depth of the j - band @xmath99 ( @xmath106 ) absorption feature indicates a spectral type later than l7 v. the spectra of gl 618.1b ( smoothed with a boxcar average of 5 pixels ) and hd 89744b are both similar to those of the late - m and early l - dwarfs . because the spectral features evolve more slowly in this regime we can not confirm a specific spectral type by eye . the spectra support a range of spectral types : m9 v @xmath107 l3.5 v and m9 v @xmath108 l2 v. at the resolution of cormass ( r @xmath109 ) quantitative comparison of spectral type indices ( i.e. ratios of average flux in specified regions , or equivalent widths , etc . ) will be necessary to improve the accuracy of spectral subclass assignments based upon nir spectra . in a subsequent paper @xcite we will define quantitative spectral indices for assignment of spectral subclass for use with the cormass spectroscopic survey of field low - mass object candidates . the assignment of ages to field stars is by its very nature imprecise due primarily to few reliable techniques , cosmic scatter , and especially slow evolution for stars with late spectral types . hopefully the analysis of systems with late - m and l - dwarf wide separation companions will motivate increased study of ` age dating ' techniques . this necessarily lengthy section gathers available evidence and applies relevant techniques to derive unfortunately imprecise age estimates for the primary systems of the l - dwarf companions . published age related properties for the three primary systems are summarized in table [ primproperties ] . a decrease in activity ( coronal and chromospheric ) with age in stars with convective envelopes is generally attributed to a reduction in the classical dynamo effect ( @xmath110-@xmath111 ) as stars spin down due to angular momentum losses in stellar winds ( see e.g. @xcite ) surface lithium abundance from high resolution spectroscopy can serve as a crude age indicator for main sequence stars with convective envelopes . primordial surface lithium in these stars is destroyed through mixing to warmer interior temperatures due to the combined action of convection in the outer envelope and more complex mechanisms not yet fully understood @xcite . only one of the three primaries in this paper have published lithium abundance measurements , namely hd 89744a . unfortunately , the measurement s usefulness as an age diagnostic is diminished because , as an f7 iv - v , it falls very close to the mid - f ` lithium dip ' region of the main sequence in which dramatic surface lithium destruction occurs where standard evolutionary models using convective mixing predict there should be no destruction ( see , e.g. @xcite ) . space motions are frequently used as indicators of age for samples of stars . but their use in assigning ages to individual stars must be done with great caution due to significant dispersion within any kinematic class . nonetheless , it is unlikely for an old star to have small velocity , but a young star can have a high velocity . we take the division between _ young and _ old to be @xmath112 gyr , the division between the young disk and old disk used by @xcite . we will refrain from placing significant weight on kinematic age when possible . _ _ in the case of hd 89744a there are several explicit age determinations already published based upon high resolution metallicity measurements and fits to stellar isochrones . these ages will be compared with our estimates derived from physical properties . gl 337ab is a double - lined spectroscopic ( sb2 ) and visual binary located @xmath113 pc away ( hipparcos , @xcite ) . astrometric , speckle and visual observations reveal an orbital semi - major axis of @xmath114 ( @xmath115 au ) and a period of 2.7 years ( see e.g. @xcite ) . it is one of the shortest - period visual pairs observed . published measurements are summarized by @xcite . they adopted a spectral type of g8 v for both members based on @xmath116 photometry . more recent lunar occultation @xcite and adaptive optics observations @xcite indicated an early k spectral type for the secondary based on @xmath117 brighter magnitudes than the primary at @xmath118 and @xmath119 bands . gl 337ab is detected as a rosat x - ray source with @xmath120 @xcite . the quantity @xmath121 is often used to describe x - ray emission in a distance and stellar radius independent logarithmic ratio . @xcite has derived a relation between @xmath122 and age for solar type stars using published empirical relations for rotational period time dependence , x - ray luminosity / rotation correlations , and the luminosity evolution of the constant - mass sun : @xmath123}\ ] ] the coefficient @xmath110 , from the rotation period time dependence relation , has published values of @xmath124 @xcite and @xmath125 @xcite . the age @xmath126 is in gyr . following @xcite we have adjusted the equation to give @xmath127 @xcite for the present solar age of 4.6 gyr . because the individual stars of gl 337ab are not resolved by rosat , and they are close in spectral type , we attribute _ ad hoc half of the observed x - ray luminosity to each star . using @xmath128 and @xmath129 @xcite , we calculate @xmath130 . equation ( [ xray ] ) can then be used to estimate an age of 1.25 gyr ( @xmath131 ) and 1.7 gyr ( @xmath132 ) for gl 337ab . _ this estimate implicitly assumes that the total x - ray luminosity measured by rosat is merely the sum of individual x - ray emission from _ non - interacting members of a binary system . this assumption is valid . @xcite reviewed physical modes of interaction in binaries with solar - mass primaries . since both members of gl 337ab are on the main sequence , interaction modes typical of evolved stars such as stellar wind accretion and roche - lobe overflow can be rejected . tidal interactions can circularize and synchronize binary orbits for systems on the order of @xmath133 days , i.e. close binaries , and this interaction is thought to cause the @xmath134 cut - off below which all close binaries have circular and presumably synchronized orbits ( see e.g. dm91 figure 5 ) . tidal interaction in close binaries nullifies the use of x - ray emission as an indicator of age since the tidal effects disrupt the spin - down , and thus rotation - activity correlation , of the individual stars . _ but timescales for orbital synchronization @xmath135 and circularization @xmath136 for stars with convective envelopes are extremely steep functions of binary fractional separation @xmath137 : @xmath138 and @xmath139 @xcite . @xcite estimated @xmath140 years , where q is the mass fraction and p is the orbital period in days . using @xmath141 and @xmath142 days for gl 337ab gives @xmath143 years . gl 337ab is therefore not a tidally interacting binary . the space motion of gl 337ab is directed opposite the galactic center : @xmath144 km @xmath145 , @xmath146 km @xmath145 , and @xmath147 km @xmath145 @xcite . ( this is with respect to the sun and with u positive towards the galactic center . ) this motion makes it a member of the kinematic old disk population with an age of @xmath148 to 10 or 12 gyr based upon its position outside of the young disk ` eggen box ' in the uv plane @xcite . the ` eggen box ' is roughly defined by @xmath149 km @xmath145 @xmath150 km @xmath145 and @xmath151 km @xmath152 km @xmath145 . since the gl 337ab system is both an sb2 and visual binary , mass and magnitude measurements for both components have been published . if we assume that the stars of the binary are coeval , plotting the observed magnitudes against model evolutionary cooling curves _ individually for the components should indicate the same age . _ we assumed the mass estimates of @xmath153 and @xmath154 @xcite and absolute magnitudes of @xmath155 and @xmath156 , inferred from lunar occultation measurements @xcite and hipparcos distances , for the components . the measured absolute magnitudes were transformed to the appropriate @xmath118 passbands for comparison with the theoretical zero - metallicity evolutionary models ( @xmath157 vs. age ) of @xcite and @xcite . surprisingly , we found disparate ages of roughly 1 gyr and 3 gyr for components a and b , respectively , when compared to both models . the most obvious discrepancy was the factor of 2 difference between the observed ( @xmath158 ) and predicted ( @xmath159 ) differential magnitudes for mass differences of @xmath160 . assuming larger published mass differentials for the components increases the discrepancy . there are no other published @xmath118 band measurements . it should be noted that the masses of the components fall within an area of extreme sensitivity to model mixing parameter , metallicity , and initial he abundance , as well as other model assumptions ( i. baraffe 2001 , private communication ) . these complexities likely contribute to the age disparity , but it is unlikely that they fully account for the discrepancy given in particular the @xcite model s success in other empirical theoretical comparisons ( see e.g. @xcite ) . while we can not use these isochrone comparisons to constrain an age for the gl 337 system , the discrepancy highlighted by this comparison is worthy of further study both from the standpoint of theoretical modelling and via observation . we adopt an age span of 0.6 to 3.4 gyr by simply assuming a factor of 2 uncertainty in the 1.25 1.7 gyr estimate derived from x - ray luminosity . our adopted age is consistent with the kinematic age if the gl 337 system is a younger member of the old disk . gl 618.1a is an m0 v star located @xmath161 pc away based on hipparcos measurements . this particular object has not been extensively studied . it is worth noting that field m - dwarfs are notoriously difficult to age date for various reasons , including : 1 ) slow evolution on the hr diagram in a hubble time , 2 ) fast depletion of photospheric lithium since full convection appears as early as m3 v , 3 ) generally slow rotation , 4 ) poorly correlated rotation and activity strength @xcite , and 5 ) intrinsic faintness . gl 618.1a does not display evidence of chromospheric or coronal activity : rosat did not detect x - ray emission from the vicinity of this object , and @xmath92 was observed in absorption on two occasions separated by a decade @xcite . it would actually be unusual if gl 618.1a displayed evidence of activity since only 8% of 98 m0 v dwarfs observed by @xcite were magnetically active ( dme ) . @xcite found that in a sample of nearly 600 nearby m - dwarfs , the dme stars ( @xmath162 of the sample ) came from a kinematically younger population than the remaining dm stars in the survey . @xcite classified over 200 low - mass stars into five populations using kinematic and photometric techniques : young disk , young / old disk , old disk , old disk / halo , and halo . the objects in the sample were first placed into populations based upon kinematics . gl 618.1a s space motion of @xmath163 km @xmath145 , @xmath164 km @xmath145 , and @xmath165 km @xmath145 met the criteria for membership in the old disk / halo based upon its eccentricity of @xmath166 in the u - v plane . @xcite also utilized infrared color - color diagrams to estimate mean metallicity ( @xmath167}$ ] ) . the kinematic populations occupied separate positions on the diagrams , albeit with scatter . despite its space motion , gl 618.1a was found to be a member of the young disk on this diagram and was assigned @xmath167 } \sim 0 $ ] . the sample photometric uncertainties gave a class membership uncertainty of @xmath168 . lastly , @xcite plotted the objects in color - absolute magnitude diagrams and again found the populations could be predicted based upon location . this diagram also supported gl 618.1a s membership in the young disk . we are confronted with conflicting and low confidence age diagnostics for this object : large total space motion ( 138 km @xmath145 ) that is unlikely to belong to a young disk ( @xmath169 gyr ) , and photometric evidence of youth @xcite . we refrain from adopting a lower limit , and simply allow the lower limit to be set by our non - detection of lithium in the spectrum of the l - dwarf companion ( [ lithium ] ) . this imposes a lower limit of @xmath166 gyr ( [ primasses ] ) . we adopt an upper limit age of 12 gyr based on the recently determined solar neighborhood age of @xmath170 gyr @xcite . hd 89744a is an f7 star at @xmath171 pc with @xmath172 based upon hipparcos measurements @xcite . @xcite summarized this star s properties when they reported the discovery of a massive planetary companion to hd 89744a . the authors adopted a mass of @xmath173 for hd 89744a based upon the average of independent results from @xcite ( @xmath174 ) and @xcite ( @xmath175 ) . the massive planet hd 89744ab ( @xmath176 ) orbits the primary every 256 days with a highly eccentric orbit ( @xmath177 ) . the rms residual uncertainty to the orbital fit of the massive planet was 20.5 m @xmath178 @xcite . this residual exceeded the instrumental uncertainties of 10 m @xmath178 and led the authors to state that they could not rule out the existence of a distant companion . a crude estimate of the velocity of the massive planet hd 89744ab induced by the wide l - dwarf companion hd 89744b is @xmath179 , where @xmath180 is the acceleration of the planet due to the gravitational force of the l - dwarf , @xmath181 is the @xmath182 year time span of @xcite radial velocity observations , @xmath183 is the mass of l - dwarf , and @xmath184 au . this gives @xmath185 m @xmath178 , thus ruling out the possibility that hd 89744b is the cause of the orbital fit residuals . at 2500 au separation , a companion must be @xmath186 to induce a residual radial velocity equal to the instrumental uncertainty of the @xcite observations . of course this does not rule out the possibility of another as yet unseen companion at a smaller orbital separation . the age of hd 89744a has been estimated from fits to evolutionary isochrones on three occasions . @xcite calculated an age of 2.09 gyr using @xmath187=0.18 $ ] . the object fell within the ` hook - region ' of the isochrones used by @xcite , so the authors noted that the age could be underestimated by 0.15 dex , corresponding to an older age of 2.95 gyr . @xcite also calculated an age with this metallicity but used more recent stellar isochrones and hipparcos distances . they gave an age of 2.04 gyr . and @xcite derived an age of 1.8 gyr from isochrone fits . all three ages are in close agreement . @xcite revised this object s luminosity class from main - sequence to subgiant using the improved @xmath188 and hence more accurate location in color magnitude diagrams afforded by hipparcos results . this change is supported by the finding of @xcite that hd 89744a was within the ` hook - region ' . @xcite assigned a spectral type of f7 iv - v using the vilnius photometric system . @xcite , using updated isochrones and hipparcos distances as well , still found the object to be in luminosity class v. due to the uncertainty in luminosity class we will refer to the object as an f7 iv - v . based on the object s proximity to the main sequence turn - off , its age can be estimated to be @xmath148 gyr , the approximate turn - off age of an @xmath189 star with our adopted metallicity ( @xmath190=0.24 $ ] ; table [ primproperties ] ) from the @xcite isochrones used by @xcite . the projected rotational velocity ( @xmath191 ) of 8 km @xmath145 @xcite for hd 89744a supports a spectral type of f7 iv - v . this rotation falls squarely in the 0 to 20 km @xmath145 range found in a study of late - f subgiants by @xcite . more recently @xcite and @xcite found a sharp rotational discontinuity at @xmath192 , or f8 iv . the rotational velocities of subgiants earlier than f8 spanned from a few to @xmath193 km @xmath145 , while subgiants later than f8 had mean rotational velocities of 6 km @xmath145 at g0 decreasing to @xmath194 km @xmath145 at k5 . abrupt magnetic braking with the deepening convective envelope in the subgiant phase is thought to be the cause of this discontinuity @xcite . while there is significant scatter in the rotational velocities of mid to late f - dwarfs , the @xmath191 of hd 89744a is compatible with a subgiant evolving redward towards the f8 iv discontinuity . @xcite reported an upper limit of @xmath195 from rosat measurements in their study of x - ray emission from early post - main sequence stars . from equation ( [ xray ] ) this gives a minimum coronal age of 1.2 gyr ( @xmath131 ) and 1.7 gyr ( @xmath132 ) . but since the x - ray measurement is an upper limit , the age indicated by this method is likely much greater . another indicator of activity is @xmath196 , where @xmath197 is the ratio of chromospheric emission in the h and k lines and the bolometric luminosity @xcite . for this indicator chromospheric emission is measured using the mount wilson hk pseudo - equivalent width @xmath198 ( see below ) . we can use a relation between age and @xmath197 from @xcite based upon observations of clusters spanning a wide range of ages : @xmath199 where the age @xmath126 is in years and @xmath200 . using equation ( [ chromage ] ) with @xmath201 @xcite gives an age of 6.4 gyr for hd 89744a . @xcite found @xmath202 and derived an age of 8.4 gyr using equation ( [ chromage ] ) as well . chromospheric emission is often measured using the mount wilson index @xmath203 $ ] , where @xmath204 and @xmath118 are counts in the passbands and @xmath205 and @xmath206 are counts in violet and red continuum bands adjacent to the h - k regions @xcite . @xcite used an empirical relationship between the mean values of @xmath207 and @xmath208 to derive an age of 9.9 gyr . this gives age estimates from activity ranging from 6.4 to 9.9 gyr , in marked contrast to the @xmath209 gyr age from isochrone fits . why are the chromospheric age estimates above so much older than ages from fits to isochrones ? interestingly , hd 89744a is known to have flat activity cycles ( i.e. no evidence of activity cycles such as the 11 year solar cycle ) . because of its flat chromospheric emission , @xcite used this object as a standard in his studies of activity variability . over a 25 year period the standard deviation of s ( @xmath210 ) for this object was 1.2% @xcite . @xcite postulated that solar - type stars that exhibit prolonged low levels of magnetic activity may be in maunder minimum - like phases . if hd 89744a is in such a phase then measures of activity will lead to old age estimates that are misleading . lastly , we note that evolutionary diagrams predict even @xmath211 stars to reach the tip of the red - giant branch in @xmath212 gyr @xcite ; the @xmath213 hd 89744a would be a white dwarf at the ages predicted by its activity . the space motion of hd 89744a is @xmath214 km @xmath145 , @xmath215 km @xmath145 , and @xmath216 km @xmath145 @xcite . this space motion falls within the uv ` eggen box ' corresponding to the young disk with ages @xmath217 gyr @xcite . based upon the well correlated published ages from isochrone fits , as well as evidence of proximity to the main sequence turn - off , we adopt an age of 1.5 - 3 gyr for hd 89744a . we ignore age estimates from activity . with estimated ages for the l - dwarf companions from [ ages ] and effective temperature estimates from spectral types ( [ os ] ) , the masses of the l - dwarf companions can be estimated using evolutionary models from @xcite . in figure [ evolutionary curves ] error boxes are plotted over the evolutionary curves to represent the range of effective temperatures and ages assigned to the objects . based on the lithium test results we also use the @xmath218 curve as the minimum mass boundary for gl 618.1b since we were unable to constrain a lower limit age from physical properties of this object s primary ( [ gl 618.1b age ] ) . masses can then be easily determined from the figure : @xmath219 , @xmath220 , and @xmath221 . thus gl 337c is a brown dwarf , hd 89744b is a very low - mass star , and gl 618.1b could be either . properties of the l - dwarf companions are summarized in table [ secparameters ] . the total mass of the sb2 binary gl 337ab is well constrained from astrometry to be @xmath222 @xcite . a mass for gl 618.1a can be derived using its absolute @xmath118 magnitude and an appropriate mass - luminosity relation ( mlr ) for sub - solar mass stars . using @xmath223 @xcite and an hipparcos distance of 30.3 pc gives @xmath224 . @xcite give an empirically determined mlr of @xmath225 for @xmath226 to @xmath227 . using equation ( [ mlr ] ) we calculate @xmath228 . @xcite adopted a mass of @xmath229 for hd 89744a from the average of published results . these authors determined @xmath230 for the extrasolar planet hd 89744ab . for a large sample @xcite have shown @xmath231 . assuming @xmath232 for the planet s orbit is reasonably close to this expectation value , the planet s mass will be negligible compared to that of the primary . hence we adopt a combined mass of @xmath233 for the primary @xmath1 planet . based on the primary and secondary masses derived above , upper limit mass ratios for the gl 337abc , gl 618.1ab , and hd 89744abb systems are 0.04 , 0.12 , and 0.06 , respectively ( table [ sysparameters ] ) . of the six previously discovered systems with wide - separation l - dwarfs , 4 of 6 have @xmath11 . it is interesting to ask how many of these systems meet the sample criteria of the @xcite g - dwarf binarity study ? @xcite ( hereafter dm91 ) specifically take as a sample f7 to g9 stars from the catalogue of nearby stars ( second edition ; @xcite ) with luminosity class iv - v , v , and vi , declinations above @xmath234 , and trigonometric parallax @xmath235 , i.e. stars within 22 pc . two of the primary systems in their study , gl 417a and gl 584ab , have wide - separation l - dwarf companions @xcite that were not found by the dm91 radial velocity studies of the primaries . one of the primaries in this paper , gl 337ab , meets the dm91 criteria except it was not added to the catalogue of nearby stars ( cns ) until the third edition in 1991 @xcite and it was assigned a combined k0 v spectral type . ( we adopted the spectral type assignment g8 v + k1 v ) . given these newly discovered systems and improvements in the already well studied cns , it would be useful to search for wide binaries around a current dm91 style g - dwarf sample to improve our understanding of the distribution of mass ratios below @xmath236 . for instance , 51 peg ( gl 882 ) , with its recently discovered extrasolar planet , is in the dm91 sample . ] returning to the three wide - companions in this paper , we find it interesting to compare these systems to other binaries in ( @xmath237 ) and ( @xmath238 ) plots . as discussed in @xcite , it is reasonable that wide binaries will have small @xmath6 ( where @xmath239 ) since system binding energies are proportional to @xmath240 . the three binaries in this paper , all with @xmath241 , follow this expectation , as do all previously discovered wide separation binaries with l dwarf companions . @xcite also find a characteristic upper limit radius for binary separation as a function of total mass in their ( @xmath238 ) plot of l dwarf and m dwarf binaries in the solar neighborhood and hst hyades and field binaries , where @xmath242 has units of au . empirically they find a cut - off defined by the log - normal relation @xmath243 the systems gl 337 and hd 89744 have total masses over @xmath244 so their separations are at least three orders of magnitude smaller than the maxima predicted by equation [ binding ] . the gl 618.1b system , an m - dwarf / l - dwarf pair with @xmath245 and @xmath246 , comes closer to the cut - off but is still less than the @xmath247 prediction from equation [ binding ] . thus all three companions have @xmath242 that fall well within the upper limits . we have discovered two wide separation l - dwarf common proper motion companions to nearby stars and identified a third candidate from 2mass . spectral types assigned from optical spectroscopy were l0 v , l2.5 v , and l8 v. nir low resolution spectra of the companions were provided as well as a grid of known objects spanning m6 v t dwarfs to support spectral type assignment for these and future l - dwarfs in the _ zjhk _ bands . using published measurements , we estimated ages of the companions from physical properties of the primaries . these crude ages allowed us to estimate companion masses using theoretical low - mass star and brown dwarf evolutionary models . we found that gl 337c is a brown dwarf , hd 89744b is a vlm star , and gl 618.1b may be either . with the addition of these new l - dwarfs there are nine wide - binary ( @xmath0 au ) l - dwarf companions of nearby stars known . these discoveries improve the statistics of binaries involving low - mass stars and brown dwarfs . in particular they support initial conclusions that the brown dwarf desert seen at small separations around main sequence stars does not extend to wide separations . the discovery of these companions in low mass ratio systems ( @xmath248 ) , along with similar discoveries previously reported , will also significantly improve our understanding of the binary mass ratio distribution for @xmath10 . we thank the anonymous referee for helpful comments that led to a more concise manuscript , john carr for a careful reading of a previous draft , and adam burrows for providing his evolutionary curves in electronic format . we thank andrea richichi and isabelle baraffe for supplying filter curves , and isabelle baraffe and france allard for helpful discussions . we also wish to thank the palomar night assistants skip staples and karl dunscombe for their expertise and support . this publication makes use of data from the two micron all sky survey , which is a joint project of the university of massachusetts and the infrared processing and analysis center , funded by the national aeronautics and space administration and the national science foundation . this research has also made use of the simbad database , operated at cds , strasbourg , france . dss images were obtained from the canadian astronomy data centre , which is operated by the herzberg institute of astrophysics , national research council of canada . allende prieto , c. & lambert , d. l. 1999 , , 352 , 555 arribas , s. & roger , c. m. 1989 , , 215 , 305 balachandran , s. 1990 , , 354 , 310 balachandran , s. 1995 , , 446 , 203 baliunas , s. & vaughan , a. 1985 , , 23 , 379 baliunas , s. & jastrow , r. 1990 , , 348 , 520 baliunas , s. , et al . 1995 , , 438 , 269 baliunas , s. , sokoloff , d. , & soon , w. 1996 , , 457 , l99 baraffe , i. , chabrier , g. , allard , f. , & hauschildt , p. h. 1998 , , 337 , 403 barnaby , d. , spillar , e. , christou , j. c. , & drummond , j. d. 2000 , , 119 , 378 barry , d. c. 1988 , , 334 , 436 bartkeviius , a. & lazauskait , r. 1996 , baltic astronomy , 5 , 1 basri , g. et al . 2000 , , 538 , 363 bertelli , g. , bressan , a. , chiosi , c. , fagotto , f. , & nasi , e. 1994 , astronomy & astrophysics supplement series , 106 , 275 binney , j. , dehnen , w. , & bertelli , g. 2000 , , 318 , 658 burgasser , a. j. , wilson , j. c. , kirkpatrick , j. d. , skrutskie , m. f. , colonno , m. r. , enos , a. t. , smith , j. d. , henderson , c. p. , gizis , j. e. , brown , m. e. , & houck , j. r. 2000 , , 120 , 1100 burrows , a. , marley , m. , hubbard , w. b. , lunine , j.i . , guillot , t. , saumon , d. , freedman , r. , sudarsky , d. , & sharp , c. 1997 , , 491 , 856 carney , b. w. 1979 , , 233 , 211 chandrasekhar , s. & mnch , g. 1950 , , 111 , 142 cutri , r. m. , et al . 2000 , explanatory supplement to the 2mass second incremental data release , http://www.ipac.caltech.edu/2mass/releases/second/doc/explsup.html de medeiros , j. r. , do nascimento , j. d. , jr . , & mayor , m. 1997 , , 317 , 701 delfosse , x. et al . 1997 , , 327 , l25 delfosse , x. , tinney , c. g. , forveille , t. , epchtein , n. , borsenberger , j. , fouqu , p. , kimeswenger , s. & tiphne , d. 1999 , a&ass , 135 , 41 delfosse , x. , forveille , t. , sgransan , d. , beuzit , j .- l . , udry , s. , perrier , c. , & mayor , m. 2000 , , 364 , 217 donahue , r. a. 1993 , phd thesis , new mexico state university duquennoy , a. & mayor , m. 1991 , , 248 , 485 , dm91 duquennoy , a. , mayor , m. , & mermilliod , j .- c . 1992 , in binaries as tracers of stellar formation , ed . m. duquennoy & m. mayor ( cambridge : cambridge univ . press ) , 52 edvardsson , b. , andersen , j. , gustafsson , b. , lambert , d. l. , nissen , p. e. , tomkin , j. 1993 , , 275 , 101 eggen , o. j. 1989b , , 101 , 366 eggen , o. j. 1998 , , 115 , 2397 epchtein , n. 1997 , in the impact of large - scale near - ir sky surveys , ed . f. garzon ( dordrecht : kluwer ) , 15 fukugita , m. , ichikawa , t. , gunn , j. e. , doi , m. , shimasaku , k. , & schneider , d. p. , 1996 , , 111 , 1748 gaidos , e. j. 1998 , , 110 , 1259 girardi , l. , bressan , a. , bertelli , g. , & chiosi , c. 2000 , astronomy & astrophysics supplementary series , 141 , 371 gizis , j. e. , monet , d. g. , reid , i. n. , kirkpatrick , j. d. , liebert , j. , & williams , r. j. 2000 , , 120 , 1085 gizis , j. e. , kirkpatrick , j. d. , burgasser , a. , reid , i. n. , monet , d. g. , liebert , j. & wilson , j. c. 2001a , , 551 , 163 gizis , j. e. , kirkpatrick , j. d. , & wilson , j. c. 2001b , , 121 , 2185 gliese , w. 1969 , catalogue of nearby stars ( 2nd ed . ; heidelberg : astron . rechen - institut ) gliese , w. & jahreiss , h. 1991 , preliminary version of the third catalogue of nearby stars ( heidelberg : astron . rechen - institut ) gonzalez , g. , laws , c. , tyagi , s. , & reddy , b. e. 2001 , , 121 , 432 hawley , s. l. , gizis , j. e. , & reid , i. n. 1996 , , 112 , 2799 henry , t. j. & mccarthy , d. w. , jr . 1993 , , 106 , 773 hoffleit , d. & warren , w. h. , jr . 1991 , the bright star catalogue , 5th revised ed . ( preliminary version ) hnsch , m. , schmitt , j. h. m. m. , sterzik , m. f. , & voges , w. 1999 , a&ass , 135 , 319 iben , i. , jr . 1967 , , 147 , 624 jones , h. r. a. , longmore , a. j. , jameson , r. f. , & mountain , c. m. 1994 , , 267 , 413 kirkpatrick , j. d. , henry , t. j. , & mccarthy , d. w. , jr . 1991 , , 77 , 417 kirkpatrick , j. d. , et al . 1999 , , 519 , 802 , k99 kirkpatrick , j. d. , reid , i. n. , liebert , j. , gizis , j. e. , burgasser , a. j. , monet , d. g. , dahn , c. c. , & nelson , b. 2000 , 120 , 447 , k00 kirkpatrick , j. d. , dahn , c. c. , monet , d. g. , reid , i. n. , gizis , j. e. , liebert , j. , & burgasser , a. j. 2001 , , 121 , 3235 klemperer , w. , norris , w. g. , & bchler , a. 1960 , , 33 , 1534 korzennik , s. g. , brown , t. m. , fischer , d. a. , nisenson , p. , & noyes , r. w. 2000 , , 533 , l147 lbre , a. , de laverny , p. , de medeiros , j. r. , charbonnel , c. , & da silva , l. 1999 , , 345 , 936 leggett , s. k. 1992 , , 82 , 351 lodders , k. 1999 , , 519 , 793 maggio , a. , sciortino , s. , vaiana , g. s. , major , p. , bookbinder , j. , golub , l. , harnden , f. r. , & rosner , r. 1987 , , 315 , 687 marcy , g. w. & butler , r. p. 2000 , , 112 , 137 mason , b. d. , mcallister , h. a. , & hartkopf , w. i. 1996 , , 112 , 276 mason , b. d. , wycoff , g. l. , & hartkopf , w. i. , the washington double star catalog , u. s. naval observatory , http://ad.usno.navy.mil/proj/wds/wds.html , updated 2001 mar 05 mclean , i. s. et al . 2000 , , 533 , l45 nelson , l. a. , rappaport , s. , & chiang , e. 1993 , , 413 , 364 ng , y. k. , & bertelli , g. 1998 , , 329 , 943 noyes , r. w. , hartmann , l. w. , baliunas , s. l. , duncan , d. k. , & vaughan , a. h. 1984 , , 279 , 763 oke , j. b. , et al . 1995 , , 107 , 375 perryman , m. a. c. et al . 1997 , , 323 , l49 pinsonneault , m. 1997 , , 35 , 557 pizzolato , n. , maggio , a. , & sciortino , s. 2000 , , 361 , 614 pourbaix , d. 2000 , a&as , 145 , 215 rebolo , r. , martn , e. l. , & magazz , a. 1992 , , 389 , l83 reid , i. n. et al . 1991 , , 103 , 661 reid , i. n. , hawley , s. l. , & gizis , j. e. 1995 , , 110 , 1838 reid , i. n. , kirkpatrick , j. d. , gizis , j. e. , dahn , c. c. , monet , d. g. , williams , r. j. , liebert , j. , & burgasser , a. j. 2000 , , 119 , 369 reid , i. n. & hawley , s. l. 2000 , new light on dark stars ( springer - verlag : berlin ) , 196 reid , i. n. , gizis , j. e. , kirkpatrick , j. d. , & koerner , d. w. 2001a , , 121 , 489 reid , i. n. , burgasser , a. j. , cruz , k. l. , kirkpatrick , j. d. , & gizis , j. e. 2001b , , 121 , 1710 richichi , a. , ragland , s. , calamai , g. , richter , s. , & stecklum , b. 2000 , , 361 , 594 saar , s. h. & brandenburg , a. 1999 , , 524 , 295 skrutskie , m. f. , et al . 1997 , in the impact of large - scale near - ir sky surveys , ed . f. garzon ( dordrecht : kluwer ) , 25 skumanich , a. 1972 , , 171 , 565 soderblom , d. r. 1985 , , 90 , 2103 testi , l. et al . 2001 , , 552 , l147 tinney , c. g. , mould , j. r. , & reid , i. n. 1993 , , 105 , 1045 tokunaga , a. t. , & kobayashi , n. 1999 , , 117 , 1010 vaughan , a. h. , preston , g. w. , & wilson , o. c. 1978 , , 90 , 267 walter , f. m. , & barry , d. c. 1991 , in the sun in time , ed . c. p. sonett , m. s. giampapa , & m. s. matthews ( tucson : univ . arizona press ) , 633 wilson , j. c. , skrutskie , m. f. , colonno , m. r. , enos , a. t. , smith , j. d. , henderson , c. p. , gizis , j. e. , monet , d. , & houck , j. r. 2001a , , 113 , 227 wilson , j. c. , gizis , j. e. , skrutskie , m. f. , miller , n. , j. davy kirkpatrick , monet , d. g. , & houck , j. r. 2001b , in preparation wilson , o. c. 1968 , , 153 , 221 york , d. g. , et al . 2000 , , 120 , 1579 zahn , j .- 1977 , , 57 , 383 2massw j0912145 + 145940 ( 2m0912 + 14 ) & 09:12:14.5 & + 14:59:40 & @xmath249 & @xmath250 & @xmath251 & @xmath252 + 2massw j1620261 - 041631 ( 2m1620 - 04 ) & 16:20:26.1 & -04:16:31 & @xmath253 & @xmath254 & @xmath255 & @xmath256 + 2massi j1022148 + 411426 ( 2m1022 + 41 ) & 10:22:14.8 & + 41:14:26 & @xmath257 & @xmath258 & @xmath259 & @xmath260 + 2m0912 + 14 ( gl 337c ) & 2000 may 10 & 4000 & palomar 60-inch & cormass + & 2000 dec 26 & 2400 & keck i & lris + 2m1620 - 04 ( gl 618.1b ) & 2000 may 9 & 3300 & palomar 60-inch & cormass + & 2000 oct 7 & 2400 & keck i & lris + 2m1022 + 41 ( hd 89744b ) & 2000 may 10 & 4550 & palomar 60-inch & cormass + & 2000 dec 26 & 900 & keck i & lris + m6 v & lhs 2034 & 1999 oct 27 & 600 & & 10.03 & & 0.97 & 1 + m8 v & vb 10 & 1999 aug 22 & 720 & & 8.80 & & 1.10 & 1 + m9 v & lhs 2924 & 2000 jun 16 & 2400 & & 10.67 & & 1.17 & 1 + m9.5 v & bri 0021 - 0214 & 1999 aug 24 & 1080 & & 10.64 & & 1.26 & 2 + l1 v & 2massw j0208183 + 254253 & 1999 aug 24 & 2700 & 12.58 & & 1.44 & & 3 + l2 v & 2massw j0015447 + 351603 & 2000 sep 18 & 3300 & 12.24 & & 1.58 & & 3 + l3.5 v & 2massw j0036159 + 182110 & 1999 aug 24 & 1440 & 11.03 & & 1.41 & & 4 + l5 v & 2massw j1507476 - 162738 & 2000 apr 20 & 2800 & 11.30 & & 1.52 & & 4 + l7 v & denis - p j0205.4 - 1159ab & 2000 sep 16 & 5100 & 12.99 & & 1.56 & & 5 + l8 v & 2massw j1632291 + 190441 & 2000 jun 14 & 5500 & 13.98 & & 1.88 & & 6 + t & 2massw j0559191 - 140448 & 1999 oct 24 & 2400 & 13.61 & & 0.22 & & 7 + lcccccccc spectral type & g8 v @xmath1 k1 v & 1 & & m0 v & 8 & & f7 iv - v & 11 + & 0.73 & 2 & & 1.4 & 9 & & 0.54 & 10 + distance ( pc ) & @xmath261 & 3 & & @xmath262 & 3 & & @xmath263 & 3 + @xmath264 & @xmath265 & 4 & & 0.67 & - & & @xmath266 & - + activity & & & & & & & & + @xmath267 & @xmath268 & 5,6 & & not detected & 6 & & @xmath269 & 12 + @xmath196 & & & & & & & @xmath270,@xmath271 & 10,13 + @xmath272 ( ew ) & & & & @xmath273 & 10 & & & + @xmath274 & & & & & & & 9 & 13 + period & & & & & & & flat & 14 + metallicity & & & & & & & & + @xmath275}$ ] & @xmath276 & 2,7 & & & & & + 0.24 & 15,16 + @xmath277}$ ] & & & & @xmath278 & 9 & & & + lithium abundance & & & & & & & & + @xmath279 & & & & & & & 2.07 & 16 + kinematics & & & & & & & & + @xmath280 & @xmath281 & 7 & & @xmath282 & 9 & & @xmath283 & 15 + @xmath284 & @xmath285 & 7 & & @xmath286 & 9 & & @xmath287 & 15 + @xmath288 & @xmath289 & 7 & & @xmath290 & 9 & & @xmath291 & 15 +

we present the ground state phase diagram of quantum - rotor chains with competing interactions ( frustration ) calculated through cluster variational mean field approaches . we consider two interaction patterns , named f and f models , between the quantum - rotor momentum and position operators , which follow exchange patterns of known one - dimensional spin-1/2 systems with a ferrimagnetic state in their phase diagrams . the spin-1/2 f model is known as the diamond chain and is related to the azurite compound , while the spin-1/2 f model was recently shown to present a frustration - induced condensation of magnons . we provide a detailed comparison between the quantum - rotor phase diagrams , in single- and multi - site mean - field approaches , and known results for the spin-1/2 models , including exact diagonalization and density matrix renormalization group data for these systems , as well as phase diagrams of the associated classical models .

1602.05383

the connection between o(@xmath3 ) quantum - rotor ( qr ) and spin models on has proved very useful in the context of phase transitions @xcite . about three decades ago , hamer , kogut and susskind @xcite mapped o(@xmath3 ) heisenberg models onto the corresponding [ ( 1 + 1 ) spatial and time dimensions ] nonlinear - sigma or qr models . the critical behavior was then inferred using strong - coupling expansion ( high - temperature , , where @xmath4 is the spin coupling ) : a kosterlitz - thouless transition for the o(2 ) model and a prediction of critical points at zero coupling ( pad@xmath5 continued ) for both o(3 ) and o(4 ) models . on the other hand , by mapping o(3 ) antiferromagnetic ( af ) heisenberg chains onto nonlinear sigma models in the semiclassical weak - coupling limit , haldane @xcite suggested that the ground state ( gs ) of chains with integral spins are gapped , while those with half - integral spins are gapless . moreover , shankar and read @xcite precisely clarified the distinction between gapped af spin models , characterized by the @xmath6 = 0 _ mod _ 2@xmath7 topological term , and gapless models for which @xmath8 _ mod _ 2@xmath7 , including the connection of the latter with a laplacian minimally coupled to the monopole potential@xcite . following the above developments , sachdev and senthil @xcite have presented a quite general mean - field and renormalization - group analysis of quantum phase transitions in magnets with the aid of _ generalized _ qr models . in particular , they showed that , under certain conditions , one can establish a mapping of double - layer antiferromagnets onto quantum rotors which sheds intuitive light on the way in which a qr can be used as an effective representation of a pair of antiferromagnetically coupled spins . still in this context , a single - site mf approximation was used to study an effective hamiltonian for spin - one bosons in an optical lattice in the presence of a magnetic field @xcite . further , a qr description of the mott - insulator transition in the bose - hubbard model within a functional - integral approach has also been elaborated in order to include particle number fluctuation effects @xcite . in this work we focus our attention on the study of the gs phase diagram of generalized quantum rotors on the frustrated ab@xmath2 chain , which is depicted in . the quantum rotors at each site are constrained , through sufficiently high values of the coupling @xmath9 ( and the coupling @xmath10 of the quartic term in the angular momentum ) , to mostly retain states with the minimum value of the angular momentum , i.e. , @xmath11 , as the frustration parameter @xmath4 is varied , thus enabling us to make a direct comparison with the corresponding _ quantum _ spin-1/2 ab@xmath2 chains . we analyze two types of frustration , as illustrated in , and try to interpret the derived phase diagrams in light of the ones of previous works on frustrated quantum spin-1/2 chains with the ab@xmath2 topology @xcite . instead of attempting to formalize a specific ( and probably rather complex ) mapping between the rotor and the spin models , we have opted to treat the rotor chain numerically by using a cluster variational mf theory , supplemented with exact diagonalization ( ed ) via lanczos algorithm@xcite and density matrix renormalization group ( dmrg ) @xcite of finite - size spin-1/2 chains . with respect to spin systems , as a motivation on the experimental side , the compound azurite @xcite has been successfully explained by the distorted diamond chain model@xcite , i.e. , a system with three spin-1/2 magnetic sites per unit cell and frustrated ferrimagnetic state . also , along with the study on the effect of frustration@xcite , for @xmath12 this class of models shares its phenomenology and unit - cell topology with quasi - one - dimensional compounds , such as the line of trimer clusters present in copper phosphates@xcite and the organic ferrimagnet pnnbno@xcite . the modeling of the ferrimagnetic phase @xcite has been mainly undertaken in the context of other models such as hubbard@xcite , @xmath13@xcite , ising@xcite , classical@xcite and quantum heisenberg@xcite , including magnetic excitations @xcite , and the quantum spherical model@xcite . the occurrence of new phases induced by hole - doping of the electronic band@xcite has also been carried out . this paper is organized as follows . in the next section we describe our qr model and numerical methods , and include in appendix a a derivation of the matrix elements of the operators acting on the single - site hilbert space represented by monopole harmonics . in we use single - site variational mf theory to study the rotor models , for the two frustration cases , and discuss the shortcomings of this semiclassical approach . then in we adopt a multi - site ( two - unit cell ) variational mf hamiltonian , which provides a substantial improvement on the treatment of quantum fluctuation effects , particularly in connection with the case of frustrated interaction between quantum rotors on b sites at the same unit cell . here we treat the respective spin-1/2 systems by making use of ed and dmrg techniques in order to pave the way for a direct comparison between rotors and spins . finally we report our conclusions in . quantum rotors can be classified according to their minimum angular momentum@xcite : rotors with @xmath14 have zero minimum angular momentum , which can be made to correspond to an even number of heisenberg spins in an underlying spin model . on the other hand , we also have rotors with @xmath15 , where @xmath16 is chosen to have one of the values : 1/2 , 1 , 3/2 , ... , as will be briefly clarified below . this is an extension of the former case , and quantum rotors with half - integer values of @xmath16 are duly suited to refer to an odd number of underlying spins-1/2 ( at least one spin remains unpaired ) . we shall focus on @xmath0-quantum rotors in view of the stated objective of comparing our results with those for the referred chains of spin-1/2 operators . the three - component unit vector ( operator ) , with @xmath17 , describes the configuration space ( @xmath3 space ) of a rotor , while stands for the canonically conjugate angular momenta . setting @xmath18 , these quantities obey the commutation relations ( operators at different sites commute ) : @xmath19 } & = & i\epsilon_{\mu\nu\lambda}\hat{l}_{\lambda } , \nonumber \\ { [ } \hat{l}_{\mu},\hat{n}_{\nu } { ] } & = & i\epsilon_{\mu\nu\lambda}\hat{n}_{\lambda } , \nonumber \\ { [ } \hat{n}_{\mu},\hat{n}_{\nu } { ] } & = & 0 , \label{comm}\end{aligned}\ ] ] where the greek letters stand for the cartesian components @xmath20 ( summation over repeated indices is subtended and @xmath21 is the levi - civita tensor ) and @xmath22 -q\hat{n}_{\mu } , \label{lmonopole}\ ] ] which incorporates the effect of a dirac monopole at the origin of @xmath3 space , whose vector potential may be conveniently chosen to satisfy @xmath23 the appropriate hilbert space is made up of _ angular section _ states for which the following are true @xcite : @xmath24 ; @xmath25 ; and the usual ladder operators ( @xmath26 ) satisfy @xmath27 . here @xmath28 , and @xmath29 . the @xmath30 are the eigensections , also called _ monopole harmonics_. an important constraint follows immediately from : @xmath31 thus , following ref . 7 , we shall consider the quite general frustrated o(3 ) qr hamiltonian : @xmath32 + \nonumber \\ & & \!\!\!\ ! \sum_{<ij>}[{\hat{\bf{n}}_{i } } \cdot \hat{\bf{n}}_{j}+{\hat{\bf{l}}_{i}}\cdot \hat{\bf{l}}_{j } + m({\hat{\bf{n}}_{i } } \cdot \hat{\bf{l}}_{j}+{\hat{\bf{n}}_{j } } \cdot \hat{\bf{l}}_{i})]+ \nonumber\\ & & \!\!\!\!\!\!\!\!\!\!\!\!\ ! \sum_{(i , j ) \in f_1 or f_2}\!\!\!\!\!\!\!\![j({\hat{\bf{n}}_{i } } \cdot \hat{\bf{n}}_{j}+{\hat{\bf{l}}_{i } } \cdot \hat{\bf{l}}_{j})+m ( { \hat{\bf{n}}_{i } } \cdot \hat{\bf{l}}_{j}+{\hat{\bf{n}}_{j } } \cdot \hat{\bf{l}}_{i } ) ] , \label{hr } \end{aligned}\ ] ] where @xmath9 and @xmath10 are positive local couplings , i.e. , associated with rotors at each site @xmath33 of the ab@xmath2 chain ( the quartic term appears with the main objective of controlling contributions of high - energy states ) ; in the second summation , @xmath34 indexes nearest - neighbor couplings between rotors on distinct sublattices which , except for @xmath35 , are all set to unity ( see ) ; in the third summation @xmath36 indexes nearest - neighbor couplings between rotors on the same sublattice which , except for @xmath35 , are set to @xmath37 . here we shall study two frustration patterns , namely , @xmath38 and @xmath39 . in @xmath38 only frustrated interactions ( @xmath4 and @xmath35 ) between rotors at b sites of the same unit cell are present , as illustrated in fig . [ figure01](a ) , whereas for f@xmath2 we consider _ all _ nearest - neighbor intra- and intercell interactions , as illustrated in fig . [ figure01](b ) . in order to isolate the effect of the coupling @xmath35 in the two above - referred cases , we take either @xmath40 or @xmath41 . before going on to the approaches described in and , we emphasize the following features about the _ stability _ of the numerical implementations carried out in this work . in our simulations we have verified that we could work safely with a minimally reduced hilbert space if the values of @xmath9 and @xmath10 were set sufficiently large . in fact , the hilbert space size and the value of @xmath9 and @xmath10 determine the stability of our problem : for small space sizes ( e.g. , @xmath42 ) and small values of @xmath9 and @xmath10 ( _ e.g. _ , @xmath43 ) , the system becomes completely unstable due to contributions of high - energy terms which cause the system to fluctuate beyond control . on the other hand , by choosing a small space size ( @xmath42 ) , but a sufficiently large value of @xmath9 , the system behaves quite stably . therefore , in this work we shall use small space size , i.e. , @xmath42 , associated with a large value of @xmath9 , in order to make computations feasible and a close contact with spin-1/2 models . we then start off by treating @xmath44 by means of a variational mf theory based on the bogoliubov theorem @xcite . thus , the variational expression of the mf energy at @xmath45 satisfies the inequality : @xmath46 where @xmath47 is the gs energy of the trial hamiltonian - here denoted by @xmath48 - and the expectation value is taken with respect to its gs wavefunction . ( [ des - bogol ] ) is then minimized with respect to its variational parameters : for chosen values of the frustration control parameter ( @xmath4 ) , minimization is carried out numerically through diagonalization of @xmath48 by way of the lanczos algorithm @xcite , and then deploying a simplex procedure @xcite . for trial hamiltonians we use both single - site and multi - site hamiltonians , as described in and , respectively . , @xmath49 , and @xmath50 . full lines indicate antiferromagnetic exchange couplings ( @xmath51 ) which give rise to the ferrimagnetic gs , while dashed lines represent exchange couplings ( @xmath52 ) which frustrate the magnetic order : ( _ a _ ) frustration pattern f@xmath1 and ( _ b _ ) frustration pattern f@xmath2.,scaledwidth=40.0% ] as a first and straightforward application of the aforementioned variational mf theory , we postulate the following trial hamiltonian , acting on one unit cell : @xmath53 , \label{hsingle}\ ] ] where and are the variational c - number fields and the subscript @xmath33 goes over the sites @xmath54 , @xmath55 , and @xmath56 . the gs wavefunction of @xmath48 and energy are given by @xmath57 and @xmath58 , where @xmath59 represents the gs energy of the respective wavefunction , such that for any pair of operators @xmath60 , with @xmath61 : @xmath62 we then get , for frustration f@xmath1 , the bogoliubov inequality for the unit cell : @xmath63 , where the @xmath64 read : @xmath65;\end{aligned}\ ] ] and @xmath66 the index @xmath33 ( @xmath67 ) visits the sites of the unit cell , with the convention : @xmath68 . for frustration f@xmath2 , a fourth term must be added to the bogoliubov inequality : @xmath69 . \end{aligned}\ ] ] the mf energy best evaluation of @xmath70 or @xmath71 is then obtained by performing the minimization with respect to variations of the fields @xmath72 and @xmath73 . to produce the results of this section , it sufficed to set @xmath74 and a space size determined by truncating the hilbert space at @xmath42 . further , we have focused only on those quantities that suffice to afford the relevant information needed for the proper interpretation of the problem at this level , i. e. , the two - point mf momentum products , defined here through the products , where @xmath75 runs over the sites of the unit cell , and the mf energy . we thereby leave out the position- and momentum - position products , for they are redundant . this is due to the fact that @xmath76 and @xmath77 have the same signature under all allowed symmetries for q@xmath78 , and so their expectation values turn out to be proportional to each other on a given site @xcite . . two - point mf momentum products [ ( _ a _ ) @xmath79 , ( _ c _ ) @xmath80 between the indicated rotors and mf energy curve [ ( _ b _ ) @xmath79 , ( _ d _ ) @xmath80 , where we have drawn straight ( full ) lines to show that at @xmath81 the system steers away from the linear regime that prevails for @xmath82 and so a phase transition takes place . full and dashed lines in ( _ a _ ) and ( _ c _ ) indicate the results of the classical vector model . dashed lines in ( _ b _ ) and ( _ d _ ) are guides to the eye.,scaledwidth=47.0% ] . same as in .,scaledwidth=47.0% ] we then proceed to discuss the results in ( frustration f@xmath1 ) and ( frustration f@xmath2 ) , which reveal some salient features . firstly , we verified that , with @xmath83 , independent of @xmath4 . the momentum products show that , _ in all cases _ , the system starts out with a magnetization plateau : and , which corresponds to the lieb - mattis@xcite phase of the analogous spin-1/2 system , before undergoing a phase transition at . this transition is of second order , as shown in and , and of first order , as shown in and . in the first case ( @xmath79 ) , the system evolves continually ( with the mf energy curve - and - smooth at the point @xmath81 ) to a stable phase where the momenta at a and b sites become uncorrelated , i.e. , , while the momenta at b sites tend to directly oppose each other with increasing @xmath4 , forming a singlet - like configuration : , for . in the second case ( @xmath84 ) , the transition takes place quite abruptly , having undoubtedly first - order characteristics , and the system immediately accommodates into the stable singlet - like phase that we have just referred to . the mf energy curves of these first - order transitions at @xmath81 are shown in and , where we notice that the cusp in the latter is less pronounced . we notice further , that the products between the momenta at a and b sites display quite sizable fluctuations around , as seen in and ( @xmath85 ) , and in lesser degree in for frustration @xmath39 and @xmath86 . the corresponding wide points occur pairwise and fairly symmetrically with respect to classical curves ( see below ) that represent the decoupling of the momenta at a and b sites , leaving the mf energy practically unaltered . in fact , with increasing @xmath4 , the system becomes more prone to wandering through near - degenerate states , which give rise to these off points . the phase , for @xmath87 , with a sites uncoupled and b sites with opposing momenta in a singlet - like configuration is much like the dimer - monomer phase of the work by takano , kubo , and sakamoto @xcite . we perceive , however , that important features in between those @xmath4 extremes of the phase diagram do not appear by way of this naive _ single - site _ mf theory . is the unique order parameter.,scaledwidth=28.0% ] we now notice , through , that with the rotor momenta being fixed at @xmath11 , we have @xmath88 , for all sites @xmath33 , independent of @xmath4 , and so all products can only vary between the extremes -0.25 and 0.25 . therefore , through this mf approach , we are led to envision the momenta on the unit cell of the ab@xmath2 chain as classical vectors of constant magnitude , such as represented in . we can thus provide a simple interpretation based on this configuration of classical vectors on the _ xy _ plane ( akin to the _ xy_-model ) . we then build our energy function for the configuration in ( @xmath89 ) on a symmetric unit of the ab@xmath2 chain centered on the a site . the classical constraints may be set as @xmath90 and @xmath91 ( we take @xmath3 constant , which is about true for small @xmath92 , as verified in the simulations , and whose value must be read off from the plots ) . we start off with the energy function for the frustration f@xmath1 , taking into account the cases @xmath40 and @xmath41 . so , for each case , up to a constant independent of @xmath6 : @xmath93 and @xmath94 . upon imposing the minimization conditions ( relative to the unique parameter @xmath6 ) , we obtain : ( i ) for we have @xmath95 , which holds for both @xmath40 and @xmath41 ; ( ii ) for @xmath96 , we have @xmath97 , which in turn implies that @xmath98 , for @xmath40 , while @xmath99 , for @xmath41 . the momentum products are accordingly given by : for , @xmath100 and @xmath101 , for both @xmath40 and @xmath41 . for j@xmath102 , @xmath103 @xmath104}^2}\!-\!\!1),\hspace{0.05 in } m=1 . \label{cclassic3}\end{aligned}\ ] ] with respect to frustration f@xmath2 , our present mf approach can only sense " a repetition of the configuration of , in that we get additional terms to the energy functions above that are independent of @xmath6 ( and therefore vanish upon minimization ) , implying the same results for the dot products . this classical description fully accounts for the momentum products in both frustration types for @xmath40 , including the nature of the phase transition at @xmath81 , as seen in and through the matching fitting of the points of the numerical implementation for the rotors ; for @xmath41 , this interpretation confirms the first - order transition at @xmath81 and offers hints at the expected behavior of these momentum products , were it not for the off points , as can be seen in the diagrams of and . in eq . ( [ cclassic3 ] ) we have used , that can be read off from plots of @xmath105 , which were not explicitly presented in this work . in the following section we try a _ more elaborate _ mf technique on a double - cell structure , as a way to circumvent ) , as well as to get a direct evaluation of the intercell two - point correlations . for simplicity , we restrict ourselves to @xmath40 . differently from the approach presented in sec . iii , we build our trial hamiltonian acting on the _ global _ space formed by the six sites of the double - cell structure made up of two contiguous unit cells , such as showed in , i.e. , we build one _ six - site _ trial hamiltonian acting , say , on the sites a@xmath1 , a@xmath2 , b@xmath1 , b@xmath2 , b@xmath106 , and b@xmath107 . in order to achieve that , we assigned to each site its own _ local _ vector subspace , and we then constructed our global space by forming the tensor product of these subspaces in one chosen order . in order to simplify the equations below , when necessary , the _ local _ operator acting on the qr located at site a@xmath1 , for instance , was denoted by @xmath108 , which may refer to either operator @xmath109 or operator @xmath110 . we now write down the trial hamiltonian acting on a given double - cell structure for frustration f@xmath1 : @xmath111 the first term ( the kinetic energy term plus effective fields ) is given by @xmath112 , \label{htrial - ext1}\ ] ] where the index @xmath33 goes over the sites a@xmath1 , a@xmath2 , b@xmath1 , b@xmath2 , b@xmath106 , and b@xmath107 ; @xmath113 and @xmath114 being the effective fields ( variational c - numbers ) due to the rest of the system ( which plays the role of a bath ) , and acting on each site @xmath33 of the double - cell cluster . the next term ( first neighbors or bonds ) reads @xmath115 , & & \label{htrial - ext2 } \end{aligned}\ ] ] while the last one ( frustration interaction ) is written as @xmath116 , \label{htrial3-ext}\ ] ] where in the first term we opted to use explicit operators . for frustration f@xmath2 , the following term ( intercell frustration interaction ) must be added to eq . ( [ htrial - ext ] ) : @xmath117 . \label{htrial4-ext}\ ] ] the direct application of yields for frustration f@xmath1 the following expression for the _ double - cell _ variational mf energy @xmath118 , where the equals sign implies that minimization with respect to the variational fields has already been carried out : @xmath119 where , @xmath47 represents the gs energy of @xmath48 and , as before , the index @xmath33 visits the sites a@xmath1 , a@xmath2 , b@xmath1 , b@xmath2 , b@xmath106 , and b@xmath107 . for frustration f@xmath2 , we get analogously @xmath120 : @xmath121 now , a given eigenfunction of @xmath48 may not necessarily be a tensor product of the eigenfunctions of the respective site subspaces as was the case in , so that , for example , @xmath122 where @xmath123 designates the gs wavefunction of @xmath124 . this is an important aspect in our approach , which differs from the standard mf result given by . thus , in principle , taking advantage of the available capability of diagonalizing more complex operators ( trial hamiltonians ) , we can produce more reliable cluster variational mf theories . the dimension of the global space is @xmath125 , where @xmath126 is the dimension of the local subspace , so due to computational implementability , this fact prompted us to limit the size of the hilbert space by deploying rotors with maximum @xmath127 . an observation about the value of g is in order . in our approach , for frustration @xmath38 , when we had set @xmath128 as in the preceding section , we verified that the momentum correlation @xmath129 remains pegged at 0.25 , even after the transition at @xmath130 , which turns out not to be true ( see below ) . therefore , in this section , we resorted to a higher value of @xmath9 ( @xmath131 ) , which inhibited more strongly the appearance of disturbing states ; however one should notice that if much greater values of @xmath9 are employed , the kinetic energy becomes overwhelmingly dominant , so that small changes in the correlations tend to go unnoticed . as before , we concentrated on the relevant quantities that can provide the information needed for the physical interpretation of the problem : namely , _ the mean - field energy _ , the _ expectation value of the total angular momentum _ , and the _ momentum correlations_. the expectation value of the total angular momentum per unit cell was calculated according to the formula , where @xmath132 ( @xmath133 ) is the respective resultant component ( component sum over all the six sites of the double - cell structure ) . in order to facilitate comparison between qr results and those for the spin-1/2 counterpart with the same type of frustration , we present in the phases obtained for the spin-1/2 diamond chain@xcite . the lieb - mattis@xcite ferrimagnetic phase ( ferri ) appears when @xmath134 ( @xmath4 is also used to indicate the frustration control parameter for the spin system ) . in the tetramer - dimer ( td ) phase , which ensues when @xmath135 , the state is precisely the regular array of singlet tetramers ( the closed loop encompasses four spins , in which the b sites form a triplet pair , and the spins on the a sites oppose those on the b sites , so that zero total spin takes place ) , and singlet dimers ( two spins within the elliptical contour ) as shown in . finally , the dimer - monomer ( dm ) state is shown in and sets in when @xmath136 ; it is composed of the regular array of singlet dimers and free spins , and vanishing total spin is also expected . because of the free spins , the dm state is macroscopically @xmath137-fold degenerate for a chain with @xmath138 sites . furthermore , both transitions are of first order @xcite . is increased from @xmath139 . ( _ a _ ) the ferrimagnetic ( ferri ) state . ( _ b _ ) the tetramer - dimer ( td ) state , where rectangles represent singlet tetramers and ellipses singlet dimers . ( _ c _ ) the dimer - monomer ( dm ) state . there are two first - order phase transitions : at @xmath140 ( ferri / td ) and @xmath141 ( td / dm).,scaledwidth=43.0% ] and @xmath142 , both of first order . dashed lines are guides to the eye.,title="fig:",scaledwidth=23.0% ] and @xmath142 , both of first order . dashed lines are guides to the eye.,title="fig:",scaledwidth=23.0% ] in order to allow a direct comparison with our mf results for quantum rotors , we have solved the spin-1/2 diamond chain ( ab@xmath2 chain with frustration between spins at sites b of the same unit cell ) for sizes up to 28 sites , using the lanczos ed procedure with open boundary conditions . the results are displayed as follows : the relevant correlations are represented by the curves plotted in ; in we plotted respectively the energy and total - spin curves ( normalized by the lieb - mattis value @xcite ) . examination of the correlation plots show clear correspondence with the phases exhibited in . the phase ferri is characterized by the following correlations : , @xmath143 , @xmath144 , @xmath145 , @xmath146 . the total spin per unit cell in shows the lieb - mattis value of 0.5 throughout . the transition to the intermediate phase td then occurs at @xmath147 , very close to the estimated value for the infinite chain @xcite : @xmath140 . we note that in this phase the chain breaks up into smaller units and quantum fluctuations within each unit do not affect the spin correlations . hence the correlations are just those calculated for the td configuration of spins in : ( triplets ) , ( singlets ) , , the other correlations being zero . with increasing @xmath4 though , quantum fluctuations become strong enough to disrupt the tetramer unit and a new phase transition to dm phase happens at @xmath130 , this point being independent of size because of the chain breakup . in this phase correlation has varying nonzero values and does not vanish , as expected in the thermodynamic limit , due to finite size effects . on the other hand , the b spins , which are interlocked in singlet units , are totally unaffected . this phase , depicted in , shows the final chain breakup as the tetramer gives way to two monomer units and another dimer , clearly indicated by the correlations in ( @xmath148 ) . the energy curve in exhibits cusps at the transition points , typical of a first - order nature , also verified through the discontinuities of the correlations at these points . the total spin per unit cell in corroborates the above phase description ; however in the last phase the apparent nonzero value is a finite - size effect . finally , getting down to the qr ab@xmath2 chain , we display our variational mf numerical results in ( momentum correlations ) and ( energy and total angular momentum ) and we proceed to a comparative examination with respect to the preceding spin results . a blow - by - blow confrontation of the correlations in both and shows that the double - cell variational mf approach is able to reproduce the three phases exhibited in ref . 10 , namely , the ferri , td , and dm phases . in the ferri phase , quantum fluctuations appear to be equally important , causing the same correlations to deviate somewhat from calculated results for stiff momenta . a closer examination shows that , up to two decimal digits , we have same correlations for @xmath149 , but slightly different correlations , namely : @xmath150 , @xmath151 , @xmath152 , and @xmath153 , for the rotor system , which should be compared with @xmath154 , @xmath155 , @xmath156 , and @xmath157 , for the spin system . phase transitions occur at @xmath158 and @xmath130 , evidently of first order ; in the first transition we have a lesser value ( @xmath158 ) than the lanczos result for the 28-site spin-1/2 chain ( @xmath147 ) , and that of ref . 10 ( @xmath159 ) . the momentum and spin correlations match one another , respectively , in both phases : in the dm phase , correlation @xmath160 shows also an erratic behavior similar to its spin counterpart , in other words , finite size effects are also at play . the minor differences in correlations in the ferri phase , as well as in the first transition point do not constitute a fundamental discrepancy between the respective phase diagrams , which are endowed with the same topological features . in the cusps in the mean - field energy at @xmath161 and @xmath130 also bespeak the occurrence of these first - order transitions . comparing the total momentum in with the total spin in , we observe similar results for the lanczos predictions for the spin model , including finite - size effects in the last phase . . one notices the phase sequence ferri@xmath162td@xmath162 dm , with first - order transitions at @xmath163 and @xmath164 . dashed lines are guides to the eye.,scaledwidth=43.0% ] : ( _ a _ ) the energy plot ( @xmath165 ) shows cusps at the first - order transition points @xmath158 and @xmath142 ; ( _ b _ ) expectation value of the total angular momentum per unit cell . dashed lines are guides to the eye.,title="fig:",scaledwidth=23.0% ] : ( _ a _ ) the energy plot ( @xmath165 ) shows cusps at the first - order transition points @xmath158 and @xmath142 ; ( _ b _ ) expectation value of the total angular momentum per unit cell . dashed lines are guides to the eye.,title="fig:",scaledwidth=23.0% ] it is instructing to study the qr system regarding the average singlet density per unit cell of the b momenta @xcite , which in our case ( _ double - cell _ cluster ) is calculated directly using @xmath166 and which is displayed in . these results permit a direct comparison with the phase diagram of , as far as the buildup of singlet pairs out of b momenta is concerned . as is the case for spins , size effects are not important here , so that one perceives that the number of singlets is very clearly a quantized quantity within each phase . : average singlet density per unit cell for the momenta of the b sites at the same unit cell . one can make out the three phases : ferri , td , and dm , as well as pertinent transitions . dashed lines are guides to the eye.,scaledwidth=30.0% ] before getting down to quantum rotors , we describe succinctly existent results@xcite for the spin-1/2 ab@xmath2 chain with the same frustration f@xmath2 pattern . the rich phase diagram of the model was studied through dmrg , exact diagonalization , and a hard - core boson model . the phase diagram thus obtained presented three transition points . the first one is continuous and occurs at @xmath167 between the lieb - mattis ferrimagnetic phase ( f1 ) and a ferrimagnetic phase ( f2 ) characterized by the condensation of the singlet component of spins at sites b@xmath1 and b@xmath2 of the same unit cell , with transverse critical antiferromagnetic correlations . at @xmath168 , a first - order transition to a phase characterized by spiral and predominantly af correlations ( _ singlet spiral _ ) takes place . the number of singlets in the lattice is quantized before this transition , but is a continuous quantity afterwards , and can be envisioned by measuring the singlet density . further , a continuous chain - ladder decoupling transition at @xmath169 is observed . above this value , the a spins present critical af correlations following the asymptotic behavior observed in a linear chain , with power - law decay , while the ladder of b spins are short - range correlated with a finite correlation length , whose value is @xmath4-dependent , and nears the two - legged - ladder configuration ( _ decoupled chain ladder _ ) . . one can distinguish three major phases : ferri , canted , and the decoupled af chain ladder system , with transitions occurring around @xmath170 and @xmath171 . dashed lines are guides to the eye.,scaledwidth=45.0% ] : ( _ a _ ) energy and ( _ b _ ) expectation value of the total angular momentum per unit cell . the inset shows details of the phase transition around @xmath172 . dashed lines are guides to the eye.,title="fig:",scaledwidth=23.0% ] : ( _ a _ ) energy and ( _ b _ ) expectation value of the total angular momentum per unit cell . the inset shows details of the phase transition around @xmath172 . dashed lines are guides to the eye.,title="fig:",scaledwidth=23.0% ] as far as quantum rotors are concerned , the examination of the momentum correlations in reveals that the system starts out with the ferri phase which is the counterpart of phase f1 of the spin system studied in ref . 12 . the double - cell variational mf energy plotted in exhibits a pattern quite similar to that of frustration f@xmath1 , shown in . but resorting to with the help of ( total average momentum per unit cell ) , we can clear up the picture : in fact , at @xmath173 its reasonable to think that a second - order transition takes place giving rise to a narrow transient phase that corresponds to the phase f2 ( condensation of singlet components of spins at sites b of the same unit cell ) for spin system and is best visualized through the inset in the latter figure , which shows the behavior of the total angular momentum . in this phase the momenta of the a sites keep their ferromagnetic configuration ( @xmath174 ) while the b momenta conform to a magnetic canted configuration . a first - order transition follows at @xmath175 to a new state that should correspond to the phase _ singlet spiral _ of the respective spin system . with respect to the momenta at the a sites , the ferromagnetic configuration also prevails in this phase . the total angular momentum of the a sublattice exactly counterbalances that of the b sublattice , so that , as it happens for the spin system , a vanishing expectation value of the total angular momentum ( spin ) per unit cell occurs . furthermore , upon inspecting the b correlations fig . [ figure11 ] , this phase appears here to have also a semiclassical canted configuration , hence the name canted that we use to designate this qr phase together with the previous one . in the same manner for the spin system @xcite , the additional intercell interactions produce nonquantized values of the b momenta ( see , for example , correlation @xmath176 ) . the nonquantization verified in the spin system , which is a coherent superposition of singlet and triplet configurations , may rather be seen as manifestation of the symmetry - breaking of the invariance of the hamiltonian under interchange of the b sites in the same cell brought about by the additional frustration . this is also clearly verified in the qr system . in the absence of the additional frustration , as is the the case for frustration @xmath38 ( ) , this symmetry stays unscathed , so that there is no singlet - triplet superposition : we have either a singlet or a triplet configuration per cell , but never both simultaneously , which was already the case for both spin and qr systems . finally , as seen in , at @xmath177 , quantum fluctuations bring the sudden decoupling of the chain through another phase transition with first - order characteristics ( in the spin system the transition is second - order ) , and the system settles into an antiferromagnetic ( af ) phase , also marked by a vanishing expectation value of the total angular momentum per unit cell , as shown in . in this phase the frustrated ab@xmath2 chain splits into two decoupled chains , namely , an af linear chain ( @xmath178 ) and an af two - legged ladder ( @xmath179 and @xmath180 ) ; the decoupling is seen through @xmath181 . in , we show a pictorial representation of the the three major phases ferri , canted and af . with respect to this af phase , our qr simulations evidently shed no light onto the criticality and short - rangedness of the linear and two - legged ladder chains , respectively . this phase corresponds to the _ decoupled chain - ladder _ system , which in turn has a vanishing total spin . the first - order transition at @xmath182 may rather be seen as a manifestation of finite - size effects of our two - cell approach : the absence of many intermediate states preclude a smooth transition . : ( _ a _ ) ferri ; ( _ b _ ) canted ; ( _ c _ ) af , which is composed of two decoupled 1d systems : a linear chain ( a sites ) and the two - legged ladder ( b sites).,scaledwidth=47.0% ] the average singlet density for this frustration pattern is shown in : the singlet number is quantized ( except for the narrow interval around @xmath183 , until the frustration reaches the value @xmath171 , wherefrom the singlet number goes on nonquantized . we see that the qr system exhibits a four - phase pattern quite similar to that of the respective spin system , with the nature of all but the last phase transitions being similar in both systems . with respect to singlet quantization , we find agreement in the first and last phases ( where singlet densities 0 and 0.7 are observed ) ; in the intermediate phases no match is observed and again we impute this naturally to finite size effects of our two - cell approach , which hinder a discrete one - by - one singlet condensation . also , because of the additional intercell frustration , it was not possible to form isolated singlet configurations as was the case for frustration f@xmath1 . in what follows , we provide a more detailed comparison between qr mf results and the spin-1/2 chain , in its quantum and classical versions . the phase diagram initially described of the spin-1/2 chain from ref . 12 is summarized in . the spiral phase can be exposed in a clear fashion through the pitch angle @xmath16 obtained from the magnetic structure factor defined as @xmath184 with @xmath185 , for @xmath186 , and @xmath187 , if @xmath67 is odd , while @xmath188 , if @xmath67 is even , and here we are labeling the sites in a more convenient way : a@xmath1 , b@xmath189 , and b@xmath190 just denote the sites a@xmath1 , b@xmath1 , and b@xmath2 of the @xmath191 unit cell . in the lieb - mattis phase the ferrimagnetic order is indicated by a sharp peak at @xmath192 ( a period-2 configuration ) ; while in the decoupled phase , in which a period-4 structure is observed ( see ) , there is a peak at @xmath193 . these two situations are magnetic configurations commensurate with the lattice , while the spiral phase is indicated by a peak at a value of @xmath16 between @xmath193 and @xmath194 . in we display the behavior of @xmath16 as a function of @xmath4 for finite systems calculated through ed and dmrg . finite size effects lead to a little shift in the transition point from the spiral phase to the decoupled phase , even though @xmath16 can be clearly used to mark the spiral phase . : average singlet density per unit cell for the momentum correlations at b sites along the same rung of the ladder.,scaledwidth=30.0% ] motivated by these results , we consider the classical model in the space spanned by two parameters ( in approach of sec . iii , one single parameter sufficed to explain the results ) : a canting angle @xmath6 between the b momenta at the same cell and the pitch angle @xmath16 between the a momentum and the sum of the b momenta , associated with the spiral order . the classical fields are accordingly written as : @xmath195\mathbf{x}+\sin[q(2l-1)]\mathbf{z};\nonumber\\ \mathbf{b}_{1l}&=&\cos(\theta)\cos(2ql)\mathbf{x}+\sin(\theta)(-1)^l\mathbf{y}\nonumber\\ & & + \cos(\theta)\sin(2ql)\mathbf{z } , \text{and}\nonumber\\ \mathbf{b}_{2l}&=&\cos(\theta)\cos(2ql)\mathbf{x}+\sin(\theta)(-1)^{l+1}\mathbf{y}\nonumber\\ & & + \cos(\theta)\sin(2ql)\mathbf{z } , \end{aligned}\ ] ] with @xmath196 , while @xmath197 are orthogonal unit vectors in the three - dimensional space . substituting these fields in the classical version of the hamiltonian , , we get the energy function @xmath198 and minimizing this function with respect to @xmath16 and @xmath6 , we find that @xmath199 and @xmath200 for @xmath201 , which is the classical version of the lieb - mattis phase found for @xmath202 in the quantum hamiltonian , for both quantum rotors ( ferri phase ) and spin system ( phase f1 of ref . for @xmath203 we obtain @xmath204 which may be seen as the classical version of the canted phase ( @xmath205 ) and of the spiral phase ( @xmath206 ) found in ref . 12 . this phase holds some similarities with the second phase observed for the quantum rotors in the first single - site approach of sec . finally , for @xmath207 the three chains are antiferromagnetically ordered with the b momenta lying in the @xmath208 direction and the a momenta ordered in the @xmath209 direction , which is the classical analog of the decoupled phase observed for @xmath210 for the quantum rotors and for @xmath211 for the spin system @xcite . such phase does not exist in the single - site approach : it is only obtained asymptotically ( @xmath212 ) . therefore , the classical solution presents two critical points : @xmath213 and @xmath214 , and the transitions remain second order . also , the first - order transition at @xmath215 ( @xmath216 , for the spin systems ) is not observed in the classical model . in fact , in the f@xmath2 phase@xcite ( @xmath217 ) the number of singlets is quantized and the spiral peak is absent , while in the classical model the two orders coexist for @xmath218 . this classical result is also indicated in . ( pitch angle ) and @xmath6 ( canting angle ) . the transition points estimated in ref . 12 are indicated . ( _ b _ ) momentum dot products ( @xmath219 , and @xmath220 denotes the unit cell ) in the minimum energy configuration of the classical vector model.,title="fig:",scaledwidth=40.0% ] ( pitch angle ) and @xmath6 ( canting angle ) . the transition points estimated in ref . 12 are indicated . ( _ b _ ) momentum dot products ( @xmath219 , and @xmath220 denotes the unit cell ) in the minimum energy configuration of the classical vector model.,title="fig:",scaledwidth=40.0% ] in , we present the results of this classical interpretation for the momentum correlations . a direct relationship with can be established : we have the classical counterparts of the ferri phase ( @xmath221 ) and the af phase ( @xmath222 ) ; the canted phase is but a gradual continuous transition between the ferri and af phases . further , the decoupling transition in the classical model is clearly observed at @xmath81 through the dot products indicated in the figure . finally , we notice that when this classical approach is applied to frustration f@xmath1 , the minimum energy configuration obtained is the same as that derived through the first classical model discussed in sec . in dealing with quantum rotors placed at the sites of an ab@xmath2 chain we resorted to a cluster variational mf theory implemented via two distinct approaches , which yielded different results . first , we learned that the size of the hilbert space could be considerably reduced without affecting results , with the proviso that the rotor states should be kept possibly nearest ( @xmath11)-momentum states . that was attained by increasing the importance of the kinetic energy term in the hamiltonian ( by setting relatively high values of the coupling @xmath9 ) , whenever needed . this was a most important fact for the computations in multiple - cell clusters in the second approach . in the first approach the natural single - site mf theory was developed . a two - phase pattern was produced with the phase transition between them being of second order for @xmath86 and of first order for @xmath223 . for @xmath224 , the lieb - mattis - like phase typical in the spin-1/2 system arose , and for @xmath87 , we observed the decoupling of the system where the momenta on the a sites tend to become uncorrelated with the momenta on the b sites , which in turn formed singlet - like pairs , while the decoupling of the a sites was a salient feature laid bare by this approach , much like the dimer - monomer phase in ref . 12 . a classical interpretation was laid down that conformed to our qr numerical findings , inclusively showing how the fixed coupling momentum - position turns the second - order transition into a first - order one , when @xmath41 . however , we were not able to provide a reasonable quantum picture that could relate to the known behavior of the corresponding quantum ab@xmath2 chain . furthermore , treated this way the system was not able to essentially tell apart frustration f@xmath1 from frustration f@xmath2 and this alone constituted a major setback . so , all this was a reminder that the main goal of our work still remained to be achieved . in our last step , we then improved the former approach by producing a cluster ( double - cell ) variational mf theory in which the trial hamiltonian acts on the space composed of the tensor product of the respective local subspaces of the six sites at two neighboring unit cells . the gist of this theory stands on the important fact that it allows the construction of the two - point correlations @xmath225 between any pairs of operators acting on sites @xmath226 of the cluster . this afforded us the observation of quantum features inherent in the system , as well as to distinguish between both frustrations f@xmath1 and f@xmath2 . for the construction of this more complex `` system '' , we relied on the availability of processing capacity to carry out the numerical implementation . for frustration f@xmath1 , besides the qr numerical simulation , we carried out ed on the spin-1/2 diamond chain using a system with 28 sites and calculated the correlation functions between spin at a central cluster , as well as other relevant physical quantities . upon confronting with the qr results , we verified that the qr phase diagrams obtained through numerical implementation of the double - cell mf variational approach exhibited a sequence of phases analogous to those of the spin chains , with phase transitions of the same nature . we therefore produced the ferri - td - dm phase sequence , with first order transitions , which is in essence the phase diagram of ref . the transition point ferri - td at @xmath163 is somewhat displaced , but the transition point at @xmath130 was exactly calculated by our approach . for frustration f@xmath2 , we obtained a phase diagram in good agreement with the results of ref . 12 on the respective spin-1/2 chains , endowed with the equivalent frustration pattern : ferri , canted , and af which are associated with the phases f1 , f2/spiral singlet , and decoupled ladder chain , respectively , of the spin model . notwithstanding , the criticality of the a spins correlations manifests itself here as an af magnetic ordering due to finite - size effects . for the same reason , we can not probe the short - rangedness of the correlation functions between b momenta . we also produced ed as well as dmrg results that helped us to visualize the spiral phase in the spin system , and derived an insightful classical interpretation . this work was supported by cnpq , facepe , capes , and finep ( brazilian agencies ) . in this appendix we provide the derivation of the matrix elements in the monopole harmonics basis representation for the @xmath110 operator , _ for any value of _ @xmath16 . this derivation was done straightforwardly based solely on the definitions @xcite and the properties of the jacobi polynomials . thus the following recurrence relations of the jacobi polynomials @xcite can be established , valid for all @xmath16 : @xmath227^{1/2}y_{q , l+1,m}(\theta,\phi ) & = & \nonumber \\ \left[l(l+1)\cos\theta + 2mq\right]\left[\frac{2l+1}{(l+m)(l - m)}\right]^{1/2}y_{q , l , m}(\theta,\phi ) & & \nonumber \\ - ( l+1)\left[\frac{(l+q)(l - q)}{2l-1}\right]^{1/2}y_{q , l-1,m}(\theta,\phi ) , & & \label{me1 } \end{aligned}\ ] ] @xmath228\left[\frac{1}{(l+m)(l+m-1)}\right]^{1/2}y_{q , l-1,m+1}(\theta,\phi ) & = & \nonumber \\ \frac{1}{l}\left[\frac{(l+q)(l - q)(l - m-1)(l - m)}{(2l+1)(2l-1)(l+m)(l+m-1)}\right]^{1/2}y_{q , l , m}(\theta,\phi ) & & \nonumber \\ - \frac{q}{l(l-1)}\left[\frac{l - m-1}{l+m-1}\right]y_{q , l-1,m}(\theta,\phi ) & & \nonumber \\ -\frac{1}{l-1}\left[\frac{(l+q-1)(l - q-1)}{(2l-3)(2l-1)}\right]^{1/2}y_{q , l-2,m}(\theta,\phi ) , & & \label{me2 } \end{aligned}\ ] ] @xmath229\left[\frac{1}{(l - m+1)(l - m)}\right]^{1/2}y_{q , l , m-1}(\theta,\phi ) & = & \nonumber \\ -\frac{q(l+m)}{l(l+1)}\left[\frac{1}{(l+m)(l - m)}\right]^{1/2}y_{q , l , m}(\theta,\phi ) & & \nonumber \\ - \frac{1}{(l+1)}\left[\frac{(l+m)(l+m+1)(l+q+1)(l - q+1)}{(2l+1)(2l+3)(l - m+1)(l - m)}\right]^{1/2}y_{q , l+1,m}(\theta,\phi ) & & \nonumber \\ + \frac{1}{l}\left[\frac{(l+q)(l - q)}{(2l+1)(2l-1)}\right]^{1/2}y_{q , l-1,m}(\theta,\phi ) . & & \label{me3 } \end{aligned}\ ] ] next we use the orthogonality relation for the monopole harmonics and specialize in the case @xmath230 to get the respective non - zero matrix elements for the operator @xmath231 , used in this work . so , we have set @xmath232 , which is defined similarly to @xmath233 , where we have the following : @xmath234 and @xmath235 , @xmath236 ( the matrix elements of @xmath237 are obtained by complex conjugation ) : @xmath238 ; @xmath239 ; ; @xmath240 ; and @xmath241 . for the sake of completeness , we write down the matrix elements for the operator @xmath242 , valid for all @xmath16 , obtained through the ladder - operator and eigenvalue relations for the monopole harmonics : , , and . 30 a. auerbach , _ interacting electrons and quantum magnetism _ ( springer - verlag , new york , 1998 ) . s. sachdev , _ quantum phase transitions _ ( cambridge university press , cambridge - uk , 2001 ) . c. j. hamer , j. b. kogut , and l. susskind , phys . d , * 19 * , 3091 ( 1979 ) . f. d. m. haldane , phys . lett . * 93a * , 464 ( 1983 ) ; f. d. m. haldane , phys.rev . lett . * 50 * , 1153 ( 1983 ) . r. shankar and n. read , nucl . b * 336 * , 457 ( 1990 ) . t. t. wu and c. n. yang , nucl . b * 107 * , 365 ( 1976 ) . s. sachdev and t. senthil , ann . phys . * 251 * , 76 ( 1996 ) . a. imambekov , m. lukin , and e. demler , phys . lett . * 93 * , 120405 ( 2004 ) . t. p. polak and t. k. kope@xmath243 , phys . b * 76 * , 094503 ( 2007 ) . k. takano , k. kubo , and h. sakamoto , j. phys . : condens . matter * 8 * , 6405 ( 1996 ) . t. tonegawa , k. okamoto , t. hikihara , y. takahashi , and m. kaburagi , j. phys . . jpn . * 69 * , 332 ( 2000 ) . r. r. montenegro - filho and m. d. coutinho - filho , phys . rev . b * 78 * , 014418 ( 2008 ) . k. sano and k. takano , j. phys . . jpn . * 69 * , 2710 ( 2000 ) ; k. okamoto , t. tonegawa , y. takahashi , and m. kaburagi , j. phys . : condens . matter * 11 * , 10485 ( 1999 ) ; h. niggemann , g. uimin , and j. zittartz , j. phys . : condens . matter * 9 * , 9031 ( 1997 ) . k. okamoto , t. tonegawa , and m. kaburagi , j. phys . : matter * 15 * , 5979 ( 2003 ) . n. b. ivanov , j. richter , and j. schulenburg , phys . b * 79 * , 104412 ( 2009 ) . c. lanczos 45 * , 255 ( 1950 ) . s. r. white , phys . b * 48 * , 10345 ( 1993 ) ; u. schollwck , rev . 77 * , 259 ( 2005 ) . h. kikuchi , y. fujii , m. chiba , s. mitsudo , t. idehara , t. tonegawa , k. okamoto , t. sakai , t. kuwai , and h. ohta , phys . . lett . * 94 * , 227201 ( 2005 ) ; see also k. c. rule , a. u. b. wolter , s. sllow , d. a. tennant , a. brhl , s. khler , b. wolf , m. lang , and j. schreuer , phys . rev . lett . * 100 * , 117202 ( 2008 ) . m. matsuda , k. kakurai , a. a. belik , m. azuma , m. takano , and m. fujita , phys . b * 71 * , 144411 ( 2005 ) . y. hosokoshi , k. katoh , y. nakazawa , h. nakano , and k. inoue , j. am . chem . soc . * 123 * , 7921 ( 2001 ) . for a review see : m. d. coutinho - filho , r. r. montenegro - filho , e. p. raposo , c. vitoriano , and m. h.oliveira , j. braz . * 19 * , 232 ( 2008 ) . a. m. s. macdo , m. c. dos santos , m. d. coutinho - filho , and c. a. macdo , phys . 74 * , 1851 ( 1995 ) ; g .- s . tian and t .- h . lin , phys . b * 53 * , 8196 ( 1996 ) . g. sierra , m. a. martn - delgado , s. r. white , d. j. scalapino and j. dukelsky , phys . b * 59 * , 7973 ( 1999 ) c. vitoriano , m. d. coutinho - filho , and e. p. raposo , j. phys . a : math . * 35 * , 9049 ( 2002 ) ; f. c. alcaraz and a. l. malvezzi , j. phys . a : math . gen . * 30 * , 767 ( 1997 ) ; e. p. raposo and m. d. coutinho - filho , phys . lett . * 78 * , 4853 ( 1997 ) ; e. p. raposo and m. d. coutinho - filho , phys . b * 59 * , 14384 ( 1999 ) ; m. a. martn - delgado , j. rodriguez - laguna , and g. sierra , phys . rev . b * 72 * , 104435 ( 2005 ) . c. vitoriano , f. b. de brito , e. p. raposo , and m. d. coutinho - filho , _ mol . cryst . _ * 374 * , 185 ( 2002 ) ; t. nakanishi and s. yamamoto , phys . rev . b * 65 * , 214418 ( 2002 ) ; s. yamamoto and j. ohara , phys . rev . b * 76 * , 014409 ( 2007 ) r. r. montenegro - filho and m. d. coutinho - filho , physica a * 357 * , 173 ( 2005 ) . m. h. oliveira , m. d. coutinho - filho , and e. p. raposo , phys . b * 72 * , 214420 ( 2005 ) . r. r. montenegro - filho and m. d. coutinho - filho , phys . b * 74 * , 125117 ( 2006 ) , and references therein . h. b. callen , _ thermodynamics and an introduction to thermostatistics _ ( john wiley & sons inc , new york - ny , 1985 ) ; r. p. feynman , _ statistical mechanics a set of lectures _ ( the benjamin cummings publishing company inc , reading - ma , 1972 ) . c. s. o. yokoi , m. d. coutinho - filho , and s. r. salinas , phys . b * 24 * , 4047 ( 1981 ) . ( cambridge university press , cambridge - uk , 1992 ) . e. h. lieb and d. mattis , j. math . phys . * 3 * , 749 ( 1962 ) . a. ederlyi ( editor ) , _ higher transcendental functions ( bateman project ) _ ( mcgraw hill book company inc , new york - ny , 1953 ) .

we study the structure of classical groups of equivalences for smooth multigerms , and extend several known results for monogerm equivalences to the case of mulitgerms . in particular , we study the group of source- and target diffeomorphism germs , and its stabilizer . for monogerms it is well - known that if is finitely-determined , then has a maximal compact subgroup , unique up to conjugacy , and is contractible . we prove the same result for finitely-determined multigerms . moreover , we show that for a ministable multigerm , the maximal compact subgroup decomposes as a product of maximal compact subgroups for suitable representatives of the monogerm components of . we study a product decomposition of in terms of and a group of target diffeomorphisms , and conjecture a decomposition theorem . finally , we show that for a large class of maps , maximal compact subgroups are small and easy to compute .

1110.1981

when studying global properties of smooth , singular maps , a typical approach is to first solve problems locally , and then glue the local solutions together . when the map @xmath9 in question is stable , the typical local situation is the following : singularities of @xmath9 are found along stratified subsets of source and target of @xmath9 . each stratum consists of points where a certain singular germ @xmath3 appears as a singularity of @xmath9 . these strata are submanifolds of source and target , respectively , and in tubular neighborhoods of these submanifolds , @xmath9 restricts to a fibered family of germs @xmath3 , as illustrated in fig . [ tubedmap ] . such decompositions are discussed in detail in @xcite . the singular germs are determined up to changes of local coordinates in source and target of the germ , also called a _ right - left equivalence _ , or , in modern terminology , an @xmath1-equivalence . two germs @xmath10 are right - left - equivalent , and hence define the same singularity , if there exist diffeomorphism germs @xmath11 and @xmath12 such that @xmath13 . the group @xmath1 of all such equivalences @xmath14 coincides with the product of groups @xmath15 , where @xmath16 consists of source diffeomorphism germs @xmath17 and @xmath18 consists of target diffeomorphism germs @xmath19 . these groups all act on the space of multigerms @xmath20 through composition of maps , e.g. , @xmath21 . and @xmath22 of the source manifold @xmath23 , we find singularities @xmath24 and @xmath25 , which are both mapped onto the submanifold @xmath26 of the target manifold @xmath27 . if @xmath9 is stable , then in tubular neighborhoods @xmath28 and @xmath29 , @xmath9 will restrict to a fibered map @xmath30 , where @xmath31 . in order to combine constructions on such local decompositions back into a global construction , we need to understand the symmetry group @xmath32.,height=132 ] the fibered family of singularities in fig . [ tubedmap ] defines a fiber bundle , whose structure group is the stabilizer @xmath2 of @xmath1-equivalences leaving @xmath3 invariant . in order to patch local constructions on such local fiber bundles together , it becomes essential to understand @xmath2 . in reality , the germ @xmath3 will often be a multigerm , where each point in the target of the map corresponds to several source points , just like in fig . [ tubedmap ] . thus , we need to understand @xmath2 in the case where @xmath3 has several source basepoints . for instance , knowledge of the structure of @xmath2 for multigerms @xmath3 is used in @xcite to prove topological stability of maps . in this article , we show that given a finitely @xmath1-determined multigerm @xmath3 , the group @xmath2 has a maximal compact subgroup , which is unique up to conjugacy . we denote an arbitrary representative of the conjugacy class by @xmath4 , and we show that the quotient @xmath5 is contractible , where contractibility is defined through a smooth map extension property . geometrically , this means that we can reduce the non - topological structure group @xmath2 of the tubular neighborhood bundle to the compact , finite - dimensional lie group given by @xmath4 . most of the results presented in this article have been proven for monogerms by k. jnich @xcite , c.t.c . wall @xcite and r. rimnyi @xcite . other groups of germ equivalences have also been studied . in his thesis ( * theorem 1.6.3 ) , rimnyi states our main theorem in the case of @xmath1-equivalences for stable multigerms , without proof . for some classical groups of germ equivalences , the step from monogerms to multigerms is trivial . it is easy to see that we can decompose the groups @xmath33 and @xmath34 , defined on p. , for multigerms @xmath35 to a product of @xmath33-groups for monogerms : @xmath36 and @xmath37 . [ unionk ] similarly , for the group of source diffeomorphisms , or @xmath16-_equivalences _ , we have @xmath38 and @xmath39 . however , in the case of @xmath1-equivalences , the multigerm case does not in any obvious way reduce to the monogerm case . by lemma [ standardstablemultigerm ] , any stable multigerm @xmath40 with @xmath41 admits a decomposition @xmath42 in suitably chosen coordinates , where the @xmath43 are ministable germs which are @xmath44-equivalent to the germs of @xmath9 at points in @xmath45 ( see @xcite ) . if @xmath9 is ministable , then @xmath46 will be @xmath47 . ( in rimnyi s terminology , the @xmath43 will be _ roots of their kinds_. ) there is a natural embedding of the product @xmath48 into @xmath49 , given by @xmath50 where @xmath51 moves the first component to the @xmath52 . however , there is no guarantee that this embedding is surjective . suppose that there are two ways of decomposing @xmath9 as in ( [ multigermdecomp ] ) , and suppose for simplicity that @xmath53 . that is , there exist some diffeomorphism germs @xmath54 , @xmath55 and @xmath56 , such that the following commutes : @xmath57 given an element @xmath58 , there is no obvious guarantee that @xmath59 is of the form on the right hand side of ( [ aemb ] ) . what is easy to show is that the axes @xmath60 are left invariant under @xmath61 , since this is where the singularities @xmath43 are found , but this is not enough . we had hoped that it might be easier to find a decomposition for compact subgroups @xmath62 , as these are conjugate to linear subgroups of @xmath1 . however , we meet yet another problem can we find coordinates that _ simultaneously _ give @xmath3 in the form ( [ multigermdecomp ] ) _ and _ linearize @xmath63 ? this problem might be solved by replacing @xmath63 by the @xmath64-jet of @xmath63 , which is isomorphic to @xmath63 but @xmath65 would not generally leave @xmath3 invariant , and hence would not be a subgroup of @xmath2 . we shall , indeed , see that for statements concerning maximal compact subgroups of @xmath2 , we _ can _ reduce to the monogerm case but this is not trivial . for statements concerning all of @xmath2 in particular concerning the contractibility of the quotient @xmath5 we need to reprove the theorems for multigerms . we shall also investigate a decomposition theorem for @xmath2 in terms of @xmath66 and a subgroup of @xmath18 . finally , we show that for a large class of multigerms @xmath3 , the maximal compact subgroups @xmath4 are very simple . some of the results presented in this article can be found in a. feragen s phd thesis @xcite , to which we refer for a higher level of detail . we denote by @xmath67 the ring of function germs @xmath68 , and denote by @xmath69 the ideal in @xmath67 consisting of germs that map @xmath47 to @xmath47 . we write @xmath70 for the set of germs @xmath71 , and note that @xmath72 . let @xmath73 be a smooth germ . define the set @xmath74 of _ vector fields along @xmath3 _ to be the set @xmath75 and set @xmath76 where we choose notation depending on how explicit we need to be . we define @xmath77 and @xmath78 by setting [ infstabpageref ] @xmath79 in this article , @xmath45 will always be a finite set @xmath80 , and we can , up to a choice of local coordinates , always write a germ @xmath0 in the form @xmath81 where @xmath82 and @xmath83 are the dimensions of @xmath23 and @xmath27 , respectively . we shall need the following standard choice of coordinates for stable multigerms : [ standardstablemultigerm ] let @xmath84 be a stable multigerm . by choosing suitable local coordinates , we may assume that @xmath3 is of the form @xmath85 where @xmath86 ; @xmath87 is ministable ; the local algebra of @xmath88 is isomorphic to the local algebra of the germ of @xmath3 at @xmath89 ; @xmath51 moves the @xmath90 coordinate to the @xmath52 in @xmath91 ; and @xmath92 is chosen to get the appropriate dimensions . here , @xmath93 denotes the source of @xmath94 , and @xmath95 denotes the target of @xmath94 . since the @xmath88 are unfoldings of the @xmath43 , the local algebras @xmath96 and @xmath97 are isomorphic , and as a consequence also the local algebras @xmath98 and @xmath99 are isomorphic ( * ? ? ? * theorem 2.1 ) . hence @xmath9 and @xmath3 are @xmath33-equivalent . but @xmath9 and @xmath3 are also both stable , hence @xmath9 and @xmath3 are @xmath1-equivalent ( * ? ? ? * chapter iii theorem 4.3 ) . suppose given a multigerm @xmath100 as in ( [ multigermnormalform1 ] ) . recall that if @xmath16 is the group of diffeomorphism germs of a source component @xmath101 , and @xmath18 is the group of diffeomorphism germs of the target @xmath102 , then @xmath1 denotes the group of equivalences on the space of multigerms @xmath103 defined by @xmath104 . we will write elements of @xmath1 either in the form @xmath105 where the @xmath106 are the diffeomorphisms in question , or sometimes in the form @xmath107 where @xmath108 for each @xmath109 . we use the latter notation when we consider @xmath19 as an element in the group @xmath33 , as defined below . note that with this definition , the @xmath1-equivalences leave the source base points fixed . the group @xmath33 of equivalences on the space of multigerms ( [ germspace ] ) is the group of diffeomorphism germs [ kdef ] @xmath110 such that the following diagram commutes : @xmath111 where @xmath112 . we will write elements of @xmath33 in the form @xmath113 where , in fact , @xmath114 . the group @xmath33 acts on germs @xmath3 in the following way : @xmath115 if @xmath116 . given a multigerm @xmath3 , we denote by @xmath2 or @xmath34 the _ stabilizer _ of @xmath1 or @xmath33 at @xmath3 ; namely @xmath117 for @xmath118 or @xmath33 . following the original definition by jnich @xcite , we define a compact subgroup of @xmath1 to be a subgroup @xmath63 of @xmath1 which is conjugate in @xmath1 to a compact subgroup of @xmath119 . the definition is reasonable : suppose that @xmath120 is isomorphic to some compact lie group @xmath121 , such that @xmath121 acts diffeomorphically on @xmath122 through the isomorphism @xmath123 , keeping the origins fixed ( compare with the definition of @xcite ) . then , in particular , @xmath121 acts diffeomorphically on each of the @xmath124 and on @xmath125 . by bochner s theorem we can choose local coordinates in @xmath126 and @xmath102 , respectively , given by diffeomorphism germs @xmath127 and @xmath128 , with respect to which @xmath63 acts linearly on the @xmath124 and on @xmath125 . these define an element @xmath129 of @xmath1 , which linearizes @xmath63 in @xmath1 . by analogy , we define a compact subgroup of @xmath33 to be a subgroup @xmath130 which is conjugate in @xmath33 to a compact linear subgroup of @xmath131 . throughout the rest of the article , we let @xmath132 denote @xmath1 or @xmath33 unless otherwise is specified . given a compact subgroup @xmath63 of a group @xmath132 of diffeomorphism - germs such as @xmath1 , @xmath33 or @xmath16 , we say that @xmath63 is a _ maximal compact subgroup _ of @xmath132 if any other compact subgroup @xmath133 of @xmath132 is conjugate in @xmath132 to a subgroup of @xmath63 . we start out by making a few basic observations concerning the groups of @xmath1-equivalences and @xmath33-equivalences : [ basicobservations ] * we have @xmath134 , so in particular @xmath135 , assuming the maximal compact subgroups exist . * suppose that @xmath20 is a monogerm , and consider @xmath1 and @xmath33 as groups acting on the source and target of @xmath3 . then @xmath136 is a subgroup of @xmath1 , so in particular , @xmath137 . * suppose that a multigerm @xmath138 is finitely @xmath139-determined , and suppose that @xmath140 is compact . then we can find an element @xmath141 and a subgroup @xmath121 of @xmath142 such that @xmath143 is a polynomial of degree @xmath144 , and @xmath121 is a linear group which is conjugate ( in @xmath132 ) to @xmath63 . * for any compact subgroup @xmath63 of @xmath132 the restriction of the @xmath64-jet map @xmath145 is injective . i ) _ is trivial . * an element @xmath146 is an element of @xmath1 if it can be decomposed as @xmath147 , where @xmath148 and @xmath149 . since @xmath146 , we know that @xmath150 , so we set @xmath151 . since @xmath133 is a linear diffeomorphism , @xmath152 for some @xmath153 implies @xmath154 . then we must have @xmath155 , for all @xmath156 , since otherwise , we would find @xmath157 with @xmath152 . * since @xmath140 is compact , we can find @xmath158 such that @xmath159 is linear , by definition . now @xmath160 , and since @xmath3 is @xmath139-determined , so is @xmath161 . in particular , @xmath161 is @xmath132-equivalent to the polynomial representative @xmath83 of its @xmath162-jet @xmath163 , by an element @xmath164 , say : @xmath165 we claim that @xmath166 . since the linear group @xmath121 leaves @xmath161 invariant , it certainly leaves the jet @xmath167 invariant in @xmath168 . but then it must leave @xmath83 invariant , since then , for any @xmath169 , the map @xmath170 is a degree @xmath162 polynomial representing @xmath167 . set @xmath171 , and we are done . * by the definition of a compact subgroup , there exists a choice of coordinates on @xmath172 such that @xmath63 acts linearly ; now in these coordinates the @xmath64-jet map is just the inclusion into @xmath131 . the topological properties of the map @xmath173 do not depend on the choice of coordinates ; hence @xmath174 is injective . now we are ready to state and prove the main theorem of the section . [ existenceofmc ] let @xmath3 be a finitely @xmath132-determined multigerm as in ( [ multigermnormalform1 ] ) . the group @xmath175 has a maximal compact subgroup , which is unique up to conjugation in @xmath175 . the monogerm version of this theorem was proven by jnich @xcite ( for @xmath176 ) and wall @xcite ( for @xmath118 or @xmath33 ) with some completing comments by du plessis and wilson @xcite , who proved similar results for actions of @xmath16 , but without finite @xmath16-determinacy . just like in the monogerm case we will need an equivariant finite determinacy condition , as formulated in the lemma below . the proof of the lemma follows the standard proof of @xmath162-determinacy @xcite , using vector fields but averaging the vector fields over the group @xmath63 using the haar integral . [ equivdeterminacy ] let @xmath177 be @xmath139-determined , and suppose that @xmath63 is a compact linear subgroup of @xmath175 . then @xmath3 is @xmath178-determined ; that is , if @xmath179 is @xmath63-invariant such that @xmath180 , then there exists @xmath141 such that @xmath181 for all @xmath182 , and @xmath183 . in particular , @xmath184 , since @xmath185 for all @xmath182 . assume that @xmath3 is @xmath139-determined . we write @xmath186 for the families of invertible @xmath162-jets @xmath187 and write @xmath188 for the subgroup stabilizing @xmath189 ; both these groups are real algebraic . real algebraic groups have finitely many components , so we may apply iwasawa s theorem @xcite and choose a maximal compact subgroup @xmath63 of @xmath188 . following the arguments of jnich @xcite and bochner @xcite we can linearize @xmath63 in the following ways : [ stronglin ] * suppose that @xmath190 is a compact subgroup . then there exists @xmath191 such that @xmath192 . * suppose that @xmath193 is a compact subgroup . then there exists @xmath141 such that @xmath194 . suppose , furthermore , that @xmath195 . then we may assume that @xmath196 . we return to the proof of theorem [ existenceofmc ] , and to the maximal compact subgroup @xmath63 of @xmath188 . denote @xmath197 . by abuse of notation , we will also denote by @xmath198 the corresponding linear subgroup of @xmath186 . let @xmath199 such that @xmath200 . then if @xmath201 , its jet @xmath202 is @xmath198-invariant , and @xmath198 is maximal compact in @xmath203 . let @xmath133 be any compact subgroup of @xmath204 . by classical lie group theory @xmath205 is conjugate in @xmath203 to a subgroup of @xmath198 , say by a family of jets of diffeomorphisms @xmath206 . let @xmath207 be a family of diffeomorphisms @xmath208 with jet @xmath209 . then the @xmath162-jet of @xmath210 is linear . by lemma [ stronglin ] , we can find @xmath113 with the same @xmath162-jet as @xmath207 , such that @xmath211 is linear . hence @xmath212 is @xmath213-invariant , just like @xmath214 ( because @xmath215 ) . since @xmath216 , we have @xmath217 . recall that @xmath214 is @xmath218-determined , so by lemma [ equivdeterminacy ] there exists some @xmath213-equivariant family of diffeomorphisms @xmath219 such that @xmath220 . hence @xmath221 , and so @xmath222 preserves @xmath214 _ and _ conjugates @xmath133 to @xmath223 where the last equality holds because @xmath61 is @xmath198-equivariant . we have seen that @xmath198 can be viewed as a compact subgroup of @xmath204 , and that any other compact subgroup @xmath133 of @xmath204 is conjugate to a subgroup of @xmath198 , so conjugating back to @xmath175 , we see that theorem [ existenceofmc ] holds . following jnich @xcite , we define what it means for @xmath5 to be contractible . we have not specified a topology on @xmath5 , and in fact we shall define contractibility not in terms of the topology of @xmath5 as a space of its own , but through the topological properties of its action on the source and target of @xmath3 . before defining contractibility , we define what it means for a map into a quotient @xmath224 to be smooth . let @xmath63 be a subgroup of @xmath1 . given a smooth manifold @xmath225 , possibly with boundary , we say that a map @xmath226 is _ smooth _ if there exists an open covering @xmath227 of @xmath225 such that @xmath228 is represented by fibered ( over @xmath229 ) maps @xmath230 and @xmath231 , which are diffeomorphisms . equivalently , a map @xmath232 is smooth if there exists an open covering @xmath233 of @xmath225 such that @xmath61 admits a local lift @xmath234 , where the corresponding fibered map - germs @xmath106 and @xmath235 are smooth . let @xmath63 be a subgroup of @xmath2 . the quotient @xmath224 is _ contractible _ if for every smooth manifold @xmath225 with boundary , any smooth map @xmath236 can be extended to a smooth map @xmath237 . we proceed to state and prove the main result of the section : [ contractibilitythm ] suppose given a finitely @xmath1-determined multigerm @xmath238 then the quotient @xmath5 is contractible . this proof follows that of rimnyi @xcite for the monogerm case . the following proposition is crucial to the proof : [ extensionprop ] there exists an @xmath239 such that the following holds : if @xmath225 is an @xmath240-dimensional manifold with boundary ( possibly empty ) and @xmath241 are fibered ( over @xmath225 ) germs at @xmath242 satisfying the following properties : @xmath243 @xmath244 then there exist @xmath245 , @xmath246 , and @xmath247 such that @xmath248 and @xmath249 @xmath250 for all @xmath251 , @xmath246 . paraphrased , proposition [ extensionprop ] says : given a pair of smooth maps @xmath252 such that @xmath253 and @xmath254 for all @xmath251 , there exists a smooth map @xmath255 such that @xmath256 , @xmath257 and @xmath258 . we find the @xmath209 and @xmath17 by using the flows of suitably chosen vector fields in the source components and in target . let @xmath259 be the map germ defined by @xmath260 . from now on , we denote by @xmath261 , the coordinates of @xmath225 , the @xmath262 source component @xmath124 , @xmath125 , and @xmath263 , respectively . we write @xmath264 , and the notation @xmath265 will denote the composition @xmath266 , and so on . we want to construct the diffeomorphisms @xmath209 and @xmath17 through flows @xmath56 and @xmath267 in the source and target of @xmath9 such that @xmath268 \to m \times \bigsqcup_s { \mathbb{r}}^n \times { \mathbb{r } } , \quad \psi=\bigsqcup_{k=1}^s \psi^k,\\ \phi| \colon ( m \times { \mathbb{r}}^p \times { \mathbb{r } } ) \times [ 0,1 ] \to m \times { \mathbb{r}}^p \times { \mathbb{r } } , \end{array}\ ] ] with @xmath269for all @xmath270 $ ] , and @xmath271 , which holds if @xmath272}).\ ] ] suppose that we have found such flows @xmath56 and @xmath267 , and define maps @xmath273 \to m \times \bigsqcup_s { \mathbb{r}}^n,\\ \tilde{\phi } \colon m \times { \mathbb{r}}^p \times[0,1 ] \to m \times { \mathbb{r}}^p , \end{array}\ ] ] by setting @xmath274 and define @xmath275 \to m \times { \mathbb{r}}^p$ ] by setting @xmath276 where @xmath277 . note that @xmath278 , and that @xmath279 . then @xmath280 and @xmath281 . it is straightforward to prove that @xmath280 , and for the second identity we note that @xmath282 , so @xmath281 if and only if @xmath283 for all @xmath284 , which holds if and only if @xmath285 where @xmath286 . but we have @xmath287 by ( [ cond-1 ] ) , and @xmath288 by the definition of @xmath9 , while @xmath289 , also by the definition of @xmath9 , so @xmath290 and ( [ floweq ] ) holds . in particular , @xmath291 , and thus @xmath292 and @xmath293 are the conjugating diffeomorphisms sought @xmath294 and @xmath17 . suppose that we are given map germs @xmath295 such that the following conditions ( [ cond1 ] ) ( [ cond4 ] ) hold : @xmath296 for all @xmath297 , and @xmath246 ; @xmath298 @xmath299 @xmath300 consider the flows of the following vector fields : @xmath301 by ( [ cond2 ] ) , these flows exist at least in a neighborhood of @xmath302 . furthermore , we see that the condition ( [ cond1 ] ) is just the condition of being a derivative of @xmath9-related flows , that is , a pair of flows satisfying ( [ cond-1 ] ) . the maps @xmath303 and @xmath293 associated with the flows of @xmath304 and @xmath305 as described above clearly satisfy @xmath306 by ( [ cond3 ] ) and ( [ cond4 ] ) . hence , the proof of proposition [ extensionprop ] is completed by finding the vector fields @xmath307 and @xmath308 . this construction is quite long and technical , but follows the construction by rimanyi for the monogerm case @xcite . we refer to the thesis @xcite for details . we may now return to the proof of theorem [ contractibilitythm ] . denote by @xmath309 the lie group of @xmath310-jets of elements of @xmath1 and set @xmath311 for a sufficiently large @xmath239 , the image of a maximal compact subgroup of @xmath2 under @xmath312 is a maximal compact subgroup of @xmath313 . this is a consequence of the apparently more general : for @xmath3 finitely @xmath1-determined and @xmath310 sufficiently large , any maximal compact subgroup @xmath63 of @xmath314 is the image under @xmath312 of a maximal compact subgroup @xmath121 of @xmath2 . let @xmath315 be a maximal compact subgroup . up to conjugation by some element @xmath316 , @xmath63 is linear , i.e. , @xmath317 is a linear maximal subgroup of @xmath318 . but then @xmath319 ( as a matrix group ) is a subgroup of @xmath320 , where @xmath321 is the @xmath322 degree polynomial map representing @xmath143 . for @xmath310 sufficiently large , @xmath321 is @xmath1-equivalent to @xmath143 via some element @xmath323 : @xmath324 . we must have @xmath325 , since @xmath326 . that is , @xmath327 is a compact subgroup of @xmath328 such that @xmath329 is a maximal compact subgroup of @xmath330 , where @xmath331 is conjugate in @xmath309 to @xmath63 via @xmath332 . but then , if @xmath333 , @xmath121 is a maximal compact subgroup of @xmath2 and we have @xmath334 let @xmath63 be a maximal compact subgroup of @xmath2 . by replacing @xmath3 by a suitably chosen representative of its @xmath1-equivalence class , we may assume that @xmath63 acts linearly . given a manifold with boundary @xmath225 , we must show that any smooth map @xmath335 extends to a smooth map @xmath336 . [ lift ] suppose given a smooth map @xmath335 . then there exists a smooth lift @xmath337 : @xmath338{{\mathscr{a}}_f } \arrow{s , r}{\pi}\\ \node{\partial m } \arrow{ne , t}{\tilde{\alpha } } \arrow{e , b}{\alpha } \node{{\mathscr{a}}_f / g } \end{diagram}\ ] ] jnich @xcite and rimnyi @xcite claim the existence of a lift @xmath339 as a consequence of a section @xmath340 lifted from the section @xmath341 , which exists because @xmath342 is contractible . see the diagram ( [ eq1 ] ) below . we take the time to explain how such a lift is constructed . associated with the section @xmath343 is a map @xmath344 such that @xmath345 for all @xmath346 . the section @xmath347 is now given by @xmath348 for any representative @xmath349 of the orbit @xmath350 . the smoothness of @xmath351 is ensured by the fact that @xmath352 . @xmath353{{\mathscr{a}}_f } \arrow{e , t}{j^l } \arrow{s , l}{\pi } \node{{\mathscr{a}}^l_f } \arrow{s , r}{\pi^l}\\ \node{{\mathscr{a}}_f / g } \arrow{e , b}{{\textrm{id } } } \arrow{ne , t, .. }{\sigma ? } \node{{\mathscr{a}}_f / g } \arrow{e , b}{\bar{j^l } } \node{{\mathscr{a}}^l_f / g } \node{{\mathscr{a}}^l_f / g } \arrow{w , b}{{\textrm{id } } } \arrow{nw , t}{\sigma^l } \end{diagram}\ ] ] consider the composition @xmath354 . since @xmath342 is contractible , we can construct an extension @xmath355 . composing with the section @xmath343 , we obtain a map @xmath356 . [ normalization ] there exists a smooth map @xmath357 such that @xmath358 coincides with @xmath359 . construct smooth maps @xmath360 and set @xmath361 . then @xmath362 if and only if @xmath363 and it is clear from the commutative diagram below that @xmath364 , @xmath338{{\mathscr{a}}_f } \arrow{e , t}{j^l } \arrow{s , r}{\pi } \node{{\mathscr{a}}^l_f } \arrow{se , t}{\varphi } \arrow{s,,l}{\pi^l } \arrow{e , t}{h } \node{g}\\ \node{\partial m } \arrow{ne , t}{\tilde{a } } \arrow{e , b}{\alpha } \arrow{s } \node{{\mathscr{a}}_f / g } \arrow{e , b}{\bar{j}^l } \node{{\mathscr{a}}^l_f / g } \node{g \times { \mathscr{a}}^l_f / g } \arrow{n , r}{{\mathrm{pr } } } \arrow{w , b}{{\mathrm{pr}}}\\ \node{m } \arrow{nee , b}{\bar{\beta } } \end{diagram}\ ] ] while @xmath365 . thus ( [ phi ] ) is true , and this concludes the proof of lemma [ normalization ] . by lemma [ normalization ] we see that replacing the old map @xmath339 by @xmath366 , we may assume that @xmath367 . this will enable us to construct a map @xmath368 which extends @xmath339 . without the assumption @xmath367 we risk for instance , if @xmath369 $ ] that @xmath370 and @xmath371 end up in different components of @xmath314 , in which case an extension of @xmath339 is impossible . this , however , is not crucial in order to get the map into @xmath224 . as a first step , we construct a map @xmath372 , extending @xmath339 , and in particular such that @xmath373 extends our map @xmath374 , where the second map is induced by the inclusion , and such that @xmath375 . in order to do this , we must construct diffeomorphism germs at @xmath376 : @xmath377 from given germs in @xmath378 and @xmath379 , and where the @xmath310-jets are given everywhere . we go through the construction for one of the @xmath380 ; the proof for @xmath381 is similar . if we can find @xmath380 locally , then we get a global solution by using a partition of unity to add the solutions together fiberwise . this gives a diffeomorphism germ since the @xmath310-jet , and thus in particular the differential @xmath382 , is fixed everywhere . at points in the interior of @xmath225 , we can just define the local @xmath380 by taking the @xmath322 degree polynomial representative of the given jet . near points in @xmath383 , we construct the component functions of the local @xmath380 in the following way : given a polynomial @xmath27 of degree @xmath310 in the variables @xmath384 with coefficients from the ring @xmath385 of smooth functions in @xmath240 variables , and a smooth function @xmath386 such that @xmath387 , ( here the coordinates of @xmath124 are denoted by @xmath89 and the local coordinates of @xmath225 are denoted by @xmath388 , where @xmath383 is given by @xmath389 ) we construct a smooth function @xmath390 such that @xmath391 , and @xmath392 . one function which satisfies all of the above , is the combination @xmath393 . it follows that the map @xmath394 exists , and it is represented by the product @xmath395 . we use it to construct the extension sought @xmath336 : we compare the maps @xmath396 given by @xmath397 these two maps coincide on @xmath378 , and their @xmath310-jets coincide at each @xmath225-level , thus we can apply proposition [ extensionprop ] to find a smooth map @xmath398 such that @xmath399 , @xmath400 and such that @xmath401 . now the map @xmath402 , is an extension of @xmath339 and @xmath403 ( because @xmath404 ) . most importantly , it defines an extension @xmath405 of @xmath61 . from the proof of theorem [ contractibilitythm ] , we see that the following corollary also holds : [ coverextension ] suppose that @xmath63 is a maximal compact subgroup of @xmath2 . given a smooth map @xmath406 , we can find a smooth map @xmath407 and a smooth map @xmath408 such that @xmath409 . the main goal of this section is to prove that maximal compact subgroups of @xmath2 for multigerms @xmath3 decompose into products of maximal compact subgroups of @xmath410 for minimal representatives @xmath7 of their associated monogerms : [ multitomono ] we are given a ministable multigerm @xmath411 where @xmath412 where @xmath7 is a ministable unfolding of a rank @xmath47 representative @xmath413 of @xmath43 , and where @xmath414 and @xmath415 . then the maximal compact subgroup factors as @xmath416 [ s=1rem ] when @xmath417 , this is ( * ? ? ? * proposition 3.2 ) . the proof of this theorem rests on relations between maximal compact subgroups of @xmath33- and @xmath1-equivalences , the monogerm versions of which are well known . the following lemma is analogous to ( * ? ? ? * theorem 1.1 ) . [ lem8 ] let @xmath138 have rank @xmath47 and let @xmath418 be a ministable unfolding of @xmath3 . suppose that @xmath419 is a compact subgroup . then there exists a compact subgroup of @xmath34 which is isomorphic to @xmath63 . in particular , @xmath420 . since @xmath63 is compact , there exists @xmath421 such that @xmath422 is linear . if @xmath423 , then @xmath424 . we shall construct a map @xmath425 such that @xmath426 is a ministable unfolding of @xmath214 . denote by @xmath427 the graph of @xmath426 . if we write @xmath428 , then @xmath429 since @xmath198 is linear and @xmath427 is @xmath198-invariant , the tangent space @xmath430 is also @xmath198-invariant . define subspaces @xmath431 which are also @xmath198-invariant . choose @xmath198-invariant complements @xmath432 and @xmath433 of @xmath434 and @xmath435 in @xmath436 and @xmath437 , respectively . then @xmath438 , @xmath439 , @xmath440 and @xmath441 . denote by @xmath214 the map germ @xmath442 where @xmath443 for each @xmath444 . we now prove that through its action on @xmath445 , @xmath198 is a subgroup of @xmath446 . it follows from the definition of @xmath435 that @xmath447 is a submersion for each @xmath162 , so in particular the inclusion of @xmath433 in @xmath437 is transverse to the germ @xmath448 . furthermore , the diagram @xmath449 is trivially cartesian for each @xmath162 , and it follows that @xmath426 unfolds @xmath214 . the unfolding @xmath426 is ministable because it is @xmath1-equivalent to the ministable map @xmath9 . note , moreover , that since @xmath426 and @xmath9 are @xmath1-equivalent , the maps @xmath3 and @xmath214 must be @xmath33-equivalent . project the group @xmath450 onto @xmath451 then the resulting group lies in @xmath452 , so in particular it must lie in @xmath453 . equivalently , the action of @xmath198 on the @xmath434 and on @xmath433 determines the action on the @xmath432 and @xmath435 . the actions on @xmath434 , @xmath435 and @xmath433 determines that on @xmath432 ( being a germ , @xmath426 is level - preserving if the coordinates on @xmath432 are appropriately chosen ) , thus it is enough to show that the action on @xmath435 is determined by that on the @xmath434 and @xmath433 . for a general ministable multigerm @xmath454 define @xmath455 @xcite . each @xmath456 is @xmath1-equivalent to a germ @xmath457 , so that @xmath458 is ministable and has an isolated singularity . we can decompose into monogerm components : @xmath459 and see that @xmath460 in our situation , we write @xmath428 , where @xmath461 and @xmath462 . in the diagram @xmath463 there is a naturally defined @xmath198-action on each of the spaces , and both maps @xmath464 and @xmath465 are @xmath198-equivariant . hence , the action on @xmath466 determines that on @xmath437 , and in particular that on the @xmath435 . it follows that the projection ( [ projectiontarget ] ) is injective on @xmath198 . hence , @xmath198 can be viewed as a compact subgroup of @xmath453 . since the germs @xmath3 and @xmath214 are @xmath33-equivalent , the compact subgroups of @xmath34 and @xmath453 are conjugate in @xmath33 . there is thus a compact subgroup of @xmath34 which is conjugate in @xmath33 , and thus isomorphic , to @xmath198 , which again is isomorphic to @xmath63 . this concludes the proof of lemma [ lem8 ] . denote by @xmath467 the rank @xmath47 multigerm @xmath468 . now we merely put the pieces together : @xmath469 where the last inequality is most easily seen to hold by considering the form ( [ standardform ] ) and taking each @xmath470 to the element @xmath471 this concludes the proof of the theorem . for monogerms @xmath3 , du plessis and wilson @xcite have studied decomposition of @xmath4 in terms of @xmath8 and a subgroup of target diffeomorphisms which preserve the discriminant of @xmath3 , that is , the set @xmath472 . the monogerms studied in @xcite are _ critical normalizations _ , a class of germs which contains all stable map - germs , finitely @xmath1-determined germs with @xmath473 and @xmath474 , and all analytic , topologically stable map - germs . we are interested in ( * ? ? ? * theorem 1.5 ) , which says that if @xmath3 is a finitely @xmath1-determined critical normalization , then a certain subgroup , @xmath475 , of the target diffeomorphisms which preserve the discriminant , admits a maximal compact subgroup @xmath63 . moreover , the natural sequence @xmath476 should be split exact . as a consequence , the maximal compact subgroup @xmath4 should be the direct sum of @xmath8 and @xmath63 . however , a closer inspection revealed a problem with the proof of the monogerm version from @xcite . our goal is to prove the following conjecture , generalizing the claim by du plessis and wilson to multigerms : [ conjecture ] let @xmath3 be a finitely @xmath1-determined multigerm which is also a critical normalization . the group @xmath477 admits a maximal compact subgroup @xmath63 , which is unique up to conjugation . moreover , the sequence @xmath478 is split exact , so that @xmath479 . in this section we shall discuss why the original proof of the monogerm version of the conjecture does not hold . moreover , we shall see just how close to the result sought we can get with tactics similar to those in @xcite . the proof of ( * ? ? ? * theorem 1.5 ) rests on the assumption that linearizability is preserved under conjugation by any element of @xmath1 , but unfortunately , this is not true . more precisely , the problem is found in the following sentence in ( * ? ? ? * , page 270 ) : _ ... @xmath310 conjugates @xmath480 onto @xmath481 and conjugates compact lie subgroups to compact lie subgroups , preserving conjugates and preserving linearizability . _ here , @xmath482 is an arbitrary element of @xmath1 , and _ linearizability _ of a compact lie subgroup @xmath133 of @xmath480 means that for some @xmath483 with @xmath484 , @xmath485 is linear . however , this statement does not generally hold : [ nothold ] for a general element @xmath486 , the conjugate group @xmath487 is not generally linearizable . to see this , consider the following linearizable diffeomorphism germ @xmath17 : @xmath488 and suppose that @xmath489 is another diffeomorphism - germ . the question now is , whether we can find a diffeomorphism - germ @xmath490 such that @xmath491 and the following diagram commutes ? @xmath492 decomposing the diagram , we get @xmath493 giving @xmath494 . it is easy to see that @xmath491 , but we also need @xmath207 to belong to @xmath480 . in particular , we need @xmath495 . since @xmath310 and @xmath19 both belong to @xmath496 , we have @xmath497 . however , @xmath498 , which equals @xmath499 if and only if @xmath500 , but this is not generally the case . it follows that linearizability is not preserved under conjugation . as a consequence of proposition [ nothold ] , the proof of ( * * theorem 1.5 ) does not hold . the linearizability is used in @xcite to relate compact subgroups of @xmath501 to compact subgroups of @xmath2 . let us temporarily assume that there is an alternative way to do this . we can prove : let @xmath3 be a multigerm as in conjecture [ conjecture ] . suppose that for any linear compact subgroup @xmath502 of @xmath503 , we can find a compact subgroup @xmath121 of @xmath2 such that @xmath504 . then @xmath505 is a maximal compact subgroup of @xmath477 , and the sequence ( [ splitseq ] ) is split exact , so that @xmath506 . the following lemma holds , even for multigerms , with a proof similar to that in @xcite : [ 41lemma ] let @xmath507 be a cn . then the sequence @xmath508 is well - defined and exact , where @xmath509 and @xmath510 . if @xmath3 is a finitely @xmath1-determined multigerm , we know by theorem [ existenceofmc ] that @xmath4 exists and is unique up to conjugation in @xmath2 . just as in @xcite , also @xmath8 exists and is unique up to conjugation in @xmath66 . moreover , the sequence @xmath511 is split exact , where @xmath512 . it is clear that @xmath505 is a compact subgroup of @xmath513 ; what we would like to prove is that @xmath505 is a maximal compact subgroup of @xmath475 , unique up to conjugation . let @xmath63 be a maximal compact subgroup of @xmath2 , and note that if we can prove the result with @xmath3 replaced by an @xmath1-equivalent germ @xmath214 , then the result holds also for @xmath3 . pick a compact subgroup @xmath133 of @xmath475 . we would like to show that @xmath514 is a maximal compact subgroup of @xmath475 by showing that @xmath133 is conjugate in @xmath475 to a subgroup of @xmath514 . choose a germ @xmath515 which is @xmath1-equivalent to @xmath3 , where @xmath516 , such that @xmath517 is linear . by abuse of notation , we identify @xmath502 with @xmath518 for any @xmath162 . if we can find a maximal compact subgroup @xmath121 of @xmath519 such that @xmath520 , then we can prove the result as follows : since @xmath121 and @xmath63 are maximal compact subgroups of @xmath519 and @xmath2 , respectively , there exists some @xmath521 such that @xmath522 , and @xmath523 . moreover , @xmath524 since @xmath525 , and @xmath526 , so @xmath502 is conjugate to a subgroup of @xmath514 via @xmath527 . but then @xmath133 is conjugate to a subgroup of @xmath514 via @xmath528 . we are thus left with the problem of lifting the group @xmath502 to a compact group @xmath529 . let us attack the problem on the jet level . let @xmath3 be as in conjecture [ conjecture ] , and let @xmath502 be a compact , linear subgroup of @xmath530 . for any @xmath531 , we can find a compact group @xmath529 such that @xmath532 . since @xmath502 acts linearly , we have @xmath533 for any @xmath534 , and by the arguments above , @xmath535 , where @xmath536 . by ( * ? ? ? * corollary 4.4 ) , if @xmath537 is a surjective homomorphism of real algebraic groups , then every compact subgroup of @xmath538 is the image under @xmath467 of a compact subgroup of @xmath133 . hence , there exists a compact subgroup @xmath198 of @xmath539 such that @xmath540 . we would like to find a corresponding subgroup on the map level , namely a subgroup @xmath541 of @xmath519 with @xmath542 and @xmath543 . by bochner s linearization theorem , we may ( by changing source coordinates ) assume that @xmath198 acts linearly , while we still have @xmath540 . let @xmath544 denote the polynomial representative of @xmath545 . since @xmath198 acts linearly , we must also have @xmath546 . since @xmath544 has the same @xmath162-jet as @xmath214 , and these maps are finitely @xmath1-determined , we have that for sufficiently large @xmath162 , there exists @xmath547 such that @xmath548 . since @xmath546 , we have @xmath549 . we had @xmath550 , and now we must also have @xmath551 . as pointed out in ( * ? ? ? * addendum to 3.5 ) , @xmath214 is actually @xmath552-determined for any @xmath240 with @xmath162 sufficiently large . hence , we may assume that @xmath553 for any @xmath240 , and for sufficiently large @xmath162 ( depending on @xmath240 ) . thus there exist , for any @xmath554 , a @xmath555 and a @xmath556 such that the compact group @xmath557 satisfies @xmath558 . we may assume @xmath559 , so @xmath560 , and we have @xmath561 . this shows that on the jet level , we can get arbitrarily close to the lifting of @xmath502 . unfortunately , this is not immediately enough to get a lifting of @xmath502 to @xmath519 , but leaves us with the following conjecture , which will be considered further in a forthcoming paper : there exists a compact subgroup @xmath121 of @xmath2 such that @xmath562 . if this conjecure holds , then by the argument above , so does conjecture [ conjecture ] . recall from lemma [ basicobservations ] that @xmath2 is a subgroup of @xmath34 , so the size of @xmath34 is an upper bound for the size of @xmath2 . in this section we prove that @xmath34 is very small for finitely determined rank @xmath47 germs , making it easy to compute . since we have shown that maximal compact subgroups of @xmath34 for multigerms @xmath3 can be decomposed as a product of maximal compact subgroups for the corresponding monogerms , the monogerm results carry directly over to multigerms . [ maxcomp ] let @xmath20 be finitely @xmath33-determined , with @xmath563 and @xmath564 . if @xmath565 , or if @xmath566 and @xmath567 , then @xmath568 is @xmath569-dimensional , and if @xmath566 then it is @xmath47-dimensional . for @xmath566 , this is related to a theorem by p. slodowy : _ ( * ? ? ? * satz p. 169 ) _ [ slodowy ] let @xmath570 be a germ such that @xmath567 and @xmath3 is finitely @xmath16-determined . if a compact group @xmath63 acts faithfully and linearly on @xmath124 , leaving @xmath3 invariant , then @xmath63 is zero - dimensional . by ( * ? * theorem 4.6.1 ) , any @xmath33-finitely determined function germ is @xmath16-finitely determined . let @xmath3 be as in theorem [ maxcomp ] with @xmath566 , and let @xmath571 be a compact subgroup . then @xmath63 is zero - dimensional . let @xmath3 be as in theorem [ maxcomp ] with @xmath566 , and let @xmath572 be a compact subgroup . then @xmath63 is zero - dimensional . changing @xmath3 by a @xmath33-equivalence , we may assume that @xmath63 acts linearly by lemma [ basicobservations ] . note furthermore that changing @xmath3 by a @xmath33-equivalence will not change the fact that @xmath573 . linear subgroups of @xmath33 lie in @xmath1 by lemma [ basicobservations ] ; hence we can assume @xmath574 . the projections from @xmath575 onto @xmath576 and @xmath577 are continuous homomorphisms , and take @xmath63 to compact subgroups @xmath578 and @xmath579 , respectively . since @xmath580 is a compact subgroup of @xmath577 , we must have @xmath581 . having this in mind , we see that @xmath121 splits into two parts , namely @xmath582 the group @xmath583 is finite by theorem [ slodowy ] , but what about @xmath133 ? since @xmath121 is a lie group , we must either have @xmath121 discrete , or @xmath584 with @xmath585 . suppose the latter . then we can form a continuous path @xmath586 such that @xmath587 and @xmath588 for @xmath589 . then we have @xmath590 , and for any given @xmath591 we have @xmath592 , since @xmath63 is a matrix group . but by the definitions of @xmath133 and @xmath66 , we have @xmath593 when @xmath589 , while @xmath594 , so unless @xmath595 , this must be false . we have @xmath596 for @xmath597 arbitrarily close to @xmath598 , and hence we can not find such a path @xmath599 . but then @xmath121 must be discrete . being a compact discrete set , @xmath121 is finite . for @xmath600 , c.t.c . wall has proven an analogous result over the complex numbers : _ ( * ? ? ? * theorem 3.3 ) _ [ wallthm ] let @xmath601 have finite singularity type , @xmath602 , and @xmath564 . then @xmath603 where @xmath604 is a maximal complex reductive subgroup of @xmath34 . we shall pass from wall s result to the corresponding claim over the real numbers from theorem [ maxcomp ] . denote @xmath605 for short . by lemma [ basicobservations ] , we may assume , up to a change of coordinates , that @xmath3 is a polynomial and that @xmath63 is linear . in particular , @xmath606 by lemma [ basicobservations ] . there is a corresponding complex polynomial @xmath607 with the same ( real ) coefficients as @xmath3 . then @xmath608 is finitely @xmath33-determined as well ( * ? ? ? * proposition 1.7 ) , hence has fst . viewing @xmath63 as a subgroup of @xmath609 we denote by @xmath610 the zariski closure of @xmath63 in @xmath611 . by schwarz @xcite , the set @xmath610 is a reductive complex algebraic subgroup of the algebraic group @xmath611 , and if we write @xmath612 and @xmath613 for the lie algebras of @xmath63 and @xmath610 , respectively , then @xmath614 . we argue that @xmath615 , which will prove that @xmath616 by wall s theorem ( that is , theorem [ wallthm ] ) . the action of @xmath610 on @xmath617 is algebraic , and hence zariski continuous . viewing @xmath63 as a subset of @xmath610 with the induced zariski topology , the maps @xmath618 are zariski continuous for any fixed @xmath619 , and @xmath620 is a continuous extension of @xmath267 . the map @xmath267 is constant , because @xmath621 and hence @xmath622 for all @xmath182 . but @xmath63 is zariski dense in @xmath610 by @xcite , and points are closed in the zariski topology on @xmath623 ; hence @xmath620 must be constant as well . since this holds for all @xmath619 , it follows that @xmath615 , and @xmath624 by theorem [ wallthm ] . then @xmath625 , and since @xmath626 , we must have @xmath627 , and in particular @xmath628 . the results from section [ oftensmall ] allow us to efficiently compute maximal compact subgroups @xmath34 for germs that are of particular interest to us , as illustrated in the example below . the argumentation will obviously carry over to a wide range of other cases . let @xmath629 belong to the @xmath630 or @xmath631-series of singularities ; namely , let @xmath3 be one of the map - germs @xmath632 with @xmath633 and @xmath634 . then @xmath3 is a weighted homogeneous polynomial , and in particular , @xmath3 is @xmath577-equivariant , so @xmath635 is a compact subgroup of @xmath2 . we show that @xmath635 is a maximal compact subgroup of @xmath2 by showing that it is a maximal compact subgroup of @xmath34 . below , we give the argument for the germ @xmath636 ; the argument for @xmath637 is almost identical . see ( * ? ? ? * theorem 97 ) for a more detailed account . let @xmath572 be a maximal compact subgroup . suppose that @xmath638 , which acts on @xmath3 by @xmath639 we note that for elements of @xmath34 , the diffeomorphism @xmath310 is completely determined by @xmath640 ; in other words @xmath63 is determined by its action on the source space , and the projection @xmath641 restricts to an injection on @xmath34 . thus , it is enough to show that the set of pairs @xmath642 , which can be part of an element in @xmath34 , is isomorphic to @xmath635 . since @xmath643 , we can show this by investigating the @xmath64-jet of @xmath467 , denoted @xmath644.\ ] ] since @xmath467 comes from an element of @xmath34 , we must have * @xmath645 , * @xmath646 , * @xmath647 $ ] can only hold if @xmath648 , * @xmath649 $ ] never holds , * @xmath650 $ ] can only hold if @xmath83 is odd , * @xmath651 $ ] can only hold if @xmath83 is even . here , a ) follows from weighted homogeneity of @xmath3 ; b ) and c ) follows from the fact that @xmath652 is a finite matrix group so @xmath653 for some @xmath654 . to see d)-f ) , plug in polynomial expansions for @xmath655 and @xmath656 into the formula for @xmath3 and use weighted homogeneity of @xmath3 to deduce relations on the coefficients of the @xmath657 , which can only be true under the conditions d)-f ) . now we know that @xmath63 consists of the identity along with elements @xmath658 such that @xmath659 is of the form e ) if @xmath83 is odd , or f ) if @xmath83 is even . again , using the fact that @xmath63 is finite , we see that in both cases , there is only one valid value of @xmath599 . hence , @xmath660 , and @xmath661 is a maximal compact subgroup of @xmath2 . the authors wish to thank hans brodersen , maria aparecida soares ruas and david trotman , whose careful reading detected a mistake in the original version of the argument . this work was supported by the magnus ehrnrooth foundation , as well as the centre for stochastic geometry and advanced bioimaging , funded by a grant from the villum foundation . aasa feragen wishes to thank the department of mathematical sciences at the university of aarhus for its hospitality while the research leading to this article was carried out .

we present a stochastic theory for the nonequilibrium dynamics of charges moving in a quantum scalar field based on the worldline influence functional and the close - time - path ( ctp or in - in ) coarse - grained effective action method . we summarize ( 1 ) the steps leading to a derivation of a modified abraham - lorentz - dirac equation whose solutions describe a causal semiclassical theory free of runaway solutions and without pre - acceleration patholigies , and ( 2 ) the transformation to a stochastic effective action which generates abraham - lorentz - dirac - langevin equations depicting the fluctuations of a particle s worldline around its semiclassical trajectory . we point out the misconceptions in trying to directly relate radiation reaction to vacuum fluctuations , and discuss how , in the framework that we have developed , an array of phenomena , from classical radiation and radiation reaction to the unruh effect , are interrelated to each other as manifestations at the classical , stochastic and quantum levels . using this method we give a derivation of the unruh effect for the spacetime worldline coordinates of an accelerating charge . our stochastic particle - field model , which was inspired by earlier work in cosmological backreaction , can be used as an analog to the black hole backreaction problem describing the stochastic dynamics of a black hole event horizon . .5 cm _ - invited talk given by blh at the international assembly on relativistic dynamics ( iard ) , june 2004 , saas fee , switzerland . to appear in a special issue of foundations of physics ( 2005 ) .

gr-qc0501029

we give in this paper a summary of our research program ( commenced in @xcite ) on the quantum statistical dynamics of relativistic particles moving in a quantum field . the relativistic particles are not coupled to each other explicitly but only through their interactions with a quantum field . because of this we need to take into consideration both the effect of each particle on the quantum field and the backreaction of the quantum field on each particle . a self - consistent treatment of the particle - field system dynamics is thus necessary . the backreaction from the field engenders _ nonlinear _ coupling amongst the particles , and _ non - markovian _ dynamics . we present a new approach to this problem which highlights the stochastic effects of noise , decoherence , dissipation , fluctuations , and correlations , and their interconnections . in @xcite we presented the basic framework built on the concepts of quantum open systems @xcite , the model of quantum brownian motion ( qbm ) @xcite , and the methodologies of the influence functional @xcite , closed - time - path ( in - in ) @xcite coarse - grained effective action @xcite , and world line quantization @xcite . in @xcite we applied this framework to spinless relativistic moving particles in a quantum scalar field , and derived the stochastic equations of motion , known as the abraham - lorentz - dirac - langevin ( aldl ) equations , for their trajectories with self - consistent backreaction . the quantum fluctuations of the field are partially encoded as a stochastic noise in the aldl equations . the mean trajectory obtained from taking the stochastic average is governed by abraham - lorentz - dirac ( ald ) equations with modified coefficients whose time - dependence enforce causality . here we use the approach developed in @xcite to analyze a uniformly accelerated particle where the trajectory is self - consistently determined by its interaction with a quantum field ( this includes the effects of radiation reaction and vacuum fluctuations ) . this important example illustrates the interconnections between radiation , radiation reaction , dissipation , quantum fluctuations , and the unruh effect . specifically , the uniform acceleration alters the quantum vacuum which in turn induces thermal - like fluctuations in the trajectory of a particle . this manifestation of the unruh effect in terms of fluctuating trajectories is in contrast to the original formulation and more common derivations @xcite , where a particle with internal degrees of freedom ( called a detector in the literature @xcite ) moving on a uniformly accelerating prescribed trajectory feels hot at the unruh temperature @xmath0 . before delving into the technical details , in the remainder of this section we discuss the outstanding physical issues touched on in this work . in sec . 2 we describe the relation of classical radiation , radiation reaction , vacuum fluctuations and quantum radiation , correcting the misconception of directly linking vacuum fluctuations with radiation reaction . in sec . 3 we present the main structure of our approach based on the open system concept and the coarse - grained effective action techniques . we show the steps leading to a derivation of the ald - langevin equation for charges moving in a quantum field . in sec . 4 we specialize to the case of a particle moving in a background field which imparts to it uniform acceleration on the average , and derive the unruh effect . we discuss how this example can serve as an analog for studying the stochastic dynamics of black hole event horizons . in sec . 5 we step back for a broader perspective and give some discussions of a ) the special features of this new approach and its potential in overcoming the stumbling blocks present in existing approaches ; and b ) applications of these results and directions for further development . in this program of investigation we take a microscopic approach and an open - systems perspective , using quantum field theory as the tool to provide a first - principles derivation of moving particles interacting with a quantum field . we begin with the closed system of quantum particles and fields . a _ closed system _ can be meaningfully partitioned into subsystems if there exists discrepancies in the scales describing each subsystem , or in accordance to the physical scales present in their interaction strengths ( energy or time scales ) in relation to the probing scale or resolution accuracy . if we are interested in the details of one such subsystem ( e.g. , particle dynamics ) and decide to ignore certain details of the other subsystems ( e.g. , details in the quantum field configurations , such as correlations and phase information ) comprising the _ environment _ , the distinguished subsystem is rendered an _ open system_. the overall effect of the coarse - grained environment on the open - system can be captured by the influence functional technique of feynman and vernon @xcite , or the closely related closed - time - path ( ctp ) effective action method of schwinger and keldysh @xcite . these are the initial value , or in - in , formulations suitable for following the time evolution of a system , in contradistinction to the usual in - out ( schwinger - dewitt ) effective action formulation useful for s - matrix calculations . for the model of particle - field interactions under study , coarse - graining the quantum field yields a nonlocal _ coarse - grained effective action _ ( cgea ) for the particle motion @xcite . an exact expression may be found in the special case that the particle and field are initially uncorrelated , and the field state is gaussian . the cgea may be used to treat the fully nonequilibrium quantum dynamics of interacting particles . however , when higher - order quantum corrections are suppressed by decoherence , the influence of the quantum fluctuations of the field may be encoded as stochastic noise , and the cgea can be transcribed into a _ stochastic effective action _ @xcite describing the stochastic particle motion . upon further coarse - graining , taking the stochastic average gives the semiclassical particle trajectory . let us analyze the physics of nonequilibrium processes for such particle - field systems at separate levels . * classical * level : from the complete quantum microscopic description of the system ( particle ) and the environment ( field ) , a classical description is reached when the system and environment are coarse - grained to an extent that both the particle and field histories are fully decohered @xcite . the fully coarse - grained trajectory for the particle with backreaction from the fully coarse - grained field then obeys classical equations of motion , such as the ald equation . the coarse - graining length scale necessary to achieve full classicality typically far exceeds the length scale for backreaction pathologies like preacceleration and runaway solutions . * semiclassical * level : this is often defined as a classical system ( particles or detectors ) interacting with a quantum environment ( quantum field ) . it can be obtained from the complete quantum microscopic description by fully coarse - graining the environment ( the quantum field ) and finding the quantum average of the particle trajectory . in the special but important class when the field enters linearly in the particle - field interaction term , and without field self - interactions , coarse graining over the quantum field gives an exact cgea . the cgea yields the equations of motion for the system which in the leading order approximation is the semiclassical limit , followed by higher - order quantum corrections . in this particular situation , if decoherence can sufficiently suppress the nonlinear quantum corrections to the semiclassical equations of motion from the particle , then the semiclassical limit from the cgea agrees with the classical limit . * stochastic * level : going beyond a mean - trajectory ( or fully coarse - grained ) description of the particle , a stochastic component in the particle trajectory appears , as induced by the quantum field fluctuations manifesting as a classical stochastic noise counterbalancing the quantum dissipation in the system dynamics . the cgea may be transcribed into a stochastic effective action , encoding much of the quantum statistical information of the field and the state of motion of the system in a noise correlator for the particle . the stochastic effective action yields classical stochastic equations of motion for the system which embodies a quantum dissipation effect ( _ over and above _ the classical radiation reaction ) that balances the quantum fluctuations via a fluctuation - dissipation relation ( fdr ) . the classical theory of moving charges interacting with a classical electromagnetic ( em ) field , when the backreaction of the em field is included , has many controversial difficulties @xcite . the generally accepted classical equation of motion for a charged , spinless point particle including radiation - reaction is the abraham - lorentz - dirac ( ald ) equation @xcite : @xmath1 ( see sec . ii for notations and definitions . ) the timescale @xmath2 determines the relative importance of the radiation - reaction term . for electrons , @xmath3 secs , which is the time it takes light to cross the electron classical radius @xmath4 m . the ald equation is pathological . because it is a third order differential equation , it requires the specification of the initial acceleration in addition to the usual position and velocity required by second order differential equations . this leads to the existence of runaway solutions . physical ( e.g. non - runaway ) solutions may be enforced by transforming eq . ( [ classical lorentz - dirac equation ] ) to a second order integral equation with boundary condition such that the final energy of the particle is finite and consistent with the total work done on it by external forces . but the removal of runaway solutions yields acausal solutions that pre - accelerate on timescales @xmath5 this is a source of lingering questions on whether the classical theory of point particles and fields is causal . there have been many efforts to understand charged particle radiation - reaction in the classical and quantum theory . to satisfy the self - consistency and causality requirements different measures are introduced to the underlying model . examples include imposing a high - energy cutoff for the field , an extended charge distribution , special boundary conditions , particle spin , and the use of perturbation theory or order reduction techniques @xcite . previous work closely related to ours include those on quantum langevin equations @xcite , and the application of the feynman - vernon influence functional to non - relativistic particle coherence @xcite and stochastic gravity @xcite . a major distinction of our work is the focus on relativistic systems , non - equilibrium processes , and nonlinear particle - field interaction . consider the situation where a localized particle at rest has some quantum fluctuations in its position just before an external force is applied at time @xmath6 , and where , by chance , the fluctuation is one that seems to anticipate the force creating the _ _ appearance _ _ of preacceleration . or , consider a situation where a quantum fluctuation produces the _ appearance _ of runaway acceleration for some period of time . while such specific fluctuations may be highly improbable to realize , they are among the set of fine - grained solutions in a sum - over - histories formulation of quantum mechanics . these examples illustrate the observational challenges to distinguishing cause and effect on a fine - grained quantum scale . why is there such a difficulty ? the underlying microscopic theory may be consistent ( though for field theory this may not be obvious ) and causal , in the sense that the operator equations of motion are causal , but the quantum state ( for particle or field ) is inherently nonlocal . observables ( expectation values ) of the particle motion involve both the operator equations of motion and the nonlocal quantum state , and thus the question of causality for trajectories requires more careful treatment . operationally , we do not observe fine - grained histories ; instead we observe coarse - grained histories where the coarse - graining scale is set by the measurement resolution in both time and space . with coarse graining , the quantum fluctuations giving highly nonclassical trajectories are suppressed , and we therefore expect that a causal and consistent quantum theory should yield , upon suitable coarse graining , a classical or semiclassical limit that is pathology free . this is an oversimplified sketch of a more elaborate theory @xcite based on the decoherent history approach to quantum mechanics @xcite . a particularly interesting class of dynamics from the solutions to the ald equation is that of a uniformly accelerated particle , defined by @xmath7 . this class of dynamics exemplifies many of the important issues under discussion . first , it follows from the ald equations that the classical radiation reaction force vanishes despite the existence of classical ( lamor ) radiation registered as an energy flux at infinity . hence , there is no direct balance between radiation and radiation reaction . there is one existing belief that since radiation is associated with radiation reaction , the extra work done on the particle against the radiation reaction force must directly provide the energy that goes into radiation , but this is a static viewpoint that neglects the full interplay of particle , near field ( the so - called acceleration or shott field ) , and far field ( the radiation field ) dynamics . fields are dynamical entities with unusual properties ( such as nonlocality and field correlations ) much more complex than just radiation , which is a far - field definition , and energy can be attributed to a variety of sources other than radiation . for example , the acceleration field is known to contain energy and do work . one can not simply equate work done against radiation reaction forces with the energy ( instantaneously ) radiated into infinity . it would require a ` freezing out ' of the near - field s ability to exchange or adjust the form of its energy consistently in time to be commensurate with the far field behavior known as radiation . during periods of uniform acceleration , the energy transfer into radiation comes from acceleration fields , leaving zero radiation reaction . this is a special situation . during periods of nonuniform acceleration , there is radiation reaction , and one finds a mixture of field components ; but even then , the energy apportioned to radiation can not be instantaneously ascribed to work done against radiation reaction . on matters related to fields one should look carefully at the energy content locally and guard against making global statements . it has long been known that a uniformly accelerated detector ( uad ) ( a detector is defined as a particle with some internal degree of freedom ) coupled to a quantum field following a prescribed trajectory registers its quantum field vacuum as a thermal state with temperature @xmath8 this is the unruh effect @xcite , often cited as an analog of hawking effect @xcite . but there are important differences : in the case of a black hole , real radiation of thermal nature is emitted at the hawking temperature . for a uniformly accelerated detector ( uad ) there is no emitted radiation associated with the unruh effect ( see , e.g. , @xcite and references therein ) except for a transient period when the charge is coming into equilibrium with its environment @xcite . for a uniformly accelerated _ charge _ ( uac ) there is of course classical radiation , but it is different from the unruh radiation , which is a distinctly quantum effect . we demonstrate below that for a uniformly accelerated charge interacting with a quantum field , the vacuum fluctuations induce stochastic fluctuations in the particle _ trajectory _ with a thermal character . we can refer to this as the unruh effect for charges , the role of the detector is played by the particle motion itself rather than an internal degree of freedom . in the semiclassical limit ( neglecting quantum corrections from the particle ) the _ average _ emitted radiation is the usual classical radiation : there is no additional net quantum radiation from the particle . what is new and interesting in our finding is that there could be radiation associated with the stochastic component of the particle trajectory . for instance , these could conceivably generate fluctuations in the emitted radiation , perhaps with a thermal character , or they could produce other non - classical correlations in the radiation . this requires further investigation . it has been shown that for a uniformly accelerated detector on a fixed trajectory the field correlation is altered around the detector , but as these altered correlations fall off faster than radiation they can be considered as a vacuum polarization effect @xcite . similar vacuum polarization effects are also expected for a moving charge ( contrast this prediction with @xcite ) . finally , the radiation emitted by an evaporating ( shrinking ) black hole under non - equilibrium conditions is expected to have a non - thermal component @xcite over and above the usual thermal ( hawking ) part . we are using the analog with non - uniformly accelerated detectors or charges to investigate this issue @xcite . we begin with the full microscopic theory of a spinless charged particle interacting with a scalar field . the action for the closed system of particle plus field is @xmath9 = s\left [ z , h\right ] + s\left [ \varphi\right ] + s\left [ z,\varphi\right ] , \ ] ] where@xmath10 & = \int d\tau\left ( m_{0}\sqrt{\dot{z}^{\mu}\dot { z}_{\mu}}+h_{\mu}\left ( \tau\right ) \dot{z}^{\mu}\left ( \tau\right ) \right ) \\ s\left [ \varphi\right ] & = \int d^{4}x\,\frac{1}{2}\left ( \partial_{\mu } \varphi\right ) ^{2}\\ s\left [ z,\varphi\right ] & = \int d^{4}x\,j\left ( x;z\right ) \left ( \varphi\left ( x\right ) + \varphi_{ext}\left ( x\right ) \right ) .\end{aligned}\ ] ] the spacetime worldline coordinates of the particle are @xmath11 @xmath12 is the scalar field@xmath13 and @xmath14 is an external field . the functions @xmath15 are auxiliary sources introduced to generate correlation functions , after which they are set to zero . the field couples to the particle through the reparametrization invariant scalar current@xmath16 the action @xmath17 $ ] is manifestly invariant under reparametrizations @xmath18 the particle mass is @xmath19 and charge is @xmath20 we use units where @xmath21 , greek indices @xmath22 and @xmath23 consider an initial density matrix for the particle plus field at time @xmath6 that has the factorized form @xmath24 where @xmath25 is the initial density matrix for the particle subsystem @xmath26 and @xmath27 is the initial density matrix for the field subsystem @xmath28 . we note at the start that strictly speaking such factorized states are unphysical because they represent a completely decoupled particle and field . more general initial states could be considered , but treating a fully physical state remains an important challenge @xcite . for @xmath29 we assume that the initial field state is gaussian ( this includes thermal , squeezed , and coherent states , and therefore provides us with a fairly rich set of interesting and physical examples . ) a relativistic covariant particle state @xmath30 where @xmath31 is the positive frequency operator for creating a field excitation centered at @xmath32 has a localization size ( in terms of the charge - density expectation value ) of a compton wavelength @xmath33 imparting a natural _ minimum _ coarse - graining scale . typical particle wavepackets have much larger ( characteristic ) de broglie wavelengths , and thus require considerable further averaging of fine - grained trajectories to achieve a coarse - grained decoherent history . we assume that the initial particle state is a wavepacket localized around @xmath34 with width @xmath35 we will see that the degree of coarse graining does ( weakly ) effect the semiclassical trajectory via higher derivative terms that can generally be neglected . ( note that both @xmath36 and @xmath37 are far larger that the classical radius of a charged particle @xmath38 the scale at which pathological backreaction behavior manifests ) . for the closed system , the density matrix at a later time @xmath39 is@xmath40 where @xmath41 is the unitary evolution operator for the fully interacting particle plus field system . the matrix elements of @xmath41 can be given a sum over histories representation:@xmath42 \right\ } , \end{aligned}\ ] ] where the functional measures are@xmath43 subject to appropriate boundary conditions that are discussed in @xcite . the in - in or closed - time - path ( ctp ) generating functional is given by @xmath44 & = \text{tr}_{\varphi , z}\text{\ } \left [ \hat{u}_{h}\left ( t_{f},t_{i}\right ) \hat{\rho}_{i}% \hat{u}_{h^{\prime}}\left ( t_{f},t_{i}\right ) ^{\dagger}\right ] \\ & = \int d\mathbf{z}_{f}d\varphi_{f}d\varphi_{i}d\varphi_{i}^{\prime } d\mathbf{z}_{i}d\mathbf{z}_{i}^{\prime}\text{\ } \nonumber\\ & \times k_{h}\left ( \varphi_{f},\mathbf{z}_{f},t_{f};\varphi_{i}% , \mathbf{z}_{i},t_{i}\right ) k_{h^{\prime}}^{\ast}\left ( \varphi _ { f},\mathbf{z}_{f},t_{f};\varphi_{i}^{\prime},\mathbf{z}_{i}^{\prime}% , t_{i}\right ) \nonumber\\ & \times\rho\left ( \varphi_{i},\mathbf{z}_{i};\varphi_{i}^{\prime } , \mathbf{z}_{i}^{\prime};t_{i}\right ) .\nonumber\end{aligned}\ ] ] this generating functional is a tool for computing the average particle trajectory and its fluctuations ( i.e. the correlation functions for the particle worldline coordinates ) including the effects of backreaction from the coarse - grained field self - consistently . using the assumed initially factorized state and the sum - over - histories representation , the generating functional may be written as@xmath45 = \int dzdz^{\prime}e^{i\left\ { s\left [ z , h\right ] -s\left [ z^{\prime},h^{\prime}\right ] \right\ } } f\left [ z , z^{\prime}\right ] , \ ] ] where@xmath46 & = \int d\varphi d\varphi^{\prime}e^{i\left\ { s\left [ \varphi\right ] + s\left [ z,\varphi\right ] -s\left [ \varphi ^{\prime}\right ] -s\left [ z^{\prime},\varphi^{\prime}\right ] \right\ } } \rho_{i}\left ( \varphi_{i},\varphi_{i}^{\prime},t_{i}\right ) \nonumber\\ & \equiv\exp\left\ { is_{if}\left [ z , z^{\prime}\right ] \right\}\end{aligned}\ ] ] is the feynman - vernon influence functional , and @xmath47 $ ] is the influence action . from the definition of the in - in generating functional it follows that@xmath48 } { \delta h_{\mu}\left ( \tau_{n}\right ) ... \delta h_{\mu}\left ( \tau_{1}\right ) } \right\vert _ { h , h^{\prime}% = 0}=\text{tr}_{\varphi , z}\langle\hat{z}^{\mu}\left ( \tau_{n}\right ) ... \hat{z}^{\nu}\left ( \tau_{1}\right ) \hat{\rho}_{i}\rangle.\ ] ] hence , we see that @xmath49 is indeed a generating function for the expectation values of the worldline coordinates at an arbitrary proper time @xmath50 evaluating the full generating functional is virtually impossible . the interaction term @xmath51 , $ ] while linear in @xmath52 is highly non - linear in the worldline coordinate @xmath53 however , if the initial state of the field @xmath54 is gaussian , the influence functional involves purely gaussian integrands in the variables @xmath55 , and thus can be evaluated exactly , given the well - known result ( see e.g. , @xcite ) @xmath56 , \ ] ] where @xmath57 \hat{\rho}_{a}\left ( t_{i}\right ) \right ) \theta\left ( t , t^{\prime}\right ) , \\ g_{h}\left ( x , x^{\prime}\right ) & = tr_{\hat{\varphi}}\left ( \left\ { \hat{\varphi}\left ( x\right ) , \hat{\varphi}\left ( x^{\prime}\right ) \right\ } \hat{\rho}_{\hat{\varphi}}\left ( t_{i}\right ) \right)\end{aligned}\ ] ] are the retarded and hadamard green s functions , respectively . the brackets @xmath58 $ ] and @xmath59 denote commutators and anticommutators . the sum and difference currents @xmath60 are defined by @xmath61 we also define@xmath62 and@xmath63 the in - in generating functional can be compactly expressed as @xmath64 = \int dz^{\left ( + \right ) } dz^{\left ( -\right ) } e^{is_{cgea}\left [ z^{\left ( \pm\right ) } , h^{\left ( \pm\right ) } \right ] } , \label{full generating functional}%\ ] ] where @xmath65 = s_{z}\left [ z , h\right ] -s_{z}\left [ z^{\prime},h^{\prime}\right ] + s_{if}\left [ z^{\left ( \pm\right ) } \right]\ ] ] is the in - in or ctp coarse - grained effective action ( ctpcgea ) . we emphasize that @xmath66 encapsules the full effects of the coarse - grained quantum field , and @xmath67 $ ] generates , in principle , the full information about the quantum dynamics of the particle . the semiclassical description of the particle dynamics is given by the particle trajectory s expectation value @xmath68 neglecting particle quantum corrections , obtained from the equations of motion@xmath69 upon setting @xmath70 the imaginary part of @xmath71 which involves the hadamard function , gives vanishing contribution . consequently , the semiclassical equations of motion depends only on the retarded green s function and are therefore real and causal . the semiclassical trajectory is a solution of the equation derived from eq . ( [ semiclassical eq from scea ] ) @xcite @xmath72 where@xmath73}-\ddot{z}_{\mu}.\ ] ] the external potential @xmath74 provides an external force @xmath75 with two main components : a gradient force @xmath76}\varphi_{ext}=e\left ( g_{\mu\nu}-\dot{z}_{\mu}\dot{z}_{\nu}\right ) \partial^{\nu}\varphi_{ext},$ ] and an effective contribution to the particle mass , @xmath77 note that @xmath78 identically , and therefore @xmath79 as required of a relativistic force that preserves the mass - shell constraint @xmath80 the scalar field force @xmath81 is analogous to the electromagnetic force @xmath82 which satisfies @xmath83 the retarded green s function is singular when @xmath84 and so requires regularization . we choose@xmath85 where @xmath86 is a first order bessel function and @xmath87 in the limit @xmath88 consistency requires that the hadamard green s function is regulated with the same effective cutoff : @xmath89 we assume that the high - frequency mode cutoff @xmath90 comes from the limit of the measurement resolution or the preparation scale for the initial particle state , so that @xmath91 it is reasonable to assume ( though this argument needs to be made rigorous ) that on finer - grained scales the correlations of the particle and field are undisturbed by the motion of the particle . this suggests viewing the initial factorized state as representing a particle that , through the initial state preparation process at time @xmath6 , is uncorrelated only with the _ longer _ wavelength modes of field , while it remains fully dressed by the ( undisturbed ) short wavelength modes . the regulated green s function is therefore a model for the response of the long wavelength field modes to the particle motion , rather than a short wavelength ( high energy ) modification of the underlying quantum field theory . using our explicit choice for @xmath92 , the resulting equations of motion are@xmath93 where @xmath94 @xmath95 is the elapsed proper - time from the initial time @xmath96 [ recall that @xmath2 ] . the time - dependent radiation reaction term proportional to @xmath98 has been absorbed into a time - dependent mass @xmath99 at @xmath100 , corresponding to the initial time when the field and particle are uncorrelated , @xmath101 ( @xmath19 is the bare mass which includes the renormalization from the short wavelength field modes above the cutoff @xmath90 ) . the field dresses the particle , changing its effective mass , which reaches an asymptotic value @xmath102 on the cutoff timescale @xmath103 it is interesting to note that this shift is @xmath104 times the shift that results for an electromagnetic field . since the em field may be viewed as two independent scalar fields ( one for each polarization ) , the factor of @xmath105 for a scalar field is intuitively reasonable . a negative rather than positive mass shift comes about due to differences between how scalar and em fields couple to a particle . radiation reaction is given by the last two terms on the right side ( rhs ) of eq . ( [ full semiclassical limit ] ) . the second term has the usual third derivative form for ald radiation reaction , except for the time - dependent coefficient @xmath106 this function depends on the exact choice of the regulated retarded green s function , but it has the general property that @xmath107 and @xmath108 with the timescale for the transition from 0 to 1 being @xmath103 the last term on the rhs of eq . ( [ full semiclassical limit ] ) represent a set of higher derivative contributions to radiation reaction . these are suppressed at low energies by factors of @xmath109 . the functions @xmath110 vanish when @xmath111 and go to constant values on a timescale @xmath103 at the initial time @xmath100 the radiation reaction terms , including the higher derivative contributions , vanish , and this makes the solutions to eq . ( [ full semiclassical limit ] ) unique ( requiring only the usual position and velocity initial data ) , causal ( no preacceleration ) , and runaway free . the functions @xmath112 involve fourth and higher derivative contributions:@xmath113 the dependence of the equations of motion in eq . ( [ full semiclassical limit ] ) on @xmath90 and the higher derivative terms reflects the degree to which the particle subsystem ( trajectory ) is distinguished from the full system ( particle and field ) by coarse graining . the larger @xmath90 the more point - like the particle . for sufficiently point - like particles ( finer grained trajectories ) and longer time scales ( @xmath114 ) , the higher derivative terms are strongly suppressed , and the dominant contribution to the equations of motion are independent of @xmath115 giving @xmath116 thus , at low energies and long times , the radiation reaction force is of the usual ald form . ( note that the coefficient @xmath117 is @xmath105 of the value for radiation reaction in the em field . again viewing the em field as equivalent to two scalar fields , this result is natural . ) since in this limit @xmath118 is a very small parameter , in practice we can treat eq . ( [ scalar ald equation ] ) perturbatively . the lowest order solution is found by neglecting the radiation reaction term so that the trajectory is determined by the background field @xmath119 giving@xmath120}\varphi_{ext}\left ( z\right ) . \label{iterative approx to z''}%\ ] ] taking one proper - time derivative gives@xmath121}\varphi_{ext}\left ( z\right ) \right ) . \label{iterative approx to z'''}%\ ] ] these two expression for @xmath122 and @xmath123 can be substituted into the radiation reaction term in eq . ( [ scalar ald equation ] ) , yielding an order @xmath124 equation for the radiation reaction force containing only first derivatives @xmath125 . this procedure can be iterated to produce a runaway - free and causal perturbative approximation to any order in @xmath126 but clearly this expansion which contains only first derivatives @xmath125 can not converge to the third - derivative ald equation . to fully understand causality the non - markovian short - time behavior must be examined , with eq.([full semiclassical limit ] ) being an example of how non - markovian effects can enforce causality . we have analyzed the special case when the particle and field are initially uncorrelated below some coarse grained preparation scale @xmath127 the nonequilibrium time - dependence of the radiation reaction force reflects the fact that in any nonequilibrium quantum setting it takes time for the particle s self - field to adjust to changes in its motion . this lag preserves causality unless one takes the cutoff to infinity ( finest grained histories ) : but this would imply that the high - energy structure of the fundamental theory is critical to the equations of motion even at low energy . this , while in principle possible , is in contradiction with our experience that realistic physical conditions are described by low - energy effective theories . our analysis is a first step in demonstrating the full causality of the coarse - grained semiclassical limit of particle motion in quantum field theory . next steps include considering more general initial conditions , rigorously deriving a coarse - grained ( regulated ) green s function from a physical coarse - graining mechanism , and analyzing the effects of higher - order quantum corrections . we now go beyond the average trajectory and include the trajectory fluctuations induced by the quantum fluctuations of the field . to do this , we return to the full quantum generating function , eq . ( [ full generating functional ] ) . in practice , to make use of the in - in generating functional the non - linear @xmath128 path integrals must be evaluated using some approximation . to find quantum corrections to the semiclassical solution , the loop expansion is often invoked . we describe here a different approach that transforms @xmath66 into a stochastic effective action . noting that @xmath129 is real and symmetric , and therefore has positive eigenvalues , it follows that @xmath130 \right ) = \frac{1}{4}\int\int dxdx^{\prime}\jmath^{\left ( -\right ) } \left ( x\right ) g_{h}^{\omega}\left ( x , x^{\prime}\right ) j^{\left ( -\right ) } \left ( x^{\prime}\right ) \geq0,\ ] ] and@xmath131 \right\vert = \left\vert e^{is_{cgea}}\right\vert = e^{-im\left ( s_{if}\left [ z^{\left ( \pm\right ) } \right ] \right ) } < 1.\ ] ] the influence functional is suppressed for large @xmath132 , which reflects the decoherence associated with the radiation produced by the particle current . consistency with the cutoff @xmath90 requires that the current @xmath132 does not produce radiation with frequency greater than @xmath127 the contribution to @xmath66 from the imaginary part of the influence functional may be re - written as@xmath133 e^{i\int j^{\left ( -\right ) } \left ( x\right ) \tilde{\varphi}\left ( x\right ) dx},\ ] ] where@xmath134 = ne^{-\int dxdx^{\prime}\tilde{\varphi } \left ( x\right ) g_{h}^{\omega}\left ( x , x^{\prime}\right ) ^{-1}% \tilde{\varphi}\left ( x^{\prime}\right ) } % \ ] ] is a normalizable probability distribution for a _ stochastic field _ @xmath135 the two - point function for the stochastic field is@xmath136 showing that @xmath137 encodes the same information as the quantum field anti - correlation function , with noise ( or correlations ) above the cutoff assumed to be coarse grained and averaged out . the in - in generating functional may be re - expressed as@xmath64 = \int dz^{\left ( + \right ) } dz^{\left ( -\right ) } \int d\tilde{\varphi}p\left [ \tilde { \varphi}\right ] e^{is_{stoch}\left [ z^{\left ( \pm\right ) } ; \tilde { \varphi}\right ] } , \ ] ] where@xmath138 & = \operatorname{re}\left\ { s_{cgea}\left [ z^{\left ( \pm\right ) } \right ] \right\ } + \int j^{\left ( -\right ) } \left ( x\right ) \tilde{\varphi}\left ( x\right ) dx\label{stochastic effective action}\\ & = s_{z}\left [ z , h\right ] -s_{z}\left [ z^{\prime},h^{\prime}\right ] + \int dxj^{\left ( -\right ) } \left ( x\right ) \left ( \tilde{\varphi}\left ( x\right ) + \int dx^{\prime}g_{r}^{\omega}\left ( x , x^{\prime}\right ) j^{\left ( + \right ) } \left ( x^{\prime}\right ) \right ) \nonumber\end{aligned}\ ] ] is the stochastic effective action . the generating functional now involves both a sum over trajectories and a sum over stochastic field configurations . if we neglect higher order quantum corrections and take @xmath139 \simeq s_{stoch}\left [ z_{0}^{\pm};\tilde{\varphi}\right ] , $ ] where @xmath140 are the classical ( extremal solutions ) , then we obtain a stochastic generating functional:@xmath141 = \int d\tilde{\varphi}p\left [ \tilde{\varphi}\right ] e^{is_{stoch}\left [ z_{0}^{\left ( \pm\right ) } ; \tilde{\varphi}\right ] } .\ ] ] this describes the stochastic regime , where higher order quantum effects are washed out by decoherence but some measure of quantum field induced stochasticity remains . we emphasize that this is not a stochastic theory of scalar field qed where stochasticity is introduced in an _ ad hoc _ manner . rather , we have started with a microscopic quantum theory and shown the conditions ( namely strong decoherence that suppresses higher order corrections ) whereby its behavior can be captured by its stochastic effects . the stochastic effective action generates nonlinear stochastic equations of motion , defined by@xmath142 where @xmath143 is now taken to be a stochastic variable . using eqs . ( [ stochastic effective action ] ) and ( [ equations of motion from stochastic action ] ) , we find that @xmath144 or , using the regulated green s function and dropping higher derivative terms , the nonlinear stochastic equations of motion take the form ( in the late - time limit @xmath145)@xmath146 + \frac{\tau_{0}}{2}\left ( \dot { z}_{\mu}\ddot{z}^{2}+\dddot{z}_{\mu}\right ) . \label{nonlinear aldl}%\ ] ] the perturbative procedure illustrated in eqs . ( [ iterative approx to z ] ) and ( [ iterative approx to z ] ) can be similarly be applied to eq . ( [ nonlinear aldl ] ) to obtain first - derivative equations of motion . the stochastic field provides a stochastic force @xmath147 which , just like the force from @xmath119 satisfies @xmath148 and hence also preserves the mass - shell constraint . note that the stochastic field generates an effective stochastic mass term , @xmath149 in parallel with the effective mass term generated by @xmath150 in the short - time limit these equations are ( like the semiclassical equations ) modified by nonequilibrium ( time - dependent ) effects . the noise @xmath151 enters the langevin equation in a highly nonlinear way , and its correlator is conditional on the particle history , making the equation very complicated . since the notion of a semiclassical trajectory already requires substantial coarse - graining , which averages out short time and higher order quantum effects , we make the analog of the background field separation by expanding the equations around the semiclassical trajectory @xmath152 to obtain linearized langevin equations for the trajectory fluctuations @xmath153 . this gives@xmath154 the linearized noise @xmath155 has vanishing mean , @xmath156 and a two - point correlator@xmath157 that is independent of the particle fluctuations @xmath158 , but is dependent upon the self - consistently determined semiclassical trajectory @xmath159 the operator @xmath160 is evaluated in terms of the solution @xmath161 the linear force from the external field is given by @xmath162 the second term on the rhs of eq . ( [ linearized ald langevin ] ) represents the radiation reaction force , where @xmath163 is the dissipation kernel for linearized fluctuations of the trajectory around its average ( a first - derivative perturbative form of @xmath164 can be obtained using eqs . ( [ iterative approx to z ] ) and ( [ iterative approx to z ] ) to remove the @xmath122 and @xmath123 factors ) . the kernel @xmath164 represents what we have described as quantum ( or stochastic ) dissipation : the backreaction in response to the quantum field induced fluctuations . while there is no direct link between the classical ( ald ) radiation reaction force and quantum fluctuations , there is a relationship between the _ quantum dissipation _ and quantum flucutations @xcite . in a weak electric field , the vacuum of an uniformly accelerated charge ( uac ) assumes thermal attributes ( the unruh effect ) and additional fluctuations are induced in the particle motion . in a stronger field , charged particle pair - creation becomes important . these are two distinct types of particle production processes . in the semiclassical - stochastic limit ( in which we have been working ) one does not have charged - particle creation effects . in the analogous situation of semiclassical gravity and hawking radiation this entails the neglect of quantum gravitational effects . for sufficiently high intensity fields , charged particle pair creation can become a significant process , and the creation of both photons and charges must be treated as a fully quantum problem . the gravitation analogy would entail calculations based on some theory of quantum gravity . there , the emergence of classical spacetime from a theory of quantum gravity is an issue of fundamental importance . prior investigation has led to some preliminary understanding @xcite . one can use this simpler theory of particle - field interaction to gain some insight into the stochastic limit of quantum cosmology and quantum gravity . this was one of the original motivations for this line of research undertaken currently . an even more ambitious goal is to follow in detail the transition from stochastic to quantum , including quantum particle aspects . we can use the more tractable problem here of particle - field interactions to gain insight into the transition from stochastic and quantum regimes of gravity in the so - called ` bottom - up ' approach to quantum gravity @xcite . on this issue some interesting results have been obtained @xcite based on quantum brownian motion ( qbm ) models . the significant nonlinearities of locally interacting particles and fields in the present problem makes for a better model to study this issue . here we take a modest step in this direction by applying our methods to a uniformly accelerated charge in a scalar field . we choose a background field @xmath74 such that @xmath165 _ constant _ is a solution to eq . ( [ full semiclassical limit ] ) if we ignore the radiation reaction terms . it can be immediately verified that uniformly accelerating solutions satisfy @xmath166 . hence , neglecting the suppressed higher derivative terms and quantum corrections , the radiation reaction force identically vanishes , and the uniformly accelerating trajectory is a solution to eq . ( [ full semiclassical limit ] ) _ including _ radiation reaction . ( this results also holds if we use the perturbative first derivative approximation to the semiclassical equations of motion . ) we choose a coordinate frame where the average trajectory is in the @xmath167 plane , so that@xmath168 defining right - moving coordinates @xmath169 and left - moving coordinates @xmath170 the uniformly accelerated trajectory is @xmath171 and @xmath172 the future particle horizon , shown in fig . 1 ( a ) , is given by @xmath173 [ t ] figure.eps let us now examine the linearized stochastic fluctuations of the particle , induced by the stochastic force @xmath174 around the semiclassical trajectory . the structure of the field fluctuations are characterized by the anticorrelator ( the hadamard function ) along the accelerated trajectory:@xmath175 for a massless scalar field in the vacuum state , evaluated at two different times @xmath176 and @xmath177 , the hadamard function gives @xmath178 -\sin k\left [ t - r\right ] } { 8\pi^{2}r\left ( \tau,\tau^{\prime}\right ) } \right ) , \end{aligned}\ ] ] where @xmath179 and @xmath180 substituting the uniformly accelerated solution , it becomes ( re - inserting factors of @xmath181)@xmath182 where the general form for @xmath183 for a massive scalar field in @xmath184 spacetime dimensions is given in @xcite as @xmath185 @xmath186 is a modified bessel function of imaginary order . the function @xmath183 is the effective spectral density of the scalar field environment seen by the particle . in general it depends on the particle s acceleration @xmath187 and mass @xmath188 when @xmath189 the integral can be evaluated in closed form . for @xmath190 ( the case of @xmath191 spacetime which is considered here ) , the effective spectral density is independent of the particle state of motion : @xmath192 the form of the hadamard function in eq . ( [ hadamard function uniform acceleration ] ) is precisely the same as that for a bath of harmonic oscillators with temperature @xmath193 and spectral density given by @xmath194 . thus , the scalar field vacuum along the accelerated particle trajectory looks like a thermal state , with temperature proportional to the acceleration . the correlator of the linearized stochastic force on the particle , given by @xmath195 governs the _ response _ of the particle to the quantum field fluctuations . note that @xmath196 and thus the induced fluctuations act only in the @xmath197 plane : @xmath198 in the linearized response approximation . ( this results does not carry over to the nonlinear langevin equation [ eq . ( [ nonlinear aldl ] ) ] when higher order effects are included ) . because the stochastic force is given by @xmath199 the response of the particle depends on its state of motion through @xmath200 while the underlying quantum field looks thermal along the uniformly accelerated particle s average trajectory , the fluctuations that are induced in the motion are complicated by non - isotropic and relativistic effects . in fact these effects make this model more interesting . first , we note that the langevin equation describes the stochastic behavior of a particle s worldline , depicting fluctuations in both the space _ and _ time coordinates . this makes this system a simple analog for quantum cosmology with the stochasticity in time induced by quantum fluctuations . second , the particle motion is an interesting analog for stochastic black hole horizon dynamics . if we view the particle trajectory in the far future as effectively determining the horizon , the langevin equations for @xmath201 then describes a stochastic distribution of horizon positions [ see fig . we note that most of the stochastic spread of the particle trajectory ( and hence the horizon ) occurs for @xmath202 when the particle approaches the velocity of light , lorentz time dilation suppresses ( i.e. slows ) fluctuations from the perspective of external observers . the analogy with black hole horizon formation suggests that most of the stochastic spread in a black hole s horizon occurs in the early stages of its formation . we are using stochastic relativistic particle dynamics as a guide to the investigation of the horizon fluctuations in black hole dynamics . the open system concept coupled with the coarse - grained effective action and the influence functional techniques have been applied to quantum brownian motion @xcite , interacting quantum fields @xcite , and stochastic semiclassical gravity @xcite . the new task we have undertaken in recent years is the adoption of the worldline quantization method to this existing framework for the treatment of relativistic charged particles moving in a quantum field @xcite . in this alternative formulation of qed the particle worldline structure is highly efficient for describing the particle - like degrees of freedom ( such as spacetime position ) , while the field structure most easily describes processes like radiation and pair creation . here in the discussions , we summarize the special features and merits of this approach and then mention its applications and extensions . [ [ particle - versus - field - formulation - fixed - versus - dynamic - background ] ] particle versus field formulation , fixed versus dynamic background ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ the contrasting paradigms of particles versus fields give very different representations and descriptions for the same physical processes . the success of quantum field theory for the past century has catapulted the field concept to the forefront , with the particle interpreted as excitations of field degrees of freedom . in recent years , the use of the quantum - mechanical path integral in string theory has inspired a renewed effort towards particle - centric quantum formalisms . the so - called worldline quantization method where the particle s spacetime coordinate @xmath203 is quantized is especially useful for calculating higher - loop processes in non - dynamical classical background fields . background field treatment is a useful and well - defined method in conditions where the quantum fluctuations of the fields are small . however , many problems involving the dynamical degrees of freedom of the field require the inclusion of field dynamics beyond the fixed - background approximation . a simple example is radiation reaction . more examples are found in semiclassical gravity @xcite , where spacetime playing the role of ` particle ' is treated classically while interacting with the quantum fields . the proper treatment of radiation - reaction requires , at the least , consistency between the particle s ( quantum ) average trajectory and the mean - field equations of motion . when one treats the particle as quantum mechanical , but coupled only to the mean - field background ( and not the quantum field fluctuations ) one is working within the semiclassical regime @xcite . ( when quantum fluctuations of the particle motion are sufficiently small , one recovers the classical regime @xcite ) . in our program , we go beyond the semiclassical approximation to include the influence of quantum field fluctuations on the moving particle , thus reaching over to the stochastic regime . to do this properly , we must begin with the full quantum theory for both particle and field , since a particle can never fully decouple from the field . also , since most field degrees of freedom remain unobserved , we treat the particle as a quantum open system with the field acting as an environment . this is how nonequilibrium and stochastic mechanics ideas enter , and why the self - consistent quantum particle evolution is aptly described by stochastic quantum dynamics . one major improvement of our approach to the problem of moving charges in a quantum field is the consideration of self - consistent backreaction of the quantum field on the particle in the determination of its trajectory . we also find that conceptual issues are easier to resolve if we deal with such problems at four distinct levels : quantum , stochastic , semiclassical and classical , as explained earlier . confusion will arise when one mixes physical processes of one level with the other without knowing their interconnections , such as drawing the equivalence between radiation reaction with vacuum fluctuations . before summarizing our thoughts for processes under nonequilibrium conditions , which cover most cases save a few special yet important situations , such as uniform acceleration , let us remark that these well - known cases are what we would call ` test field ' or prescribed ( trajectory ) approximations . they are not self - consistently determined with backreaction considerations . these cases are easier to study because they possess some special symmetry , such as is present for the uniform acceleration case ( rindler spacetime ) , inertial case ( minkowski ) , or the eternal black hole spacetime . they are so legitimatized meaning , physically relevant only if the backreaction of the field on the particle permits such solutions ( which may not be difficult to attain if the field strength is weak ) . it is under these special conditions that a detector will detect thermality . these very special conditions are the presuppositions of the unruh and hawking effects . in more general situations the noise kernel is nonlocal , which means that the noise in the detector is colored and temperature is no longer a strictly viable concept . quantum field fluctuations , while responsible for the decoherence in the quantum particle ( system ) leading to the emergent semiclassical and classical behaviors , also impart an effective classical stochastic noise in the equations of motion for the particle trajectory @xcite . these stochastic differential equations feature colored noise that encodes the influence of quantum fluctuations in the field . when there is sufficient decoherence , these equations give excellent approximations to the particle motion . even in the case of weak decoherence they can provide a reasonable approximation to the particle correlation functions . more accurate results can be obtained from the n - particle - irreducible ( npi ) closed - time - path effective action under a large n approximation , which is a nice way to systemize the hierarchy of correlation functions from the cgea @xcite . this scheme has been applied to stochastic gravity @xcite . finally we want to connect our work with existing work in other areas , mention applications of the present results to related problems , and show directions where this new approach can stimulate further developments . in terms of theory , the world - line quantization formalism has been applied to qed , qcd and string theory @xcite . our interest in this present problem and its particular formulation stemmed partly from stochastic gravity @xcite , which is the stochastic generalization of semiclassical gravity . the relation of semiclassical and stochastic regimes was made clear in the investigation of quantum brownian motion @xcite . more interesting is the relation between the quantum and the stochastic regimes . some work has appeared addressing this issue @xcite but more insight is needed . this is of particular interest in the ` bottom - up ' approach to quantum gravity @xcite : what features of quantum gravity can one detect from the lower energy ( mesoscopic ) phenomena as manifested in stochastic gravity ? we do nt have a theory of quantum gravity but we can learn from a known and well - proven one such as qed . the investigations in this program offers a concrete example of how a full quantum theory can yield stochastic and semiclassical features . this could partially illuminate the common pathways from stochastic semiclassical gravity to quantum gravity . in terms of applications our results are useful for a range of problems , from the consideration of the quantum aspects of beam physics ( see , e.g. , @xcite ) to the consideration of possible schemes for the detection of unruh radiation , e.g. , in linear accelerators or storage rings ( note however our critique for certain proposals of the former @xcite and misconceptions in the so - called ` circular unruh effect ' @xcite ) . the stochastic trajectory from the backreaction of quantum fields studied in the last section is an interesting analog model to quantify black hole event horizon fluctuations . the new approach as illustrated in this report can also serve as a launching pad for the exploration of nonequilibrium dynamics of charges undergoing arbitrary ( self - consistently determined , not prescribed ) motions , strongly correlated quantum systems , relativistic quantum kinetic and stochastic theories . it is currently being generalized to curved spacetime ( see also @xcite ) and applied to gravitational radiation reaction problems . recently a field - theoretical derivation of the mstqw equation was given and a new stochastic radiation reaction equation proposed @xcite . this approach is also being applied to the study of quantum entanglement and teleportation of relativistic detectors @xcite ( see also @xcite ) and environmental decoherence and entanglement @xcite issues . a. o. caldeira and a. j. leggett , physica a * 121 * , 587 ( 1983 ) ; ann . ( n.y . ) * 149 * , 374 ( 1983 ) ; a. j. leggett _ et al . _ , rev . phys . * 59 * , 1 ( 1987 ) ; v. hakim and v. ambegaokar , phys . a * 32 * , 423 ( 1985 ) ; f. haake and r. reibold , phys . a * 32 * , 2462 ( 1985 ) ; h. grabert , p. schramm , and g. l. ingold , phys . rep . * 168 * , 115 ( 1988 ) ; b. l. hu , j. p. paz , and y. zhang , phys . d * 45 * , 2843 ( 1992 ) ; d * 47 * , 1576 ( 1993 ) . j. schwinger , j. math . phys . * 2 , * 407 ( 1961 ) ; l. v. keldysh , zh . fiz . * 47 * , 1515 ( 1964 ) [ sov . jetp * 20 * , 1018 ( 1965 ) ] ; g. zhou , z.b . su , b. hao , and l. yu , phys . rep . * 118 * , 1 ( 1985 ) ; z.b . su , l. y. chen , x.t . yu , and k.c . chou , phys . b * 37 * , 9810 ( 1988 ) ; b. s. dewitt , in _ quantum concepts in space and time _ , edited by r. penrose and c. j. isham ( claredon , oxford , 1986 ) ; r. d. jordan , phys . rev . d * 33 * , 444 ( 1986 ) ; e. calzetta and b. l. hu , phys . d * 35 * , 495 ( 1987 ) ; * 37 * , 2878 ( 1988 ) ; * 49 * , 6636 ( 1994 ) ; a. campos and e. verdaguer , phys . d * 49 * , 1861 ( 1994 ) . b. l. hu and y. zhang , coarse - graining , scaling , and inflation " univ . maryland preprint 90 - 186 ( 1990 ) ; b. l. hu , in _ relativity and gravitation : classical and quantum , _ proc . silarg vii , cocoyoc , mexico 1990 . j. c. d olivo et al ( world scientific , singapore 1991 ) ; sukanya sinha and b. l. hu , phys . d * 44 * , 1028 ( 1991 ) ; f. lombardo and f. d. mazzitelli , phys . d * 53 , * 2001 ( 1996 ) . esteban a. calzetta , bei - lok hu , francisco d. mazzitelli , coarse - grained effective action and renormalization group theory in semiclassical gravity and cosmology " physics report . 352 ( 2001 ) 459 - 520 [ hep - th/0102199 ] . barut and i.h . duru , phys . rev . * 172 * , 1 ( 1989 ) ; m.j . strassler , nucl . b * 385 * , 145 ( 1992 ) ; d.g.c . mckeon , ann . ( n.y . ) * 224 * , 139 ( 1993 ) ; m.j . schmidt and c. schubert , phys . lett . * 331 * , 69 ( 1994 ) ; phys . b * 318 * , 438 ( 1993 ) ; h. kleinert , _ path integrals in quantum mechanics , statistical and polymer physics _ ( world scientific , london , 1995 ) ; d. fliegner , m. g. schmidt , and c. schubert , nucl . b * 51c * , 174 ( 1996 ) ; j. jalilian - marian _ et al . d * * 62 , * * 045020 ( 2000 ) ; r. venugopalan and j. wirstam , phys . rev . d * 63 * , 125022 ( 2001 ) . b. l. hu , in _ proceedings of the third international workshop on thermal fields and its applications _ , cnrs summer institute , banff , august 1993 , edited by r. kobes and g. kunstatter ( world scientific , singapore , 1994 ) [ gr - qc/9403061 ] ; yuhong zhang , ph.d . thesis , university of maryland 1990 ( unpublished ) . r. b. griffiths , j. stat . 36 * , 219 ( 1984 ) ; r. omns , j. stat phys . * 53 * , 893 , 933 , 957 ( 1988 ) ; ann . ( ny ) * 201 * , 354 ( 1990 ) ; rev . phys . * 64 * , 339 ( 1992 ) ; _ the interpretation of quantum mechanics _ , ( princeton university press , princeton , 1994 ) ; m. gell - mann and j. b. hartle , in _ complexity , entropy and the physics of information _ , ed . by w. h. zurek ( addison - wesley , reading , 1990 ) ; j. b. hartle , quantum mechanics of closed systems in _ directions in general relativity _ vol . 1 , edited by b. l. hu , m. p. ryan and c. v. vishveswara ( cambridge univ . , cambridge , 1993 ) . m. abraham and r. becker , _ electricity and magnetism _ ( blackie , london , 1937 ) ; h.a . lorentz , _ the theory of electrons _ ( dover , new york , 1952 ) , pp . 49 and 253 ; p.a.m . dirac , proc . london a * 167 * , 148 ( 1938 ) . kampen , k. dan . selsk . mat . * 26 * ( 1951 ) ; e. moniz and d. sharp , phys . d * 10 * , 1133 ( 1974 ) ; h. levine , e. moniz , and d. sharp , am . * 45 * , 75 ( 1977 ) ; j. wheeler and r. feynman , rev . * 17 * , 157 ( 1945 ) ; f. e. low , annals phys . * 266 * , 274 ( 1998 ) ; v. krivitskii and v. tsytovich , sov . usp . * 34 * , 250 ( 1991 ) . h. callen and t. welton , phys . rev . * 83 * , 34 ( 1951 ) ; h. mori , prog . * 33 * , 423 ( 1965 ) ; g. w. ford , m. kac , and p. masur , j. math . phys . * 6 * , 504 ( 1965 ) ; h. haken , rev . * 80 * , 67 ( 1975 ) ; h. dekker , phys . rep . * 80 * , 1 ( 1981 ) . g. w. ford , j. t. lewis , and r. f. oconnell , phys . lett . * 55 * , 2273 ( 1985 ) ; phys . a * 37 , * 4419 ( 1988 ) ; g. w. ford and r. f. oconnell , phys . lett . a * 157 * , 217 ( 1991 ) ; phys . a * 174 * , 182 ( 1993 ) ; g. w. ford and r. f. oconnell , phys . rev . a * 57 * , 3112 ( 1998 ) . l. disi , found . phys . * 20 * , 63 ( 1990 ) ; p.m.v.b . barone and a.o . caldeira , phys . a * 43 * , 57 ( 1991 ) ; l. h. ford , phys . d * 47 * , 5571 ( 1993 ) ; phys . a * 56 * , 1812 ( 1997 ) ; h - p . breuer and f. petruccione , in _ relativistic quantum measurement and decohernce _ , edited by h - p . breuer and f. petruccione ( springer - verlag , berlin , germany , 2000 ) . for reviews see , e.g. , b. l. hu and e. verdaguer , stochastic gravity : a primer with applications `` , class . quant . grav . 20 ( 2003 ) r1-r42 [ gr - qc/0211090 ] ; stochastic gravity : theory and applications '' , in living reviews in relativity 7 ( 2004 ) 3 ; article number lrr-2004 - 3 [ gr - qc/0307032 ] . b. l. hu and a. raval , is there radiation in unruh effect ? " , in _ quantum aspects of beam physics _ , 18th advanced ifca deam dynamics workship , edited by pisin chen , world scientific , singapore ( 2001 ) [ quant - ph/0012135 ] . b. l. hu and p. r. johnson , beyond unruh effect : nonequilibrium quantum dynamics of moving charges , in _ quantum aspects of beam physics _ , 18th advanced ifca deam dynamics workship , edited by pisin chen , world scientific , singapore ( 2001 ) . lin pointed out ( _ private communication ) _ that the factors involving @xmath204 were missing in ref @xcite . these factors did nt play a role in the specific application treated in @xcite , but are necessary for the full consistency of the equations of motion . we thank him for this observation . ya . zeldovich and a. starobinsky , zh . fiz * 61 * , 2161 ( 1971 ) [ sov . phys.- jetp * 34 * , 1159 ( 1971 ) ] ; b. l. hu and l. parker , phys . d * 17 * , 933 ( 1978 ) ; f. v. fischetti , j. b. hartle and b. l. hu , phys . d * 20 * , 1757 ( 1979 ) ; j. b. hartle and b. l. hu , phys . d * 20 * , 1772 ( 1979 ) ; p. a. anderson , phys . d * 28 * , 271 ( 1983 ) ; * 29 * , 615 ( 1984 ) ; e. calzetta and b. l. hu , phys . d * 35 * , 495 ( 1987 ) ; a. campos and e. verdaguer , phys . d * 49 * , 1861 ( 1994 ) . e. calzetta and b. l. hu , correlations , decoherence , disspation and noise in quantum field theory , in _ heat kernel techniques and quantum gravity _ , edited by s. fulling ( texas a & m press , college station 1995 ) [ hep - th/9501040 ] ; e. calzetta and b.l . hu , phys . d * 61 * , 025012 ( 2000 ) . m. b. green , j. h. schwarz and e. witten , _ superstring theory _ ( cambridge university , cambridge , 1990 ) ; e. witten , physics today 49 , 24 ( 1996 ) ; j. polchinsky , _ superstring theory _ ( cambridge university press , cambridge 1998 )

the nearby ( ) cluster of galaxies abell 2199 was observed by suzaku in x - rays , with five pointings for ks each . from the xis data , the temperature and metal abundance profiles were derived out to kpc ( 0.4 times virial radius ) . both these quantities decrease gradually from the center to peripheries by a factor of , while the oxygen abundance tends to be flat . the temperature within ( kpc ) is kev , and the 0.510 kev x - ray luminosity integrated up to is erg s , in agreement with previous xmm - newton measurements . above this thermal emission , no significant excess was found either in the xis range below kev , or in the hxd - pin range above kev . the 90%-confidence upper limit on the emission measure of an assumed 0.2 kev warm gas is ( 3.77.5 ) arcmin , which is 3.77.6 times tighter than the detection reported with xmm - newton . the 90%-confidence upper limit on the 2080 kev luminosity of any power law component is erg s , assuming a photon index of 2.0 . although this upper limit does not reject the possible 2.1 detection by the bepposax pds , it is a factor of 2.1 tighter than that of the pds if both are considered upper limits . the non - detection of the hard excess can be reconciled with the upper limit on diffuse radio emission , without invoking the very low magnetic fields ( g ) which were suggested previously .

0911.5205

the intra - cluster medium ( icm ) in clusters of galaxies , i.e , hot thermal plasmas in collisional ionization equlibria confined within their gravitational potential , constitutes the most dominant form of cosmic baryons that has ever been detected . although the icm radiates predominantly optically - thin thermal x - ray emission , some clusters have been reported to exhibit excess signals above the thermal emission at the lowest or highest energy ends of their x - ray spectra . one interpretation of the soft excess is emission from warm hot intergalactic medium ( whim ) , which has been predicted from cosmological simulations ( e.g. @xcite ) and is expected to solve the so - called missing baryon problem ( e.g @xcite ) . another interpretation of the spectral soft and/or hard excess is non - thermal emission from accelerated particles in galaxy clusters ( e.g. @xcite ; @xcite ) : a power - low shaped spectrum can exceed the dominant thermal emission at both sufficiently high and low energies . this provides a possibility that galaxy clusters are giant accelerators in the universe . thus , the search for soft and hard excess signals in clusters forms a very important research subject . however , the instrumental sensitivity in these photon energies has been insufficient , and hence the existence of these excess components has remained controversial ( e.g. @xcite ; @xcite ; @xcite ; @xcite ) . among those galaxy clusters from which the excess emission has been reported , abell 2199 is particularly interesting , because it has been suspected to exhibit both the soft and hard excess components . in fact , kaastra et al . ( 1999 ) reported the detection of both , based on the bepposax , extreme ultraviolet explorer , and rosat observations . the two excess components were simultaneously expressed by a single power law with a photon index @xmath18 which is superposed on the thermal emission spectrum , and interpreted as inverse compton ( ic ) photons produced when the cosmic microwave background ( cmb ) photons are scattered up by relativistic electrons in the cluster . however , as argued by @xcite , the magnetic field in abell 2199 would have to be unusually weak ( @xmath19 g ) in order for the relativistic electrons not to produce synchrotron emission beyond measured upper limits of diffuse radio emission . as an alternative interpretation , @xcite proposed non - thermal bremsstrahlung by supra - thermal electrons which are being accelerated in the cluster . in the era of xmm - newton , the possibility of thermal emission from a warm gas also became another popular idea to explain the soft excess in several clusters . kaastra et al . ( 2004 ) reported a redshifted o vii emission line from a 0.2 kev warm gas in abell 2199 , along with coma , abell 1795 , sersic 159 - 03 , mkw 3s , abell 2052 , and abell 3112 , and hence ascribed their soft excess to emission from whim . at the same time , chandra has made progress on the icm physics of abell 2199 especially in the central region , such as an asymmetric temperature distribution in the direction perpendicular to the jets within @xmath20 of the center ( kawano et al . 2003 ) , and a weak isothermal shock associated with the central active galactic nucleus ( sanders & fabian 2006 ) . in order to examine the soft and hard excess phenomena of abell 2199 with a better sensitivity , we observed the object with suzaku ( @xcite ) . the excellent low - energy capability of xis - bi ( back illuminated ccd , @xcite ) and the low background of the silicon pin detectors in the hxd ( @xcite ; @xcite ) have provided the best data in the soft and hard x - ray bands , respectively . as a result , we have obtained tight upper limits on both excess components , and strengthened the dominance of the thermal emission from this cluster . in the present paper , quantities quoted from the literature are rescaled to the hubble constant of @xmath21 , and errors are given at 90 % confidence range unless otherwise stated . as shown in figure [ fig : mosaic ] and table [ table : observation ] , we observed abell 2199 with suzaku in the beginning of october 2006 . the observation was conducted in 5 separate pointings ( `` center '' and `` offset1 '' through `` offset4 '' ) , for @xmath1 ks each , with the xis field of view ( fov ) partially overlapping . throughout the 5 pointings , all the xis sensors were in normal clock mode without window or burst options . these observations became one of the first ones in which spaced - row charge injection ( sci ) was applied to the xis detectors ( nakajima et al . 2008 ) . in the present paper , we use data of xis0 , xis1 , xis3 and the hxd - pin . although xis2 was still operational , we do not use its data because a database of non x - ray background ( nxb ) for sci - on xis2 is too small , which is available only until the anomaly in xis 2 on 2009 november 9 . between the offset2 and offset4 observations , operations of changing the hxd - pin bias voltage were carried out in response to flare - like events in a pin diode ( w10p0 ) . as summarized in table [ table : hv - ope ] , data of all the 64 pins were available throughout the center and offset1 pointings , and for a part of offset2 and offset4 . in contrast , only one fourth of the pin diodes were operational in offset3 . therefore , we do not analyze the offset3 hxd - pin data in the present paper . for the other observations , we exclude the data of the four pin diodes in w10 unit , and use the remaining 60 channels . backgrounds and responses excluding the four pin diodes were prepared with a help of the hxd team . the present paper utilizes the xis and hxd - pin data , prepared through version 2.0 pipe - line data processing , and obtained from the suzaku ftp site . we re - processed unscreened xis event files using suzaku software version 11 in heasoft 6.6 , together with the calibration datebase ( caldb ) released on 2009 january 9 . we applied ` xiscoord ` , ` xisputpixelquality ` , ` xispi ` , and ` xistime ` in this order , and performed the standard event screening . bad pixels were rejected with ` cleansis ` , employing the option of chipcol = segment . we selected events with grade 0,2,3,4 , and 6 . good time intervals ( gti ) were detemined by criteria as ` " saa_hxd==0 ` ` & & ` ` t_saa_hxd``>436 & & cor>6 & & elv>5 ` ` & & dye_elv>20 & & aocu_hk_cnt3_nml_p==1 & & sn_dtrate<3 & & ang_dist<1.5 " ` . details of the processing and screening are the same as those described in the suzaku data reduction guide . the redistribution matrix files ( rmfs ) of the xis were produced by ` xisrmfgen ` , and auxillary responce files ( arfs ) by ` xissimarfgen ` ( @xcite ) . as an input to the arf generator , we prepared an x - ray surface brightness profile of abell 2199 using a double-@xmath22 model ( sum of two @xmath22 models ; @xcite ) of which the parameters were determined through a least chi - square fit to the background subtracted and vignetting corrected 0.510 kev xmm - newton mos1 image of the abell 2199 cluster . the employed parameters are ( @xmath23 , @xmath22)@xmath24(@xmath25 kpc , @xmath26 ) for the narrower component and ( @xmath27 kpc , @xmath28 ) for the wider component , where @xmath23 is the core radius . absorption below 2 kev , caused by a carbon - dominated contamination material on the xis optical blocking filters , is included in the arfs . there , differences of the contaminant thickness among the xis sensors are taken into account , along with its radial dependences and secular changes ( @xcite ) . non x - ray background ( nxb ) of each xis sensor was created using ` xisnxbgen ` , which sorts spectra of night earth observations according to the geomagnetic cut - off - rigidity ( cor ) and makes an averaged spectrum weighted by residence times for which suzaku stayed in each cor interval ( @xcite ) . we assumed systematic errors ( 90% confidence range ) of 6.0% and 12.5% for the nxb from xis - fi ( front illuminated ccd , @xcite ) and xis - bi ( @xcite ; table 5 ( c ) ) , respectively , and added them in quadrature to the corrensponding statistical errors . the x - ray background was estimated as described in subsection [ subsection : bgd : xis ] . unscreened event files of hxd - pin were re - processed by ` hxdtime ` , ` hxdpi ` , and ` hxdgrade ` in this order . using ` hxdgtigen ` , we rejected the time intervals when `` fifo full '' , `` bfsh '' , or `` tlmrjct '' happened , and further narrowed the gti by imposing criteria of ` " saa_hxd==0 & & ` ` t_saa_hxd>500 & & ` ` tn_saa_hxd>180 & & ` ` cor>6 & & ` ` elv>5 & & ` ` aocu_hk_cnt3_nml_p==1 & & ` ` hxd_dtrate<3 & & ` ` ang_dist<1.5 & & ` ` hxd_hv_wn_cal>700 & & ` ` hxd_hv_tn_cal>700 " ` . for the center , offset1 , offset2 , and offset4 pointings , we made special hxd - pin rmfs which exclude contributions of the four pins in w10 unit . this was constructed by summing rmfs from the remaining 60 pin diodes . since ` hxdarfgen ` did not support extended sources yet , we created pin arfs for each @xmath29 pixel by ` hxdarfgen ` , assuming the same x - ray surface brightness as used for xis arfs ( subsection [ subsection : obs : xis ] ) . the arfs were then averaged by weighting with counts contained in the corresponding pixels . as the hxd - pin nxb , the `` tuned '' pin nxb files published by the hxd team were used ( @xcite ) . in the present paper , we used 2.3% as systematic error ( 90% confidence range ) of the pin nxb . details of its derivation is described in appendix [ appen : nxb - sys ] . the pin cxb was estimated from an observation of a nearby region in the same manner as the xis cxb ( subsection [ subsection : bgd : pin ] ) . to analyze the xis data of an extended object in general , we must subtract the cosmic x - ray background ( cxb ) and galactic foreground emission ( gfe ) . since the emission from abell 2199 itself hampers their direct estimation using our own data , we used 73.6 ks suzaku observation of a nearby blank sky region named `` high latitude diffuse a '' ( hereafter hld ; observation i d = 500027010 ) , which is @xmath30 away from abell 2199 . the difference of their galactic latitudes is only @xmath31 , and hence the hydrogen column densities ( @xcite , weighted average over @xmath32 cone radius ) toward abell 2199 ( @xmath33 cm@xmath14 ) and hld ( @xmath35 cm@xmath14 ) are similar . after applying the standard processing described in subsection [ subsection : obs : xis ] to the unscreened xis data of hld , we removed circular regions ( @xmath36 in radius ) around six point sources that are visible in the 0.510.0 kev xis images , and extracted spectra of the remaining region from the four xis detectors . the corresponding detection limit for the point sources is @xmath37 erg s@xmath9 cm@xmath14 in 210 kev . then , the nxb - subtracted xis - bi and xis - fi ( averaged over the three xis - fi detectors ) spectra in energy ranges of 0.35.5 kev and 0.55.5 kev , respectively , were fitted simultaneously with a cxb+gfe model ( see below ) using xspec12 version 12.4.0ad , incorporating the xis response for a source of uniform brightness . in the analyses described below , metal abundances refer to anders & grevesse ( 1989 ) , and the photoelectric absorption cross - sections to balucinska - church & mccammon ( 1992 ) with a new he cross section ( @xcite ) . as the cxb model , we used a power law with a fixed photon index of 1.41 ( @xcite ) and a free normalization . to represent the gfe , we employed model 1 of @xcite , which consists of three _ apec _ components ( @xcite ) : a non - absorbed 0.08 kev component , an absorbed 0.11 kev one , and an absorbed 0.27 kev one , representing the emission from local bubble ( lb ) , a cool halo , and a hot halo , respectively . the three _ apec _ temperatures were all fixed . the hydrogen column density was fixed to the galatic value of @xmath35 cm@xmath14 , and the metal abundance and redshift were also fixed to one solar and zero , respectively . relative normalizations of the two halo components were tied as cool halo : hot halo @xmath38 ( @xcite ) . thus , the gfe model has only two free parameters ; normalizations of the lb and halo . as summarized in table [ table : fit - bgd ] , this cxb+gfe model reproduced the xis spectra from hld successfully . the 2.010.0 kev surface brightness of the cxb component is @xmath39 erg s@xmath9 cm@xmath14 sr@xmath9 ( 90% statistical and systematic errors ) , where the systematic error , 10% , refers to appendix 1 of @xcite . using asca , @xcite derived an absolute 2.010.0 kev cxb surface brightness to be @xmath40 erg s@xmath9 cm@xmath14 sr@xmath9 ( 90% statistical and systematic errors ) . this value becomes @xmath41 @xmath42 erg s@xmath9 cm@xmath14 sr@xmath9 when the surface brightness is integrated to our detection limit of @xmath43 erg s@xmath9 cm@xmath14 ( equation 6 of @xcite ) . thus , the cxb level obtained from the hld observation is consistent with the asca result . we also analyzed the hxd - pin data of the hld observation which were processed by the same standard procedure as described in subsection [ subsection : obs : pin ] . ignoring the gfe which is negligible at @xmath44 kev , we fitted the nxb - subtracted @xmath45@xmath46 kev hxd - pin spectrum with a wide - band cxb model . namely , following the heao-1 result ( @xcite ) , we expressed the cxb spectral surface brightness as @xmath47 where @xmath48 , a normalization , is a free parameter in the fit . when reproducing the cxb signals in hxd - pin , equation [ eqn : pin - cxb ] was corrected for the 13% systematic difference of cross - normalization between the xis and pin . the result we obtained , @xmath49 , is consistent with the heao result of @xmath50 , which can in turn reproduce , within 1% , blank - sky pin spectra which have higher statistics ( @xcite ) . the contribution of the six point sources to the hxd - pin spectrum is an order of magnitude lower than the nxb - subtracted signal in 1240 kev . therefore , the contribution of point sources affects the cxb level only by @xmath51 , which is much smaller than the statistical error . although we have thus adopted the two separate models to estimate the cxb in the xis and hxd - pin , the derived two surface brightness results agree within 1% in the connecting energy range of 6.012.0 kev . therefore , our two modelings of the cxb are consistent . if we constrain the two cxb models so that the 6.012.0 kev surface brightness becomes the same , and fit the xis and hxd - pin spectra simultaneously , we can constrain the cxb above 10 kev better with @xmath52 . this case is shown in figure [ fig : hld - spec ] . in the following spectral analysis , we fit the abell 2199 data simultaneously with the hld data , in order to determine the x - ray background . however , the true x - ray background of the abell 2199 field may be different from that of the hld field . in order to examine the gfe brightness for possible differences between them , we employed x - ray count rates of these two fields from the rosat all - sky survey ( rass ) which are available via the nasa s heasarc website . since the rass data right on the abell 2199 field is contaminated , of course , with the cluster emission , we obtained a count rate averaged over an annular region of which the inner radius is @xmath53 ( @xmath54 mpc ) and the outer radius is @xmath55 , centered on the cd galaxy ngc 6166 . since the virial radius of abell 2199 is estimated to be @xmath56 mpc ( @xmath57 ) from the icm temperature of 4 kev ( @xcite ) , this annular region can be considered to be free from the icm emission . table [ table : rass - cnt ] summarizes the rass count rates of the abell 2199 annular region and the hld field ( circular region , @xmath53 in radius ) . the 3/4 kev ( 0.471.21 kev ) and 1.5 kev ( 0.762.04 kev ) rates are the same between the two fields within statistical errors . thus , the hld region is confirmed to be a good background estimator for our abell 2199 observations in these energy bands . the 1/4 kev ( 0.120.284 kev ) band rates differ by @xmath1% , suggesting a spectral difference betweeen the abell 2199 and hld fields in the softest energy band . however , the lb component , which dominates the spectrum in the 1/4 kev band , is poorly determined with the xis spectra , and the 20% discrepancy is within the statistical error of @xmath58% associated with the lb normalization in the following spectral analysis ( [ subsection : results : xis - ana ] ) . the cxb surface brightness is known to fluctuate from direction to direction ( @xcite ) . in the case of suzaku , this fluctuation is estimated to be 10% and 18% for the xis ( full fov ) and pin , respectively ( see appendix of @xcite for details ) . we take these systematic errors into account in the spectral analyses described in section [ section : results ] . in order to quantify the thermal emission of abell 2199 , we defined seven concentric annular regions as shown in figure [ fig : mosaic ] , each with a radial width of @xmath59 . the center is chosen to be the x - ray emission centroid at ( @xmath60)=(@xmath61 ) in j2000.0 coordinates . this is @xmath62 offset from the nucleus of the cd galaxy , ngc 6166 , but well inside its optical extent of @xmath63 in diameter . in analyzing a given annulus , we extracted xis - fi ( averaged over xis 0 and xis 3 ) and xis - bi spectra from corresponding regions in the five pointings ( but excluding those which have no intersection ) , and subtracted the nxb as described in subsection [ subsection : obs : xis ] . for each annulus , the utilized observations are given in table [ table : fit - reg ] . then , the spectra from the same annulus ( but different fields of view ) were read into xspec together with the nxb - subtracted hld spectra to constrain the x - ray background . the annular spectra were fitted simultaneously with a common _ vapec _ model with a free photoelectric absorption , together with the x - ray background model described in subsection [ subsection : bgd : xis ] , incorporating the responses created for the abell 2199 icm emission ( subsection [ subsection : obs : xis ] ) and for a uniform x - ray source , respectively . all the parameters of the x - ray background model were tied between the annular and hld spectra , except for the thickness of xis contamination which depends on the date of observation ( @xcite ) . among the different observations of abell 2199 , all the _ vapec _ parameters , except the overall normalization , were constrained to be the same . the cross normalization between xis - fi and xis - bi was set free . metal abundances , @xmath64 , in the _ vapec _ model were tied in three groups as @xmath65 , @xmath66 , and @xmath67 . an example of this fitting in the @xmath68@xmath59 region is shown in figure [ fig : xis_hld - spec ] . we also changed the normalization of cxb model by @xmath69% and studied the effect of cxb fluctuation described in subsection [ subsection : bgd : sys ] . although the x - ray background model is determined separately for the seven individual regions , the normalizations of cxb , lb , and halo are consistent among the regions within errors ( see figure [ fig : bgd - norm ] in appendix [ appen : bgd - norm ] ) . the results of this analysis are summarized in figure [ fig : vapec - prof ] ( for details , see table [ table : fit - apec ] in appendix [ appen : fit - result ] ) . the single - temperature _ vapec _ model thus reproduced the spectra successfully , and the cxb fluctuation did not affect the results significantly . the temperature and iron abundance were determined out to @xmath70 kpc , which is about 0.4 times the virial radius of a @xmath71 kev cluster . the temperature gradually decreases toward cluster outskirts from @xmath6 kev to @xmath72 kev . the temperature within @xmath4 , @xmath6 kev , is consistent with that derived with asca ( 4.1 kev ; @xcite ) . the iron and silicon abundances also decrease outwards by a factor of @xmath3 , while the oxygen abundance tends to be flat . the hydrogen column density obtained from the fit became ( @xmath73@xmath74 cm@xmath14 , which is significantly higher than the galacitic value . we examine the cause of this effect in subsection [ subsection : results : soft - excess ] . the 0.5 10 kev x - ray luminosity integrated out to @xmath7 is @xmath8 erg s@xmath9 , which is consistent with the result of xmm - newton ( @xcite ) . next , we attempted a two - temperature fit using a model expressed as _ vapec_+_vapec_. metal abundances of the two components were tied together . as shown in figure [ fig:2vapec - prof]a and figure [ fig:2vapec - prof]f ( for details , see table [ table : fit-2apec ] in appendix [ appen : fit - result ] ) , the spectra in the innermost region ( @xmath68@xmath59 ) were reproduced better with the two - temperature model of 2 kev and 5 kev , while the improvement was not significant in the other regions . this is reasonable , because @xcite detected , with chandra , a temperature drop from 5 kev to 2 kev toward the center within a central region of radius 100 kpc ( @xmath75 ) . the iron and silicon abundance profiles did not change significantly between the single - temperature and two - temperature models . however , the oxygen abundance in the inntermost region decreased and the profile became flatter . the high values of @xmath76 , found with the single - temperature analysis , persisted ; this effect can not be regarded as an artifact caused by an inappropriate temperature modeling . a soft excess may manifest itself as a hydrogen column density @xmath76 that is significantly lower than the galactic hi columnn density toward the abell 2199 field ( weighted - averaged value over @xmath32 radius from ngc 6166 ) , @xmath77 cm@xmath14 from @xcite or @xmath78 cm@xmath14 from @xcite . the measured values of @xmath76 in figure [ fig : vapec - prof ] and figure [ fig:2vapec - prof ] , however , are factor of @xmath73@xmath79 _ higher _ than the galactic value in all regions . while this result generally argues against the presence of any flux excess in lower energies , we need to understand the nature of the higher absorption before discussing the soft excess issue . there are five possibilities to explain this effect : ( 1 ) excess absorption within the abell 2199 system , ( 2 ) an overestimate of the background which was subtracted , ( 3 ) inadequate modeling of the icm emission , ( 4 ) presence of high metallicity clouds at the direction of abell 2199 , and ( 5 ) underestimate of the xis contaminant thickness which is included in the arf . the first possibility is unlikely , because the measured @xmath76 is high even in the outer regions such as @xmath4@xmath80 . the second alternative is rejected , because we already incorporated the uncertainties of nxb and cxb in the analysis , and because within @xmath81 , the background level is more than an order of magnitude lower than the signal below 1 kev band . the third is also unrealistic , because the values of @xmath76 remained unchanged between the single- and two - temperature modelings . the forth is probably unphysical because metal abundances in such clouds would have to be as large as @xmath72 solar in order to explain the excess x - ray absorption , while keeping @xmath76 to the galactic hi value . then , we focus on the last possibility . when we obtain @xmath76 separately with xis - fi and xis - bi in the three inntermost regions where the values of @xmath76 are constrained well ( figure [ fig : vapec - prof](b ) ) , the results with xis - fi become systematically higher than those with xis - bi by @xmath1% . this difference suggests an instrumental cause such as ( 5 ) . then , we evaluated an excess thickness of contaminant , assuming ( 5 ) is the cause of the high @xmath76 . we included an additional absorption by the xis contamination in the single temperature model as _ vapec@xmath82phabs@xmath82varabs_. here , the _ varabs _ factor represents the additional absorption which is added to that already included in the arf , and can handle separately column densities of elements from hydrogen to nickel . the _ varabs _ column densities of all the elements except carbon and oxygen were fixed to zero , and the relative column - number - density ratio of carbon to oxygen was fixed to @xmath83 ( @xcite ) . by introducing the additional _ varabs _ factor , the best - fit model parameters did not change significantly except @xmath76 in _ phabs_. figure [ fig : conf - contour ] shows the confidence contours between @xmath76 in _ phabs _ and the additional carbon column density @xmath84 in _ varabs_. as easily expected , these two quantities correlate negatively with each other , and we can not determine the true @xmath76 accurately . fixing @xmath76 at the galactic hi value in figure [ fig : conf - contour ] ( dotted magenta lines ) yields @xmath85 cm@xmath14 ( dashed cyan lines in figure [ fig : conf - contour ] ) , which is comparable to an upper limit allowed by its systematic uncertainty . therefore , we may regard the excess absorption observed in figure [ fig : vapec - prof]b and figure [ fig:2vapec - prof]b as due to an underestimation of the xis contaminant by @xmath85 cm@xmath14 , and presume that @xmath76 is equal to the galactic hi value . from figure [ fig : conf - contour ] of the innermost annuli , we can reject any @xmath76 value which is significantly lower than the galactic hi value ( dotted magenta lines in figure [ fig : conf - contour ] ) , because the extra carbon thickness must be @xmath86 cm@xmath14 . in figure [ fig : resi - spec ] , we show data - to - model ratios in the 0.31.0 kev range , when the data are fitted by the single temperature _ vapec _ model with @xmath87 cm@xmath14 and @xmath88 cm@xmath14 both fixed . the value of @xmath84 is thus set to the allowed maximum , and the combination of @xmath84 and @xmath76 means a slight ( @xmath89 ) overestimate of the overall absorption specified by the data in figure [ fig : conf - contour ] of the innermost annulus . furthermore , the data are divided by the single - temperature model which was determined over a wide band of @xmath90@xmath91 kev . therefore , soft excess emission , if any , should be seen in these ratio spectra . however , no significant systematic excess is seen therein . then , to obtain an upper limit on the soft excess due to a thermal warm gas , we added to the model a 0.2 kev _ apec _ component with 0.2 solar metallicity , following the result of @xcite . the derived 90% upper limit on surface brightness of the 0.2 kev component is summarized in table [ table : warm - upper ] . in table [ table : warm - upper ] , we also give xis upper limits when a power law with photon index of 2.0 is added to the model . the upper limit on the power law in terms of the 0.210 kev luminosity becomes @xmath92 erg s@xmath9 ( @xmath93 erg s@xmath9 in 2080 kev range ) when integrated to @xmath7 . in subsection [ subsection : results : hard - excess ] , we independently constrain such a power law component using the hxd - pin data . as a cross confirmation , we repeated the same analysis by choosing @xmath94 and @xmath95 cm@xmath14 , the latter suggested by figure [ fig : vapec - prof ] . however , the ratio spectra remained essentially unchanged . in figure [ fig : pin - spec ] , we show a 1540 kev nxb - subtracted hxd - pin spectrum summed over center , offset1 , offset2 , and offset4 . by this summation , the statistical error associated with the 1540 kev hxd - pin signal reduced to @xmath96% of the nxb , which is considerably smaller than the systematic nxb error of @xmath97% ( 90% confidence range ) adopted in the present paper . after subtracting the nxb , the hxd - pin signal ( black crosses in figure [ fig : pin - spec ] ) was thus detected significantly up to @xmath98 kev . for comparison , we also show in figure [ fig : pin - spec ] spectra of the cxb and the thermal emission . here , the thermal emission was modeled by summing the single temprature _ vapec _ models determined for individual annular regions . as to thermal emission outside the xis fov which contributes @xmath99% , we assumed that the surface brightness obeys the double-@xmath22 profile which was described in [ subsection : obs : xis ] , and that the temperature and metal abundances therein is the same as those in the outermost region covered by the xis . after the nominal xis vs. hxd - pin cross normalization ( @xcite , see also a suzaku calibration document named suzakumemo-2007 - 11 ) , the model normalization of the thermal component was multiplied by a factor of 1.13 when calculating its predicted contribution to the hxd - pin data . using this cross normalization , intensities of the two cxb models in 6.012.0 kev agree within 1% as described in [ subsection : bgd : pin ] . this assures that the systematic errors associated with the cross normalization is @xmath100% , which is much smaller than the statistic erros in the hxd - pin spectrum . thus , the nxb - subtracted spectrum is consistent , within statistical errors , with a sum of the cxb and the extrapolated thermal emission ( cyan points in figure [ fig : pin - spec ] ) . even when the nxb is set to the minimum that is allowed by the 90% systematic uncertainty range , the hxd - pin signal does not exceed the ( cxb+thermal ) emission significantly . thus , the hard x - ray excess was not detected . in order to obtain an upper limit on hard excess emission , we lowered the nxb within its 90% systematic uncertainty , and also lowered the cxb model within its fluctuation of 18% ( 90%-confidence ) . the estimation of the cxb fluctuation is described in appendix [ appen : cxb - sys ] . then , we fitted the nxb - subtracted spectrum with a power law ( with its photon index fixed at 2.0 and its normalization left free ) plus the fixed cxb and thermal models , and obtained upper limits . this very conservative method gave a 90%-confidence upper limit of @xmath15 erg s@xmath9 on the 2080 kev luminosity of the power - law component . if this component is uniformly distributed over a circular region of radius @xmath101 , the implied 0.210 kev luminosity per solid angle becomes @xmath102 erg s@xmath9 arcmin@xmath14 . this limit is comparable to those obtained with the xis using the softest specctral end . using the five pointing observations with suzaku , we measured the temperature and metal abundances of abell 2199 out to @xmath70 kpc ( 0.4 times virial radius ) , and searched the xis and hxd - pin data for soft and hard excess emissions , respectively , both above the thermal component . we however found no significant excess emissions in either energy bands , and derived their upper limits . over a broad energy band of @xmath103 kev to @xmath98 kev , the emission is dominated by the thermal components with temperatures of a few kev . the upper limits on soft excess , expressed as the emission measure per unit solid angle from a 0.2 kev warm gas ( table [ table : warm - upper ] ) , are more stringent , typically by more than a factor of 3 , than the reported positive detection by xmm - newton ( @xcite ) , @xmath104 @xmath13 arcmin@xmath14 . one likely cause of this inconsistency between suzaku and xmm - newton is confusion with solar wind charge exchange ( swcx ) emission , which has been observed with chandra ( @xcite ) , xmm - newton ( @xcite ) , and suzaku ( @xcite ) . in fact , proton flux near the earth , observed with wind - swe , enhanced by a factor of 34 for a part of the xmm - newton observatons compared with that of suzaku . the relevant proton light curves are shown in appendix [ appen : proton - flux ] . similarly , in suzaku observations of outskirt regions of the coma cluster , takei et al . ( 2008 ) found no significant excess o vii or o viii emissions which were previously reported with xmm - newton ( @xcite ) , and concluded that the excess emission is likely due to an enhancement in the swcx emission during the xmm - newton observations . the suzaku data have constrained the 2080 kev luminosity of any excess hard emission to be @xmath105 erg s@xmath9 , assuming that it has a power - law spectrum of photon index 2.0 and employing the most conservative estimates of the cxb . using the bepposax pds , @xcite detected non - thermal emission from abell 2199 at 2.1@xmath16 level , with 2080 kev luminosity of @xmath106 erg s@xmath9 ( 90%-confidence errors ) assuming a power - law spectrum of photon index 2.0 . this detection by the pds is not rejected by our result . however , the bepposax results , with a significance of only 2.1@xmath16 , would be considered as an upper limit , like the present result . in this case , our upper limit is a factor of 1.7 more stringent than that of the pds . from a calibration analysis using the crab nebula , the 2080 kev flux derived from the pds is known to be systematically lower by 21% than that derived from hxd - pin ( @xcite ) . if we take into account this systematic effect , the difference in the upper limits between hxd - pin and pds increases to a factor of 2.1 . considering the difference between the fovs of hxd - pin ( @xmath107 in fwhm ) and pds ( @xmath108 in fwhm ) , the claimed detection by the pds becomes consistent with our upper limit if the excess hard x - ray emission is much more extended than the hxd - pin fov . however , the total fov of hxd - pin in our pointing observations ( @xmath109 mpc radius in fwhm ) mostly cover the whole cluster region , @xmath110 times the virial radius of abell 2199 ( @xmath56 mpc , see [ subsection : bgd : sys ] ) . therefore , to reconcile the two results , we would have to assume that the emission is spatially distributed much beyond the virial radius . according to kempner & sarazin ( 2000 ) , a radio synchrotron flux density at 327 mhz , @xmath111 , expected from abell 2199 , is related to the 10100 kev ic x - ray flux , @xmath112 , as @xmath113 assuming the photon index of the hard x - ray emission to be 1.81 . here , @xmath114 is the cluster magnetic field , and @xmath115 is the observed frequency . the upper limit they quoted on the diffuse radio flux , @xmath116 jy at 327 mhz , and the hard x - ray emission detected by bepposax , required a very weak magnetic field ( @xmath19 g ) . however , given our non - detection of such an excess hard x - ray flux , this is no longer the case . when the hxd - pin data are used to constrain a power - law component with a photon index of 1.81 ( instead of 2.0 ) , its 10100 kev luminosity becomes @xmath117 erg cm@xmath14 s@xmath9 . as a result , effectively any strength of magnetic field is allowed as shown in figure [ fig : mag - limit ] . regardless of the magnetic field strength , the present result constrain the number of relativistic electrons in the system , so that their ic emission should not exceed the present hard x - ray upper limit . if we assume a power - law spectrum of synchrotron emission with a photon index 2.0 , the corresponding electron number spectrum has a power - law index 3.0 , and is written as @xcite @xmath118 where @xmath119 is the lorentz factor , @xmath120 is the total number of electrons , and @xmath121 is a lower - cutoff in @xmath119 which is conservatively estimated to be @xmath122 ( @xcite ) . these relativistic electrons produce ic emission with a 2080 kev luminosity of @xmath123 where @xmath124 is the redshift . from equation [ eqn : hxr - lumi ] and our upper limit , we obtain @xmath125 . then , the integrated electron kinetic energy @xmath126 for @xmath127 in abell 2199 becomes @xmath128 where @xmath129 is the electron mass . since electrons with @xmath130 , which contribute to the @xmath131@xmath132 kev ic emission , have a liftime of @xmath133 yr in typical icm conditions ( @xcite ) , the acceleration luminosity in abell 2199 can be constrained as @xmath134 erg s@xmath9 . this upper limit is rather loose , and is three times higher than the 0.510 kev luminosity of the thermal emission . if we assume that the magnetic field of abell 2199 is @xmath135 g as suggested by faraday rotation measurements ( @xcite ) , the radio upper limit gives the dominant constraint on the relativistic electrons , with the predicted @xmath112 two orders of magnitude below the present upper limit ( figure [ fig : mag - limit ] ) . in this case , the constraint of acceleration luminosity becomes @xmath136 erg s@xmath9 which is about an order of magnitude lower than the 0.510 kev luminosity . the total mass of abell 2199 is @xmath137 when integrated to 0.8 mpc ( @xcite ) . if abell 2199 was formed via a major merger between two clusters with similar masses @xmath138 ( @xmath139 each ) at a relative speed of @xmath140 km s@xmath9 , the kinetic energy deposited onto the merged system becomes @xmath141 since shocks produced in cluster mergers are thought to last typically @xmath142 yr ( @xcite ) , the energy input rate to non - thermal electrons becomes @xmath143 erg s@xmath9 , where @xmath144 is the fraction of energies given to them . values of @xmath145 erg s@xmath9 are also expected from theoretical models of merging clusters such as coma ( e.g. @xcite ) . then , our upper limit on the acceleration luminosity is an order of magnitude lower than @xmath146 , assuming @xmath147 . we hence infer that abell 2199 have been free from major merger events for more than @xmath148 yr , which is a typical time scale on which the ic emission decays ( @xcite ) . mk acknowledges support from a grant based on the special postdoctoral researchers program of riken . the present work is supported in part by the grant - in - aid for scientific research ( s ) , no . @xmath149 systematic erros of pin 1540 kev band , using earth occultation data and e0102 - 72 observations which has been regularly scheduled for the xis calibration , is 2.3% ( @xcite ) . since the statistical error is 1.8% ( @xmath149 ) , the systematic uncertainty of the pin nxb model in 1540 kev band is estimated to @xmath97% in 90% confidence region . the cxb has uncertainty because the number of unresolved faint sources in fov statistically fluctuate . this cxb fluctuation scales as @xmath150 , where @xmath151 is the fov and @xmath152 is the detection threshold flux for point sources ( @xcite ) . by scaling from the cxb fluctuation of heao - i , 2.8% for @xmath153 deg@xmath154 and @xmath155 erg s@xmath9 cm@xmath14 ( @xcite ) , that of the hxd is estimated to be 18% , assuming @xmath156 deg@xmath154 and @xmath157 erg s@xmath9 cm@xmath14 in the 1040 kev band . this corresponds to 1.3% of the nxb level . the background model ( gfe+cxb ) are determined separately for the seven annular regions . we show the normalizations of background components in figure [ fig : bgd - norm ] . the normalizations of background models are consistent among the regions within errors ( 90% confidence range ) . figure [ fig : wind - lc ] is proton flux of wind - swe , when xmm - newton and suzaku observed abell 2199 . these plots are created by multiplying proton speed and proton density of which data are public in the wind - swe web site . anders , e. , & grevesse , n. 1989 , gecoa , 53 , 197 balucinska - church , m. , & mccammon , d. 1992 , apj , 400 , 699 boldt , e. 1987 , phr , 146 , 215 bregman , j. n. , & lloyd - davies , e. j. 2006 , apj , 644 , 167 buote , d. a. , & canizares , c. r. 1996 , apj , 457 , 565 cen , r. , & ostriker , j. p. 1999 , apj , 514 , 1 cox , d. p. 1998 , in lecture notes in physics 506 , the local bubble and beyond , ed . d. breitschwerdt , m. j. freyberg , & j. truemper ( new york : springer ) , 121 cravens , t. e. 2000 , apj , 532 , l153 dickey , j. m. , & lockman , f. j. 1990 , ara&a , 28 , 215 evrard , a. e. , metzler , c. a. , & navarro , j. f. 1996 , apj , 469 , 494 finoguenov , a. , briel , u. g. , & henry , j. p. 2003 , a&a , 410 , 777 fujimoto , r. , mitsuda , k. , mccammon , d. , takei , y. , bauer , m. , ishisaki , y. , porter , s. f. , yamaguchi , h. , et al . 2007 , pasj , 59 , 133 fukazawa , y. , makishima , k. , & ohashi , t. 2004 , pasj , 56 , 965 fukazawa , y. , mizuno , t. , watanabe , s. , kokubun , m. , takahashi , h. , kawano , n. , nishino , s. , sasada , m. , et al . 2009 , pasj , in press fukugita , m. , hogan , c. j. , & peebles , p. j. e. 1998 , apj , 503 , 518 fusco - femiano , r. , dal fiume , d. , feretti , l. , giovannini , g. , grandi , p. , matt , g. , molendi , s. , & santangelo , a. 1999 , apj , 513 , l21 fusco - femiano , r. , landi , r. , & orlandini , m. 2007 , apj , 654 , l9 ge , j. , & owen , f. n. 1994 , aj , 108 , 1523 henley , d. b. , & shelton , r. l. 2008 , apj , 676 , 335 ishisaki , y. 1997 , ph.d . thesis , university of tokyo ishisaki , y. , maeda , y. , fujimoto , r. , ozaki , m. , ebisawa , k. , takahashi , t. , ueda , y. , ogasaka , y. , et . al 2007 , pasj , 59 , 113 johnstone , r. m. , allen , s. w. , fabian , a. c. , & sanders , j. s. 2002 , mnras , 336 , 299 kaastra , j. s. , lieu , r. , mittaz , j. p. d. , bleeker , j. a. m. , mewe , r. , colafrancesco , s. , lockman , f. j. 1999 , apj , 519 , l119 kaastra , j. s. , lieu , r. , tamura , t. , paerels , f. b. s. , & den herder , j. w. a. 2004 , in soft x - ray emission from clusters of galaxies and related phenomena , ed . r. lieu & j. mittaz . ( dordrecht : kluwer ) , 37 kalberla , p. m. w. , burton , w. b. , hartmann , d. , arnal , e. m. , bajaja , e. , morras , r. , & pppel , w. g. l. 2005 , a&a , 440 , 775 kawano , n. , ohto , a. , & fukazawa , y. 2003 , pasj , 55 , 585 kempner , j. c. , & sarazin , c. l. 2000 , apj , 530 , 282 king , i. r. 1962 , aj , 67 , 471 kokubun , m. , makishima , k. , takahashi , t. , murakami , t. , tashiro , m. , fukazawa , y. , kamae , t. , madejski , g. m. , et al . 2007 , pasj , 59 , 53 koutroumpa , d. , acero , f. , lallement , r. , ballet , j. , & kharchenko , v. 2007 , a&a , 475 , 901 koyama , k. , tsunemi , h. , ; dotani , t. , bautz , m. w. , hayashida , k. , tsuru , t. g. , matsumoto , h. , ogawara , y. , et al . 2007 , pasj , 59 , 23 kushino , a. , ishisaki , y. , morita , u. , yamasaki , n. y. , ishida , m. , ohashi , t. , & ueda , y. 2002 , pasj , 54 , 327 lieu , r. , ip , w .- h . , axford , w. i. , & bonamente , m. 1999 , 510 , l25 mitsuda , k. , bautz , m. , inoue , h. , kelley , r. l. , koyama , k. , kunieda , h. , makishima , k. , ogawara , y. , et al . 2007 , pasj , 59 , 1 nakajima , h. , yamaguchi , h. , matsumoto , h. , tsuru , t. g. , koyama , k. , tsunemi , h. , hayashida , k. , torii , k. , et al . 2008 , pasj , 60 , s1 nakazawa , k. , sarazin , c. l. , kawaharada , m. , kitaguchi , t. , okuyama , s. , makishima , k. , kawano , n. , fukazawa , y. , et al . 2009 , pasj , 61 , 339 nevalainen , j. , oosterbroek , t. , bonamente , m. , & colafrancesco , s. 2004 , apj , 608 , 166 . nevalainen , j. , bonamente , m. , & kaastra , j. s. 2007 , apj , 656 , 733 ota , n. , murase , k. , kitayama , t. , komatsu , e. , hattori , m. , matsuo , h. , oshima , t. , suto , y. , et al . 2008 , a&a , 491 , 363 petrosian , v. 2001 , apj , 557 , 560 petrosian , v. , madejski , g. , & luli , k. 2006 , apj , 652 , 948 robertson , i. p. , & cravens , t. e. 2003a , georl , 30 , 22 robertson , i. p. , & cravens , t. e. 2003b , jgra , 108 , lis 6 rossetti , m. , & molendi , s. 2004 , a&a , 414 , l41 sanders , j. s .. , & fabian , a. c. 2006 , mnras , 371 , l65 shafer , r. a. 1983 , phd . thesis , university of maryland smith , r. k. , brickhouse , n. s. , liedahl , d. a. , & raymond , j. c. 2001 , apj , 556 , l91 snowden , s. l. , collier , m. r. , & kuntz , k. d. 2004 , apj , 610 , 1182 snowden , s. l. , mushotzky , r. f. , kuntz , k. d. , & davis , d. s. 2008 , a&a , 478 , 615 takahashi , t. , abe , k. , endo , m. , endo , y. , ezoe , y. , fukazawa , y. , hamaya , m. , hirakuri , s. , et al . 2007 , pasj , 59 , 35 takei , y. , miller , e. d. , bregman , j. n. , kimura , s. , ohashi , t. , mitsuda , k. , tamura , t. , yamasaki , n. y. , et al . 2008 , apj , 680 , 1049 takizawa , m. 2000 , apj , 532 , 183 takizawa . m 2004 , pasj , 54 , 363 tawa , n. , hayashida , k. , nagai , m. , nakamoto , h. , tsunemi , h. , yamaguchi , h. , ishisaki , y. , miller , e. d. , et . al 2008 , pasj , 60 , 11 wargelin , b. j. , markevitch , m. , juda , m. , kharchenko , v. , edgar , r. , & dalgarno , a. 2004 , apj , 607 , 596 yan , m. , sadeghpour , h. r. , & dalgarno , a. 1998 , apj , 496 , 1044 cm@xmath14 ) and carbon ( @xmath158 cm@xmath14 ) column densities . from top to bottom , the ratio spectra of the @xmath68@xmath59 , @xmath59@xmath81 , @xmath81@xmath159 , @xmath159@xmath4 , @xmath4@xmath80 , @xmath80@xmath160 , and @xmath160@xmath161 regions are shown . colors specify different observations . see the electronic version of the paper for a colour figure . , scaledwidth=30.0% ] @xmath162 : surface brightness of 0.2 kev gas , expressed as emission measure per unit solid angle in units of @xmath163 @xmath13 arcmin@xmath14 + @xmath164 : 0.210 kev luminiosity per unit solid angle of a power law with photon index of 2.0 , in units of @xmath165 erg s@xmath9 arcmin@xmath14 . ccccccc [ table : fit - apec ] + region & @xmath166 & @xmath167 & @xmath168 & @xmath169 & @xmath170 & @xmath171/d.o.f . + + region & @xmath166 & @xmath167 & @xmath172 & @xmath169 & @xmath170 & @xmath171/d.o.f . + + + + + @xmath173@xmath174 & @xmath175 & @xmath176 & @xmath177 & @xmath178 & @xmath179 & @xmath180 + @xmath181@xmath182 & @xmath183 & @xmath184 & @xmath185 & @xmath186 & @xmath187 & @xmath188 + @xmath189@xmath190 & @xmath191 & @xmath192 & @xmath193 & @xmath194 & @xmath195 & @xmath196 + @xmath197@xmath198 & @xmath199 & @xmath200 & @xmath201 & @xmath202 & @xmath203 & @xmath204 + @xmath4@xmath205 & @xmath206 & @xmath207 & @xmath208 & @xmath209 & @xmath210 & @xmath211 + @xmath80@xmath212 & @xmath213 & @xmath214 & @xmath215 & @xmath216 & @xmath217 & @xmath218 + @xmath160@xmath219 & @xmath220 & @xmath221 & @xmath222 & @xmath223 & @xmath224 & @xmath225 + + @xmath173@xmath174 & @xmath175 & @xmath176 & @xmath177 & @xmath178 & @xmath179 & @xmath226 + @xmath181@xmath182 & @xmath183 & @xmath184 & @xmath185 & @xmath186 & @xmath187 & @xmath227 + @xmath189@xmath190 & @xmath228 & @xmath229 & @xmath193 & @xmath230 & @xmath231 & @xmath232 + @xmath197@xmath198 & @xmath233 & @xmath234 & @xmath235 & @xmath236 & @xmath237 & @xmath238 + @xmath4@xmath205 & @xmath239 & @xmath240 & @xmath241 & @xmath242 & @xmath243 & @xmath244 + @xmath80@xmath212 & @xmath245 & @xmath246 & @xmath247 & @xmath216 & @xmath248 & @xmath218 + @xmath160@xmath219 & @xmath249 & @xmath250 & @xmath251 & @xmath223 & @xmath252 & @xmath253 + + @xmath173@xmath174 & @xmath175 & @xmath176 & @xmath177 & @xmath178 & @xmath179 & @xmath254 + @xmath181@xmath182 & @xmath183 & @xmath184 & @xmath185 & @xmath186 & @xmath187 & @xmath255 + @xmath189@xmath190 & @xmath256 & @xmath257 & @xmath193 & @xmath230 & @xmath258 & @xmath259 + @xmath197@xmath198 & @xmath260 & @xmath261 & @xmath235 & @xmath262 & @xmath263 & @xmath264 + @xmath4@xmath205 & @xmath265 & @xmath266 & @xmath186 & @xmath267 & @xmath268 & @xmath269 + @xmath80@xmath212 & @xmath270 & @xmath271 & @xmath272 & @xmath273 & @xmath274 & @xmath275 + @xmath160@xmath219 & @xmath276 & @xmath277 & @xmath278 & @xmath279 & @xmath280 & @xmath281 + ccccccc [ table : fit-2apec ] + region & @xmath166 & @xmath167 & @xmath168 & @xmath169 & @xmath170 & @xmath171/d.o.f . + + region & @xmath166 & @xmath167 & @xmath172 & @xmath169 & @xmath170 & @xmath171/d.o.f . + + + + + @xmath173@xmath174 & @xmath282 & @xmath283 & @xmath284 & @xmath285 & @xmath286 & @xmath287 + @xmath181@xmath182 & @xmath288 ( fix ) & @xmath289 & @xmath290 & @xmath186 & @xmath291 & @xmath292 + @xmath189@xmath190 & @xmath293 & @xmath294 & @xmath295 & @xmath296 & @xmath297 & @xmath298 + @xmath197@xmath198 & @xmath299 ( fix ) & @xmath300 & @xmath301 & @xmath302 & @xmath303 & @xmath304 + @xmath4@xmath205 & @xmath305 ( fix ) & @xmath306 & @xmath307 & @xmath308 & @xmath309 & @xmath310 + @xmath80@xmath212 & @xmath311 & @xmath312 & @xmath313 & @xmath314 & @xmath315 & @xmath316 + @xmath160@xmath219 & @xmath317 & @xmath318 & @xmath319 & @xmath320 & @xmath321 & @xmath322 + + @xmath173@xmath174 & @xmath282 & @xmath323 & @xmath284 & @xmath285 & @xmath286 & @xmath324 + @xmath181@xmath182 & @xmath325 ( fix ) & @xmath326 & @xmath290 & @xmath186 & @xmath291 & @xmath327 + @xmath189@xmath190 & @xmath328 & @xmath329 & @xmath295 & @xmath330 & @xmath297 & @xmath331 + @xmath197@xmath198 & @xmath332 ( fix ) & @xmath333 & @xmath301 & @xmath334 & @xmath303 & @xmath335 + @xmath4@xmath205 & @xmath336 ( fix ) & @xmath337 & @xmath307 & @xmath338 & @xmath339 & @xmath340 + @xmath80@xmath212 & @xmath341 & @xmath342 & @xmath313 & @xmath343 & @xmath344 & @xmath345 + @xmath160@xmath219 & @xmath346 & @xmath347 & @xmath348 & @xmath349 & @xmath350 & @xmath351 + + @xmath173@xmath174 & @xmath282 & @xmath352 & @xmath284 & @xmath285 & @xmath286 & @xmath353 + @xmath181@xmath182 & @xmath354 ( fix ) & @xmath184 & @xmath290 & @xmath355 & @xmath291 & @xmath356 + @xmath189@xmath190 & @xmath357 & @xmath358 & @xmath295 & @xmath296 & @xmath297 & @xmath359 + @xmath197@xmath198 & @xmath360 ( fix ) & @xmath361 & @xmath362 & @xmath363 & @xmath303 & @xmath364 + @xmath4@xmath205 & @xmath365 ( fix ) & @xmath366 & @xmath307 & @xmath367 & @xmath368 & @xmath369 + @xmath80@xmath212 & @xmath370 & @xmath371 & @xmath313 & @xmath349 & @xmath372 & @xmath373 + @xmath160@xmath219 & @xmath374 & @xmath375 & @xmath376 & @xmath377 & @xmath378 & @xmath379 +

in this paper , we propose a factorization of a fourth - order harmonic tensor into second - order tensors . we obtain moreover explicit equivariant reconstruction formulas , using second - order covariants , for transverse isotropic and orthotropic harmonic fourth - order tensors , and for trigonal and tetragonal harmonic fourth - order tensors up to a cubic fourth order covariant remainder .

1612.08281

interest in coordinate - free representations formulas for linear anisotropic elastic materials has been an active research area , starting by the formulation of elastic energy functionals in the framework of finite strains @xcite and having a key role in the classification of linear elastic or piezoelectric materials @xcite . it has also been of main importance in the continuum mechanics representation of cracked / damaged media @xcite . one underlying difficulty is that several elasticity tensors may represent the same linear anisotropic elastic material in different orientations . more precisely , any change of orientation of a material specified by some rotation @xmath0 defines a new elasticity tensor @xmath1 deduced from the previous one by some _ group action _ @xmath2 where einstein convention on repeated indices is used . as the material is rotated the tensor @xmath3 moves on its _ orbit _ in the space @xmath4 of elasticity tensors . thus , a linear elastic material is represented by the _ orbit _ of an elasticity tensor under this group action , and any intrinsic parameter is necessary an _ invariant _ of this group action . in a seminal paper , boehler kirilov onat @xcite emphasized the fundamental role played by polynomial _ invariants _ of the elasticity tensor and which have been used by auffray kolev petitot @xcite to classify the orbits of the elasticity tensor . the old problem of finding a basis of polynomial invariants for the fourth order elasticity tensor was finally solved by olive in 2014 @xcite after several attempts @xcite . a _ minimal integrity basis of @xmath5 invariants _ for this space was definitively obtained in @xcite . note that in 2d , this integrity basis is composed by only @xmath6 invariants @xcite , which shows the incredible complexity of 3d linear elasticity compared to 2d case . in 3d , the problem can not be reduced as finding an integrity basis of second - order tensor - valued functions . two main tools have been used to solve this extremely difficult computational problem . the first one is the so - called _ harmonic decomposition _ which corresponds to the splitting of the elasticity tensor into _ irreducible pieces _ @xcite which was first achieved by backus in 1970 @xcite . such a decomposition is useful to compute the _ material symmetry classes _ @xcite . as mentioned , it has been used also in the study of effective elastic properties of cracked media @xcite and without explicit reference to it in the homogenization techniques of laminated composites @xcite . note that in the later case , the polar decomposition method for 2d media has been used @xcite , while its equivalence with 2d harmonic decomposition has been shown in @xcite . the second one is a purely mathematical tool , _ binary forms _ , which was the cornerstone of invariant theory in the nineteenth century @xcite and is connected to spinors . this tool has been brought first to the knowledge of the mechanical community by backus @xcite , and then by boehler kirilov onat @xcite but its deep power did not seem to have been widely considered so far . it happens to be extremely useful for solving problems in tensor analysis for higher order tensors in 3d . for instance , this tool was used by auffray olive @xcite to produce a minimal integrity basis for a traceless and totally symmetric third order tensor , which appears in piezoelectricity @xcite and second - gradient strain elasticity theory @xcite . note , furthermore , that applications of binary forms are not just bounded to questions in continuum mechanics but are also related to other fields such as quantum computation @xcite and cryptography @xcite . the harmonic decomposition of the elasticity tensor produces a quintuple @xmath7 where @xmath8 are scalar invariants , @xmath9 are second order traceless tensors and @xmath10 is a _ fourth order harmonic tensor_. the scalar and second order tensor components are linearly related to the voigt and the dilatation tensors , the two independent traces of the elasticity tensor . the irreducible fourth order component remains problematic in the study the elasticity tensor . in the present work , we try to go further in the representation of the harmonic fourth order component @xmath10 . we propose to introduce a secondary decomposition of this tensor , which was first discovered by sylvester @xcite and now known as _ maxwell multipoles _ ( see also @xcite ) , through a binary operation called _ harmonic product _ ( a commutative and associative product directly inherited from the product of binary forms ) . this decomposition aims ideally to reduce @xmath10 to a family of vectors , the maxwell multipoles , or , as detailed next , to reduce @xmath10 to a family of second order tensors . in 2d , the factorization of the harmonic component @xmath10 by means of second order tensors was already established in @xcite , it results from the decomposition of elasticity tensor by verchery s polar method @xcite . besides , this factorization has the good taste to be _ equivariant _ , which means that it commutes with the action of the rotation group . however , the solution in 3d of such a decomposition _ i.e. _ the four maxwell multipoles or in an equivalent manner the two second order tensors that factorize @xmath10 is far from being unique and constructive and is thus of poor value in practice . moreover , there are no global _ equivariant section _ of such mappings . we concentrate therefore on producing an _ equivariant explicit solution _ of the problem that we call a _ reconstruction _ and which are limited to specific symmetry classes . this process uses the concept of _ covariants _ , which generalizes the idea of invariants but which are tensors ( rather than scalars ) depending in an equivariant manner of the original tensor . for obvious reasons , no such reconstruction can be globally defined . as it is well known @xcite , only triclinic , monoclinic , orthotropic , transversely isotropic and isotropic classes have a chance to be reconstructed using second - order covariants . in this paper , we obtain for the first time explicit reconstruction formulas in the orthotropic and the transversely isotropic cases , using _ polynomial _ second - order harmonic covariants and rational invariants . moreover , to overpass the geometrical constraints in the trigonal and in the tetragonal cases , we establish for these symmetry classes reconstruction formulas by means of second order covariants up to a single _ cubic fourth order covariant remainder_. the considered reconstruction problem is closely related to the so - called _ isotropic extension _ of anisotropic constitutive functions via structural tensors developed by boehler @xcite and liu @xcite independently ( see also @xcite ) . in the case of a _ linear constitutive law _ , an isotropic extension is just an _ equivariant reconstruction _ _ limited to a given symmetry class _ of the constitutive tensor . furthermore , our approach is more constructive since we are able to give _ explicit equivariant formulas _ in which the `` structural tensors '' are polynomial covariant tensors of the constitutive tensor . in section [ sec : tensorial - representations ] , we recall basic materials on tensorial representations of the orthogonal group , recalling the link between totally symmetric tensor and homogeneous polynomials and the harmonic decomposition . in section [ sec : harmonic - factorization ] , we introduce the _ harmonic product _ between harmonic polynomials and formulate an harmonic factorisation theorem which proof is postponed in appendix [ sec : annexe_sylvester - theorem - proof ] . the general reconstruction problem is formulated in section [ sec : reconstruction ] , where a geometric obstruction for an equivariant reconstruction of fourth - order tensors , by means of second order polynomial covariants is explained . explicit reconstructions formulas , using rational invariants and second order polynomial covariants are obtained in section [ sec : second - order - covariants ] for transversely isotropic and orthotropic tensors . in section [ sec : other - results ] , we propose similarly equivariant reconstructions for tetragonal and trigonal fourth order harmonic tensors up to a cubic covariant remainder . the following spaces are involved : * @xmath4 : the space of _ elasticity tensors _ ; * @xmath11 : the space of @xmath12-th order tensors on @xmath13 ; * @xmath14 : the space of @xmath12-th order _ totally symmetric tensors _ on @xmath13 ; * @xmath15 : the space of @xmath12-th order _ harmonic tensors _ on @xmath13 ; * @xmath16 : the space of homogeneous polynomials of degree @xmath12 on @xmath13 ; * @xmath17 : the space of _ harmonic polynomials _ of degree @xmath12 on @xmath13 ; * @xmath18 : the space of binary forms of degree @xmath12 on @xmath19 ; * @xmath20 : the space of binary forms of degree @xmath21 which correspond to real harmonic tensors of degree @xmath12 . in addition , we adopt the following conventions : * @xmath22 : coordinates on @xmath13 ; * @xmath23 : coordinates on @xmath19 ; * @xmath24 , @xmath25 , @xmath26 , @xmath27 , @xmath28 : vectors of @xmath13 ; * @xmath29 : a vector in @xmath19 ; * @xmath30 , @xmath31 , @xmath32 , @xmath33 , @xmath34 : second - order tensors ; * @xmath35 : second - order dilatation tensor ; * @xmath36 : second - order voigt tensor ; * @xmath37 : second order unit tensor ; * @xmath3 : elasticity tensor ; * @xmath38 : fourth - order unit tensor ; * @xmath39 , @xmath40 : generic tensors ; * @xmath10 : fourth - order harmonic tensor ; * @xmath41 : polynomial on @xmath13 ; * @xmath42 : harmonic polynomial on @xmath13 ; * @xmath43 : binary forms ; * @xmath44 : element of @xmath45 ; * @xmath46 : the scalar product ; * @xmath47 : the tensor product ; * @xmath48 , @xmath49 : the symmetric tensor product ; * @xmath50 : the harmonic product ; * @xmath51 : the action ; * @xmath52 : the euclidean norm ; * @xmath53 : the transpose ; * @xmath54 : the deviatoric part ; * @xmath55 : the harmonic projection ; * @xmath56 : the complex conjugate . in this section , we recall classical facts about tensorial representations of the rotation group @xmath45 . more details on the subject can be found in @xcite or @xcite . we consider thus tensors on the euclidean space @xmath13 and , thanks to the euclidean product , we do not have to distinguish between _ upper _ and _ lower _ indices . therefore , a @xmath12-th order tensor may always be considered as a @xmath12-linear mapping @xmath57 let @xmath11 be the space of @xmath12-th order tensors . the rotation group @xmath45 acts on @xmath11 , by the rule @xmath58 where @xmath59 and @xmath60 . usually , we are not interested in the whole tensor space @xmath11 but rather in a subspace @xmath61 defined by some particular _ index symmetries _ and _ stable _ ( or invariant ) under the action of the rotation group @xmath45 , which means that @xmath62 [ ex : symmetric - tensors ] the permutation group @xmath63 acts on @xmath11 by the rule @xmath64 and this action commutes with the action of @xmath45 . the space of _ totally symmetric tensors _ of order @xmath12 , denoted by @xmath14 , is the space of tensors @xmath39 which are invariant under this action , in other words , such that @xmath65 [ ex : antisymmetric - tensors ] the space of _ alternate tensors _ of order @xmath12 , denoted by @xmath66 , is the space of tensors @xmath39 such that @xmath67 where @xmath68 is the signature of the permutation @xmath69 . the representation @xmath61 is _ irreducible _ if the only stable subspaces are @xmath70 and @xmath61 . the irreducible representations of the rotation group @xmath45 are known ( up to isomorphism ) @xcite . explicit models for these irreducible representations are given by _ harmonic tensor spaces _ which are described below . contracting two indices @xmath71 on a totally symmetric tensor @xmath39 does not depend on the particular choice of the pair @xmath71 . thus , we can refer to this contraction without any reference to a particular choice of indices . we will denote this contraction as @xmath72 , which is a totally symmetric tensor of order @xmath73 and is called the _ trace _ of @xmath39 . iterating the process leads to @xmath74 which is a totally symmetric tensor of order @xmath75 . a _ harmonic tensor _ of order @xmath12 is a totally symmetric tensor @xmath76 such that @xmath77 . the space of harmonic tensors of order @xmath12 will be denoted by @xmath15 . it is a sub - vector space of @xmath14 of dimension @xmath78 . the sub - space @xmath15 of @xmath14 is invariant under the action of @xmath45 and is irreducible ( see for instance @xcite ) . moreover , every finite dimensional irreducible representation of @xmath45 is isomorphic to some @xmath15 . as a special case of the peter weil theorem @xcite , every @xmath45-representation @xmath61 splits into a direct sum of _ harmonic tensor spaces _ @xmath15 , and such a direct sum is called the _ harmonic decomposition _ of @xmath61 . the harmonic decomposition of a second order symmetric tensor @xmath79 corresponds to its decomposition @xmath80 into its deviatoric and spheric parts @xmath81 there is a well - known correspondence between totally symmetric tensors of order @xmath12 and homogeneous polynomial of degree @xmath12 on @xmath13 , which extends the well - known correspondence between a symmetric bilinear form and a quadratic form via polarization . indeed , to each symmetric tensor @xmath76 corresponds a homogeneous polynomial of degree @xmath12 given by @xmath82 this correspondence defines a linear isomorphism @xmath83 between the tensor space @xmath14 and the polynomial space @xmath16 of homogeneous polynomials of degree @xmath12 on @xmath13 . the linear isomorphism @xmath83 satisfies moreover @xmath84 it is said to be an _ equivariant _ mapping . the rotation group @xmath45 acts on the polynomial space @xmath16 , by the rule @xmath85 the inverse @xmath86 can be recovered by _ precisely , the expression @xmath87 is a homogeneous polynomial in the variables @xmath88 and we get @xmath89 note that the _ laplacian operator _ of @xmath90 writes as @xmath91 thus , _ totally symmetric tensors with vanishing trace _ correspond _ via _ @xmath83 to _ harmonic polynomials _ ( polynomials with vanishing laplacian ) . this justifies the denomination of _ harmonic tensors _ for elements of @xmath15 . more generally , for any non - negative integer @xmath92 , we get @xmath93 the equivariant isomorphism @xmath83 sends @xmath15 to @xmath17 , the space of homogeneous harmonic polynomials of degree @xmath12 . thus , the spaces @xmath17 provide alternative models for _ irreducible representations _ of the rotation group @xmath45 . a third model of these irreducible representations is given by the spaces of _ binary forms _ @xmath20 , whose construction is detailed in appendix [ sec : annexe_binary - forms ] . the harmonic decomposition of a homogeneous polynomial of degree @xmath12 is described by the following proposition . [ prop : symmetric - harmonic - decomposition ] every homogeneous polynomial @xmath94 can be decomposed uniquely as @xmath95 where @xmath96 , @xmath97 $ ] with @xmath98 $ ] integer part and @xmath99 is a harmonic polynomial of degree @xmath75 . the harmonic polynomial @xmath100 is the orthogonal projection of @xmath41 onto @xmath17 ( for the induced euclidean structure of @xmath13 ) . its explicit expression can be obtained recursively as follows ( see @xcite ) . let first define @xmath101 then @xmath99 is obtained inductively by the formula @xmath102 with @xmath103 which leads to @xmath104 after @xmath105 iterations . introducing the symmetric tensor product @xmath106 for @xmath107 and @xmath108 , we can recast proposition [ prop : symmetric - harmonic - decomposition ] in tensorial form : @xmath109 where @xmath97 $ ] . more precisely , every totally symmetric tensor @xmath76 of order @xmath12 can be decomposed uniquely as @xmath110 where @xmath111 is a harmonic tensor of degree @xmath75 and @xmath112 means the symmetrized tensorial product of @xmath92 copies of @xmath113 . the harmonic part @xmath114 is the orthogonal projection of @xmath39 onto @xmath15 . using , the tensorial form of reads @xmath115 which is a scalar for even @xmath12 or a vector for odd @xmath12 . finally , the tensorial form of is given by : @xmath116\ ] ] note that in this equation , @xmath117 $ ] is the orthogonal projection of @xmath118 on @xmath119 . harmonic decomposition of the space @xmath4 of elasticity tensors was first obtained by backus @xcite and is given by @xmath120 more precisely , an elasticity tensor @xmath3 admits the following harmonic decomposition @xcite : @xmath121 in this formula , @xmath113 is the euclidean canonical bilinear 2-form represented by the components @xmath122 in any orthonormal basis and the young - symmetrized tensor products @xmath49 and @xmath123 , between symmetric second - order tensors @xmath124 , are defined as follows : @xmath125 where @xmath49 is the totally symmetric tensor product also denoted by @xmath48 , and @xmath126 in the harmonic decomposition , @xmath127 are scalars , @xmath128 are second order harmonic tensors ( also known as symmetric deviatoric tensors ) and @xmath10 is a fourth - order harmonic tensor . in this decomposition , @xmath127 and @xmath128 are related to the _ dilatation tensor _ @xmath129 and the _ voigt tensor _ @xmath130 by the formulas @xmath131 and @xmath132 where @xmath133 and @xmath134 are the deviatoric parts of @xmath135 and @xmath136 respectively . in this section , we introduce a product between harmonic tensors ( which is associative and commutative ) . then , we show that every harmonic tensor @xmath137 can be factorized as a harmonic product of lower order tensors . the product of two harmonic polynomials is not harmonic in general . however , we define the following binary operation on the space of harmonic polynomials : @xmath138 this product is commutative : @xmath139 and is moreover associative , which means that @xmath140 the fact that this binary operation is associative is essential in practice . it means that the final result does not depend on the way we have chosen to put the brackets " to compute it . we can therefore write @xmath141 and omit the brackets . endowed with this product , the vector space @xmath142 of harmonic polynomials becomes a _ commutative algebra_. the unity ( for the multiplicative operation @xmath50 ) is represented by the constant harmonic polynomial @xmath143 . [ rem : h1odoth2 ] using the equivariant isomorphism @xmath83 between @xmath15 and @xmath17 , this algebraic structure is inherited on harmonic tensors . more precisely , given two harmonic tensors @xmath144 and @xmath145 of order @xmath146 and @xmath147 respectively , we define a harmonic tensor @xmath148 of order @xmath149 by the following formula : @xmath150 [ ex : v - ast - w ] for two vectors @xmath151 , we get @xmath152 [ ex : a - ast - b ] for two second order harmonic tensors @xmath153 , we get @xmath154 we now introduce the following @xmath45-equivariant mapping , from the direct sum @xmath155 of @xmath12 copies of @xmath13 to @xmath17 : @xmath156 this mapping is obviously not _ one - to - one_. it is however _ surjective_. this result is not trivial and seems to have been demonstrated for the first time by sylvester @xcite ( see also ( * ? ? ? * appendix 1 ) ) . we provide our own proof of sylvester s theorem in appendix [ sec : annexe_sylvester - theorem - proof ] , which relies on _ binary forms_. note however , that the use of binary forms for the proofs limits the validity of these theorems to dimension @xmath157 ( see @xcite for a counter - example in dimension @xmath158 ) . [ thm : sylvester ] let @xmath159 , then there exist @xmath12 vectors @xmath160 in @xmath13 such that @xmath161 [ ex : multipoles - determination ] let @xmath162 and @xmath163 be the corresponding binary form ( see example [ ex : order2-harmonic - tensors ] ) . the four roots of @xmath164 are @xmath165 with the possible couple @xmath166 among them . using the stereographic projection @xmath167 ( see remark [ ex : order1-harmonic - tensors ] ) , we set @xmath168 note that @xmath169 then , we have @xmath170 for some constant @xmath171 . the lack of injectivity for the mapping @xmath172 is detailed in the following proposition , which proof is provided in appendix [ sec : annexe_sylvester - theorem - proof ] . [ prop : fiber ] we have @xmath173 if and only if @xmath174 where @xmath175 is a permutation of the vectors @xmath176 for some @xmath177 and @xmath178 are real numbers such that @xmath179 . it is tempting to restrict the mapping @xmath172 defined in to unit vectors , in order to try to reduce the indeterminacy in the choice of an @xmath12-uple of vectors @xmath176 . however , this does not really solve the problem of indeterminacy . of course , the indeterminacy due to the scaling by the @xmath180 is replaced by signs @xmath181 , but the indeterminacy due to the permutation persists anyway . sylvester s theorem is equivalent to the following ( in appearance , more general ) theorem which leads to other important factorization results for harmonic polynomials ( and thus harmonic tensors ) . [ thm : harmonic - factorization - theorem ] let @xmath182 . then there exist harmonic polynomials @xmath183 such that @xmath184 [ rem : tensorial - hkn ] the tensorial counterpart of theorem [ thm : harmonic - factorization - theorem ] is that for every harmonic tensor @xmath185 there exist harmonic tensors @xmath186 such that @xmath187 a corollary of theorem [ thm : harmonic - factorization - theorem ] is the following result , which applies to every even order harmonic tensor . [ cor : diffcarre_tensor ] let @xmath188 . then there exist @xmath189 such that @xmath190 let @xmath191 corresponding to the harmonic tensor @xmath192 . using theorem [ thm : harmonic - factorization - theorem ] , we can find @xmath193 such that @xmath194 . but @xmath195 hence , if we set @xmath196 and @xmath197 , we get @xmath198 which tensorial form is precisely . this result has already been obtained in @xcite for monoclinic harmonic fourth - order tensors , using direct elimination techniques . in 2d , the factorization of the harmonic part @xmath199 of the elasticity tensor has already been computed in @xcite , using previous results obtained in @xcite . in that case , we get @xmath200 with second order tensor @xmath33 harmonic ( _ i.e. _ symmetric deviatoric ) , and where @xmath201 and @xmath202 . as in 3d , we can define a harmonic product between 2d harmonic tensors @xmath203 @xmath204 and equation can be rewritten as @xmath205 hence , any 2d harmonic fourth - order tensor can be expressed as the harmonic square @xmath206 , _ i.e. _ the projection of diadic symmetric square @xmath207 onto the vector space of fourth - order harmonic tensors @xmath208 . in 3d , this result is generally false . indeed , if a fourth - order harmonic tensor @xmath10 is a harmonic square , _ i.e. _ @xmath209 , then @xmath10 is either orthotropic or transversely isotropic ( see section [ sec : reconstruction ] ) . although these various decompositions are interesting by themselves , there is however something frustrating . these decompositions are not unique in general and the solution is not enough constructive ( they require root s extraction of possible high degree polynomials ) . in the next sections , we provide explicit geometric reconstructions . in this section , we consider the general problem of reconstructing a fourth - order harmonic tensor @xmath10 using lower order _ covariants _ ( i.e. lower order tensors which depend on @xmath10 in an equivariant manner ) . this reconstruction is useful when applied to the fourth - order harmonic part @xmath10 of the elasticity tensor ( [ eq : harme ] ) , since its other components @xmath210 are already decomposed by means of second order covariants tensors and invariants . from a mechanical point of view the definition of covariants of a given effective tensor is strongly related to the tensorial nature of the thermodynamics variables allowing to properly model a specific state coupling with elasticity @xcite . in the case of effective elasticity of cracked / damaged media for instance , either second order @xcite or fourth order @xcite damage variables are considered , depending on the authors choice @xcite . considering the reconstruction by means of lower order covariants of the harmonic part of an effective elasticity tensor in this case would answer the question : do we have to use a fourth order damage variable a fourth order covariant in next derivations in order to reconstruct a given harmonic effective elasticity tensor ? in 2d the answer is negative , as shown in @xcite : only one second order tensor @xmath33 is needed ( see eq . ) . to illustrate the problem , let us first come back to theorem [ thm : harmonic - factorization - theorem ] . applied to a fourth - order harmonic tensor @xmath10 , this theorem states that @xmath10 can be factorized as the harmonic product of two second order harmonic tensors @xmath211 in other words , the equivariant mapping @xmath212 is surjective . what would be more interesting is to have an explicit mapping ( i.e. an explicit construction of the solution ) @xmath213 such that : @xmath214 and with geometric properties . a mapping like @xmath215 is called a _ section _ of @xmath216 . a geometric property would be for instance for @xmath215 to be equivariant , in which case each second order tensors @xmath217 is a covariant tensor . the main problem is that there does not exist any _ equivariant section _ @xmath215 for @xmath216 defined on the whole space @xmath218 : for instance , for a tensor @xmath10 with cubic symmetry , all second order harmonic covariants vanish , which forbids the existence of such a general equivariant section . we will provide now the following definitions . let @xmath39 be some @xmath12th order tensor . a _ decomposition _ of a @xmath39 into other _ tensors _ ( possibly scalars ) @xmath219 ( @xmath220 ) is a mapping @xmath221 the decomposition is _ equivariant _ if @xmath216 is an equivariant mapping . given a decomposition @xmath222 of a tensor @xmath39 into other _ tensors _ ( possibly scalars ) , a _ reconstruction _ of @xmath39 is a section @xmath215 of @xmath216 . the reconstruction is _ equivariant _ if both @xmath216 and @xmath215 are equivariant mappings . in other words , a reconstruction is given by a mapping @xmath223 such that @xmath224 if the reconstruction is equivariant the @xmath225 are _ covariant tensors _ ( or invariants if there are scalars ) , which means that @xmath226 the existence of an equivariant reconstruction may not be possible for certain symmetry classes . more precisely , we have the following result . [ lem : obstruction - equivariant - reconstruction ] consider an equivariant decomposition @xmath227 and suppose that there exists an equivariant reconstruction @xmath215 , so that @xmath228 are covariants tensors . then @xmath229 where @xmath230 is the _ symmetry group _ of the tensor @xmath39 . consider an equivariant decomposition @xmath227 . then , we have @xmath231 if moreover @xmath232 are covariant tensors , we get thus @xmath233 for all @xmath234 , and hence @xmath235 , which achieves the proof . according to lemma [ lem : obstruction - equivariant - reconstruction ] , the existence of an equivariant reconstruction associated to an equivariant decomposition @xmath236 requires some conditions on the symmetry groups of the involved covariants @xmath232 . for a decomposition involving second - order covariant tensors , we have the following result ( for details on closed subgroups of @xmath45 , see appendix [ sec : symmetry - classes ] ) . [ cor : second - order - obstruction ] if @xmath227 is an equivariant decomposition of @xmath39 into _ second - order symmetric covariant tensors _ @xmath228 , then @xmath237\in { \left\{[{\mathds{1}}],[{\mathbb{z}}_{2}],[{\mathbb{d}}_{2}],[{\mathrm{o}}(2)]\right\}},\ ] ] where @xmath238 $ ] means the conjugacy class of a subgroup @xmath239 . the proof is easily achieved using the _ clips operator _ introduced in @xcite . this binary operator , noted @xmath240 is defined on conjugacy classes ( or finite families of conjugacy classes ) of closed subgroups of @xmath45 : @xmath241\circledcirc [ h_{2}]={\left\ { [ h_{1}\cap ( gh_{2}g^{-1})],\quad g\in { \mathrm{so}}(3,{\mathbb{r}})\right\}}.\ ] ] it is an associative and commutative operation and @xmath242 \circledcirc [ { \mathds{1 } } ] = { \left\{[{\mathds{1}}]\right\ } } , \qquad [ h ] \circledcirc [ { \mathrm{so}}(3,{\mathbb{r } } ) ] = { \left\{[h]\right\}},\ ] ] for all closed subgroup @xmath243 . moreover , we have in particular ( see @xcite ) : [ cols="^,^,^,^",options="header " , ] thus , if each @xmath228 is a second - order symmetric covariant ( or an invariant ) , then @xmath244 $ ] is either @xmath245 $ ] , @xmath246 $ ] or @xmath247 $ ] ( see ) and we deduce that @xmath237\in [ h_{1}]\circledcirc [ h_{2 } ] \dotsc \circledcirc [ h_{n}],\quad [ h_{i}]\in { \left\{[{\mathbb{d}}_{2}],[{\mathrm{o}}(2)],[{\mathrm{so}}(3,{\mathbb{r}})]\right\}},\ ] ] which achieves the proof . corrollary [ cor : second - order - obstruction ] implies in particular that there is no equivariant reconstruction by means of second order symmetric covariant tensors of a harmonic fourth order tensor @xmath10 ( and thus of the elasticity tensor @xmath3 ) for the cubic , the trigonal and the tetragonal symmetry classes . for a cubic tensor , the situation is even worse since all its second order symmetric covariants are spheric , whereas its first order covariants vanish . the geometric framework introduced here is connected to the so - called _ isotropic extension _ of anisotropic constitutive functions via structural tensors developed by boehler @xcite and liu @xcite independently ( see also @xcite ) . in the case of a linear constitutive function , an isotropic extension is just an equivariant decomposition _ limited to a given symmetry class _ @xmath248 $ ] and for which the arguments @xmath249 ( the structural tensors ) of the decomposition satisfy . in that case , our approach is however more constructive since we are able to give an _ explicit equivariant formula _ in which the `` structural tensors '' are covariant tensors of the constitutive tensor . moreover , condition is a corollary of the theory and does not need to be imposed _ a priori _ as an hypothesis . in this section , we provide explicit equivariant reconstructions for transverse - isotropic and orthotropic fourth order harmonic tensors . these reconstructions use the following second - order ( polynomial ) covariants introduced in @xcite : @xmath250 where the following notations have been used . * for two fourth - order tensors @xmath251 and @xmath252 , @xmath253 is the fourth - order tensor @xmath254 * for a fourth - order tensor @xmath251 and a second - order tensor @xmath255 , @xmath256 is the second - order tensor @xmath257 we also make use of the following invariants : @xmath258 which constitute an _ integrity basis _ for @xmath218 ( see @xcite ) . the second order covariants @xmath259 ( @xmath260 ) are always _ symmetric _ covariants , while in general @xmath259 ( @xmath261 ) are neither symmetric nor antisymmetric . finally , using the kelvin representation , a fourth order harmonic tensor @xmath262 is represented by the symmetric matrix @xmath263 with only 9 independent components due to the traceless property , namely : @xmath264 a harmonic tensor @xmath265 is transverse isotropic if and only if there exists @xmath0 such that @xmath266 where @xmath267 has the following normal matrix form @xmath268 where @xmath269 . it was moreover established in ( * ? ? ? * section 5.2 ) that @xmath270 thus , by a direct computation of the second order covariant @xmath271 ( see eq . ) and its deviatoric part @xmath272 , we get the following result : [ thm : o2 ] for any transversely isotropic harmonic fourth - order tensor @xmath10 , we have @xmath273 [ rem : o2-harmonic - square ] if @xmath274 , then @xmath10 is a perfect harmonic square . the converse is also true and can be easily established , using binary forms , where the harmonic product corresponds to the ordinary product of polynomials . a harmonic tensor @xmath265 is orthotropic if and only if there exists @xmath0 such that @xmath266 where @xmath267 has the following normal matrix form @xmath275 where @xmath276 are three _ distinct _ real numbers . note that this normal form is however not unique : any permutation of the @xmath277 provides an alternative normal form , but this is the only ambiguity . it is therefore useful to introduce the three elementary symmetric functions @xmath278 which are independent of a particular normal form , as being invariant under any permutation of the @xmath279 . moreover , it was shown in ( * ? ? ? * section 5.5 ) that the _ discriminant _ @xmath280 which is _ strictly positive _ , is a polynomial invariant , and that the @xmath281 are themselves _ rational invariants_. more precisely : @xmath282 and @xmath283 consider now the matrix @xmath284 since both @xmath267 and @xmath285 have the same symmetry group , namely @xmath286 ( as defined in ) , the following definition @xmath287 does not depend on the rotation @xmath44 and thus defines a covariant mapping from the orthotopic class in @xmath218 to the space of symmetric second order tensors . we have moreover the following result , which can be checked by a direct evaluation on @xmath267 and @xmath288 . [ lem : rational - expression - lambda ] let @xmath10 be an orthotropic harmonic fourth - order tensor . then , the deviatoric part @xmath289 of the second order covariant @xmath290 can be written as @xmath291 where @xmath292 note that @xmath293 and we can thus introduce the positive invariant @xmath294 and the _ lode invariant _ defined by @xmath295 since moreover @xmath296 and @xmath297 , we deduce that @xmath298 . with these notations , we obtain the following result which can be checked by a direct computation on the normal form . [ thm : d2 ] for any orthotropic harmonic fourth - order tensor @xmath10 , we have @xmath299 where @xmath289 is defined by and the three invariants @xmath300 are given by @xmath301 the three invariants @xmath300 are in fact rational invariants of @xmath10 . using invariants @xmath281 ( eq . ( [ eq : sym - func ] ) ) , we have indeed @xmath302 as in the transversely orthotropic case , the following theorem characterized orthotropic harmonic tensors which are a perfect harmonic squares . [ thm : orth_square ] an orthotropic harmonic fourth - order tensor @xmath10 is a perfect harmonic square @xmath303 if and only if @xmath304 in that case the harmonic second order tensor @xmath33 is is given by @xmath305 let @xmath10 be an orthotropic fourth order harmonic tensor . suppose first that conditions are satisfied . then , the coefficients @xmath306 , @xmath307 , @xmath308 in satisfy @xmath309 and we have thus a factorization @xmath303 , where @xmath33 is defined by . conversely , suppose that @xmath10 can be factorized as @xmath303 . without loss of generality , we can assume that @xmath33 is diagonal , with diagonal elements @xmath310 , @xmath311 and @xmath312 . thus @xmath10 is fixed by @xmath286 and is represented by matrix normal form . consider now the associated binary form @xmath313 of @xmath33 ( resp . @xmath164 of @xmath10 ) see . we get then @xmath314 and @xmath315 now , @xmath316 leads to @xmath317 and we get @xmath318 therefore , we have @xmath319 if @xmath320 , we get also @xmath321 and thus @xmath322 , which leads to a contradiction ( since @xmath10 is supposed to be orthotropic ) . this concludes the proof . in this section , we consider the problem of reconstruction of harmonic tensors for which an equivariant reconstruction into second - order tensors is not possible , namely the harmonic tensors within the trigonal and the tetragonal classes . for these two classes , we show that such a reconstruction is possible up to a _ fourth - order harmonic tensor with cubic symmetry_. all the closed @xmath45 subgroups used in this section are defined in . a harmonic tensor @xmath265 is tetragonal if and only if there exists @xmath0 such that @xmath266 where @xmath267 has the following normal matrix form @xmath323 where @xmath324 and @xmath325 . note that this normal form is however not unique : the change @xmath326 , provides an alternative normal form . nevertheless , the choice @xmath327 allows to fix this ambiguity . note also that @xmath328 where @xmath329 is the rotation by angle @xmath330 around the @xmath331 axis . for tetragonal harmonic tensors , it was shown in ( * ? ? ? * section 5.4 ) that the following polynomial invariants do not vanish @xmath332 and that @xmath333 and @xmath334 are rational invariants , given by @xmath335 the choice @xmath327 in the normal form allows to write @xmath69 as follows : @xmath336 note that the condition @xmath337 is equivalent to @xmath338 with @xmath339 and corresponds to the degeneracy when @xmath10 has at least a cubic symmetry . on the other hand , the condition @xmath340 ( with @xmath341 ) is equivalent to @xmath342 and corresponds to the degeneracy when @xmath10 has at least a transverse isotropic symmetry . the normal form corresponds to the subspace @xmath343 , where given a subgroup @xmath344 of @xmath345 the _ fixed point set _ @xmath346 is defined as @xmath347 note now that there is only one subgroup in the conjugacy class @xmath246 $ ] which contains @xmath348 , it is @xmath349 itself . its fix point set , @xmath350 , is the one - dimensional subspace spanned by @xmath351 however , there are two different subgroups @xmath352 and @xmath353 in the conjugacy class @xmath354 $ ] which contains @xmath348 and which are defined as follows . the first one , @xmath352 , is the symmetry group of the cube with edges @xmath355 . the second one , @xmath353 , is obtained by rotating @xmath352 around the @xmath331 axis by angle @xmath330 . in other words , @xmath356 , where @xmath105 is the rotation by angle @xmath330 around the @xmath331 axis . the fixed point sets @xmath357 and @xmath358 are one - dimensional subspaces spanned respectively by @xmath359 and @xmath360 moreover , both @xmath357 and @xmath358 are supplementary subspaces of @xmath350 in @xmath343 , and thus we can write either @xmath361 or @xmath362 let @xmath10 be a tetragonal harmonic fourth order tensor . then there exists @xmath60 ( defined up to a right composition by an element of @xmath348 ) and a unique couple of real numbers @xmath363 with @xmath364 such that : @xmath365 since @xmath366 , the following definitions @xmath367 do not depend on the rotation @xmath44 , leading to well - defined covariant mappings @xmath368 , @xmath369 from the tetragonal class in @xmath218 to the transversely isotropic classes . similarly , since @xmath352 and @xmath353 contains @xmath348 , the same is true for @xmath370 let @xmath329 be the rotation by angle @xmath330 around the @xmath331 axis . we have then @xmath371 now if @xmath372 , we get @xmath373 [ thm : d4 ] for any tetragonal harmonic fourth - order tensor @xmath10 , we have @xmath374 where @xmath375 is a transversely isotropic covariant , and @xmath376 is a cubic covariant . all the formulas can be checked on the normal form , for which @xmath377 . since the formulas are covariant , this is enough to achieve the proof . in theorem [ thm : d4 ] , each formula , i.e. @xmath378 or @xmath379 , is self - sufficient . it is equivariant and applies in any basis to any tetragonal harmonic fourth order tensor @xmath10 . the two reconstruction formulas correspond to the choice of a frame and therefore of a cube , rotated by the angle @xmath330 or not around @xmath331 . they use the invariants @xmath327 and @xmath333 defined in and . the decomposition in theorem [ thm : d4 ] can be rewritten in terms of the invariants @xmath380 , @xmath381 , @xmath382 and @xmath383 , rather than @xmath384 by using their rational expressions and . [ cor : d4 ] for any tetragonal harmonic fourth - order tensor @xmath10 , we have @xmath385 where @xmath386 is a cubic fourth order covariant and @xmath387 a harmonic tensor @xmath265 is trigonal if and only if there exists @xmath0 such that @xmath266 where @xmath267 has the following normal matrix form @xmath388 where @xmath389 and @xmath325 . note that this normal form is however not unique . changing @xmath390 , provides an alternative normal form . nevertheless , the choice @xmath327 allows to fix this ambiguity . note also that @xmath391 where @xmath392 is the rotation by angle @xmath393 around the @xmath331 axis . for trigonal harmonic tensors , it was shown in ( * ? ? ? * section 5.3 ) that the polynomial invariants @xmath381 and @xmath382 , defined by , are strictly positive and that @xmath333 , @xmath334 are rational invariants , given by @xmath394 the choice @xmath327 in the normal form allows to write @xmath69 as follows : @xmath395 note that the condition @xmath396 is equivalent to @xmath397 with @xmath398 and corresponds to the degeneracy case when @xmath10 has at least a cubic symmetry . on the other hand , with @xmath399,condition @xmath340 is equivalent to @xmath342 and corresponds to the degeneracy case when @xmath10 has at least a transverse isotropic symmetry . the geometric picture is similar to the tetragonal case : @xmath400 is contained in only one subgroup in the conjugacy class @xmath246 $ ] and two different subgroups @xmath401 , @xmath402 in the conjugacy class @xmath354 $ ] , which are defined as follows . @xmath403 where @xmath352 is the symmetry group of the cube with edges @xmath355 ( already defined in section [ subsec : tetragonal ] ) and the rotations introduced are @xmath404 the fixed point sets @xmath405 and @xmath406 are one - dimensional subspaces spanned respectively by @xmath407 and @xmath408 the normal form can be decomposed as the sum of a transversely isotropic harmonic tensor and of a cubic harmonic tensor , either as @xmath409 or as @xmath410 where @xmath411 has been defined in . as in the tetragonal case , we can give coherent definitions of transversely isotropic parts @xmath412 and cubic parts @xmath413 [ thm : d3 ] for any trigonal harmonic fourth - order tensor @xmath10 , we have @xmath414 where @xmath415 is a transversely isotropic covariant and @xmath416 is a cubic covariant . all the formulas can be checked on the normal form , for which @xmath417 . since the formulas are covariant , this is enough to achieve the proof . in theorem [ thm : d3 ] , each formula , i.e. @xmath378 or @xmath379 , is self - sufficient . it is equivariant and applies in any basis to any trigonal harmonic fourth order tensor @xmath10 . the two reconstruction formulas correspond to the choice of a frame and therefore of a cube , rotated by the angle @xmath393 or not around @xmath331 . they use the invariants @xmath327 and @xmath333 defined in eq . ( [ eq : coeff - trigonal ] ) and ( [ eq : coeff - trigonal - sigma ] ) . the decomposition in theorem [ thm : d4 ] can be rewritten in terms of the invariants @xmath380 , @xmath381 , @xmath382 and @xmath418 , rather than @xmath384 by using their rational expressions and . [ cor : d3 ] for any trigonal harmonic fourth - order tensor @xmath10 , we have @xmath419 where @xmath420 is a cubic fourth order covariant , and @xmath421 in this article , we have defined the _ harmonic projection _ @xmath422 of a totally symmetric @xmath12-th order tensor @xmath39 , and the _ harmonic product _ @xmath423 of two harmonic tensors @xmath424 , @xmath425 . using sylvester s theorem [ thm : sylvester ] , we have shown that the fourth order harmonic part @xmath10 of an elasticity tensor @xmath3 can be expressed as the harmonic product @xmath426 of two second order harmonic tensors @xmath427 ( _ i.e. _ symmetric deviatoric ) . this decomposition is independent of any reference frame , even for triclinic materials . nevertheless , such a factorization is non - unique and not really constructive . moreover , a _ globally defined _ solution @xmath428 can never be _ we have therefore introduced an _ equivariant reconstruction _ problem of a tensor by means of _ lower order covariants _ , restricted to a given symmetry class . finally , we have obtained explicit formulas for the reconstruction of the fourth order harmonic part @xmath10 of an elasticity tensor : * for the _ transversely isotropic _ and _ orthotropic _ symmetry classes , by means of second order covariants ( theorems [ thm : o2 ] and theorem [ thm : d2 ] ) . moreover , necessary and sufficient conditions for such tensors to be perfect harmonic squares , which means @xmath303 , have also been obtained ( remark [ rem : o2-harmonic - square ] and theorem [ thm : orth_square ] ) ; * for the _ tetragonal _ and _ trigonal _ symmetry classes , by means of second order covariants and a fourth order cubic covariant ( theorem [ thm : d4 ] and theorem [ thm : d3 ] ) . these reconstruction formulas are only valid inside each symmetry class , this is illustrated by the fact that they contain denominators which vanish when a degeneracy into a higher symmetry class occurs . we have already described two models for the irreducible representations of the rotation group @xmath45 : the space of _ harmonic tensors _ @xmath15 and the space of harmonic polynomials @xmath17 . in this section , we shall derive a third model : the space of _ binary forms _ @xmath20 , whose construction is slightly more involved . to start , let us recall that there is a well - known relation between the rotation group @xmath45 and the group of _ special unitary complex matrices _ @xmath429 let @xmath430 be a linear mapping from @xmath13 to the space of traceless , hermitian matrices of order @xmath431 ( @xmath432 , @xmath433 and @xmath434 are the famous _ pauli matrices _ ) . the group @xmath435 acts on this space by conjugacy @xmath436 and preserves the quadratic form @xmath437 it can be checked , moreover , that @xmath438 for all @xmath439 . therefore , we deduce a group morphism @xmath440 for @xmath441 we get @xmath442 in the basis @xmath443 . the group @xmath435 acts naturally on @xmath19 and more generally on complex polynomials in two variables @xmath444 by the rule @xmath445 let @xmath18 be the space of complex , homogeneous polynomials of degree @xmath12 in two variables . an element of @xmath18 @xmath446 is called a _ binary form _ of degree @xmath12 . the main observation , due to cartan ( see @xcite for the details ) , is that there is an _ equivariant isomorphism _ between the space @xmath447 of binary forms of degree @xmath21 and the space @xmath448 of complex , harmonic polynomials of degree @xmath12 in three variables @xmath449 . this linear isomorphism @xmath450 is given explicitly by @xmath451 it satisfies moreover the _ equivariant relation _ @xmath452 the proof that @xmath453 is an equivariant isomorphism can be found in @xcite . the inverse @xmath454 is obtained explicitly as followed . let @xmath455 for each @xmath92 make the substitution @xmath456 we obtain this way a homogeneous polynomial in three variables @xmath41 of degree @xmath12 such that @xmath457 now , let @xmath458 be the @xmath12-th order harmonic component of @xmath41 in the harmonic decomposition of @xmath41 . then @xmath459 . under this isomorphism , the space @xmath17 of _ real _ harmonic polynomials corresponds to the _ real subspace _ @xmath460 of binary forms @xmath164 which satisfy @xmath461 these binary forms @xmath462 are also characterized by the following property : @xmath463 the space @xmath20 is invariant under the action of @xmath435 and is moreover _ irreducible _ for this action . [ rem : multiplicative - property ] note that if @xmath464 and @xmath465 , then @xmath466 and @xmath467 [ ex : order1-harmonic - tensors ] an order-@xmath143 harmonic tensor @xmath33 on @xmath13 is just a linear form @xmath42 on @xmath13 , which can be written as @xmath468 with @xmath469 . the corresponding binary form @xmath470 is given by @xmath471 where @xmath472 conversely , given @xmath473 , the corresponding real @xmath474-vector @xmath25 is given by @xmath475 [ rem : stereographic - projection ] for order-@xmath143 tensors , there is a closed relation between the cartan - map and the stereographic projection @xmath476 where @xmath477 is the unit sphere and @xmath478 . the stereographic projection can be extended to a bijection @xmath479 by setting @xmath480 . its inverse , @xmath481 , is given by @xmath482 now let @xmath483 . then we can write @xmath484 , where @xmath485 and @xmath486 . if moreover @xmath487 , we can write @xmath488 as @xmath489 [ ex : order2-harmonic - tensors ] let @xmath490 be a second order real harmonic tensor and @xmath42 be the corresponding real harmonic polynomial , then the corresponding binary form @xmath491 is given by @xmath492 where @xmath493 conversely , given @xmath494 , the corresponding harmonic tensor @xmath33 is given by @xmath495 [ ex : order4-harmonic - tensors ] let @xmath496 be a fourth - order real harmonic tensor and @xmath42 be the corresponding real harmonic polynomial , then the corresponding binary form @xmath497 is given by @xmath498 where the independent components are @xmath499 conversely , given @xmath500 , the corresponding harmonic tensor in @xmath501 is given by : @xmath502 the key point to prove sylvester s theorem [ thm : sylvester ] is a _ root characterization _ of binary forms in @xmath20 . [ lem : root - characterization ] let @xmath503 . then @xmath504 if and only if it can be written as @xmath505 where @xmath506 and @xmath507 . let @xmath504 which does not vanish identically . due to , we deduce that for each @xmath508 , @xmath509 if and only if @xmath510 . thus , there exists a non - negative integer @xmath511 such that @xmath512 where @xmath513 with leading term @xmath514 . let @xmath515 then @xmath41 is a polynomial of degree @xmath516 with no vanishing root . due to , we deduce that if @xmath517 is a root of @xmath41 , then @xmath518 is another root of @xmath41 . thus @xmath519 and we get @xmath520 now , using , we have @xmath521 and thus @xmath522 is real . the converse can be immediately established using the functional characterization of @xmath20 . [ rem : about - non - unicity ] this factorization is not unique . permuting a couple of paired roots @xmath523 , leads to the transformation @xmath524 and may result in a different value for @xmath525 . we can now give the proof of sylvester s theorem [ thm : sylvester ] and of proposition [ prop : fiber ] . using the isomorphism between @xmath17 and @xmath20 described in , we set @xmath526 . using lemma [ lem : root - characterization ] , we can write @xmath527 let @xmath528 each @xmath529 is in @xmath530 so @xmath531 is a polynomial of degree @xmath143 on @xmath13 and can be written as @xmath532 for some vector @xmath533 in @xmath13 . we get thus @xmath534 which achieves the proof . let @xmath535 and @xmath536 be unit vectors . we can assume , rotating if necessary this configuration of @xmath21 vectors on the sphere , that none of the @xmath537 are equal to @xmath538 or @xmath539 . set @xmath540 and note that the @xmath541 belong to @xmath542 . using , the relation @xmath543 can be recast as the binary form identity @xmath544 therefore , there exists a permutation @xmath69 of @xmath545 , such that either @xmath546 or @xmath547 . in the first case we deduce that @xmath548 while in the second case , we have @xmath549 we obtain therefore that @xmath550 where @xmath551 for @xmath552 with @xmath553 . if the vectors @xmath554 are not unit vectors , a normalisation process lead to the conclusion that @xmath555 where @xmath556 and @xmath179 , which achieves the proof . the action of the rotation group @xmath45 on the space of @xmath12-order harmonic tensors @xmath15 partitions this space into _ symmetry classes _ , where two tensors @xmath557 belong to the same symmetry class if their respective symmetry group @xmath558 and @xmath559 are conjugates , that is @xmath560 symmetry classes are in correspondence with conjugacy classes @xmath561 : = { \left\ { gkg^{-1};\ ; g\in { \mathrm{so}}(3,{\mathbb{r}})\right\}}\ ] ] of closed subgroups @xmath562 of @xmath45 ( several conjugacy classes may however correspond to empty symmetry classes ) . moreover , symmetry classes are partially ordered by the following partial order defined on conjugacy classes @xmath563\preceq [ k_{2 } ] \qquad \text{if } \qquad \exists g\in { \mathrm{so}}(3,{\mathbb{r}}),\quad k_{1}\subset gk_{2}g^{-1}.\ ] ] * @xmath349 is the subgroup generated by all the rotations around the @xmath565-axis and the order 2 rotation @xmath566 around the @xmath567-axis . * @xmath568 is the subgroup of all the rotations around the @xmath565-axis . * @xmath569 is the unique cyclic subgroup of order @xmath12 of @xmath568 , the subgroup of rotations around the @xmath565-axis . * @xmath570 is the _ dihedral _ group . it is generated by @xmath569 and @xmath571 . * @xmath572 is the _ tetrahedral _ group , the ( orientation - preserving ) symmetry group of a tetrahedron . it has order 12 . * @xmath573 is the _ octahedral _ group , the ( orientation - preserving ) symmetry group of a cube or octahedron . it has order 24 . * @xmath574 is the _ icosahedral _ group , the ( orientation - preserving ) symmetry group of a icosahedra or dodecahedron . it has order 60 . * @xmath575 is the trivial subgroup , containing only the unit element . the symmetry classes of a second order symmetric tensor are described by the following partially ordered set : @xmath576 $ } \end{array } \longrightarrow \begin{array}{c } \textrm{transversely isotropic } \\ \textrm{$[{\mathrm{o}}(2)]$ } \end{array } \longrightarrow \begin{array}{c } \textrm{isotropic } \\ \textrm{$[{\mathrm{so}}(3,{\mathbb{r}})]$ } \end{array}\ ] ] for @xmath218 , it is known that there are only @xmath577 symmetry classes ( the same classes as for the elasticity tensor @xcite ) . the corresponding partial ordering is illustrated in figure [ fig : lattice ] , with the convention that a subgroup at the starting point of an arrow is conjugate to a subgroup of the group pointed by the arrow . j. betten and w. helisch . integrity bases for a fourth - rank tensor . in _ anisotropy , inhomogeneity and nonlinearity in solid mechanics ( nottingham , 1994 ) _ , volume 39 of _ solid mech . _ , pages 3742 . kluwer acad . , dordrecht . introduction to the invariant formulation of anisotropic constitutive equations . in _ applications of tensor functions in solid mechanics _ , volume 292 of _ cism courses and lectures _ , pages 1330 . springer , vienna , 1987 . le concept de contrainte effective appliqu llasticit et la viscoplasticit en prsence dun endommagement anisotrope . in j .- boehler , editor , _ colloque int . cnrs 295 , villard de lans _ , pages 737760 . martinus nijhoff publishers and editions du cnrs , 1982 , 1979 . p. gordan . . chelsea publishing co. , new york , second edition , 1987 . erster band : determinanten . i : determinants ] , zweiter band : binre formen . ii : binary forms ] , edited by georg kerschensteiner . g. verchery . les invariants des tenseurs dordre 4 du type de llasticit . in j .- boehler , editor , _ colloque int . cnrs 295 , villard de lans _ , pages 93104 . martinus nijhoff publishers and editions du cnrs , 1982 , 1979 .

given a unitary representation of a lie group on a hilbert space , we develop the theory of-invariant self - adjoint extensions of symmetric operators both using von neumann s theorem and the theory of quadratic forms . we also analyze the relation between the reduction theory of the unitary representation and the reduction of the-invariant unbounded operator . we also prove a-invariant version of the representation theorem for quadratic forms . the previous results are applied to the study of-invariant self - adjoint extensions of the laplace - beltrami operator on a smooth riemannian manifold with boundary on which the group acts . these extensions are labeled by admissible unitaries acting on the-space at the boundary and having spectral gap at . it is shown that if the unitary representation of the symmetry group is traceable , then the self - adjoint extension of the laplace - beltrami operator determined by is-invariant if and commute at the boundary . various significant examples are discussed at the end .

1402.5537

symmetries of quantum mechanical systems are described by a group of transformations that preserves its essential structures . they play a fundamental role in studying the properties of the quantum system and reveal fundamental aspects of the theory which are not present neither in the dynamics involved nor in the forces . space or time symmetries , internal symmetries , the study of invariant states or spontaneously broken symmetries are standard ingredients in the description of quantum theories . in many cases quantum numbers or superselection rules are labels characterizing representations of symmetry groups . the publication of the seminal books of weyl , wigner and van der waerden ( cf . @xcite ) in the late twenties also indicates that quantum mechanics was using group theoretical methods almost from its birth . we refer , e.g. , to ( * ? ? ? * chapter 12 ) or @xcite for a more thorough introduction to various symmetry notions in quantum mechanics . it was shown by wigner that any symmetry transformation of a quantum system preserving the transition probabilities between two states must be implemented by a semi - unitary ( i.e. , by a unitary or an anti - unitary ) operator ( see , e.g. , ( * ? ? ? * introduction ) or ( * ? ? ? * chapters 2 ) ) . the action of a symmetry group @xmath0 on a system is given in terms of a semi - unitary projective representation of @xmath0 on the physical hilbert space , that can be described in terms of semi - unitary representations of @xmath6-central extensions of the group or by means of an appropriate representation group ( see , for instance , @xcite ) . since the main examples of symmetries considered in this article will be implemented in terms of unitary operators we will restrict here to this case . moreover , anti - unitary representation appear rarely in applications ( typically implementing time reversal ) and restrict to discrete groups . the situation with an anti - unitary representation of a discrete symmetry group can also be easily incorporated in our approach . in order to motivate how the symmetry can be implemented at the level of unbounded operators , consider a self - adjoint hamiltonian @xmath7 on the hilbert space @xmath8 and let @xmath9 be the strongly continuous one - parameter group implementing the unitary evolution of the quantum system . then , if @xmath0 is a quantum symmetry represented by the unitary representation @xmath10 it is natural to require that @xmath5 and @xmath2 commute , i.e. , @xmath11 at the level of self - adjoint generators and , recalling that the domain of @xmath7 is given by @xmath12 we have that ( [ commute ] ) implies @xmath13 of course , the requirement that the unitary representation @xmath5 of the symmetry group @xmath0 commutes with the dynamics of the system as in eq . ( [ commute ] ) is restrictive . for example , if @xmath5 is a strongly continuous representation of a lie group , then ( [ commute ] ) implies the existence of conserved quantities that do not depend explicitly on time . nevertheless , the previous comments justify that in the context of a single unbounded symmetric operator @xmath7 ( not necessarily a hamiltonian ) it is reasonable to define @xmath0-invariance of @xmath7 as in eq . ( [ eq : first - inv ] ) ( see section [ sec : vonneumannsymmetry ] for details ) . in the study of quantum systems it is standard that some heuristic arguments suggest an expression for an observable which is only symmetric on an initial dense domain but not self - adjoint . the description of such systems is not complete until a self - adjoint extension of the operator has been determined , e.g. , a self - adjoint hamiltonian operator @xmath7 . only in this case a unitary evolution of the system is given . this is due to the one - to - one correspondence between densely defined self - adjoint operators and strongly continuous one - parameter groups of unitary operators @xmath14 provided by stone s theorem . the specification of a self - adjoint extension is typically done by choosing suitable boundary conditions and this corresponds to a global understanding of the system ( see , e.g. , @xcite and references therein ) . accordingly , the specification of the self - adjoint extension is not just a mathematical technicality , but a crucial step in the description of the observables and the dynamics of the quantum system ( see , e.g. , ( * ? ? ? * chapter x ) for further results and motivation ) . we refer also to @xcite for recent textbooks that address systematically the problem of self - adjoint extension from different points of view ( see also the references therein ) . the question of how does the process of selecting self - adjoint extensions of symmetric operators intertwine with the notion of quantum symmetry arises . this question is at the focus of our interest in this article . we provide here natural characterizations of those self - adjoint extensions that are compatible with the given symmetries . concretely , if a symmetric operator is @xmath0-invariant in the sense of eq . ( [ eq : first - inv ] ) , then it is clear that not all self - adjoint extensions of the operator will also be @xmath0-invariant . this is evident if one fixes the self - adjoint extension by selecting boundary conditions . in general , these conditions need not preserve the underlying symmetry of the system . we present in section [ sec : vonneumannsymmetry ] and section [ sec : invariantqf ] the characterization of @xmath0-invariant self - adjoint extensions from two different point of views : first , in the most general context of deficiency spaces provided by von neumann s theorem . second , using the representation theorem of quadratic forms in terms of self - adjoint operators . we prove in theorem [ qainvariant ] a @xmath0-invariant version of the representation theorem for quadratic forms . in section [ sec : reduction ] we give an alternative notion of @xmath0-invariance in terms of the theory of von neumann algebras ( cf . , proposition [ pro : affiliate ] ) . we relate also here the irreducible sub - representations of @xmath5 with the reduction of the corresponding @xmath0-invariant self - adjoint extension @xmath7 . in particular , we show that if @xmath7 is unbounded and @xmath0-invariant , then the group @xmath0 must act on the hilbert space via a highly reducible representation @xmath5 . finally , we apply the theory developed to a large class of self - adjoint extensions of the laplace - beltrami operator on a smooth , compact manifold with smooth boundary on which a group is represented with a traceable unitary representation ( see definition [ def : traceable ] ) . in particular , self - adjoint extension of the laplace - beltrami operator with respect to groups acting by isometries on the manifold are discussed . in this context the extensions are labeled by suitable unitaries on the boundary of the manifold ( see @xcite for details ) . concrete manifolds like a cylinder or a half - sphere with @xmath15 or @xmath16 actions , respectively , will also be analyzed . apart from the previous considerations there are many instances where , though only partially , the previous problem has been considered . just to mention a few here we refer to the analysis of translational symmetries and the study of self - adjoint extensions of the laplacian in the description of a scalar quantum field in 1 + 1 dimensions in a cavity @xcite . in a different vein we quote the spectral analysis of hamiltonians in concentric spherical shells where the spherical symmetry is used in a critical way @xcite . in an operator theoretic context we refer , for example , to the notion of periodic weyl - titchmarsh functions or invariant operators with respect to linear - fractional transformations @xcite . even from a purely geometric viewpoint we should mention the analysis of isospectral manifolds in the presence of symmetries @xcite . we also refer to ( * ? ? ? * section 13.5 ) for the analysis of self - adjoint extensions commuting with a conjugation . this article is organized as follows : in section [ sec : preliminaries ] we summarize well - known results on the theory of self - adjoint extensions , including the theory of scales of hilbert spaces . in the next section we introduce the main definitions concerning @xmath0-invariant operators and give an explicit characterization of @xmath0-invariant self - adjoint extensions in the most general setting , i.e. , using the abstract characterization due to von neumann @xcite . in section [ sec : invariantqf ] we introduce the notion of @xmath0-invariant quadratic forms and show that the self - adjoint operators representing them will also be @xmath0-invariant operators . in the following section we present first steps of a reduction theory for @xmath0-invariant self - adjoint operators . for this we use systematically the notion of an unbounded operator affiliated to a von neumann algebra . in section [ sec : invariantlb ] we analyze the quadratic forms associated to the laplace - beltrami operator when there is a lie group acting on the manifold . thus , we provide a characterization of the self - adjoint extensions of the laplace - beltrami operator that are @xmath0-invariant . * notation * : in this article all unbounded , linear operators @xmath7 that act on a separable , complex hilbert space @xmath8 are densely defined and we denote the corresponding domain by @xmath17 . for convenience of the reader and to fix our notation we will summarize here some standard facts on the theory of self - adjoint extensions of symmetric operators , representation theorems for quadratic forms and the theory of rigged hilbert spaces . we refer to standard references , e.g. , @xcite , for proofs , further details and references . let @xmath7 be an unbounded , linear operator on the complex , separable hilbert space @xmath8 and with dense domain @xmath17 . recall that the operator @xmath7 is called symmetric if @xmath18 moreover , @xmath7 is self - adjoint if it is symmetric and @xmath19 , where the domain of the adjoint operator @xmath20 is the set of all @xmath21 such that there exists @xmath22 with @xmath23 in this case we define @xmath24 . if @xmath7 is symmetric then @xmath25 is a closed extension of @xmath7 , @xmath26 , i.e. , @xmath27 and @xmath28 the relation between self - adjoint and closed , symmetric operators is subtle and extremely important , specially from the physical point of view . it is thus natural to ask if given a symmetric operator one can find a closed extension of it that is self - adjoint and whether or not it is unique . von neumann addressed this issue in the late 20s and answered the question in an abstract setting , cf . , @xcite . we recall the main definition and results needed later ( see ( * ? ? ? * theorem x.2 ) ) . [ def : deficiencyspaces ] let @xmath7 be a closed , symmetric operator . the * deficiency spaces * @xmath29 are defined to be @xmath30 the * deficiency indices * are @xmath31 [ thmvonneumann ] let @xmath7 be a closed , symmetric operator . the self - adjoint extensions of @xmath7 are in one - to - one correspondence with the set of unitaries ( in the usual inner product ) of @xmath32 onto @xmath33 . if @xmath34 is such a unitary then the corresponding self - adjoint operator @xmath35 has domain @xmath36 and @xmath37 * the preceding definition and theorem can be also stated without assuming that the symmetric operator @xmath7 is closed ( see , e.g. , ( * ? ? ? * section xii.4 ) ) . in view of corollary [ cor : gclosure ] and that in the context of von neumann algebras of section [ sec : reduction ] the closure of @xmath7 is essential , we make this simplifying assumption here . * we refer to @xcite for a recent article that characterizes the class of all self - adjoint extensions of the symmetric operator obtained as a restriction of a self - adjoint operator to a suitable subspace of its domain . in particular , the explicit relation of the techniques used to the classical result by von neumann is also worked out . finally , we recall that the densely defined operator @xmath38 is * semi - bounded from below * if there is a constant @xmath39 such that @xmath40 the operator @xmath7 is * positive * if the lower bound satisfies @xmath41 . note that semi - bounded operators are automatically symmetric . in this section we introduce the notion of closed and closable quadratic forms . standard references are , e.g. , ( * ? ? ? * chapter vi ) , ( * ? ? ? * section viii.6 ) or ( * ? ? ? * section 4.4 ) . let @xmath42 be a dense subspace of the hilbert space @xmath8 and denote by @xmath43 a sesquilinear form ( anti - linear in the first entry and linear in the second entry ) . the quadratic form associated to @xmath44 with domain @xmath45 is its evaluation on the diagonal , i.e. , @xmath46 , @xmath47 . the sesquilinear form is called * hermitean * if @xmath48 the quadratic form is * semi - bounded from below * if there is an @xmath49 such that @xmath50 the smallest possible value @xmath51 satisfying the preceding inequality is called the * lower bound * for the quadratic form @xmath44 . in particular , if @xmath52 for all @xmath47 then we call @xmath44 * positive*. note that if @xmath44 is semi - bounded with lower bound @xmath51 , then @xmath53 is positive on the same domain . we need to recall also the notions of closable and closed quadratic forms as well as the fundamental representation theorems that relate closed , semi - bounded quadratic forms with self - adjoint , semi - bounded operators . let @xmath44 be a semi - bounded quadratic form with lower bound @xmath49 and dense domain @xmath54 . the quadratic form @xmath44 is * closed * if @xmath42 is closed with respect to the norm @xmath55 if q is closed and @xmath56 is dense with respect to the norm @xmath57 , then @xmath58 is called a * form core * for @xmath44 . conversely , the closed quadratic form @xmath44 with domain @xmath42 is called an * extension * of the quadratic form @xmath44 with domain @xmath58 . a quadratic form is said to be * closable * if it has a closed extension . @xmath59[remclosable ] 1 . the norm @xmath57 is induced by the following inner product on the domain : @xmath60 2 . it is always possible to close @xmath61 with respect to the norm @xmath57 . the quadratic form is closable iff this closure is a subspace of @xmath8 . the following representation theorem shows the deep relation between closed , semi - bounded quadratic forms and self - adjoint operators . this result goes back to the pioneering work in the 50ies by friedrichs , kato , lax and milgram , and others ( see , e.g. , comments to section viii.6 in @xcite ) . the representation theorem can be extended to the class of sectorial forms and operators ( see ( * ? ? ? * section vi.2 ) ) , but we will only need here its version for self - adjoint operators . [ fundteo ] let @xmath44 be an hermitean , closed , semi - bounded quadratic form defined on the dense domain @xmath54 . then it exists a unique , self - adjoint , semi - bounded operator @xmath7 with domain @xmath62 and the same lower bound such that 1 . @xmath63 iff @xmath64 and it exists @xmath65 such that @xmath66 in this case we write @xmath67 and @xmath68 for any @xmath69,@xmath70 . 2 . @xmath62 is a core for @xmath44 . following ( * ? ? ? * theorem 4.4.2 ) we get the following characterization of representable quadratic forms . [ pro : rep ] let @xmath44 be a semi - bounded , quadratic form with lower bound @xmath51 and domain @xmath42 . then the following conditions are equivalent : 1 . there is a lower semi - bounded operator @xmath7 with lower bound @xmath51 representing the quadratic form @xmath44 . the domain @xmath42 is complete with respect to the norm @xmath57 . one of the most common uses of the representation theorem is to obtain self - adjoint extensions of symmetric , semi - bounded operators . given a semi - bounded , symmetric operator @xmath7 one can consider the associated quadratic form @xmath71 these quadratic forms are always closable , cf . , ( * ? ? ? * theorem x.23 ) , and therefore their closure is associated to a unique self - adjoint operator . even if a symmetric operator has uncountably many possible self - adjoint extensions , the representation theorem above allows to select a particular one given a suitable quadratic form . this extension is called the friedrichs or _ hard _ extension and is in a natural sense maximal ( see chapters 10 and 13 in @xcite for a relation to krein - von neumann or _ soft _ extensions ) . the approach that we shall take in section [ sec : invariantqf ] and section [ sec : invariantlb ] uses this kind of friedrichs type extension . the theory of scales of hilbert spaces , also known as theory of rigged hilbert spaces , has been used in many ways in mathematics and mathematical physics . one of the standard applications of this theory appears in the proof of the representation theorems mentioned above . we state next the main results , ( see ( * ? ? ? * chapter i ) , ( * ? ? ? * chapter 2 ) for proofs and more results ) . + let @xmath8 be a hilbert space with scalar product @xmath72 and induced norm @xmath73 . let @xmath74 be a dense , linear subspace of @xmath8 which is a complete hilbert space with respect to another scalar product that will be denoted by @xmath75 . the corresponding norm is @xmath76 and we assume that @xmath77 any vector @xmath78 generates a continuous linear functional @xmath79 as follows . for @xmath80 define @xmath81 continuity follows by the cauchy - schwartz inequality and eq . . since @xmath82 is a continuous linear functional on @xmath74 it can be represented , according to riesz theorem , using the scalar product in @xmath74 . namely , it exists a vector @xmath83 such that @xmath84 and the norm of the functional coincides with the norm in @xmath74 of the element @xmath85 . one can use the above equalities to define an operator @xmath86 this operator is clearly injective since @xmath74 is a dense subset of @xmath8 and therefore it can be used to define a new scalar product on @xmath8 @xmath87 the completion of @xmath8 with respect to this scalar product defines a new hilbert space , @xmath88 , and the corresponding norm will be denoted accordingly by @xmath89 . it is clear that @xmath90 , with dense inclusions . since @xmath91 , the operator @xmath92 can be extended by continuity to an isometric bijection . [ def : scales ] the hilbert spaces @xmath74 , @xmath8 and @xmath88 introduced above define a * scale of hilbert spaces*. the extension by continuity of the operator @xmath92 is called the * canonical isometric bijection*. it is denoted by : @xmath93 [ proppairing ] the scalar product in @xmath8 can be extended continuously to a pairing @xmath94 let @xmath78 and @xmath80 . using the cauchy - schwartz inequality we have that @xmath95 . we begin now analyzing the question of how the process of finding a self - adjoint extension of a symmetric operator intertwines with the notion of a quantum symmetry . we will denote by @xmath0 a group and let @xmath96 be a fixed unitary representation of @xmath0 on the complex , separable hilbert space @xmath1 . we will introduce the notion of @xmath0-invariance by which we mean invariance under the fixed representation @xmath5 . [ def : g - inv ] let @xmath7 be a linear operator with dense domain @xmath97 and consider a unitary representation @xmath98 . the operator @xmath7 is said to be * @xmath0-invariant * if @xmath99 , i.e. , if @xmath100 for all @xmath101 and @xmath102 due to the invertibility of the unitary operators representing the group we have the following immediate consequence on @xmath0-invariant subspaces @xmath103 of @xmath8 which we will use several times : @xmath104 [ adjointinvariant ] let @xmath105 be a g - invariant , symmetric operator . then the adjoint operator @xmath106 is @xmath0-invariant . let @xmath107 . then , according to the definition of adjoint operator there is a vector @xmath22 such that @xmath108 using the @xmath0-invariance we have @xmath109 the preceding equalities hold for any @xmath110 and therefore @xmath111 . moreover , we have that @xmath112 [ cor : gclosure ] let @xmath105 be a @xmath0-invariant and symmetric operator on @xmath1 . then its closure @xmath113 is also @xmath0-invariant . the operator @xmath7 is symmetric and , therefore , closable . from @xmath114 and since @xmath7 is @xmath0-invariant , we have by the preceding proposition that @xmath25 is @xmath0-invariant , hence also @xmath115 . the preceding result shows that we can always assume without loss of generality that the @xmath0-invariant symmetric operators are closed . we begin next with the analysis of the @xmath0-invariance of the self - adjoint extensions given by von neumann s classical result ( cf . , theorem [ thmvonneumann ] ) . [ cor : invariantdeficiency ] let @xmath105 be a closed , symmetric and @xmath0-invariant operator . then , the deficiency spaces @xmath29 , cf . , definition [ def : deficiencyspaces ] , are invariant under the action of the group , i.e. , @xmath116 let @xmath117 . then @xmath118 and we have from proposition [ adjointinvariant ] that @xmath119 this shows that @xmath120 for all @xmath121 and by ( [ subsetequal ] ) we get the equality . similarly for @xmath33 . [ ginvariantoperator ] let @xmath105 be a closed , symmetric and @xmath0-invariant operator with equal deficiency indices ( cf . , definition [ def : deficiencyspaces ] ) . let @xmath35 be the self - adjoint extension of @xmath7 defined by the unitary @xmath122 . then @xmath35 is @xmath0-invariant iff @xmath123 for all @xmath124 , @xmath101 . to show the direction `` @xmath125 '' recall that by theorem [ thmvonneumann ] the domain of @xmath35 is given by @xmath126 let @xmath70 and @xmath127 . then we have that @xmath128 by assumption @xmath129 , and by corollary [ cor : invariantdeficiency ] , @xmath130 . hence @xmath131 . moreover , we have that for @xmath132 @xmath133 where we have used proposition [ adjointinvariant ] . to prove the reverse implication `` @xmath134 '' suppose that we have the self - adjoint extension defined by the unitary @xmath135 if we consider the domain @xmath136 defined by this unitary we have that @xmath137 where we have used again proposition [ adjointinvariant ] , corollary [ cor : invariantdeficiency ] and ( [ subsetequal ] ) . now von neumann s result stated in theorem [ thmvonneumann ] establishes a one - to - one correspondence between isometries @xmath122 and self - adjoint extensions of the operator @xmath7 . therefore @xmath138 and the statement follows . as mentioned in the first two sections the relation between closed , semi - bounded quadratic forms and self - adjoint operators is realized through the so - called representation theorems . we present here a notion of @xmath0-invariant quadratic form and prove a representation theorem for @xmath0-invariant structures . let @xmath44 be a quadratic form with domain @xmath42 and let @xmath139 be a unitary representation of the group @xmath0 . we will say that the quadratic form is * @xmath0-invariant * if @xmath140 for all @xmath101 and @xmath141 it is clear by the polarization identity that if the quadratic form @xmath44 is @xmath0-invariant , then the associated sesquilinear form also satisfies @xmath142 , @xmath101 . we will now relate the notions of @xmath0-invariance for self - adjoint operators and for quadratic forms . [ qainvariant ] let @xmath44 be a closed , semi - bounded quadratic form with domain @xmath42 and let @xmath7 be the representing semi - bounded , self - adjoint operator . the quadratic form @xmath44 is @xmath0-invariant iff the operator @xmath7 is @xmath0-invariant . to show the direction `` @xmath134 '' recall from theorem [ fundteo ] that @xmath70 iff @xmath143 and there exists @xmath22 such that @xmath144 then , if @xmath70 , and using the @xmath0-invariance of the quadratic form , we have that @xmath145 this implies that @xmath129 and from @xmath146 we show the @xmath0-invariance of the self - adjoint operator @xmath7 . for the reverse implication `` @xmath125 '' we use the fact that @xmath62 is a core for the quadratic form . for @xmath147 we have that @xmath148 these equalities show that the @xmath0-invariance of @xmath44 is true at least for the elements in the domain of the operator . now for any @xmath143 there is a sequence @xmath149 such that @xmath150 . this , together with the equality above , implies that @xmath151 is a cauchy sequence with respect to @xmath152 . since @xmath44 is closed , the limit of this sequence is in @xmath42 . moreover it is clear that @xmath153 , hence @xmath154 so far we have proved that @xmath140 . now for any @xmath155 consider sequences @xmath156 that converge respectively to @xmath155 in the topology induced by @xmath152 . then the limit @xmath157 concludes the proof . the preceding result and theorem [ pro : rep ] allow to give the following characterization of representable @xmath0-invariant quadratic forms . [ invariantrepresentation ] let @xmath44 be a @xmath0-invariant , semi - bounded quadratic form with lower bound @xmath51 and domain @xmath42 . the following statements are equivalent : 1 . there is a @xmath0-invariant , self - adjoint operator @xmath7 on @xmath17 with lower semi - bound @xmath51 and that represents the quadratic form , i.e. , @xmath158 2 . the domain @xmath42 of the quadratic form is complete in the norm @xmath152 . to conclude this section we make contact with the theory of scales of hilbert spaces introduced in section [ sec : scaleshs ] let @xmath44 be a closed , semi - bounded quadratic form with domain @xmath54 . we will show that if @xmath44 is @xmath0-invariant then one can automatically produce unitary representations @xmath159 on the natural scale of hilbert spaces @xmath160 where @xmath161 . [ thm : unit - scales ] let @xmath44 be a closed , semi - bounded , @xmath0-invariant quadratic form with lower bound @xmath51 . then 1 . @xmath5 restricts to a unitary representation @xmath162 on @xmath163 with scalar product given by @xmath164 2 . @xmath5 extends to a unitary representation @xmath165 on @xmath166 and we have , on @xmath166 , @xmath167 where @xmath168 is the canonical isometric bijection of definition [ def : scales ] . \(i ) to show that the representation @xmath169 is unitary with respect to @xmath75 note that by definition of @xmath0-invariance of the quadratic form we have for any @xmath101 that @xmath170 and @xmath171 since any @xmath172 is invertible we conclude that @xmath5 restricts to a unitary representation on @xmath173 . \(ii ) to show that @xmath5 extends to a unitary representation @xmath165 on @xmath88 consider first the following representation of @xmath0 on @xmath88 : @xmath174 we show first that this representation is unitary : since @xmath165 is invertible it is enough to check the isometry condition using part ( i ) . indeed , for any @xmath175 we have @xmath176 the restriction of @xmath177 , @xmath101 , to @xmath8 coincides with @xmath172 . indeed , consider the pairing @xmath178 of proposition [ proppairing ] and let @xmath179 . then for all @xmath80 @xmath180 since @xmath177 is a bounded operator in @xmath88 and @xmath8 is dense in @xmath88 , @xmath177 is the extension of @xmath172 to @xmath88 . the aim of this section is to provide an alternative point of view for the notion of @xmath0-invariance of operators ( cf . , section [ sec : vonneumannsymmetry ] ) in terms of von neumann algebras . based on this approach we will address some reduction issues of the unbounded operator in terms of the reducibility of unitary representation @xmath5 implementing the quantum symmetry . recall that a von neumann algebra @xmath181 is a unital * -subalgebra of @xmath182 ( the set of bounded linear operators in @xmath8 ) which is closed in the weak operator topology . even if a von neumann algebra consists only of bounded linear operators acting on a hilbert space , this class of operator algebras can be related in a natural way to closed , unbounded and densely defined operators . in fact , von neumann introduced these algebras in 1929 and proved the celebrated bicommutant theorem , when he extended the spectral theorem to closed , unbounded normal operators in a hilbert space ( cf . , @xcite ) . since then , the notion of affiliation of an unbounded operator to an operator algebra has been applied in different situations ( see , e.g. , @xcite ) . let @xmath183 be a self - adjoint subset of @xmath182 , i.e. , if @xmath184 , then @xmath185 . we denote by @xmath186 the commutant of @xmath183 in @xmath182 , i.e. , the set of all bounded and linear operators on @xmath8 commuting with all operators in @xmath183 . it is a fact that @xmath186 is a von neumann algebra and that the corresponding bicommutant @xmath187 is the smallest von neumann algebra containing @xmath183 , i.e. , @xmath188 is the von neumann algebra generated by the set @xmath189 . we refer to sections 4.6 and 5.2 of @xcite for further details and proofs . the definition of commutant of a densely defined unbounded operator @xmath7 is more delicate since one has to take into account the domains . the following definition generalizes the notion of commutant mentioned before . [ def : com ] let @xmath190 be a closed , densely defined operator . the commutant of @xmath7 is given by @xmath191 i.e. , @xmath192 if @xmath193 . since @xmath7 is a closed operator we have that @xmath194 is a von neumann algebra in @xmath195 . we denote the von neumann algebra associated to the bounded components of @xmath7 as @xmath196 in particular , if @xmath7 is self - adjoint , the spectral projections of @xmath7 are contained in @xmath197 . [ def : affi ] the closed , densely defined operator @xmath190 is * affiliated to a von neumann algebra * @xmath198 ( and we write this as @xmath199 ) if @xmath200 there are different equivalent characterizations of the notion of affiliation : @xmath201 iff @xmath202 . in particular , this implies that @xmath203 , @xmath204 , i.e. , @xmath205 and @xmath206 for all @xmath204 , @xmath207 . if @xmath7 is an ( unbounded ) self - adjoint operator , then @xmath197 coincides with the von neumann algebra @xmath208 generated by the cayley transform of @xmath7 . recall that the cayley transform @xmath209 is a unitary that can be associated with @xmath7 . we conclude that @xmath210 iff @xmath211 . finally , note that if @xmath7 is a bounded operator then , @xmath212 is the von neumann algebra generated by @xmath213 and @xmath199 iff @xmath214 . in the following result we will give a useful characterization of @xmath0-invariance for symmetric operators in terms of the affiliation to the commutant of the quantum symmetry . [ pro : affiliate ] let @xmath190 be a closed , symmetric operator . then , @xmath7 is @xmath0-invariant iff @xmath215 , where @xmath216 is the von neumann algebra generated by @xmath217 ( i.e. , @xmath218 ) and @xmath219 its commutant . moreover , any @xmath0-invariant self - adjoint extension of @xmath7 is also affiliated to @xmath219 . if @xmath215 , then it is immediate that @xmath7 is @xmath0-invariant , since @xmath220 and therefore the generators @xmath217 of the von neumann algebra @xmath221 satisfy @xmath222 . this gives the @xmath0-invariance of @xmath7 ( cf . , definition [ def : com ] ) . to show the reverse implication assume that @xmath7 is @xmath0-invariant according to definition [ def : g - inv ] , i.e. , @xmath223 . from proposition [ adjointinvariant ] we also have that @xmath224 hence @xmath225 which implies that @xmath215 . the same argument shows that any @xmath0-invariant , self - adjoint extension of @xmath7 is also affiliated to @xmath219 . we begin next with the analysis of the relation between the reducing subspaces of the quantum symmetry @xmath5 and those of the self - adjoint operator @xmath7 defined on the dense domain @xmath226 . the following result and part ( ii ) of theorem [ thm : disjointirrep ] are straightforward consequences of schur s lemma . [ lem : irrep ] let @xmath190 be a self - adjoint operator . if @xmath7 is @xmath0-invariant with respect to a unitary , irreducible representation @xmath5 of @xmath0 on the hilbert space @xmath227 , then @xmath7 must be bounded and @xmath228 schur s lemma and the irreducibility of @xmath5 imply that @xmath229 . moreover , by proposition [ pro : affiliate ] and since the cayley transform is a unitary and @xmath230 we have @xmath231 the case @xmath232 is not possible since @xmath233 is an isometry of @xmath234 onto @xmath235 and @xmath226 is dense . therefore @xmath236 for some @xmath237 . this implies that for any @xmath207 we have @xmath238 since the right hand side of the previous equation is bounded we can extend the formula for @xmath7 to the whole hilbert space . to continue our analysis we have to define first in which sense an unbounded operator can be reduced by a closed subspace . roughly speaking , the reduction means that we can write @xmath7 as the sum of two parts : one acting on the reducing subspace and one acting on its orthogonal complement . the following definition generalizes the standard one for bounded operators and uses the notion of commutant of an unbounded operator as in definition [ def : com ] . let @xmath190 be a self - adjoint operator and @xmath239 be a closed subspace of the hilbert space @xmath8 . we denote by @xmath240 the orthogonal projection onto @xmath239 . the subspace @xmath239 ( or @xmath240 ) reduces @xmath7 if @xmath241 , i.e. , if @xmath242 and @xmath243 , @xmath110 . the previous definition implies that if @xmath239 is reducing for @xmath7 , then @xmath244 is also reducing and @xmath245 . moreover , the subspace @xmath239 is invariant in the sense that @xmath246 if @xmath7 is self - adjoint , then the spectral projections @xmath247 ( with @xmath248 borel on the spectrum @xmath249 ) reduce @xmath7 . [ thm : disjointirrep ] let @xmath0 be a group and consider a unitary , reducible representation @xmath5 which decomposes as @xmath250 where the sub - representations @xmath251 , @xmath252 are irreducible and mutually inequivalent . let @xmath190 be a self - adjoint and @xmath0-invariant operator with respect to the representation @xmath5 . then * any projection @xmath253 onto @xmath254 , @xmath252 , is central in @xmath216 , ( i.e. , @xmath255 ) , and reduces the operator @xmath7 , ( i.e. , @xmath256 ) . * if @xmath257 , then @xmath7 must be a bounded operator and there exist @xmath258 , @xmath252 , such that @xmath259 \(i ) since the @xmath251 s are all irreducible and mutually inequivalent it follows by schur s lemma that @xmath260 moreover , since any @xmath253 reduces @xmath5 it is immediate that the projections are central . to show that @xmath256 , @xmath252 , consider the spectral projections @xmath261 of @xmath7 and define for any @xmath262 the following positive finite measure on the borel sets of @xmath249 : @xmath263 now , any @xmath253 is central for @xmath216 and , by @xmath0-invariance , proposition [ pro : affiliate ] implies that @xmath264 for any borel set @xmath248 . therefore we have @xmath265 and this implies that @xmath266 . similarly , using the spectral theorem one can show that @xmath267 hence @xmath256 . \(ii ) since @xmath268 we have that the cayley transform is unitary and @xmath269 therefore , there is a @xmath270 , @xmath252 , such that @xmath271 as in the proof of lemma [ lem : irrep ] we exclude first the case @xmath272 , @xmath252 . if @xmath272 and since the projection @xmath253 is reducing we have for any @xmath273 that @xmath274 therefore @xmath275 and we can omit the @xmath276th - summand in the decomposition of @xmath7 . hence without loss of generality we can assume that @xmath258 , @xmath252 and a similar reasoning on each block as in lemma [ lem : irrep ] gives the formula for @xmath7 . part ( ii ) of the previous theorem says that any representation of @xmath5 implementing a quantum symmetry of an unbounded , self - adjoint operator must be highly reducible . note that only if @xmath277 may @xmath7 be unbounded . e.g. , consider the case where @xmath278 . in the particular case of a compact group acting on an infinite dimensional hilbert space , we know that the decomposition of @xmath5 into irreducible representations must have infinite irreducible components . in this sense the representations considered in the examples of the following sections are meaningful . the following remark and proposition [ prop : equivalentirrep ] show that this is so even if we consider equivalent irreducible representations . if the irreducible representations are not mutually inequivalent , then the corresponding projections need not be reducing . in fact , consider the example @xmath279 on @xmath280 with @xmath281 , i.e. , there is a unitary @xmath282 such that @xmath283 . then @xmath284 and @xmath285 this shows that @xmath286 and , in fact if @xmath5 is a quantum symmetry for @xmath7 , then @xmath253 need not be reducing for @xmath7 . it can be shown that @xmath7 must also be bounded in this later case . take into account that it is not assumed that the irreducible representations are finite dimensional . below we show this in the simple case that the representation @xmath5 is a composition of two equivalent representations . the generalization to a finite number of equivalent representations is straightforward . [ prop : equivalentirrep ] let @xmath0 be a group and consider a unitary , reducible representation @xmath5 which decomposes as a direct sum of two equivalent , irreducible representations . let @xmath287 be a @xmath0-invariant , self - adjoint operator with respect to the representation @xmath5 . then @xmath7 must be a bounded operator . by assumption we have that @xmath279 with @xmath288 where @xmath289 is the unitary operator representing the equivalence and @xmath280 . according to the previous remark we have that the cayley transform of the operator @xmath7 is @xmath290 moreover , since @xmath291 is a unitary operator , the coefficient matrix @xmath292 is a @xmath293 unitary matrix , i.e. , @xmath294 . therefore it exists a unique , unitary matrix @xmath295 that diagonalizes @xmath296 , i.e. , @xmath297 , @xmath298 . consider the unitary operator @xmath299 this unitary transformation satisfies that @xmath300 where the new block structure represents a different decomposition of @xmath301 . with respect to this decomposition there are associated two proper subspaces of @xmath291 with proper projections @xmath302 and @xmath303 . notice that these projections reduce @xmath7 . these projections do not coincide in general with @xmath240 and @xmath304 , the projections associated to the decomposition @xmath279 . the same arguments as in the proof of theorem [ thm : disjointirrep ] lead us to exclude the cases @xmath305 and we can consider that @xmath306 . hence @xmath7 is a bounded operator . the following result is a straightforward consequence of the main results in this section . let @xmath7 be an unbounded self - adjoint operator on a hilbert space @xmath1 which is @xmath0-invariant with respect to a unitary representation @xmath307 . then @xmath5 can not be a direct sum of finitely many irreducible representations . the present section is a first step to analyze the relation between the reduction theory of a quantum mechanical symmetry and the reduction of the unbounded @xmath0-invariant operator . in particular we consider only when @xmath5 decomposes as a direct sum of irreducible representations . this is enough for the applications we have in mind in the following sections , where mainly compact groups act as a quantum symmetry . for a systematic and general theory of reduction one has to address , among other things , the type decomposition of the von neumann algebras corresponding to the intertwiner spaces of the representation @xmath5 and the corresponding direct integral decomposition of the self - adjoint operator @xmath7 ( see , e.g. , @xcite ) . as an application of the previous results we analyze the class of self - adjoint extensions of the laplace - beltrami operator on a riemannian manifold introduced in @xcite according to their invariance properties with respect to a symmetry group , in particular with respect to a group action on the manifold . throughout the rest of this and the next section we will consider a smooth , compact , riemannian manifold with boundary @xmath308 . the boundary @xmath309 of the riemannian manifold @xmath308 has itself the structure of a riemannian manifold without boundary @xmath310 . the riemannian metric at the boundary is just the pull - back of the riemannian metric @xmath311 , where @xmath312 is the canonical inclusion map . the spaces of smooth functions over the manifolds verify that @xmath313 the sobolev spaces of order @xmath314 over the manifolds @xmath308 and @xmath310 are going to be denoted by @xmath315 and @xmath316 , respectively . there is an important relation between the sobolev spaces defined over the manifolds @xmath317 and @xmath309 . this is the well known lions trace theorem ( see , e.g. , ( * ? ? ? * theorem 7.39 ) and theorem 9.4 of chapter 1 in @xcite ) . [ lmtracetheorem ] let @xmath318 and let @xmath319 be the trace map @xmath320 . there is a unique continuous extension of the trace map such that 1 . @xmath321 , @xmath322 . the map is surjective . + we recall here some results from @xcite that describe a large class of self - adjoint extensions of the laplace - beltrami operator . the extensions are parameterized in terms of suitable unitaries on the boundary hilbert space . this class is constructed in terms of a family of closed , semi - bounded quadratic forms via the representation theorem ( cf . , theorem [ fundteo ] ) . before introducing this family we shall need some definitions . [ defgap ] let @xmath323 be unitary and denote by @xmath324 its spectrum . the unitary @xmath2 on the boundary * has spectral gap at @xmath4 * if one of the following conditions hold : 1 . @xmath325 is invertible . 2 . @xmath326 and @xmath4 is not an accumulation point of @xmath324 . the eigenspace associated to the eigenvalue @xmath4 is denoted by @xmath327 . the corresponding orthogonal projections will be written as @xmath328 and @xmath329 . [ partialcayley ] let @xmath2 be a unitary operator acting on @xmath330 with spectral gap at @xmath4 . the * partial cayley transform * @xmath331 is the operator @xmath332 [ def : admissible ] let @xmath2 be a unitary with spectral gap at @xmath4 . the unitary is said to be * admissible * if the partial cayley transform @xmath333 leaves the subspace @xmath334 invariant and is continuous with respect to the sobolev norm of order @xmath335 , i.e. , @xmath336 consider a manifold with boundary given by the unit circle , i.e. , @xmath337 , and define the unitary @xmath338 , @xmath339 . if @xmath340 and @xmath341\cup [ \pi+\delta,2\pi)$ ] , for some @xmath342 , then @xmath343 has gap at @xmath4 . if , in addition , @xmath344 , then @xmath343 is admissible . [ defqu ] let @xmath2 be a unitary with spectral gap at @xmath4 , @xmath333 the corresponding partial cayley transform and @xmath345 the trace map considered in theorem [ lmtracetheorem ] . the hermitean quadratic form @xmath346 associated to the unitary @xmath2 is defined by @xmath347 on the domain @xmath348 here @xmath349 stands for the canonical scalar product among one - forms on the manifold @xmath317 . it is worth to mention the reasons behind definitions [ defgap ] and [ def : admissible ] . the spectral gap condition ensures that the partial cayley transform becomes a bounded , self - adjoint operator on the subspace @xmath350 and this guarantees that the quadratic form @xmath346 is lower semi - bounded . notice that we are dealing with unbounded quadratic forms and thus they are not continuous mappings of the hilbert space . the admissibility condition is an analytic requirement to ensure that @xmath346 is a closable quadratic form . in the next theorem we give a class of self - adjoint extensions of the minimal laplacian operator @xmath351 on the domain @xmath352 . we refer to ( * ? ? ? * section 4 ) for a complete proof and additional motivation . all the extensions are labeled by suitable unitaries @xmath2 at the boundary . [ char_u ] let @xmath353 be an admissible unitary operator with spectral gap at @xmath4 . then the quadratic form @xmath346 of definition [ defqu ] is semi - bounded from below and closable . its closure is represented by a semi - bounded , self - adjoint extension of the minimal laplacian @xmath354 . we will use next the results of section [ sec : invariantqf ] to give necessary and sufficient conditions on the characterization of the unitary @xmath2 in order that the corresponding quadratic form @xmath355 is @xmath0-invariant . in particular , from theorem [ qainvariant ] we conclude that the self - adjoint operator representing its closure will also be @xmath0-invariant . we need to consider first the quadratic form corresponding to the neumann extension of the laplace - beltrami operator : @xmath356 we will also call it neumann quadratic form . note that it corresponds to the quadratic form @xmath346 of the previous section with admissible unitary @xmath357 . moreover , @xmath2 has spectral gap at @xmath4 and for the corresponding orthogonal projection we have @xmath358 , hence @xmath359 . let @xmath0 be a lie group and @xmath360 be a continuous unitary representation of @xmath0 , i.e. , for any @xmath361 the map @xmath362 is continuous in the @xmath363-norm . we will assume that @xmath364 is @xmath0-invariant , that is , @xmath365 and @xmath366 for all @xmath367 and @xmath101 . then the following lemma shows that @xmath5 defines also a continuous unitary representation on @xmath368 with its corresponding sobolev scalar product . let @xmath5 a strongly continuous unitary representation of the lie group @xmath0 on @xmath369 such that neumann s quadratic form @xmath364 is @xmath0-invariant . then @xmath5 leaves invariant the subspace @xmath368 and defines a strongly continuous unitary representation on it . since @xmath172 is invertible it is enough to show that @xmath172 is an isometry with respect to the sobolev norm @xmath370 ( see also the proof of theorem [ thm : unit - scales ] ) . but this is immediate since @xmath5 is unitary on @xmath369 and @xmath364 is @xmath0-invariant this is trivial if we realize that @xmath371 , then because of the @xmath0-invariance of @xmath364 we get that @xmath372 for all @xmath373 , @xmath121 . finally , to prove strong continuity on @xmath368 use the invariance @xmath374 and a standard weak compactness argument . [ def : traceable ] the representation @xmath375 * has a trace ( or that it is traceable ) along the boundary * @xmath376 if there exists another continuous , unitary representation @xmath377 such that @xmath378 for all @xmath367 and @xmath101 or , in other words , that the following diagram is commutative : @xmath379 we will call @xmath380 the trace of the representation @xmath5 . notice that if the representation @xmath5 is traceable , its trace @xmath380 is unique . now we are able to prove the following theorem : [ repcommutation ] let @xmath0 a lie group , and @xmath381 a traceable continuous , unitary representation of @xmath0 with unitary trace @xmath382 along the boundary @xmath383 . denote by @xmath384 the closable and semi - bounded quadratic form of definition [ defqu ] with an admissible unitary @xmath385 having spectral gap at @xmath4 . assume that the corresponding neumann quadratic form @xmath364 is @xmath0-invariant . then we have the following cases : * if @xmath386=0 $ ] for all @xmath101 , then @xmath346 is @xmath0-invariant . its closure is also @xmath0-invariant and the self - adjoint extension of the minimal laplacian representing the closed quadratic form will also be @xmath0-invariant . * consider the decomposition of the boundary hilbert space @xmath387 , where @xmath327 is the eigenspace associated to the eigenvalue @xmath4 of @xmath2 and denote by @xmath328 the orthogonal projection onto @xmath327 . if @xmath346 is @xmath0-invariant and @xmath388 continuous , then @xmath386=0 $ ] for all @xmath101 . \(i ) assume first that @xmath386=0 $ ] , @xmath101 . to show that @xmath346 is @xmath0-invariant we have to analyze first the block structure of @xmath2 and @xmath389 with respect to the decomposition @xmath390 . since @xmath391 the commutation relations imply that @xmath392 and @xmath393=0 $ ] . but since @xmath2 has spectral gap at @xmath4 , then @xmath394 is invertible on @xmath350 and we must have @xmath395 . the unitarity of @xmath389 implies @xmath396 and @xmath389 has block diagonal structure . by assumption @xmath364 is @xmath0-invariant , so it is enough to show that the boundary quadratic form @xmath397 is also @xmath0-invariant , i.e. , @xmath398 and @xmath399 , @xmath400 . to show the first inclusion , note that for any @xmath400 we have @xmath401 where we have used that @xmath5 is traceable along @xmath309 , @xmath389 has diagonal block structure and @xmath402 . finally , for any @xmath101 and @xmath400 we check @xmath403 note that all scalar products refer to @xmath350 and that for the last equation we used @xmath404 iff @xmath405 and , again , all commutants are taken with respect to @xmath350 ( cf . , section [ sec : reduction ] ) . since @xmath346 is @xmath0-invariant and closable and the representation @xmath5 unitary it is straightforward to show that its closure is also @xmath0-invariant ( see , e.g. , theorem vi.1.17 in @xcite ) . by theorem [ qainvariant ] it follows that the self - adjoint extension of the minimal laplacian representing the closed quadratic form will also be @xmath0-invariant . \(ii ) by assumption we have that @xmath406 is continuous in the fractional sobolev norm and , therefore , @xmath407 is dense in @xmath350 . since @xmath346 is @xmath0-invariant we have @xmath398 , hence for any @xmath400 @xmath408 from the density of @xmath407 in @xmath350 we conclude that @xmath409 and we only need to show @xmath393=0 $ ] on @xmath350 . but this follows from the @xmath0-invariance of the boundary quadratic form @xmath410 and , again , the density of @xmath407 in @xmath350 . we can now consider the following immediate consequences . [ cor:1 ] let @xmath411 be admissible and such that @xmath325 is invertible . then @xmath346 on @xmath412 is @xmath0-invariant iff @xmath386=0 $ ] for all @xmath101 . we only need to show that the @xmath0-invariance implies that the unitaries on the boundary commute . note that by assumption @xmath413 and , with the notation in the preceding theorem , we have that @xmath414 . therefore @xmath415 and the corresponding trace gives @xmath416 which is dense in @xmath330 . the rest of the reasoning is litteraly as in the proof of part ( ii ) of theorem [ repcommutation ] . we conlcude giving a characterization of @xmath0-invariance of the quadratic form @xmath346 that uses the point spectrum of the unitary @xmath2 . recall that @xmath417 is in the point spectrum of an operator @xmath7 if @xmath418 is not injective , i.e. , @xmath417 is an eigenvalue of @xmath7 . we denote the set of all eigenvalues by @xmath419 . [ lem:2 ] consider a unitary @xmath411 with spectral gap at @xmath4 . assume that @xmath334 is invariant for @xmath2 and that its restriction @xmath420 is continuous with respect to the sobolev @xmath335-norm . if @xmath421 has only point spectrum , i.e. , @xmath422 , then @xmath2 is admissible , i.e. , the partial cayley transform @xmath333 leaves the fractional sobolev space @xmath334 invariant and is continuous with respect to the sobolev @xmath335-norm . the orthogonal projection @xmath328 leaves also the sobolev space @xmath334 invariant and is continuous with respecto to the @xmath335-norm . note first that @xmath423 and , therefore , if @xmath2 has spectral gap at @xmath4 , then @xmath421 has also spectral gap at @xmath4 . then by cauchy - riesz functional calculus ( cf . , * chapter vii ) ) we have that @xmath424 where @xmath425 are closed , simple and positively oriented curves . the curve @xmath426 encloses @xmath427 and @xmath428 encloses only @xmath429 . note that the the gap condition is essential here . it is clear that both @xmath333 and @xmath328 are bounded operators in @xmath334 as required . [ cor:2 ] consider a unitary @xmath411 with spectral gap at @xmath4 . assume that @xmath334 is invariant for @xmath2 , that its restriction @xmath430 is continuous with respect to the sobolev @xmath335-norm and that @xmath422 . then @xmath346 on @xmath412 is @xmath0-invariant iff @xmath386=0 $ ] for all @xmath101 . by the preceding lemma we have that @xmath2 is admissible and that @xmath328 is continuous in the @xmath335-norm the statement follows then directly from theorem [ repcommutation ] . in the final section we will present examples with unitaries which satisfy the conditions mentioned in the statements above . we will discuss now the important instance when the unitary representation is determined by an action of the group @xmath0 on @xmath317 by isometries . thus , assume that the group @xmath0 acts smoothly by isometries on the riemannian manifold @xmath308 . any @xmath101 specifies a diffeomorphism @xmath431 that we will denote with the same symbol for simplicity of notation . moreover , we have that @xmath432 where @xmath433 stands for the pull - back by the diffeomorphism @xmath434 . these diffeomorphisms restrict to isometric diffeomorphisms on the riemannian manifold at the boundary @xmath310 ( see , e.g. , ( * ? ? ? * lemma 8.2.4 ) ) , @xmath435 these isometric actions of the group @xmath0 induce unitary representations of the group on @xmath317 and @xmath309 . in fact , consider the following representations : @xmath436 @xmath437 then a simple computation shows that , @xmath438 where we have used the change of variables formula and the fact that isometric diffeomorphisms preserve the riemannian volume , i.e. , @xmath439 . the result for the boundary is proved similarly . the induced actions are related with the trace map as @xmath440 and therefore the unitary representation @xmath5 is traceable along the boundary of @xmath317 with trace @xmath380 . moreover we have that the quadratic form @xmath364 is @xmath0-invariant . [ prop : dphiinvariant ] let @xmath0 be a lie group that acts by isometric diffeomorphisms on the riemannian manifold @xmath308 and let @xmath441 be the associated unitary representation . then , neumann s quadratic form @xmath442 with domain @xmath443 is @xmath0-invariant . first notice that the pull - back of a diffeomorphism commutes with the action of the exterior differential . then we have that @xmath444 hence [ dphiinvariant ] @xmath445 where in the second inequality we have used that @xmath431 is an isometry and therefore @xmath446 the equations guaranty also that @xmath447 since @xmath172 is a unitary operator in @xmath369 and the norm @xmath448 is equivalent to the sobolev norm of order 1 . before making explicit the previous structures in concrete examples we summarize the the main ideas in this section as follows : given a group acting by isometric diffeomorphisms on a riemannian manifold , then any operator at the boundary , that satisfies the conditions in definition [ defgap ] and definition [ def : admissible ] , and that verifies the commutation relations of theorem [ repcommutation ] ( i ) describes a @xmath0-invariant quadratic form . the closure of this quadratic form characterizes uniquely a @xmath0-invariant self - adjoint extension of the laplace - beltrami operator ( cf . theorems [ fundteo ] and [ qainvariant ] ) . in this section we introduce two particular examples of @xmath0-invariant quadratic forms . in the first example we are considering a situation where the symmetry group is a finite , discrete group . in the second one we consider @xmath0 to be a compact lie group . let @xmath317 be the cylinder @xmath449\times[-1,1]/\negthickspace\sim$ ] , where @xmath450 is the equivalence relation @xmath451 . the boundary @xmath309 is the disjoint union of the two circles @xmath452/\negthickspace\sim\bigr\}$ ] and @xmath453/\negthickspace\sim\bigr\}$ ] , ( see figure [ fig : cilindro ] ) . let @xmath454 be the euclidean metric on @xmath317 . now let @xmath455=\{e , f } be the discrete , abelian group of two elements and consider the following action in @xmath317 : @xmath456 the induced action at the boundary is @xmath457 clearly @xmath0 transforms @xmath317 onto itself and preserves the boundary . moreover , it is easy to check that @xmath458 . since the boundary @xmath309 consists of two disjoints manifolds @xmath459 and @xmath460 , the hilbert space of the boundary is @xmath461 . any @xmath462 can be written as @xmath463 with @xmath464 . the only nontrivial action on @xmath330 is given by @xmath465 the set of unitary operators that describe the closable quadratic forms as defined in the previous section is given by suitable unitary operators @xmath466 with @xmath467 . according to theorem [ repcommutation ] ( i ) the unitary operators commuting with @xmath468 will lead to @xmath0-invariant quadratic forms . imposing @xmath469 we get the conditions @xmath470 obviously there is a wide class of unitary operators , i.e. , boundary conditions , that will be compatible with the symmetry group @xmath0 . we will consider next two particular classes of boundary conditions . first , consider the following unitary operators @xmath471 where @xmath472\cup\{\pi\}\right)$ ] for some @xmath473 . it is showed in ( * ? ? ? * sections 3 and 5 ) that this class of unitary operators have spectral gap at -1 and are admissible ( see also subsection [ subsec:6 - 2 ] ) . moreover , this choice of unitary matrices corresponds to select robin boundary conditions of the form : @xmath474 the @xmath0-invariance condition above imposes @xmath475 . notice that when @xmath476 we can obtain meaningful self - adjoint extensions of the laplace - beltrami operator that , however , will not be @xmath0-invariant . we can also consider unitary operators of the form @xmath477 where @xmath478)$ ] . again , it is proved in @xcite that this class of unitary operators have spectral gap at @xmath4 and are admissible . in this case the unitary matrix corresponds to select so - called quasi - periodic boundary conditions , cf . , @xcite , i.e. , @xmath479 @xmath480 the @xmath0-invariance condition imposes @xmath481 and therefore among all the quasi - periodic conditions only the periodic ones , @xmath482 , are compatible with the @xmath0-invariance . [ ex : upperhemisphere ] let @xmath317 be the unit , upper hemisphere centered at the origin . its boundary @xmath309 is the unit circle on the @xmath483 plane . let @xmath454 be the induced riemannian metric from the euclidean metric in @xmath484 . consider the compact lie group @xmath485 that acts by rotation around the @xmath486-axis . if we use polar coordinates on the horizontal plane , then the boundary @xmath309 is isomorphic to the interval @xmath487 $ ] with the two endpoints identified . we denote by @xmath488 the coordinate parameterizing the boundary and the boundary hilbert space is @xmath489 . let @xmath490 and consider the action on the boundary by a group element @xmath491 , @xmath492 $ ] , given by @xmath493 to analyze what are the possible unitary operators that lead to @xmath0-invariant quadratic forms it is convenient to use the fourier series expansions of the elements in @xmath330 . let @xmath494 , then @xmath495 where the coefficients of the expansion are given by @xmath496 we can therefore consider the induced action of the group @xmath0 as a unitary operator on @xmath497 , the hilbert space of square summable sequences . in fact we have that : @xmath498 this shows that the induced action of the group @xmath0 is a unitary operator in @xmath499 that acts diagonally in the fourier series expansion . more concretely , we can represent it as @xmath500 . from all the possible unitary operators acting on the hilbert space of the boundary , only those whose representation in @xmath501 commutes with the above operator will lead to @xmath0-invariant quadratic forms ( cf . , theorem [ repcommutation ] ( i ) ) . since @xmath502 acts diagonally on @xmath501 it is clear that only operators of the form @xmath503 , @xmath504 , will lead to @xmath0-invariant quadratic forms . as a particular case we can consider that all the parameters are equal , i.e. , @xmath505 , @xmath506 . in this case it is clear that @xmath507 , which gives the following admissible unitary with spectral gap at @xmath4 : @xmath508 this shows that the unique robin boundary conditions compatible with the @xmath16-invariance are those that are defined with a constant parameter along the boundary , i.e. , @xmath509\;,\ ] ] where @xmath510 stands for normal vector field pointing outwards to the boundary . , _ quantum dynamics : from automorphisms to hamiltonians _ , in _ studies in mathematical physics . essays in honor of valentine bargman _ , e.h . lieb , b. simon and a.s . wightman eds . , princeton university press , princeton , 1976 .

recently , the temperature and luminosity of the hot gas halos of early type galaxies have been derived with unprecedented accuracy from data , for a sample of 30 galaxies covering a wider range of galactic luminosity ( and central velocity dispersion ) than before . this work investigates the origin of the observed temperatures , by examining the relationship between them and the galaxy structure , the gas heating due to type ia supernovae ( snia s ) and the gravitational potential , and the dynamical status of the gas flow . in galaxies with km s , the s are close to a fiducial average temperature for the gas when in outflow ; at 200(km s , the s are generally lower than this , and unrelated with , which requires a more complex gas flow status ; at larger , the s may increase as , as expected for infall heating , though heating from snia s , independent of , should be dominant . all observed s are larger than the virial temperature , by up to kev . this additional heating can be provided in the x - ray brightest galaxies by snia s and infall heating , with a snia s energy input even lower than in standard assumptions ; in the x - ray fainter ones it can be provided by snia s , whose energy input would be required close to the full standard value at the largest . this same energy input , though , would produce temperatures larger than observed at low , if entirely thermalized . the values of the observed s increase from outflows to inflows ; the gas is relatively hotter in outflows , though , if the s are rescaled by the virial temperature . for km s , lower values tend to correspond to lower s , which deserves further investigation .

1106.2898

the advent of the @xmath2 x - ray observatory , with its unprecedented sub - arcsecond resolution , allowed to study better than ever before the main contributors to the total x - ray emission of early - type galaxies ( hereafter etgs ) : the low - mass x - ray binaries ( lmxbs ; fabbiano 2006 ) , a population of weak sources as late type stellar coronae , cataclismic variables , and coronally active binaries ( pellegrini & fabbiano 1994 , revnivtsev et al . 2008 ) , the nuclear emission due to a supermassive black hole ( mbh ; e.g. , gallo et al . 2010 , pellegrini 2010 ) , and a hot interstellar medium ( ism ) with a temperature of a few million degrees . after careful subtraction of the stellar ( resolved and unresolved ) and nuclear emissions , the properties of the hot ism could be characterized with unprecedented accuracy . recently , this has been done for a sample of 30 normal ( non - cd ) etgs observed with @xmath2 to a depth ensuring the detection of bright lmxbs ( boroson et al . 2011 , hereafter bkf ) . this is the first x - ray sample of etgs covering a wide range of galactic luminosity , central velocity dispersion @xmath3 , and hot gas emission @xmath1 , and with the x - ray properties of the hot gas ( e.g. , luminosity @xmath1 and average temperature @xmath0 ) derived in a homogeneous way , using a complete and accurate procedure to subtract all kinds of non - gaseous emission ( nucleus , detected and undetected lmxbs , and unresolved weak stellar sources ) . this approach resulted in a larger fraction of hot gas - poor galaxies than in previous samples , with @xmath1 extending down to much lower values than before ( @xmath12 erg s@xmath5 ) , and showing a variation of up to @xmath13 orders of magnitude at the same galactic luminosity ( see also david et al . 2006 , diehl & statler 2007 , memola et al . such a wide variation , even larger than previously found , had been linked to the origin and evolution of the hot ism , and had provided evidence for the effectiveness of an internal heating mechanism ( as from type ia supernovae , hereafter snia s ) to regulate the gas evolution and produce its very different content in etgs at the present epoch ( loewenstein & mathews 1987 , david et al . 1990 , ciotti et al . 1991 ) ; the action of external agents ( gas stripping , confinement , accretion ) to reduce or enhance the gas content was also invoked ( e.g. , white & sarazin 1991 , brown & bregman 2000 , sun et al . 2007 ) . with this new characterization of the hot gas , bkf revisited the relationships between fundamental properties of the hot gas and of the host galaxy , as the @xmath14 , @xmath15 and @xmath16 relations , where @xmath3 is a representative measure of the depth of the galactic potential well ( eskridge et al . 1995 , osullivan et al . 2001 , 2003 ) . @xmath1 correlates positively with @xmath0 and @xmath3 , though with a wide variation at fixed @xmath8 and @xmath0 . interestingly , the best fit relation @xmath17 , close to what already known for x - ray luminous etgs ( osullivan et al . 2003 ) , is still moderately strong among etgs with low @xmath0 and @xmath1 ; also , in the @xmath15 relation , etgs with @xmath18 kev are the x - ray brightest ( with one exception ) , while those with @xmath19 kev are the x - ray faintest . the least gas rich etgs are then the coolest ones , which seemed contrary to expectation , if low @xmath1 etgs loose their ism in an outflow ( e.g. , david et al . 1990 , ciotti et al . 1991 ) , and the hotter the gas , the stronger is the outflow ( bkf ) . on average @xmath0 increases with @xmath3 , and most etgs lie above a rough estimate of the gas virial temperature ( @xmath20 ) , suggesting the presence of additional heating . etgs with a moderate to high gas content ( @xmath21 erg s@xmath5 ) follow a trend roughly parallel to that of @xmath22 ; instead , etgs with little hot gas ( @xmath23 erg s@xmath5 ) have a similar temperature for @xmath3 ranging from 160 to 250 km s@xmath5 . this lack of correlation was attributed to a different dynamical state of the hot ism in gas - poor with respect to gas - rich etgs , though a full explanation of this aspect remained to be found ( bkf ) . this work takes advantage of the new accurate measurements of the hot gas properties , and of the fundamental relations @xmath15 and @xmath16 , derived down to galaxy masses and x - ray luminosities smaller than ever before ( bkf ) , to investigate the relationship between @xmath0 , the galaxy structure , the internal gas heating mechanisms ( snia s , and those linked to the gravitational potential ) , and the dynamical status of the gas flow . to this purpose , a few characteristic temperatures are introduced , depending on the nature of the gas heating sources and the galaxy structure , and relevant for the various gas flow phases ; these characteristic temperatures are then compared with the observed @xmath0 values . in doing so , galaxy mass models are built according to the most recent understanding of the etgs structure , such as their stellar mass profile and their dark matter content and distribution , as indicated by detailed modeling of optical observations and by the main scaling laws ( e.g. , cappellari et al . 2006 , weijmans et al . 2009 , auger et al . 2010 , napolitano et al . 2010 , shen & gebhardt 2010 ) . the aims are to address the following questions : can the gas heating sources above account for the observed @xmath0 s ? how are the various input energy sources for the gas used in the different flow phases ? is there any relation between @xmath0 and the flow phase ? we present in sect . [ heat ] the sources of mass and heating for the hot ism , in sect . [ esc ] the conditions for the gas to escape from the galaxy , in sect . [ mass ] the galaxy mass models , in sect . [ disc ] the comparison between observed and predicted temperatures , in sect . [ temp ] the relation between gas temperature and flow status , and in sect . [ concl ] the conclusions . in etgs the hot gas comes from stellar mass losses produced by evolved stars , mainly during the red giant , asymptotic giant branch , and planetary nebula phases , and by snia s , that are the only ones observed in an old stellar population ( e.g. , cappellaro et al . 1999 ) . the first , more quiescent , type of losses originates ejecta that initially have the velocity of the parent star , then individually interact with the mass lost from other stars or with the hot ism , and mix with it ( mathews 1990 , parriott & bregman 2008 ) . for a galaxy of total stellar mass @xmath24 , the evolution of the stellar mass loss rate @xmath25 can be calculated using single burst stellar population synthesis models ( maraston 2005 ) , for a salpeter and for a kroupa initial mass function ( imf ) , assuming for example solar abundance . so doing , at an age of 12 gyrs , a rate is recovered of @xmath26= b @xmath27yr@xmath5 , where @xmath28 is the galactic b - band luminosity at an age of 12 gyr , and b=1.8 or b=1.9 for the salpeter or kroupa imf ( see also pellegrini 2011 ) . this value is in reasonable agreement with the average derived for nine local etgs from @xmath29 data ( athey et al . 2002 ) of @xmath30 m@xmath31yr@xmath5 , an estimate based on individual observed values that vary by a factor of @xmath32 , though , which was attributed to different ages and metallicities . the total mass loss rate of a stellar population @xmath33 is given by the sum @xmath34 , where @xmath35 is the rate of mass loss via snia events for the whole galaxy . @xmath36 is given by @xmath37 , where @xmath38 is the ejected mass by one event , and @xmath39 is the explosion rate . @xmath40 has been determined for local etgs to be @xmath41 , where @xmath42 is the hubble constant in units of km s@xmath5 mpc@xmath5 ( cappellaro et al . 1999 ) . more recent measurements of the observed rates of supernovae in the local universe ( li et al . 2010 ) give a snia s rate in etgs consistent with that of cappellaro et al . ( 1999 ) . for @xmath43 km s@xmath5 mpc@xmath5 , one obtains @xmath44 = 2.2@xmath45 m@xmath31 yr@xmath5 , that is @xmath46 times smaller than @xmath26 derived above for an age of 12 gyr ; therefore , the main source of mass for the hot gas is provided by @xmath47 . a reasonable assumption is that the gas is shed by stars with a radial dependence that follows that of the stellar distribution , so that the density profile of the injected gas is @xmath48 , where @xmath49 is the stellar density profile . this assumption is adopted hereafter , and the characteristic temperatures presented below apply to a gas distribution following @xmath48 ( but see also sect . [ disc ] ) . the material lost by stars is ejected at a velocity of few tens of km s@xmath5 and at a temperature of @xmath50 k ( parriott & bregman 2008 ) , and is subsequently heated to high , x - ray emitting temperatures by the thermalization of the stellar velocity dispersion , as it collides with the mass lost from other stars , or with the ambient hot gas , and is shocked . another source of heating for the stellar mass losses is provided by the thermalization of the kinetic energy of snia s events . the internal energy given by these heating processes to the unit mass of injected gas is @xmath51 ( with @xmath52 the boltzmann constant , @xmath53 the proton mass , @xmath54 the mean particle mass , with @xmath55 for solar abundance ) ; @xmath56 is determined by the heating due to thermalization of the motions of the gas - losing stars ( @xmath57 ) , and of the velocity of the snia s ejecta ( @xmath58 ) , and is written as ( e.g. , gisler 1976 , white & chevalier 1983 ) : @xmath59 here @xmath60 is the equivalent temperature of the stellar motions ( see below ) , and @xmath61 is the equivalent temperature of the kinetic energy @xmath62 of the snia s ejecta , with @xmath63 erg for one event ( e.g. , larson 1974 ) . @xmath64 can be calculated assuming that a factor @xmath65 of @xmath62 is turned into heat ; @xmath66 , since radiative energy losses from expanding supernova remnants may be important , and values down to @xmath67 have been adopted ( larson 1974 , chevalier 1974 ) ; a value of @xmath68 could be not too far off for the hot diluted ism of etgs ( e.g. , tang & wang 2005 ) . in this way , @xmath69k . from approximating @xmath70 , and using the estimates of sect . [ loss ] for @xmath71 and @xmath47 ( for a kroupa imf ) at the present epoch , one obtains @xmath72 k. the injection temperature @xmath56 is then the sum of two parts : one ( @xmath58 ) is independent of the position within the galaxy where the gas is injected ( e.g. , independent of radius in spherical symmetry ) , and is also constant from galaxy to galaxy ( for fixed imf and age of the stellar population , and snia s rate ) ; for each etg it can , though , evolve with time , if @xmath71 and @xmath47 evolve differently with time ( ciotti et al . 1991 ) . the other part ( @xmath57 ) is instead basically independent of time , but has a radial dependence , and changes with the galaxy structure , i.e. , with the total mass and its distribution . an average @xmath73 is obtained calculating the gas mass - weighted temperature gained by the thermalization of the stellar random motions , @xmath74 : @xmath75 where @xmath76 is the one - dimensional velocity dispersion of the stars . the integral term in eq . [ eq : tvir ] is the same that gives the kinetic energy associated with the stellar random motions [ @xmath77 , and that enters the virial theorem for the stellar component ; the mass - weighted temperature in eq . [ eq : tvir ] is then often called `` gas virial temperature '' . for a galaxy mass model made of stars and dark matter , characterized by @xmath78 , where @xmath79 is the total dark mass , and @xmath80 , with @xmath81 and @xmath82 the scale radii of the two mass distributions , @xmath74 can be expressed using the central velocity dispersion @xmath3 as @xmath83 ( e.g. , ciotti & pellegrini 1992 ) . the function @xmath84 increases mildly for larger @xmath85 and for lower @xmath86 , that is for a larger amount of gravitating mass or a higher mass concentration , but always @xmath87 , since @xmath88 has in general a negative radial gradient ( e.g. , sect . [ mass ] and fig . [ f1 ] below ) . @xmath74 is then proportional to @xmath9 , and a simplified version of the virial temperature in eq . [ eq : tvir ] that is often used is @xmath89 ; @xmath22 of course overestimates the true @xmath74 . the mass - averaged injection temperature is finally given by @xmath90 where in general the second term dominates , as is shown in sect . [ disc ] below . in case of mass losses flowing to the galactic center , the gas can be heated due to infall in the galactic potential and adiabatic compression ; this process is sometimes referred to as `` gravitational heating '' . the average change in gravitational energy per unit gas mass inflowing through the galactic potential down to the galactic center is @xmath91 dr , \label{eq : lgp}\ ] ] for galaxy mass distributions with a finite value of @xmath92 ( see also ciotti et al . one can define a temperature equivalent to the energy in eq . [ eq : lgp ] as @xmath93 . as @xmath74 , also @xmath94 is @xmath95 , and increases for larger @xmath96 and smaller @xmath86 , which , for inflowing gas , can be understood as a larger gas heating by compression during infall for a larger dark matter amount or its higher concentration . not all of @xmath97 can be available for heating , though . if the inflow keeps quasi - hydrostatic , then , by the virial theorem , the energy radiated away is roughly one - half of the change in the gravitational potential energy , and that available for the heating of the gas is the remaining half for a self - gravitating gas ; for gas in an external potential , the result should remain roughly valid ( binney & tremaine 1987 ) . ] ( i.e. , @xmath98 ) . actually , the energy available for heating will be much less than this . inflows are caused by the radiative losses produced by the accumulation of the stellar mass return , that makes the cooling time lower than the galactic age ; in the central regions , within a radius of @xmath99 kpc , the cooling time can be as short as @xmath100 yr , even shorter than the infall time ( e.g. , sarazin & white 1988 , pellegrini 2011 ) . in these conditions , the gas departs from a slow inflow , becomes very dense and supersonic close to the center , and cools rapidly down to low temperatures , so that @xmath101 is radiated away or goes into kinetic energy of condensations ( sarazin & ashe 1989 ) . furthermore , there is the possibility that not all the gas reaches hot the galactic central region , if thermal instabilities develop and produce drop - outs from the flow ; if gas cools and condenses out of the flow at large radii , then @xmath97 can be much lower than in the definition above , and heating due to infall in the gravitational potential is `` lost '' ( sarazin & ashe 1989 ) . in conclusion , without a precise knowledge of how to compute @xmath102 ( which depends on the radius at which the injected gas drops below x - ray emitting temperatures ) , and about what fraction of @xmath102 is radiated or goes into kinetic energy of the condensations , @xmath103 remains a reference value ; a more direct use can instead be made of the analogous temperature for escape @xmath104 introduced in sect . [ esc ] below . due to the presence of a central mbh in etgs , another potential source of heating for the gas could be provided by nuclear accretion . this subject has been studied intensely recently , through both observations and modeling , and it appears that the energy provided by accretion is of the order of that needed to offset cyclically the cooling of the inflowing gas in the central regions of gas - rich etgs ( e.g. , brzan et al . 2004 , million et al . 2010 ) . therefore , the central mbh is believed to be a heat source that balances the radiative losses of the gas , acting mostly in the central cooling region . in gas poor etgs the nuclear accretion energy is far lower due to the very small mass accretion rate , if present ( pellegrini et al . 2007 ) , and the absorption of the energy output from accretion is likely not efficient . given the role of the mbh outlined above , possible energy input from the mbh will not be considered as a source of global heating for the gas . another characteristic temperature for comparison with observed @xmath0 values is the temperature with which the gas can escape from the galaxy . assuming that the flow is stationary and adiabatic , the bernoulli constant on each streamline along which the gas flows out of the galaxy must be positive . the bernoulli equation with the minimum energy for escape is written as @xmath105 , where @xmath106 is the enthalpy per unit gas mass , @xmath107 is the ratio of specific heats , @xmath108 is the sound velocity , and @xmath109 is the flow velocity . integrating over the galaxy volume and gas - mass averaging , this condition becomes : @xmath110 where the escape energy @xmath111 is the average energy required to remove from the galaxy the unit gas mass . @xmath112 gives a minimum energy requirement , since energy losses due to cooling may be present ; but these are not important for outflows that typically have a low density ( e.g. , sect . [ sna ] below ) . the escape temperature equivalent to @xmath113 is @xmath114 . the condition for the minimum energy for escape can then be translated into a condition for the injection temperature of the gas @xmath115 to be larger than @xmath116 . in the simple case that @xmath117 is due only to one ( stellar ) mass component , from @xmath118 and the virial theorem ( sect . [ heat ] ) , one derives that @xmath119 . the estimates of @xmath112 in eq . [ eq : lm ] and then of @xmath116 will be calculated below for a general mass model ( e.g. , made by the superposition of stars and dark matter , with a different radial distribution ) . in previous works , the sufficient condition for the existence of a galactic wind was that the injection temperature exceeded an `` escape temperature '' , defined as @xmath120 ( white & chevalier 1983 ) , or `` twice the equivalent dark halo temperature '' ( loewenstein & mathews 1987 ) , coupled with the request for the radiative cooling time in the central part of the galaxy to be longer than the time required to flow out of this region . these conditions are similar to imposing that @xmath115 exceeds @xmath121 as derived above . in principle the gas can escape with different combinations of @xmath109 and @xmath0 , and the observed @xmath0 should be close . ] to that entering @xmath122 in eq . [ eq : bern ] . there are two extreme cases for the value of the flow velocity @xmath109 with respect to @xmath108 ( i.e. , to the temperature ) . one is when the material is brought to infinity keeping a subsonic velocity , then the minimum energy requirement becomes @xmath123 ; neglecting the kinetic term in eq . [ eq : bern ] , one then obtains a characteristic gas - mass averaged subsonic escape temperature : @xmath124 in a general case , @xmath123 gives for @xmath0 a larger requirement than obtained when @xmath109 is not neglected ; also the partition of the gas energy between enthalpy and kinetic energy can vary with radius within a galaxy . all this means that @xmath125 represents a fiducial upper limit to the observed temperatures of outflowing gas : if the kinetic energy of the flow is important , then the actual gas temperature will be lower ( the stronger the outflow , with respect to @xmath108 , the cooler the gas ) . at the opposite extreme case where the `` temperature '' contribution to the gas energy is minor and that of the velocity is dominant , the bernoulli equation reduces to @xmath126 ; this gives the usual escape velocity of a unit mass from a potential well : @xmath127 . finally , @xmath116 and @xmath125 have the same dependence as @xmath74 on @xmath128 . for the representative galaxy mass models used here ( sect . [ mass ] , fig . [ f2 ] ) , @xmath129 ; since @xmath130 can be converted into heat , then the corresponding temperature gained by infall will be @xmath131 . in this work the exact values of @xmath74 , @xmath125 and @xmath104 are calculated as a function of @xmath3 , for a series of representative 3-component galaxy mass models , made by the superposition of a stellar distribution and a dark matter halo , to which a central mbh is added . the stellar density profile is given by the deprojection of a srsic law with index @xmath132 or 5 , as appropriate for etgs of the luminosities considered in this work ( e.g. , kormendy et al . the mass of the mbh is @xmath133 , in agreement with the magorrian et al . ( 1998 ) relation . the dark halo has a navarro et al . ( 1997 ; nfw ) profile [ @xmath134 , with @xmath81 the scale radius , and truncated at large radii ] , and a total mass @xmath79 . for each @xmath3 , the defining parameters of the stellar mass model were chosen accordingly to the observational constraints that @xmath28 follows the faber - jackson relation , and that @xmath28 , @xmath3 and the effective radius @xmath135 lie on the fundamental plane of etgs ( e.g. , bernardi et al . 2003 ) . the free parameters defining the dark matter were chosen in agreement with the results from dynamical modeling of the observed motions of stars , planetary nebulae and globular clusters at small and large radii ; these indicate that the dark matter begins to be dynamically important at 23@xmath135 ( e.g. , saglia et al . 1992 , cappellari et al . 2006 , weijmans et al . 2009 , shen & gebhardt 2010 ) . this requires that @xmath136 , and @xmath137 or 5 ( the latter value corresponding to the baryon - to - total mass ratio of wmap , komatsu et al . 2009 ) . by solving numerically the jeans equations for the three mass components in the isotropic orbits case ( e.g. , binney & tremaine 1987 ) , these choices produce @xmath24 , @xmath79 and @xmath138 ( the dark - to - luminous mass ratio within @xmath135 ) . reasonable values of @xmath139 , and @xmath140 are obtained . the main properties of a few representative mass models are shown in fig . [ f1 ] . for a consistent comparison between observed @xmath0 s and the characteristic temperatures derived for the mass models , the central stellar velocity dispersion @xmath3 must be the same for observed etgs and models . typically , for nearby well observed etgs , the value of @xmath3 is that of the projected and luminosity - weighted average within an aperture of radius @xmath141 . therefore , when defining a mass model , the chosen value of @xmath3 was assigned to this quantity . finally , streaming motions as stellar rotation are not considered in these models ( possible heating from these motions is discussed by ciotti & pellegrini 1996 ) . we investigate here the relationship between the observed @xmath0 s and those expected from the various sources of heating ( stellar motions , gravitational potential , snia s ) , or during the escape of the hot gas . for this purpose , [ f2 ] shows the run with @xmath3 of the various temperatures defined in sects . [ inj ] , [ egp ] , and [ esc ] , together with the distribution of the observed @xmath0 values from bkf ( sect . [ intro ] ) ; for the bkf sample , the value of @xmath3 is the luminosity weighted average within @xmath141 , and has been taken from sauron studies for 12 etgs ( kuntschner et al . 2010 ) , for the remaining cases from the references in the hyperleda catalog ( see tab . [ tab1 ] ) . the temperatures defined in sects . [ inj ] , [ egp ] , and [ esc ] are mass - weighted averages , which is required when discussing energetic aspects of the gas ( e.g. , the energy required for escape as measured by @xmath142 compared with the input energy from snia s ) . when a direct comparison is made with observed @xmath0 s , it must be noted that the latter coincide with mass - weighted averages only if the ism has everywhere one single temperature value ; if the gas is multi - phase , or its temperature profile has a gradient , a single @xmath0 value measured from the spectrum of the integrated emission will be close to an emission - weighted average ( e.g. , ciotti & pellegrini 2008 , kim 2011 ) . this means that , since the densest region is the central one , the measured @xmath0 s tend to be closer to the central values than the mass - weighted ones . the temperature profiles observed with @xmath2 change continuously in shape , as the emission - weighted average @xmath0 decreases from @xmath143 kev to @xmath144 kev : they switch from a flat central profile that increases outward of @xmath145 , to a quasi - isothermal profile , to a profile with a negative gradient ( diehl & statler 2008 , nagino & matsushita 2009 ) . therefore , the lowest observed @xmath0 s , presumably associated with the last category , may be larger than mass - weighted values , intermediate @xmath0 s may be the closest to mass - weighted averages , while the largest observed @xmath0 s may be lower than mass - weighted averages . another aspect to recall is that the temperatures defined in sects . [ inj ] , [ egp ] , and [ esc ] refer to a gas distribution with @xmath146 ; this is appropriate for the continuously injected gas ( e.g. , for @xmath115 ) , while it may be less accurate when comparing observed @xmath0 s with @xmath125 , or when discussing the energetics of the whole gas content of an etgs by means of @xmath147 and @xmath142 , since the bulk of the hot ism may have a different distribution from the stars . for example , the observed x - ray brightness profile of gas - rich etgs was found to follow the optical one , which was taken as evidence that roughly @xmath148 ( e.g. , sarazin & white 1988 , fabbiano 1989 ) . for gas - poor etgs hosting galactic winds , the modeling shows that the profile @xmath149 will again be shallower than @xmath150 , though not as much as in the previous case ( see , e.g. , white & chevalier 1983 ) . if @xmath151 has a flatter radial profile than @xmath152 , then it is easy to show that its mass - weighted @xmath147 will be larger than derived using eq . [ eq : lgp ] , and its mass - weighted @xmath153 and @xmath125 will be lower than derived using eqs . [ eq : lm ] and [ eq : tent ] . in conclusion , the comparison of observed @xmath0 s with mass - weighted expectations is the best that can be done currently , in a general analysis as that of the present work , though with the warnings above . note , however , that all arguments and conclusions below remain valid or are strenghtened , when taking into account the above considerations about observed @xmath0 s , or about the modifications to @xmath154 , @xmath142 and @xmath125 . in the left panel of fig . [ f2 ] the observed @xmath0 s are compared with approximate estimates of the stellar temperature @xmath22 , of @xmath56 , and of the escape temperature @xmath155 ( sect . [ esc ] ) . the gas luminosity is also indicated with different colors , having grouped the @xmath1 values in three ranges , chosen to have a roughly equal number of etgs in each range . this grouping gives an indication of the gas flow status , based on previous works : a galactic wind leaving the galaxy with a supersonic velocity has @xmath156 erg s@xmath5 ( e.g. , mathews & baker 1971 , trinchieri et al . 2008 ) , global subsonic outflows and partial winds can reach @xmath157 erg s@xmath5 ( ciotti et al . 1991 , pellegrini & ciotti 1998 ) , and a central inflow becomes increasingly more important in etgs of increasingly larger @xmath1 . magenta etgs ( @xmath158 erg s@xmath159 erg s@xmath5 ) should then host winds , subsonic outflows , and partial winds with a very small inflowing region of radius @xmath160 pc ; cyan etgs ( @xmath161 erg s@xmath162 erg s@xmath5 ) should host subsonic outflows and partial winds with an increasingly larger inflowing region ( of radius up to a few hundreds pc ) ; black etgs are hot gas - rich and mostly inflowing . is @xmath163 times the integrated emission from the population of weak unresolved stellar sources ( sect . [ intro ] ) , generally referred to as ab@xmath164cvs . bkf derived an ab@xmath164cv emission model by jointly fitting the m31 and m32 spectra , and then tested how the measurement of the gas properties may be affected by the adopted ab@xmath164cv model ( their sect . by fitting with different ab@xmath164cv models , all within the uncertainties in the adopted one , the measured @xmath0 changed only negligibly , for the 6 lowest-@xmath1 etgs of their sample ; a systematic uncertainty of 10% 20% was found for the gas flux ( much lower for the x - ray brighter etgs ) . ] the best fit found for x - ray bright etgs is also shown in the left panel ( osullivan et al . 2003 ) , and gives a good representation of the distribution of observed @xmath0 s down to a range of low temperatures and gas contents never explored before . the slope of the fit ( @xmath165 ) and that of the @xmath166 relation are similar , with the fit being shallower ; this could be due to the fact that not all heating sources depend on @xmath9 , see , e.g. , the important snia s contribution in @xmath56 , that produces a much flatter run of @xmath56 with @xmath3 ( fig . [ f2 ] ) . the fit was mostly based on gas - rich etgs , whose @xmath0 s show a trend with @xmath3 closer to that of @xmath22 ( an aspect further addressed below in sects . [ gravhea ] and [ temp ] ) , while gas - poor etgs depart most from it , since their @xmath0 s change little for largely varying @xmath167 ( as found by bkf ; see sect . [ reduc ] below ) . the right panel of fig . [ f2 ] shows the stellar temperature @xmath74 , the injection temperature @xmath115 , the escape temperature @xmath142 , and the characteristic temperature for slowly outflowing gas @xmath125 , calculated for a set of representative galaxy mass models ( sect . [ mass ] ) . at any fixed @xmath8 and srsic index @xmath168 , @xmath74 , @xmath142 and @xmath125 are larger for larger galaxy mass ( @xmath96 ) , and mass concentration ( smaller @xmath86 ) , due to the galaxy being more massive to reproduce the same @xmath3 , since the stellar mass profile is less steep ( e.g. , fig . [ f1 ] ) . ] . the dashed lines represent a reasonable upper limit to the values of each of the characteristic temperatures , since they correspond to the most massive model etgs , with the most concentrated dark matter allowed for by recent studies ( sect . [ mass ] ) . the @xmath142 curves lie below the simple approximation of the escape temperature given by @xmath155 ; @xmath169 , for @xmath170 , and @xmath171 , for the three cases with @xmath172 . as expected , all the @xmath74 curves lie below @xmath22 , that overestimates the kinetic energy associated with the stellar random motions ( sect . [ inj ] ) . note that , from the virial theorem , @xmath74 is independent of orbital anisotropy , that just redistributes differently the stellar heating within a galaxy ; the presence of ordered rotation in the stellar motions , instead , requires a more careful consideration . for any fixed galaxy mass model , this rotation would leave the total stellar heating unchanged or lower it , depending on whether the whole stellar streaming motion is converted into heat , or just a fraction of it ( ciotti & pellegrini 1996 ) . for the worst case that the stellar rotational motion is not thermalized at all , and the galaxy is a flat isotropic rotator , @xmath74 in fig . [ f2 ] should be an overestimate of @xmath173% of the temperature corresponding to the stellar heating ( ciotti & pellegrini 1996 ) ; the possible reduction of @xmath74 to be considered should be lower than this , as far as the massive etgs in fig . [ f2 ] are less flattened and more pressure supported systems ( e.g. , emsellem et al . 2011 ) . all observed @xmath0 s are located above @xmath74 ; thus , additional heating with respect to the thermalization of the stellar kinetic energy is needed , as noticed previously using @xmath22 ( e.g. , davis & white 1996 , bkf ) . the gas could retain memory of its injection temperature , and have the additional infall heating , as examined in sects . [ gravhea ] and [ sna ] below . finally , the values of @xmath115 for @xmath68 are by far the largest temperatures of fig . [ f2 ] , larger than @xmath142 up to @xmath174 km s@xmath5 ; therefore , snia s should cause the escape of the gas for all etgs up to this @xmath3 , since the gas at every time is injected with an energy larger than required to leave the galaxy potential . this expectation is fulfilled by all etgs with @xmath175 km s@xmath5 : their x - ray properties ( a low @xmath1 , and @xmath0 s of the order of @xmath125 ) agree well with what expected if outflows are important in them . this result had been suggested previously based on the low observed @xmath1 ; now for the first time it can be confirmed based on the observed @xmath0 values . at @xmath176 km s@xmath5 , instead , etgs may have @xmath1 far larger than expected for outflows ( black symbols ) , and most etgs where likely outflows are important ( magenta or cyan symbols ) have @xmath0 much lower than @xmath125 ; these findings are discussed in sect . [ reduc ] below . we examine here the possibility that the additional heating with respect to the thermalization of the stellar kinetic energy is provided by infall heating and snia s . davis & white ( 1996 ) assumed that in all etgs the hot gas is inflowing , and suggested that the observed temperatures are larger than @xmath22 just due to the luminous parts of etgs being embedded in dark matter halos dynamically hotter than the stars ; i.e. , a form of `` gravitational potential '' way for the gas heating was invoked . this way can consist of an effect of the dark halo on the stellar motions , that are then thermalized , or directly on the gas during infall ( e.g. , via @xmath97 ) . the first possibility is excluded by the @xmath74 curves in fig . [ f2 ] , that are always lower than @xmath22 , and that , through the jeans equations , include the effect of a massive dark halo consistent with the current knowledge of the etgs structure . in the second possibility of heating from gas infall , @xmath97 is indeed potentially an important source of heating , that increases with the amount and concentration of the dark matter . this can be judged from fig . [ f2 ] , after considering that @xmath177 , and that the temperature possibly attainable from infall was estimated to be @xmath178 ( end of sect . [ esc ] ) . note that @xmath179 ( if it behaves as @xmath180 ) could be @xmath95 , a trend close to that shown by the @xmath0 s of gas - rich etgs ( bkf ; see also sect . [ temp ] below ) . inflowing etgs can also benefit of the snia s energy input ; for the unit mass of injected gas , this is written as @xmath181 . both @xmath97 and @xmath182 then contribute to the required additional thermal energy with respect to that gained from the stellar random motions , i.e. , to @xmath183 . @xmath97 and @xmath182 can be in large part radiated in gas - rich etgs , but they seem to far exceed the required @xmath184 . for example , for the highest-@xmath1 of fig . [ f2 ] , @xmath185 erg @xmath186 ( i.e. , @xmath1870.20.4 kev ) , when adopting an average galaxy mass model as that of the thick black line in fig . [ f2 ] . the energy spent in radiation can be estimated , in a stationary situation , as @xmath188 ( per unit injected gas mass ) ; using @xmath1 from bkf and deriving @xmath47 as in sect . [ loss ] , for the same distances in bkf and galactic b - magnitudes given by hyperleda , for the gas - rich etgs one finds that @xmath188 ranges between @xmath189 erg @xmath186 . the energy available is far larger than the sum of @xmath184 and @xmath188 : @xmath190 erg @xmath186 , and @xmath191 ranges from @xmath192 erg @xmath186 ( @xmath193 km s@xmath5 ) to @xmath194 erg @xmath186 ( @xmath195 km s@xmath5 ) , for the mass model with the thick black line in fig . these results are detailed in fig . [ f3 ] , where the values of @xmath184 and @xmath188 for each galaxy are shown , together with various combinations of @xmath97 and @xmath182 . in conclusion , additional input energy for the gas to account for the observed @xmath0 s of gas - rich etgs seems available in a sufficient amount , even if @xmath65 were to be @xmath196 . the gas is not mostly inflowing in all etgs , while it is hotter than @xmath74 in all of them . when in outflow , the radiative losses are far smaller , but energy is spent in extracting the gas from the galaxy and giving it a bulk velocity . we discuss here the possibility of heating from the snia s energy input to account for the observed @xmath197 of etgs where outflows are likely important ( those with low / medium @xmath1 , magenta and cyan symbols in fig . [ f2 ] ) . we assume that the snia s energy is used for the uplift of the gas and the kinetic energy with which it escapes from the galaxy , and , neglecting radiative losses , that all the remaining part is available to account for the observed @xmath184 . then , the energy balance per unit mass of injected gas is @xmath198 , where @xmath184 is the same as in sect . [ gravhea ] , and @xmath199 is the mass - averaged kinetic energy of the escaping material per unit gas mass . figure [ f4 ] shows @xmath184 derived from this balance , for @xmath200 , and @xmath108 calculated for @xmath201 and @xmath202 kev , a temperature of the order of that observed for etgs likely in outflow ( fig . [ f2 ] ) . adopting @xmath203 ( independent of @xmath3 ) produces an @xmath204 on the upper end of those expected , due to their @xmath1 ( e.g. , trinchieri et al . 2008 ) . ] , and then the estimate of @xmath184 may be biased low . with @xmath203 , @xmath204 is just @xmath205 . the energy needed for gas extraction @xmath113 , for the same galaxy mass model used in sect . [ gravhea ] , varies instead from @xmath206 for @xmath207 km s@xmath5 , to @xmath208 for @xmath209 km s@xmath5 ; this explains the strong dependence of the predicted @xmath184 on @xmath3 in fig . it is clear from this figure that for @xmath68 snia s can account for the needed heating in all etgs with low / medium @xmath1 ; for @xmath210 , instead , the temperature increase would fall short of what required for all etgs . in fig . [ f4 ] , the energy losses due to radiation are also shown ; their small size supports the hypothesis that in most cases they do not affect significantly the energy budget of the gas . given the flat distribution of the observed points in fig . [ f4 ] , and the steep behavior of the curves predicting @xmath184 , the value of @xmath65 required to account for the observed @xmath184 increases with @xmath3 . in particular , the value of @xmath211 that is required at high @xmath3 would produce an expected @xmath184 at low @xmath3 that is larger than observed . a possible solution could reside in the efficiency of the snia s energy mixing process . in massive , gas - rich etgs , snia s bubbles should disrupt and share their energy with the local gas within @xmath212 yr ( mathews 1990 ) ; for a milky way - size bulge in a global wind , instead , 3d hydrodynamical simulations of discrete heating from snia s suggest a non - uniform thermalization of the snia s energy , with overheated gas by a snia explosion at the bulge center that is advected outwards , carrying a large fraction of the snia energy with it ( tang et al . 2009 ) . for subsonic outflows the mixing is expected to be more local and more complete ( lu & wang 2011 ) . the magenta and cyan etgs in fig . [ f4 ] have gas densities and luminosities larger than those considered by tang et al . ( 2009 ) ; however , if a discrete heating effect were still present at @xmath213 km s@xmath5 , it could qualitatively explain a lower @xmath65 for these galaxies . there is also the possibility that @xmath74 has been overestimated ( and then the observed @xmath184 underestimated ) at the low @xmath3 , if these etgs are less pressure supported systems ( e.g. , emsellem et al . 2011 ) , and the stellar rotational streaming is not all thermalized ( sect . [ obs ] ) . another possible explanation could be that etgs with @xmath214 km s@xmath5 are less outflow - dominated than those at lower @xmath3 ( though this is not supported just based on @xmath1 , since magenta etgs are found over the whole @xmath3 range in fig . [ f4 ] ) , so that their @xmath113 would be lower than assumed by the curves in fig . [ f4 ] , and more snia s energy would be available for heating . in this way , @xmath65 could have a value @xmath196 , possibly similar for all etgs . finally , a comparison of figs . [ f3 ] and [ f4 ] shows that the average @xmath184 is slightly larger for the x - ray brightest etgs ( for which it ranges between 0.10.5 kev ) than for the x - ray faintest ones ( 00.3 kev ) ; moreover , while the @xmath184 in fig . [ f3 ] can be explained even with @xmath215 , it is required that @xmath211 for the x - ray faintest etgs with the largest @xmath3 in fig . both facts are the consequence of the large fraction of the snia s energy input that is used in gas extraction where outflows dominate , while all the snia s energy remains within the galaxies where inflow dominates . in conclusion , even for etgs with low / medium @xmath1 , a fundamental x - ray property as @xmath0 can be accounted for by simple arguments , just based on realistic galaxy mass models , and reasonable snia s heating capabilities . there may be , though , more energy available for the gas in etgs with @xmath213 km s@xmath5 than can be accounted for by the present simple scenario . the observed x - ray properties ( low @xmath1 and @xmath216 ) , and the energy budget of the gas ( e.g. , @xmath142 vs. @xmath115 ) , for etgs with @xmath217 km s@xmath5 are all consistent with the expectations for outflows ; the large @xmath1 and @xmath218 larger than @xmath115 of etgs with @xmath219 km s@xmath5 agree with the gas being mostly inflowing . for @xmath11 km s@xmath7 , instead , etgs show very different @xmath1 and @xmath0 , whose values seem unrelated to @xmath125 and to the relative size of @xmath142 and @xmath115 ( fig . for example , for @xmath68 , @xmath115 exceeds @xmath142 , but most @xmath0 values lie well below @xmath125 , and even high @xmath1 values ( incompatible with outflows ) are common . one first explanation could be that @xmath215 ; for example , for @xmath210 , @xmath115 becomes lower than @xmath142 at @xmath220 km s@xmath5 ( fig . four etgs with @xmath214 km s@xmath5 and a very low @xmath1 ( fig . [ f2 ] , magenta symbols ) , though , require that @xmath221 in them , and then that @xmath65 varies from galaxy to galaxy , or that their gas was removed by other processes as an agn outburst ( e.g. , machacek et al . 2006 , ciotti et al . 2010 ) , or a merging or an interaction ( read & ponman 1998 , sansom et al . 2006 , brassington et al . an event like the latter two in the recent past is unlikely for three of these etgs ( ngc1023 , ngc3115 , ngc3379 ) , that are very regular in their stellar morphological and kinematic properties , while in the other etg ( ngc4621 ) that hosts a counter - rotating core ( wernli et al . 2002 ) it is possible . a second explanation could be that @xmath115 exceeds @xmath222 only at the present epoch : while @xmath74 and @xmath222 are independent of time , @xmath58 may have been lower in the past ( eq . [ eq : tinj ] ) , to the point that @xmath115 may have been lower than @xmath142 for more etgs than in fig . [ f2 ] ( that represents a snapshot of the present epoch ) . the gas then could have accumulated and radiative losses have become important , even for the gas injected in later epochs . in fact , the population synthesis models of sect . [ heat ] predict that @xmath47 was larger at early times ( e.g. , by @xmath223 times at an age of 3 gyr ) , and then to keep @xmath58 high in the past , from eq . [ eq : tinj ] , @xmath71 must decrease with time @xmath224 at a rate similar to or steeper than that of the stellar mass losses ( @xmath225 ; ciotti et al . recent observational estimates indicate instead a snia s rate decaying close to @xmath226 ( maoz et al . 2010 , sharon et al . 2010 ) , thus @xmath58 and @xmath115 should be increasing with time , reaching the values of fig . [ f2 ] at the present epoch . then , a `` cooling effect of the past '' would explain a moderate or high @xmath1 even where @xmath115 exceeds @xmath142 in fig . [ f2 ] . both explanations , @xmath215 and/or a lower @xmath115 in the past , can account for the lack of a widespread presence of outflows for @xmath227(km s@xmath228 . both , though , require a mechanism different from the snia s energy input to cause degassing in some low @xmath1 etgs in the same @xmath3 range . another possibility is that partial winds become common for @xmath214 km s@xmath5 : these etgs host an inner inflow and an outer outflow ( e.g. , macdonald & bailey 1978 ) , with variations in the galactic structure causing different sizes for the inflowing region , and then different @xmath1 ( pellegrini & ciotti 1998 ) . this possibility holds even for @xmath68 , and for both kinds of time evolution of @xmath115 ; in fact , if the flow is decoupled , etgs may host a central inflow even if @xmath115 is larger than @xmath142 . similarly , the observed @xmath0 s can be lower than @xmath125 , in etgs where the outflow is only external ; in this case , the observed @xmath0 s may also be lowered by radiative losses in the central inflowing region . in fig . [ f2 ] , the hottest gas is in etgs with the highest @xmath1 , and the coolest one in etgs with the lowest @xmath1 ( as also found by bkf ) . this feature is also present in the @xmath15 relation , where etgs with @xmath18 kev are the x - ray brightest ( with one exception ) , while those with @xmath19 kev are the x - ray faintest ( bkf ) . all this may seem contrary to the simple expectation that hotter gas is needed for escape , and that the hotter the gas , the stronger the outflow , the lower the gas content . we re - examine below this point , first across the whole @xmath3 range , and then at fixed @xmath3 . a proper consideration of whether outflowing etgs possess hotter or colder gas than inflowing ones requires that all @xmath0 s are rescaled by a temperature equivalent to the depth of the potential where the gas resides ( for example , by @xmath22 ) . is there a trend then of the distance of the observed @xmath0 s from @xmath125 , or @xmath74 ? this is examined by fig . [ f5 ] , where temperatures are rescaled by @xmath22 , and the @xmath9 dependency of all the curves in fig . [ f2 ] is removed . figure [ f5 ] shows that for @xmath229 km s@xmath5 the observed points reach @xmath230 , a result similar to that of fig . [ f2 ] , and that they fall below it with increasing @xmath3 ( with a transition region of large dispersion in @xmath231 ) . therefore etgs with @xmath232 km s@xmath5 are indeed the hottest , relatively to the virial temperature ; since in these etgs outflows are important ( sect . [ obs ] ) , indeed the flow is relatively hotter in outflows , and @xmath0 increases going from outflows to inflows only in an absolute sense . the @xmath0 s of the x - ray brightest etgs should show a dependence on @xmath9 , if gravitational heating of the gas dominates over snia s heating , and then they should lie within a horizontal zone in fig . [ f5 ] . the observed distribution does not disagree with this kind of dependence , but more cases are needed to firmly establish its presence ; such a dependence is not expected , though , since the snia s heating should easily dominate over the gravitational one ( sect . [ gravhea ] ) . we finally compare the @xmath0 s at similar @xmath3 , in the most populated region of fig . [ f5 ] , for @xmath11 km s@xmath7 . here the variation of @xmath231 is the largest , and is covered by etgs of all x - ray emission levels ; the x - ray brightest etgs are found at @xmath233 , while the lowest @xmath231 values belong to the x - ray fainter etgs . for example , two of the three lowest @xmath231 values of the figure ( those of ngc3379 and ngc4621 ) belong to the x - ray faintest group . while heating sources seem abundant in gas - rich etgs to account for their @xmath0 s ( see , e.g. , fig . [ f3 ] , and the additional possibility of mbh heating , sect . [ mbh ] ) , even after taking into account their radiative losses , this result remains more difficult to explain for etgs of low / medium @xmath1 , and may require ad hoc solutions . it may be another representation of what mentioned in sect . [ sna ] , that @xmath184 can be larger for the x - ray brightest etgs than for the x - ray faintest ones , due to the different employment of the snia s input energy ; or it could be that @xmath234 in these etgs so that snia s can not make their gas hotter than this ( fig . [ f4 ] ) ; or their galaxy structure may be much different from an average one , so that dividing all @xmath0 s for the same @xmath22 produces a biased view ; or the evolutionary history of the gas may have been peculiar . certainly , this trend needs further investigation and , if confirmed , it will provide the basis for further theoretical work . this work has focussed on the origin of the hot gas temperatures recently derived for a sample of etgs observed with @xmath2 down to galaxy masses and x - ray luminosities smaller than ever before . a few characteristic mass - weighted average temperatures have been defined for a gas distribution @xmath48 , as for the gas shed by stars : the virial temperature @xmath74 ; the injection temperature @xmath115 , as the sum of @xmath74 and of a temperature equivalent to the snia s kinetic energy input ( with a factor @xmath65 allowing for its uncertain thermalization ) ; the escape temperature @xmath142 , defined as the temperature equivalent of the energy required for escape from the gravitational potential ; a fiducial value for the temperature of escaping gas , evaluated on a streamline of very subsonic velocity ( @xmath235 ) ; and finally , the temperature equivalent to the energy liberated by the gas inflow to the galactic center , @xmath147 . these temperatures were then calculated for a set of representative galaxy mass models , made by the superposition of a central mbh , and a stellar and a dark mass density distributions , with parameters constrained from the fundamental scaling laws of etgs and recent observational findings . the main properties of the characteristic temperatures are that : @xmath236 all temperatures scale as @xmath9 ( except for @xmath115 ) , and increase for larger and/or more concentrated mass content . for the adopted set of representative galaxy mass models , @xmath74 is lower than @xmath22 ( by @xmath237 kev ) , @xmath238 , and @xmath239 ; the temperature that can be produced by infall heating , though , will be much lower than @xmath142 , due to energy losses in radiation , kinetic energy of mass condensations , and mass drop - outs from the flow . @xmath236 @xmath115 is by far the largest of the characteristic temperatures , due to the important snia s contribution ( independent of @xmath3 ) ; for @xmath68 , it is larger than the minimum injection temperature for global escape up to @xmath240 km s@xmath5 . @xmath236 the best fit @xmath16 relation previously found for x - ray bright etgs reproduces the average trend of the observed @xmath0 down to low temperatures , and low @xmath1 . etgs with low / medium @xmath1 show the largest departures from this fit , which can be explained by the variety of gas flow phases possible in them ( winds , subsonic outflows , partial winds ) , where the main input energies ( from snia s and gas infall ) are used in different ways . @xmath236 all observed @xmath0 s are larger than @xmath74 ; the additional heating of the gas @xmath184 , with respect to that provided by the thermalization of the stellar motions , is @xmath241 kev for the x - ray faintest etgs , and @xmath242 kev for the x - ray brightest ( for a representative galaxy mass model ) . @xmath236 in a stationary situation , @xmath184 of the x - ray brightest etgs can be accounted for by the energy input of snia s and gas infall , even if they are much reduced with respect to standard assumptions ( i.e. , @xmath65 can be @xmath196 ) . the gravitational heating produces a @xmath243 trend , that may be present in the x - ray brightest etgs ; the snia s heating , though , is expected to be dominant . @xmath236 @xmath184 can be provided by snia s in x - ray fainter etgs , where outflows are important ; most of the snia s energy is needed for gas extraction , and less for the kinetic energy of the escape . the value of @xmath65 to account for the observed @xmath184 increases with @xmath3 , until the whole snia s energy ( @xmath244 ) is required at the highest @xmath3 . with this @xmath65 , though , at low @xmath3 the observed @xmath184 are lower than expected . possible solutions require a different efficiency of the snia s energy mixing process , or an overestimate of @xmath74 at low @xmath3 if these etgs are less pressure supported systems , or a more complex flow status than in the simple scheme adopted . @xmath236 at low @xmath217 km s@xmath5 , @xmath115 is larger than @xmath142 , the @xmath1 values are low and the @xmath0 s are of the order of @xmath125 : all this agrees well with what expected for outflows . at high @xmath219 km s@xmath5 , @xmath115 is lower than @xmath142 , and the high @xmath1 and @xmath0 can be explained with the gas mostly inflowing . for @xmath11 km s@xmath7 , instead , there is a large variation in @xmath1 and @xmath0 . possible explanations could be that the snia s energy input varies from galaxy to galaxy , and/or that @xmath115 was lower in the past , due to the different time evolution of the mass loss and the snia s rate ; or that partial winds become common , with the flow status less related to the values of @xmath115 , @xmath142 , and @xmath125 . @xmath236 when measured relatively to the depth of the potential well , the observed temperatures @xmath231 are larger for @xmath245 km s@xmath5 ( outflows ) , and lower for @xmath246 km s@xmath5 ( inflows ) . the observed @xmath0 s then increase from outflows to inflows only in an absolute sense , and the gas is relatively hotter in outflows . in the intermediate region of @xmath11 km s@xmath7 , lower @xmath1 values tend to correspond to lower @xmath0 and @xmath231 , which requires ad hoc explanations , and then deserves further observational and theoretical investigation . column ( 1 ) : galaxy name . cols . ( 2 ) , ( 3 ) and ( 4 ) : the hot gas temperature , its uncertainty , and the 0.38 kev gas luminosity , from bkf . ( 5 ) : the stellar velocity dispersion , as the luminosity - weighted average within an aperture of radius @xmath141 , with its reference in col . ( 6 ) . stars@xmath164dark matter ) , for two representative etgs with isotropic orbits and an aperture velocity dispersion within @xmath141 of @xmath247 km s@xmath5 ( solid lines ; @xmath248 , and @xmath249 kpc ) , and of @xmath250 km s@xmath5 ( dashed lines ; @xmath251 , and @xmath252 kpc ) . red lines refer to a stellar srsic profile with index @xmath132 , blue ones with @xmath253 . the dark halo in the upper panel ( black , with the same line type as the corresponding stellar profile ) follows the nfw profile , with @xmath254 , and @xmath170 ( for @xmath132 ) or @xmath255 ( for @xmath253 ) , and then @xmath138=0.24 or 0.41 , from the jeans equations ( see sect . [ mass]).,title="fig:",scaledwidth=85.0% ] -7.6truecm stars@xmath164dark matter ) , for two representative etgs with isotropic orbits and an aperture velocity dispersion within @xmath141 of @xmath247 km s@xmath5 ( solid lines ; @xmath248 , and @xmath249 kpc ) , and of @xmath250 km s@xmath5 ( dashed lines ; @xmath251 , and @xmath252 kpc ) . red lines refer to a stellar srsic profile with index @xmath132 , blue ones with @xmath253 . the dark halo in the upper panel ( black , with the same line type as the corresponding stellar profile ) follows the nfw profile , with @xmath254 , and @xmath170 ( for @xmath132 ) or @xmath255 ( for @xmath253 ) , and then @xmath138=0.24 or 0.41 , from the jeans equations ( see sect . [ mass]).,title="fig:",scaledwidth=85.0% ] of the energies provided by snia s ( @xmath182 ) and gas infall ( @xmath97 ) ; points with errorbars show , for the etgs with the largest @xmath1 in fig . [ f2 ] , the additional thermal energy ( @xmath184 ) with respect to that gained from the thermalization of the stellar random motions , required to explain the observed @xmath0 s ( see sect . [ gravhea ] ) ; all these energies ( @xmath182 , @xmath102 , and @xmath184 ) , that are defined per unit mass in the text , have been multiplied by @xmath256 to obtain their temperature - equivalent in kev plotted here . solid lines show the sum of the energies provided by the snia s and infall , for two cases of @xmath65 , and for @xmath97 rescaled by a factor of 0.1 , and calculated for the galaxy mass model described by the thick black line in fig . [ f2 ] ( right panel ) . the dashed line gives for reference the value of @xmath257 , and the dot - dashed line the value of @xmath182 for @xmath210 . the points with errorbars are obtained subtracting to the observed @xmath258 the value of @xmath259 corresponding to its @xmath3 , for the mass model adopted for @xmath260 ; each point is linked by a red line to a point including the energy spent in radiation observed for that etg ( i.e. , the upper point measures @xmath261 ) . see sect . [ gravhea ] for more details.,scaledwidth=65.0% ] of the energy provided by snia s ( @xmath182 , for three cases of @xmath65 ) , after subtraction of the energy needed for the removal of the gas from the galaxy ( @xmath113 ) , and for escape ( @xmath204 ) at an average @xmath200(0.3 kev ) ; all energies have been computed as in fig . the adopted galaxy mass model is that corresponding to the thick black line in fig . [ f2 ] ( right panel ) . points with errorbars show the additional thermal energy required to explain the observed @xmath0 s , calculated as in fig . [ f3 ] , for etgs with low / medium @xmath1 in fig . [ f2 ] . for the cyan etgs , a red line connects each point with the value including the radiated energy , as in fig . [ f3 ] ; the red lines of the magenta etgs , whose @xmath1 values are the lowest , would be included within the colored circle , if shown . see sect . [ sna ] for more details.,scaledwidth=65.0% ]

we introduce a representation of electron operators as a product of a spin - carrying fermion and of a phase variable dual to the total charge ( slave quantum rotor ) . based on this representation , a new method is proposed for solving multi - orbital anderson quantum impurity models at finite interaction strength . it consists in a set of coupled integral equations for the auxiliary field green s functions , which can be derived from a controlled saddle - point in the limit of a large number of field components . in contrast to some finite- extensions of the non - crossing approximation , the new method provides a smooth interpolation between the atomic limit and the weak - coupling limit , and does not display violation of causality at low - frequency . we demonstrate that this impurity solver can be applied in the context of dynamical mean - field theory , at or close to half - filling . good agreement with established results on the mott transition is found , and large values of the orbital degeneracy can be investigated at low computational cost .

cond-mat0206571

the anderson quantum impurity model ( aim ) and its generalizations play a key role in several recent developments in the field of strongly correlated electron systems . in single - electron devices , it has been widely used as a simplified model for the competition between the coulomb blockade and the effect of tunnelling @xcite . in a different context , the dynamical mean - field theory ( dmft ) of strongly correlated electron systems replaces a spatially extended system by an anderson impurity model with a self - consistently determined bath of conduction electrons @xcite . naturally , the aim is also essential to our understanding of local moment formation in metals , and to that of heavy - fermion materials , particularly in the mixed valence regime @xcite . it is therefore important to have at our disposal quantitative tools allowing for the calculation of physical quantities associated with the aim . the quantity of interest depends on the specific context . many recent applications require a calculation of the localized level green s function ( or spectral function ) , and possibly of some two - particle correlation functions . many such `` impurity solvers '' have been developed over the years . broadly speaking , these methods fall in two categories : numerical algorithms which attempt at a direct solution on the computer , and analytical approximation schemes ( which may also require some numerical implementation ) . among the former , the quantum monte carlo approach ( based on the hirsch - fye algorithm ) and the numerical renormalization group play a prominent role . however , such methods become increasingly costly as the complexity of the model increases . in particular , the rapidly developing field of applications of dmft to electronic structure calculations @xcite requires methods that can handle impurity models involving orbital degeneracy . for example , materials involving f - electrons are a real challenge to dmft calculations when using quantum monte carlo . similarly , cluster extensions of dmft require to handle a large number of correlated local degrees of freedom . the dimension of the hilbert space grows exponentially with the number of these local degrees of freedom , and this is a severe limitation to all methods , particularly exact diagonalization and numerical renormalization group . the development of fast and accurate ( although approximate ) impurity solvers is therefore essential . in the one - orbital case , the iterated perturbation theory ( ipt ) approximation @xcite has played a key role in the development of dmft , particularly in elucidating the nature of the mott transition @xcite . a key reason for the success of this method is that it becomes exact both in the limit of weak interactions , and in the atomic limit . unfortunately however , extensions of this approach to the multi - orbital case have not been as successful @xcite . another widely employed method is the non - crossing approximation , or nca @xcite . this method takes the atomic limit as a starting point , and performs a self - consistent resummation of the perturbation theory in the hybridization to the conduction bath . it is thus intrinsically a strong - coupling approach , and indeed it is in the strong - coupling regime that the nca has been most successfully applied . the nca does suffer from some limitations however , which can become severe for some specific applications . these limitations are of two kinds : * the low - energy behaviour of nca integral equations is well - known to display non - fermi liquid power laws . this can be better understood when formulating the nca approach in terms of slave bosons @xcite . it becomes clear then that the nca actually describes accurately the overscreened regime of multichannel models . this is in a sense a remarkable success of nca , but also calls for some care when applying the nca to fermi - liquid systems in the screened regime . it must be noted that recent progresses have been made to improve the low - energy behaviour in the fermi liquid case @xcite . * a more important limitation for practical applications has to do with the finite-@xmath0 extensions of nca equations . standard extensions do not reproduce correctly the non - interacting @xmath1 limit . in contrast to ipt , they are not `` interpolative solvers '' between the weak - coupling and the strong coupling limit . furthermore , the physical self - energy tends to develop non - causal behaviour at low - frequency ( _ i.e. _ @xmath2 ) below some temperature ( the `` nca pathology '' ) at half - filling and large @xmath0 , this only happens at a rather low - energy scale , but away from half - filling or for smaller @xmath0 , this scale can become comparable to the bandwith , making the finite-@xmath0 nca of limited applicability . in this article , we introduce new impurity solvers which overcome some of these difficulties . our method is based on a rather general representation of strongly correlated electron systems , which has potential applications to lattice models as well @xcite . the general idea is to introduce a new slave - particle representation of physical electron operators , which emphasizes the _ phase variable dual to the total charge _ on the impurity . this should be contrasted to slave - boson approaches to a multi - orbital aim : there , one introduces as many auxiliary bosons as there are fermion states in the local hilbert space . this is far from economical : when the local interaction depends on the total charge only , it should be possible to identify a collective variable which provides a minimal set of collective slave fields . we propose that the phase dual to the total charge precisely plays this role . this turns a correlated electron model ( at finite @xmath0 ) into a model of spin- ( and orbital- ) carrying fermions coupled to a _ quantum rotor degree of freedom_. various types of approximations can then be made on this model . in this article , we emphasize an approximate treatment based on a sigma - model representation of the rotor degree of freedom , which is then solved in the limit of a large number of components . this results in coupled integral equations which share some similarities to those of the nca , but do provide the following improvements : * the non - interacting ( @xmath1 ) limit is reproduced exactly . for a fully symmetric multi - orbital model at half - filling and in the low temperature range , the atomic limit is also captured exactly , so that the proposed impurity solver is an interpolative scheme between weak and strong coupling . the sigma - model approximation does not treat as nicely the atomic limit far from half - filling however ( though improvements are possible , see section [ sec : doping ] ) . * the physical self - energy does not display any violation of causality ( even at low temperature or small @xmath0 ) . this is guaranteed by the fact that our integral equations become exact in the formal limit of a large number of orbitals and components of the sigma - model field . it should be emphasized , though , that the low - energy behaviour of our equations is similar to the ( infinite-@xmath0 ) nca , and characterized by non - fermi liquid power laws below some low - energy scale . as a testing ground for the new solver , we apply it in this paper to the dmft treatment of the mott transition in the multi - orbital hubbard model . we find an overall very good agreement with the general aspects of this problem , as known from numerical work and from some recently derived exact results . we also compare to other impurity solvers ( ipt , exact diagonalization , qmc , nca ) , particularly regarding the one - electron spectral function . this article is organized as follows . in section [ sec : rotor ] , we introduce a representation of fermion operators in terms of the phase variable dual to the total charge ( taking the finite @xmath0 , multi - orbital anderson impurity model as an example ) . in section [ sec : nca ] , a sigma - model representation of this phase variable is introduced , along with a generalization from @xmath3 to @xmath4 . in the limit of a large number of components , a set of coupled integral equations is derived . in section [ sec : siam ] , we test this `` dynamical slave rotor '' ( dsr ) approach on the single - impurity anderson model . the rest of the paper ( sec . [ sec : dmft ] ) is devoted to applications of this approach in the context of dmft , which puts in perspective the advantages and limitations of the new method . the present article emphasizes the role played by the total electron charge , and its conjugate phase variable . we introduce a representation of the physical electron in terms of two auxiliary fields : a fermion field which carries spin ( and orbital ) degrees of freedom , and the ( total ) charge raising and lowering operators which we represent in terms of a phase degree of freedom . the latter plays a role similar to a slave boson : here a `` slave '' @xmath3 quantum rotor is used rather than a conventional bosonic field . we first explain this construction on the simple example of an atomic problem , consisting in an @xmath5-fold degenerate atomic level subject to a local @xmath6-symmetric coulomb repulsion : @xmath7 ^ 2\ ] ] we use here @xmath8 as a convenient redefinition of the impurity level , and we recast the spin and orbital degrees of freedom into a single index @xmath9 ( for @xmath10 , we have a single orbital with two possible spin states @xmath11 ) . we note that @xmath12 is zero at half filling , due to particle - hole symmetry . the crucial point is that the spectrum of the atomic hamiltonian ( [ eq : hlocal ] ) depends only on the total fermionic charge @xmath13 and has a simple quadratic dependence on @xmath14 : @xmath15 ^ 2\ ] ] there are @xmath16 states , but only @xmath17 different energy levels , with degeneracies @xmath18 . in conventional slave boson methods @xcite , a bosonic field is introduced for each atomic state @xmath19 ( along with spin - carrying auxiliary fermions @xmath20 ) . hence , these methods are not describing the atomic spectrum in a very economical manner . the spectrum of ( [ eq : hlocal ] ) can actually be reproduced by introducing , besides the set of auxiliary fermions @xmath20 , a single additional variable , namely the angular momentum @xmath21 associated with a quantum @xmath3 rotor @xmath22 ( _ i.e. _ an angular variable in @xmath23 $ ] . indeed , the energy levels ( [ eq : atomic_spectrum ] ) can be obtained using the following hamiltonian @xmath24 a constraint must be imposed , which insures that the total number of fermions is equal to the @xmath3 angular momentum ( up to a shift ) : @xmath25\ ] ] this restricts the allowed values of the angular momentum to be @xmath26 , while in the absence of any constraint @xmath27 can be an arbitrary ( positive or negative ) integer . the spectrum of ( [ eq : hloc_f_theta ] ) is @xmath28 , with @xmath29 thanks to ( [ eq : contrainte ] ) , so that it coincides with ( [ eq : atomic_spectrum ] ) . to be complete , we must show that each state in the hilbert space can be constructed in terms of these auxiliary degrees of freedom , in a way compatible with the pauli principle . this is achieved by the following identification : @xmath30 in which @xmath31 denotes the antisymmetric fermion state built out of @xmath32 and @xmath33 fermions , respectively , and @xmath34 denotes the quantum rotor eigenstate with angular momentum @xmath27 , _ i.e. _ @xmath35 . the creation of a physical electron with spin @xmath36 corresponds to acting on such a state with @xmath20 as well as raising the total charge ( angular momentum ) by one unit . since the raising operator is @xmath37 , this leads to the representation : @xmath38 let us illustrate this for @xmath10 by writing the four possible states in the form : @xmath39 , @xmath40 , @xmath41 and @xmath42 , and showing that this structure is preserved by @xmath43 . indeed : @xmath44 the key advantage of the quantum rotor representation is that the original quartic interaction between fermions has been replaced in ( [ eq : hloc_f_theta ] ) by a simple kinetic term ( @xmath45 ) for the phase field . we now turn to an @xmath6-symmetric anderson impurity model in which the atomic orbital is coupled to a conduction electron bath : @xmath46 ^ 2\\ \nonumber & & + \sum_{k,{\sigma } } \epsilon_k c^\dagger_{k,{\sigma } } c_{k,{\sigma } } + \sum_{k,{\sigma } } v_k \left ( c^\dagger_{k,{\sigma } } { d_{\sigma}}+ { d_{\sigma}^{\dagger}}c_{k,{\sigma } } \right)\end{aligned}\ ] ] using the representation ( [ eq : repres ] ) , we can rewrite this hamiltonian in terms of the @xmath47 fields only : @xmath48 we then set up a functional integral formalism for the @xmath20 and @xmath22 degrees of freedom , and derive the action associated with ( [ eq : hinit ] ) . this is simply done by switching from phase and angular - momentum _ operators _ @xmath49 to _ fields _ @xmath50 depending on imaginary time @xmath51 $ ] . the action is constructed from @xmath52 $ ] , and an integration over @xmath53 is performed . it is also necessary to introduce a complex lagrange multiplier @xmath54 in order to implement the constraint @xmath55 . we note that , because of the charge conservation on the local impurity , @xmath54 can be chosen to be independent of time , with @xmath56 $ ] . this leads to the following expression of the action : @xmath57\end{aligned}\ ] ] we can recast this formula in a more compact form by introducing the hybridization function : @xmath58 and integrating out the conduction electron bath . this leads to the final form of the action of the @xmath6 anderson impurity model in terms of the auxiliary fermions and phase field : @xmath59 in this section , we present an alternative derivation of the expression ( [ eq : action2 ] ) of the action which does not rely on the concept of slave particles . this has the merit to give a more explicit interpretation of the phase variable introduced above , by relating it to a hubbard - stratonovich decoupling field . this section is however not essential to the rest of the paper , and can be skipped upon first reading . let us start with the imaginary time action of the anderson impurity model in terms of the physical electron field for the impurity orbital : @xmath60 ^ 2 \\ & \!\!+ & \!\!\!\!{\int_0^\beta \!\!\!\ ! ! d\tau } ' \delta(\tau\!-\!\tau ' ) \sum_{{\sigma } } { d_{\sigma}^{\dagger}}(\tau ) { d_{\sigma}}(\tau')\end{aligned}\ ] ] because we have chosen a @xmath6-symmetric form for the coulomb interaction , we can decouple it with only one bosonic hubbard - stratonovitch field @xmath61 : @xmath62 hence , a linear coupling of the field @xmath61 to the fermions has been introduced . the idea is now to eliminate this linear coupling for all the fourier modes of the @xmath63-field , except that corresponding to zero - frequency : @xmath64 $ ] . this can be achieved by performing the following gauge transformation : @xmath65 the reason for the second phase factor in this expression is that it guarantees that the new fermion field @xmath20 also obeys antiperiodic boundary conditions in the path integral . it is easy to check that this change of variables does not provide any jacobian , so that the action simply reads : @xmath66 } \end{aligned}\ ] ] we now set : @xmath67\right)\ ] ] and notice that the field @xmath68 has the boundary condition @xmath69 $ ] . it therefore corresponds to an @xmath3 quantum rotor , and the expression of the action finally reads : @xmath70 this is exactly expression ( [ eq : action2 ] ) , with the identification : @xmath71 . this , together with ( [ eq : int_theta ] ) , provides an explicit relation between the quantum rotor and lagrange multiplier fields on one side , and the hubbard - stratonovich field conjugate to the total charge , on the other . instead of using a phase field to represent the @xmath3 degree of freedom , one can use a constrained ( complex ) bosonic field @xmath72 with : @xmath73 the action ( [ eq : action2 ] ) can be rewritten in terms of this field , provided a lagrange multiplier field @xmath74 is used to implement this constraint : @xmath75 hence , the anderson model has been written as a theory of auxiliary fermions coupled to a non - linear @xmath3 sigma - model , with a constraint ( implemented by @xmath54 ) relating the fermions and the sigma - model field @xmath76 . a widely - used limit in which sigma models become solvable , is the limit of a large number of components of the field . this motivates us to generalize ( [ eq : action_de_phase ] ) to a model with an @xmath4 symmetry . the bosonic field @xmath77 is thus extended to an @xmath78-component complex field @xmath79 ( @xmath80 ) with @xmath81 . the corresponding action reads : @xmath82 let us note that this action corresponds to the hamiltonian : @xmath83 in this expression , @xmath84 denotes the angular momentum tensor associated with the @xmath79 vector . the hamiltonian ( [ eq : anderson_om ] ) is a generalization of the @xmath85 anderson impurity model to an @xmath86 model in which the total electronic charge is associated with a specific component of @xmath53 . it reduces to the usual anderson model for @xmath87 . in the following , we consider the limit where both @xmath5 and @xmath78 become large , while keeping a fixed ratio @xmath88 . we shall demonstrate that exact coupled integral equations can be derived in this limit , which determine the green s functions of the fermionic and sigma - model fields ( and the physical electron green s function as well ) . the fact that these coupled integral equations do correspond to the exact solution of a well - defined hamiltonian model ( eq . ( [ eq : anderson_om ] ) ) guarantees that no unphysical features ( like e.g violation of causality ) arise in the solution . naturally , the generalized hamiltonian ( [ eq : anderson_om ] ) is a formal extension of the anderson impurity model of physical interest . extending the charge symmetry from @xmath3 to @xmath4 is not entirely inocuous , even at the atomic level : as we shall see below , the energy levels of a single @xmath4 quantum rotor have multiple degeneracies , and depend on the charge ( angular momentum ) quantum number in a way which does not faithfully mimic the @xmath3 case . nevertheless , the basic features defining the generalized model ( a localized orbital subject to a coulomb charging energy and coupled to an electron bath by hybridization ) are similar to the original model of physical interest . in this section , we derive coupled integral equations which become exact in the limit where _ both _ @xmath78 and @xmath5 are large with a fixed ratio : @xmath89 following @xcite ( see also @xcite ) , the two body interaction between the auxiliary fermions and the sigma - model fields is decoupled using ( bosonic ) bi - local fields @xmath90 and @xmath91 depending on two times . hence , we consider the action : @xmath92 in the limit ( [ eq : limit ] ) , this action is controlled by a saddle - point , at which the lagrange multipliers take static expectation values @xmath54 and @xmath93 , while the saddle point values of the @xmath14 and @xmath94 fields are translation invariant functions of time @xmath95 and @xmath96 . introducing the imaginary - time green s functions of the auxiliary fermion and sigma - model fields as : @xmath97 we differentiate the effective action ( [ eq : actionlargenm ] ) with respect to @xmath98 and @xmath99 , which leads to the following saddle - point equations : @xmath100 and @xmath101 . the functions @xmath102 ( @xmath103 ) and @xmath104 ( @xmath105 ) define fermionic and bosonic self - energies : @xmath106 where @xmath107 ( _ resp . _ @xmath108 ) is a fermionic ( _ resp . _ bosonic ) matsubara frequency . the saddle point equations read : @xmath109 together with the constraints associated with @xmath54 and @xmath93 : @xmath110\end{aligned}\ ] ] there is a clear similarity between the structure of these coupled integral equations and the infinite-@xmath0 nca equations @xcite . we note also significant differences , such as the constraint equations . furthermore , the finite value of the coulomb repulsion @xmath0 enters the bosonic propagator ( [ eq : gx ] ) in a quite novel manner . the two key ingredients on which the present method are based is the use of a slave rotor representation of fermion operators , and the use of integral equations for the frequency - dependent self - energies and green s functions . for this reason , we shall denote the integral equations above under the name of `` dynamical slave rotor '' method ( dsr ) in the following . we make here some technical remarks concerning these integral equations . first , we clarify how the interaction parameter @xmath0 was scaled in order to obtain the dsr equations above . this issue is related to the manner in which the atomic limit ( @xmath111 ) is treated in this method . in the original @xmath3 atomic hamiltonian ( [ eq : hlocal ] ) , the charge gap between the ground - state and the first excited state is @xmath112 at half - filling ( @xmath113 ) . in the dsr method , the charge gap is associated with the gap in the slave rotor spectrum . if the @xmath4 generalization of ( [ eq : hlocal ] ) is written as in ( [ eq : anderson_om ] ) : @xmath114 the spectrum reads : @xmath115 . as a result , the energy difference from the ground state to the first excited state is @xmath116 at large @xmath78 , whereas it is @xmath112 at @xmath87 . in order to use the dsr method in practice as an approximate impurity solver , the parameter @xmath0 should thus be normalized in a different way than in ( [ eq : anderson_om ] ) , so that the gap is kept equal to @xmath112 in the large - m limit as well . technically this can be enforced by choosing the following normalization : @xmath117 instead of ( [ eq : rotor_om ] ) . note that this scaling coincides with ( [ eq : rotor_om ] ) for @xmath87 , but does yield @xmath118 for large - m , as desired . this definition of @xmath0 was actually used when writing the saddle - point integral equations ( [ eq : gx ] ) , although we postponed the discussion of this point to the present section for reasons of simplicity . let us elaborate further on the accuracy of the dsr integral equations in the atomic limit . in appendix [ app : atomic ] , we show that the physical electron spectral function obtained within dsr in the atomic limit coincides with the exact @xmath3 result at half - filling and at @xmath119 . this is a non - trivial result , given the fact that the constraint is treated on average and the above remark on the spectrum spectrum . in contrast to nca , the dsr method ( in its present form ) is not based by construction on a strong - coupling expansion around the exact atomic spectrum , so that this is a crucial check for the applicability of this method in practice . in the context of dmft for example , it is essential in order to describe correctly the mott insulating state @xcite . however , the dsr integral equations fail to reproduce exactly the o(2 ) atomic limit off half - filling , as explained in appendix [ app : atomic ] . deviations become severe for too high dopings , as discussed in sec . [ sec : doping ] . this makes the present form of dsr applicable only for systems in the vicinity of half - filling . we now discuss some general spectral properties of the dsr solver . from the representation of the physical electron field @xmath120 , and from the convention chosen for the pseudo - particules green s functions ( [ eq : gfdef]-[eq : gxdef ] ) , the one - electron physical green s function is simply expressed as : @xmath121 therefore eq . ( [ eq : col3 ] ) , combined with the fact that @xmath20 has a @xmath122 discontinuity at @xmath123 ( which is obvious from ( [ eq : gf ] ) ) , shows that @xmath124 possesses also a @xmath122 jump at zero imaginary time . this ensures that the physical spectral weight is unity in our theory , and thus that physical spectral function are correctly normalized . because the dsr integral equations result from a controlled large @xmath125 limit , it also insures that the physical self - energy always have the correct sign ( _ i.e. _ @xmath126 ) . this is _ not _ the case @xcite for the finite @xmath0 version of the nca @xcite which is constructed as a resummation of the strong - coupling expansion in the hybridization function @xmath127 ( strong coupling resummations generically suffer from non - causality , see also @xcite for an illustration . ) finally , we comment on the non - interacting limit @xmath128 . this is a major failure of the usual nca , which limits its applicability in the weakly correlated regime . in the dsr formalism , this limit is _ exact _ as can be noticed from equation ( [ eq : gx ] ) . indeed , as @xmath0 vanishes , only the zero - frequency component of @xmath129 survives , so that @xmath130 simply becomes a constant . because of the constraint ( [ eq : col3 ] ) , we get correctly @xmath131 at @xmath1 . from ( [ eq : col2 ] ) and ( [ eq : green ] ) , this proves that @xmath132 is the non - interacting green s function : @xmath133 we finally acknowledge that an alternative dynamical approximation to the finite @xmath0 anderson model @xcite was recently developed as an extension of nca by kroha , wlfle and collaborators ( a conventional slave boson representation was used in this work ) . many progresses have been made following this method , but , to the authors knowledge , this technique has not yet been implemented in the context of dmft ( one of the reasons is its computational cost ) . by developing the dsr approximation , we pursue a rather complementary goal : the aim here is not to improve the low - energy singularities usually encountered with integral equations , but rather to have a fast and efficient solver which reproduces correctly the main features of the spectral functions and interpolates between weak and strong coupling . in that sense , it is very well adapted to the dmft context . we now discuss the application of the dsr in the simplest setting : that of a single impurity hybridized with a fixed bath of conduction electrons . for simplicity , we focus on the half - filled , particle - hole symmetric case , which implies @xmath134 . the doped ( or mixed valence ) case will be addressed in the next section , in the context of dmft . as the strength of the coulomb interaction @xmath0 is increased from weak to strong coupling , two well - known effects are expected ( see e.g @xcite ) . first the width of the low - energy resonance is reduced from its non - interacting value @xmath135 . as one enters the kondo regime @xmath136 ( with @xmath137 the conduction electrons bandwidth ) , this width becomes a very small energy scale , of the order of the kondo temperature : @xmath138 a ( local ) fermi liquid description applies , with quasiparticles having a large effective mass and small weight : @xmath139 . the impurity spin is screened for @xmath140 . second , the corresponding spectral weight is transferred to high energies , into `` hubbard bands '' associated to the atomic - like transitions ( adding or removing an electron into the half - filled impurity orbital ) , broadened by the hybridization to the conduction electron bath . the suppression of the low - energy spectral weight corresponds to the suppression of the charge fluctuations on the local orbital . these satellites are already visible at moderate values of the coupling @xmath141 . as temperature is increased from @xmath140 to @xmath142 , the kondo quasiparticle resonance is quickly destroyed , and the missing spectral weight is added to the hubbard bands . the aim of this section is to investigate whether the integral equations introduced in this paper reproduce these physical effects in a satisfactory manner . we have solved numerically these integral equations by iteration , both on the imaginary axis and for real frequencies . working on the imaginary axis is technically much easier . a discretization of the interval @xmath143 $ ] is used ( with typically 8192 points , and up to 32768 for reaching the lowest temperatures ) , as well as fast fourier transforms for the green s functions . searching by dichotomy for the saddle - point value of the lagrange multiplier @xmath93 ( eq . [ eq : col3 ] ) is conveniently implemented at each step of the iterative procedure . technical details about the analytic continuation of ( [ eq : col1]-[eq : col4 ] ) to real frequencies , as well as their numerical solution are given in appendix [ appendix1 ] . -level spectral function @xmath144 for fixed @xmath145 and inverse temperatures @xmath146 , with the conduction electron bath described in the text , width=264 ] on fig . [ fig : pousse_kondo ] , we display our results for the impurity - orbital spectral function @xmath147 , at three different temperatures ( the density of states of the conduction electron bath is chosen as a semi - circle with half - width @xmath148 , the resonant level width is @xmath149 and we take @xmath145 ) . the growth of the kondo resonance as the temperature is lowered is clearly seen . the temperatures in fig . [ fig : pousse_kondo ] have been chosen such as to illustrate three different regimes : for @xmath150 , no resonance is seen and the spectral density displays a `` pseudogap '' separating the two high - energy bands ; for @xmath151 transfer of spectral weight to low energy is seen , resulting in a fully developed kondo resonance for @xmath152 . we have not obtained an analytical determination of the kondo temperature within the present scheme . in the case of nca equations , it is possible to derive a set of differential equations in the limit of infinite bandwith which greatly facilitate this . this procedure can not be applied here , because of the form ( [ eq : gx ] ) of the boson propagator . nevertheless , we checked that the numerical estimates of the width of the kondo peak is indeed exponentially small in @xmath0 as in formula ( [ eq : tk ] ) , see fig . [ fig : compare_tk ] . however , because @xmath0 is normalized as in ( [ eq : normalization ] ) , ( which gives the correct atomic limit ) , the prefactor inside the exponential appears to be twice too small . from the exact formula ( line ) and from the numerical solution of the dsr equations ( dots ) . units of energy are such that @xmath153.,width=264 ] on fig . [ fig : analyse_tk ] , we display the spectral function for a fixed low temperature and increasing values of @xmath0 . the strong reduction of the kondo scale ( resonance width ) upon increasing @xmath0 is clear on this figure . we note that the high - energy peaks have a width which remains of order @xmath154 , independently of @xmath0 , which is satisfactory . however , we also note that they are not peaked exactly at the atomic value @xmath155 , which might be an artefact of these integral equations . the shift is rather small however . at low temperature ( @xmath156 ) and for increasing @xmath157,width=264 ] fig . [ fig : pseudo ] illustrates how the high- and low- energy features in the d - level spectral functions are associated with corresponding features in the auxiliary particle spectral functions @xmath158 and @xmath159 . in particular , the sigma - model boson ( slave rotor ) is entirely responsible for the hubbard bands at high energy ( as expected , since it describes charge fluctuations ) . and the kondo resonance is visible in @xmath160 , broken curve , whereas @xmath161 displays higher energy features , plain curve , width=264 ] as stressed in the introduction , an advantage of our scheme is that the d - level self - energy is always causal , even for small @xmath0 or large doping . this is definitely an improvement as compared to the usual @xmath0-nca approximation . this is illustrated by fig . [ fig : sd ] , from which it is also clear that @xmath162 decreases ( and eventually vanishes ) as @xmath0 goes to zero ( for a more detailed discussion and comparison to nca , see sec . [ sec : nca ] ) . however , it is also clear from this figure that the low - energy behavior of the self - energy is not consistent with fermi liquid theory . this is a generic drawback of nca - like integral equation approaches , that we now discuss in more details . ( upper curve ) and @xmath163 ( lower curve ) at @xmath156,width=264 ] we discuss here the low - energy behavior of the integral equations ( in the case of a featureless conduction electron bath ) , for both the d - level and auxiliary field green s functions . as explained below , this low - energy behavior depends sensitively on the ratio @xmath164 which is kept fixed in the limit considered in this paper . the calculations are detailed in appendix [ app : low ] , where we establish the following . * _ friedel sum rule_. the zero - frequency value of the d - level spectral function at @xmath119 is independent of @xmath0 and reads : @xmath165 this is to be contrasted with the exact value for the @xmath3 model ( @xmath87 ) , which is independent of @xmath5 and reads : @xmath166 this is the _ friedel sum rule _ @xcite , which in this particle - hole symmetric case simply follows from the fermi - liquid requirement that the ( inverse ) lifetime @xmath167 should vanish at @xmath119 ( since @xmath168 ) . as a result , @xmath169 is pinned at its non - interacting value . the integral equations discussed here yield a non - vanishing @xmath167 ( albeit always negative in order to satisfy causality ) , and hence do not describe a fermi liquid at low energy . the result ( [ eq : friedel ] ) is identical to that found in the nca for @xmath170 , but holds here for arbitrary @xmath0 . there is actually no contradiction between this remark and the fact that our integral equations yield the exact spectral function in the @xmath171 limit . indeed , the limit @xmath172 at finite @xmath173 does not commute with @xmath174 at finite @xmath0 ( in which ( [ eq : friedel ] ) holds ) . * _ low - frequency behavior . _ the auxiliary particle spectral functions have a low - frequency singularity characterized by exponents which depend continuously on @xmath175 ( as in the @xmath170 nca ) : @xmath176 , @xmath177 , with : @xmath178 . these behavior are characterized more precisely in appendix [ app : low ] . a power - law behavior is also found for the physical self - energy @xmath179 at low frequency ( as evident from fig . [ fig : sd ] ) . let us comment on the origin of these low - energy features , as well as on their consequences for the practical use of the present method . first , it is clear from expression ( [ eq : anderson_om ] ) that the anderson impurity hamiltonian generalized to @xmath180 actually involves _ @xmath78 channels of conduction electrons_. hence , the non - fermi liquid behaviour found when solving the integral equations associated with the @xmath181 limit simply follows from the fact that multi - channel models lead to _ overscreening _ of the impurity spin , and correspond to a non - fermi liquid fixed point . in that sense , these integral equations reproduce very accurately the expected low - energy physics , as previously studied for the simplest case of the kondo model in @xcite . naturally , this means that the use of such integral equations to describe the one - channel ( exactly screened ) case becomes problematic in the low - energy region . in particular , the exact friedel sum rule is violated , the d - level lifetime remains finite at low energy and non - fermi liquid singularities are found . while the approach is reasonable in order to reproduce the overall features of the one - electron spectra , it should not be employed to calculate transport properties at low energy for example . we note however that the deviation from the exact friedel sum rule vanishes in the @xmath182 limit . this is expected from the fact that in this limit the number of channels ( @xmath78 ) is small as compared to orbital degeneracy ( @xmath5 ) . the violation of the sum rule remains rather small even for reasonable values of @xmath175 . this parameter can actually be used as an adjustable parameter when using the present method as an approximate impurity solver . there is no fundamental reason for which @xmath183 should provide the best approximate description of the spectral functions of the one - channel case . we shall use this possibility when applying this method in the dmft context in the next section : there , we choose @xmath184 in order to adjust to the known critical value of @xmath0 for a single orbital . we also used this value in the calculations reported above . the zero - frequency value of the spectral density is thus @xmath185 , while the friedel sum rule would yield @xmath186 ( cf . [ fig : analyse_tk ] ) . hence the violation of the sum rule is a small effect ( of the order of @xmath187 ) , comparable to the one found with `` numerically exact '' solvers , due to discretization errors @xcite . also , we point out that the pinning of @xmath169 at a value independent of @xmath0 ( albeit not that of the friedel sum - rule ) is an important aspect of the present method , which will prove to be crucial in the context of dmft in order to recover the correct scenario for the mott transition . finally , we emphasize that increasing the parameter @xmath175 also corresponds to increasing the orbital degeneracy of the impurity level . this will be studied in more details in sec . [ sec : multi - orbital ] . in particular , we shall see that correlation effects become weaker as @xmath175 is increased ( for a given value of @xmath0 ) , due to enhanced orbital fluctuations . dynamical mean - field theory has led to significant progress in our understanding of the physics of a correlated metal close to the mott transition @xcite . the detailed description of this transition itself within dmft is now well established @xcite . in this section , we use these established results as a benchmark and test the applicability of the method introduced in this paper in the context of dmft , with very encouraging results . as explained above , this is particularly relevant in view of the recent applications of dmft to electronic structure calculations of correlated solids @xcite , which call for efficient multi - orbital impurity solvers . as is well - known @xcite , dmft maps a lattice hamiltonian onto a self - consistent quantum impurity model . we discuss first the half - filled hubbard model , and address later the doped case . we then have to solve a particle - hole symmetric anderson impurity model : @xmath188 ^ 2 \\ \nonumber & + & { \int_0^\beta \!\!\!\ ! d\tau}{\int_0^\beta \!\!\!\ ! d\tau } ' \ ; \delta(\tau-\tau ' ) \sum_{{\sigma } } { d_{\sigma}^{\dagger}}(\tau ) { d_{\sigma}}(\tau')\end{aligned}\ ] ] subject to the self - consistency condition : @xmath189 in this expression , a semi - circular density of states with half - bandwith @xmath190 has been considered , corresponding to a infinite - connectivity bethe lattice ( @xmath191 ) with hopping @xmath192 . in the following , we shall generally express all energies in units of @xmath193 ( @xmath194 ) . in practice , one must iterate numerically the `` dmft loop '' : @xmath195 , using some `` impurity solver '' . here , we make use of the integral equations ( [ eq : col1]-[eq : col4 ] ) . the hybridization function @xmath127 being determined by the self - consistency condition ( [ eq : self ] ) , there are only two free parameters , the local coulomb repulsion @xmath0 and the temperature @xmath196 ( normalized by @xmath193 ) . we display in fig . [ fig : transition ] the spectral functions obtained at low temperature , for increasing values of @xmath0 , and in fig . [ fig : phase ] the corresponding phase diagram . the value of the parameter @xmath175 has been adapted to the description of the one - orbital case ( see below ) . the most important point is that we find a _ coexistence region _ at low - enough temperature : for a range of couplings @xmath197 , both a metallic solution and an insulating solution of the ( paramagnetic ) dmft equations exist . the mott transition is thus first - order at finite temperatures . this is in agreement with the results established for this problem by solving the dmft equations with controlled numerical methods @xcite , as well as with analytical results @xcite . in particular , the spectral functions that we obtain ( fig . [ fig : transition ] ) display the well - known separation of energy scales found within dmft : there is a gradual narrowing of the quasiparticle peak , together with a preformed mott gap at the transition . in the next section , we compare these spectral functions to those obtained using other approximate solvers . as pointed out there , despite some formal similarity in the method , it is well - known that the standard @xmath0-nca does not reproduce correctly this separation of energy scales close to the transition @xcite . and @xmath198 for the half - filled hubbard model within dmft , as obtained with the dsr solver , width=264 ] let us explain how the parameter @xmath175 has been chosen in these calculations . as demonstrated below , the values of the critical couplings @xmath199 and @xmath200 ( and hence the whole phase diagram and coexistence window ) strongly depends on the value of this parameter this is expected , since @xmath175 is a measure of orbital degeneracy . what has been done in the calculations displayed above is to choose @xmath175 in such a way that the known value @xcite @xmath201 of the critical coupling at which the @xmath119 metallic solution disappears in the single - orbital case , is accurately reproduced . we found that this requires @xmath202 ( note that @xmath203 in the one - orbital case , so that the best agreement is not found by a naive application of the large @xmath125 limit ) . this value being fixed , we find a critical coupling @xmath204 in good agreement with the value from ( adaptative ) exact diagonalizations @xmath205 . the whole domain of coexistence in the @xmath206 plane is also in good agreement with established results ( in particular we find the critical endpoint at @xmath207 , while qmc yields @xmath208 ) . these are very stringent tests of the applicability of the present method , since we have allowed ourselves to use only one adjustable parameter ( @xmath175 ) . in sec.[sec : multi - orbital ] , we study how the mott transition depends on the number of orbitals , which further validates the procedure followed here . let us now compare the spectral functions obtained by the present method with other impurity solvers commonly used for solving the dmft equations . we start with the iterated perturbation theory approximation ( ipt ) and the exact diagonalization method ( ed ) . both methods have played a major role in the early developments of the dmft approach to the mott transition @xcite . a comparison of the spectral functions obtained by the present method to those obtained with ipt and ed is displayed in fig . [ fig : compare_dos_u2.4 ] , for a value of @xmath0 corresponding to a correlated metal close to the transition . and @xmath209,width=264 ] the overall shape and characteristic features of the spectral function are quite similar for the three methods . a narrow quasiparticle peak is formed , together with hubbard bands , and there is a clear separation of energy scales between the width of the central peak ( related to the quasiparticle weight ) and the `` preformed '' mott ( pseudo-)gap associated with the hubbard bands : this is a distinguishing aspect of dmft . it is crucial for an approximate solver to reproduce this separation of energy scales in order to yield a correct description of the mott transition and phase diagram . there are of course some differences between the three methods , on which we now comment . first , we note that the ipt approximation has a somewhat larger quasiparticle bandwith . this is because the transition point @xmath200 is overestimated within ipt ( cf . [ fig : phase ] ) , so that a more fair comparison should perhaps be made at fixed @xmath210 . it is true however , that the dsr method has a tendency to underestimate the quasiparticle bandwith , and particularly at smaller values of @xmath0 . accordingly , the hubbard bands have a somewhat too large spectral weight , but are correctly located in first approximation . the detailed shape of the hubbard bands is not very accurately known , in any case . ( the ed method involves a broadening of the delta - function peaks obtained by diagonalizing the impurity hamiltonian with a limited number of effective orbitals , so that the high - energy behaviour is not very accurate on the real axis . this is also true , actually , of the more sophisticated numerical renormalization group ) . we emphasize that , since the dsr method does not have the correct low - frequency fermi liquid behaviour , the quasiparticle bandwith should be interpreted as the width of the central peak in @xmath147 ( while the quasiparticle weight @xmath211 can not be defined formally ) . -nca spectral function for @xmath212 at low temperature.,width=264 ] in fig [ fig : unca ] and fig . [ fig : u_petit ] , we display the spectral functions obtained by using the nca method ( extended at finite @xmath0 in the simplest manner ) . it is clear that the @xmath0-nca underestimates considerably the quasiparticle bandwith ( and thus yields a mott transition at a rather low value of the coupling ) . this is not very surprising , since this method is based on a strong coupling expansion around the atomic limit , and one could think again of a comparison for fixed @xmath213 . more importantly , the @xmath0-nca misses the important separation of energy scales between the central peak and the mott gap . as a result , it does not reproduce correctly the phase diagram for the mott transition within dmft ( in particular regarding coexistence ) . a related key observation is that the @xmath0-nca does not have the correct weak - coupling ( small @xmath0 ) limit . to illustrate this point , we display in fig . [ fig : u_petit ] the @xmath0-nca and dsr spectral functions for a tiny value of the interaction @xmath214 : the dsr result is shown to approach correctly the semi - circular shape of the non - interacting density of states , while the @xmath0-nca displays a characteristic inverted v - shape : in this regime the violation of the friedel sum rule becomes large and the negative lifetime pathology is encountered within @xmath0-nca . in contrast , the dsr yields a pinning of the spectral density at a @xmath0-independent value and does not lead to a violation of causality . it should be emphasized however that , even though it yields the exact @xmath1 density of states at @xmath119 , the dsr method is not quantitatively very accurate in the weak - coupling regime . showing how @xmath0-nca overshoots the friedel s sum rule . the slave rotor method is correctly converging towards the free density of states ( semi - circular),width=264 ] finally , we demonstrate that two - particle correlators in the charge sector can also be reliably studied with the dsr method , taking the fraction of doubly occupied sites @xmath215 as an example . using the constraint ( [ eq : contrainte ] ) , we calculate the connected charge susceptibility in the following manner : @xmath216\end{aligned}\ ] ] the double occupancy is obtained by taking the equal - time value : @xmath217 in fig . [ fig : double ] , we plot the double occupancy obtained in that manner , in comparison to the ed and ipt results . within ed , @xmath218 is calculated directly from the charge correlator . within ipt , @xmath219 is not approximated very reliably @xcite , but the double occupancy can be accurately calculated by taking a derivative of the internal energy with respect to @xmath0 . [ fig : double ] demonstrates that the dsr method is very accurate in the mott transition region and in the insulator . in particular , the hysteretic behaviour is well reproduced . in the weak coupling regime however , the approximation deteriorates . this issue actually depends on the quantity : the physical green s function has the correct limit @xmath128 , as emphasized above , but this is not true for the charge susceptibility ( hence for @xmath218 ) . the mathematical reason is that the constraint ( [ eq : contrainte ] ) is crucial for writing ( [ eq : chicharge ] ) , but this constraint is only treated on average within our method . this shows the inherent limitations of slave - boson techniques for evaluating two - particle properties . the frequency dependence of two - particle correlators will be dealt with in a forthcoming publication . as a function of @xmath0 . the ed result is indicated with squares . dsr overshoots the exact result on the left , whereas ipt is slightly displaced to the right.,width=264 ] in this section , we apply the dsr method to study the dependence of the mott transition on orbital degeneracy @xmath5 . we emphasize that there are at this stage very few numerical methods which can reliably handle the multi - orbital case , specially when @xmath5 becomes large . multi - orbital extensions of ipt have been studied @xcite , but the results are much less satisfactory than in the one - orbital case . the ed method is severely limited by the exponential growth of the size of the hilbert space . the mott transition has been studied in the multi - orbital case using qmc , with a recent study @xcite going up to @xmath220 ( 4 orbitals with spin ) . furthermore , we have recently obtained @xcite some analytical results on the values of the critical couplings in the limit of large-@xmath5 . those , together with the qmc results , can be used as a benchmark of the dsr approximation presented here . as explained above , a value of @xmath221 was found to describe best the single orbital case ( @xmath10 ) . hence , upon increasing @xmath5 , we choose the parameter @xmath175 such that @xmath222 . , @xmath223 , and coexistence region for one , two , three and four orbitals ( @xmath224 ) as obtained by dmft ( dsr solver),width=264 ] in fig . [ fig : lignes ] , we display the coexistence region in the @xmath206 plane for increasing values of @xmath5 , as obtained with dsr . the values of the critical interactions grow with @xmath5 . a fit of the transition lines @xmath225 ( where the insulator disappears ) and @xmath223 ( where the metal disappears ) yields : @xmath226 these results are in good accordance with both the qmc data @xcite , and the exact results established in @xcite . when increasing the number of orbitals , the coexistence region widens and the critical temperature associated with the endpoint of the mott transition line also increases . we display on fig . [ fig : dos_n ] the dsr result for the spectral function for @xmath10 and @xmath227 , at a fixed value of @xmath0 and @xmath196 . two main effects should be noted . first , correlations effects in the metal become weaker as @xmath5 is increased ( for a fixed @xmath0 ) , as clear e.g from the increase of the quasiparticle bandwith . this is due to increasing orbital fluctuations . second , the hubbard bands also shift towards larger energies , an effect which can be understood in the way atomic states broadens in the insulator @xcite . to conclude this section , we have found that the dsr yields quite satisfactory results when used in the dmft context for multi - orbital models . in the future , we plan to use this method in the context of realistic electronic structure calculations combined with dmft , in situations where `` numerically exact '' solvers ( e.g qmc ) become prohibitively heavy . there are however some limitations to the use of dsr ( at least in the present version of the approach ) , which are encountered when the occupancy is not close to @xmath228 . we examine this issue in the next section . , @xmath209 and for one ( full line ) and two orbitals ( broken line),width=264 ] up to now , we studied the half - filled problem ( _ i.e. _ containing exactly @xmath228 electrons ) , with exact particle - hole symmetry . in that case , @xmath229 and the lagrange multiplier @xmath54 enforcing the constraint ( [ eq : col4 ] ) could be set to @xmath230 . in realistic cases , particle - hole symmetry will be broken ( so that @xmath231 ) , and we need to consider fillings different from @xmath228 . hence , the lagrange multiplier @xmath54 must be determined in order to fulfill ( [ eq : col4 ] ) , which we rewrite more explicitly as : @xmath232 in this expression , @xmath233 is the average occupancy per orbital flavor . ( note that the number of auxiliary fermions and physical fermions coincide , as clear from ( [ eq : green ] ) ) . @xmath234 is related to the d - level position ( or chemical potential , in the dmft context ) by : @xmath235 and @xmath209 ) with @xmath236 ( corresponding to @xmath237 respectively),width=264 ] we display in fig . [ fig : dope_isolant ] the spectral function obtained with the dsr solver when doping the mott insulator away from half - filling . a critical value of the chemical potential ( _ i.e. _ of @xmath238 ) is required to enter the metallic state . the spectral function displays the three expected features : a lower and upper hubbard bands , as well as a quasiparticle peak ( which in this case where the doping is rather large , is located almost at the top of the lower hubbard band ) . however , it is immediately apparent from this figure that the dsr method in the doped case overestimates the spectral weight of the upper hubbard band . this can be confirmed by a comparison to other solvers ( e.g ed ) . note that the energy scale below which a causality violation appears within u - nca becomes rapidly large as the system is doped , while no such violation occurs within dsr . the dsr method encouters severe limitations however as the total occupancy becomes very different from @xmath228 . this is best understood by studying the dependence of the occupancy upon @xmath238 , in the atomic limit . as explained in appendix . [ app : atomic ] , the use of sigma - model variables @xmath77 ( in the large-@xmath78 limit ) results in a poor description of the @xmath234 vs. @xmath238 dependence ( `` coulomb staircase '' ) . as a result , it is not possible to describe the mott transitions occurring in the multi - orbital model at integer fillings different from @xmath228 using the dsr approximation . it should be emphasized however that this pathology is only due to the approximation of the @xmath3 quantum rotor @xmath37 by a @xmath239 sigma - model field . a perfect description of the coulomb staircase is found for all fillings when treating the constraint ( [ eq : contrainte ] ) on average while keeping a true quantum rotor @xcite . hence , it seems feasible to overcome this problem and extend the practical use of the ( dynamical ) slave rotor approach to all fillings . we intend to address this issue in a future work . we conclude by summarizing the strong points as well as the limitations of the new quantum impurity solver introduced in this paper , as well as possible extensions and applications . on the positive side , the dsr method provides an interpolating scheme between the weak coupling and atomic limits ( at half - filling ) . it is also free of some of the pathologies encountered in the simplest finite-@xmath0 extensions of nca ( negative lifetimes at low temperature ) . when applied in the context of dmft , it is able to reproduce many of the qualitatively important features associated with the mott transition , such as coexisting insulating and metallic solutions and the existence of two energy scales in the dmft description of a correlated metal ( the quasi - particle coherence bandwith and the `` preformed gap '' ) . hence the dsr solver is quite useful in the dmft context , at a low computational cost , and might be applicable to electronic structure calculations for systems close to half - filling when the orbital degeneracy becomes large . to incorporate more realistic modelling , one can introduce different energy levels for each correlated orbital , while the extension to non - symmetric coulomb interactions ( such as the hund s coupling ) may require some additional work . the dsr method does not reproduce fermi - liquid behaviour at low energy however , which makes it inadequate to address physical properties in the very low - energy regime ( as is also the case with nca ) . the main limitation however is encountered when departing from half - filling ( i.e from @xmath228 electrons in an @xmath5-fold degenerate orbital ) . while the dsr approximation can be used at small dopings , it fails to reproduce the correct atomic limit when the occupancy differs significantly from @xmath228 ( and in particular can not deal with the mott transition at other integer fillings in the multi - orbital case ) . we would like to emphasize however that this results from extending the slave rotor variable to a field with a large number of components . it is possible to improve this feature of the dsr method by dealing directly with an @xmath3 phase variable , which does reproduce accurately the atomic limit even when the constraint is treated at the mean - field level . we intend to address this issue in a future work . another possible direction is to examine systematic corrections beyond the saddle - point approximation in the large @xmath125 expansion . finally , we would like to outline some other possible applications of the slave rotor representation introduced in this paper ( sec . [ sec : rotor ] ) . this representation is both physically natural and economical . in systems with strong coulomb interactions , the phase variable dual to the local charge is an important collective field . promoting this single field to the status of a slave particle avoids the redundancies of usual slave - boson representations . in forthcoming publications , we intend to use this representation for : i ) constructing impurity solvers in the context of _ extended _ dmft @xcite , in which the frequency - dependent charge correlation function must be calculated @xcite ii ) constructing mean - field theories of _ lattice _ models of correlated electrons ( e.g the hubbard model ) @xcite and iii ) dealing with quantum effects on the coulomb blockade in mesoscopic systems . in this appendix we prove the claim that the atomic limit of the model is exact at half - filling and at zero temperature ( at finite temperature , deviations from the exact result are of order @xmath240 and therefore negligeable for pratical purposes ) . to do this , we first extract the values of the mean - field parameters @xmath93 and @xmath54 from the saddle - point equations at zero temperature and @xmath241 . @xmath242 performing the integrals shows that @xmath243 and @xmath244 , so that @xmath245 . if @xmath246 , the equations lead actually to a solution with an empty or full valence that we show on fig . [ fig : staircase ] . we can now compute the physical green s function @xmath247 from the pseudo - propagators @xmath248 and : @xmath249 performing the convolution in imaginary frequency and taking the limit @xmath119 leads to : @xmath250 because @xmath251 this is the correct atomic limit of the single - band model ( at half - filling ) . the result for the empty or full orbital is however not accurate , as shown in figure [ fig : staircase ] . this discrepancy with the correct result ( even for one orbital ) finds its root in the large @xmath78 treatment of the slave rotor @xmath77 . -level position in the rotor description ( full curve ) and the exact result ( dot curve ) , in the two - orbital case.,width=264 ] here we show how the saddle - point equations ( [ eq : col1]-[eq : col4 ] ) can be analytically continued along the real axis . we start with @xmath252 , which can be first fourier transformed into : @xmath253 where we used the spectral representation @xmath254 for each green s function . the notations here are quite standard : @xmath255 , @xmath256 is the fermi factor , and @xmath257 denotes the bose factor . it is then immediate to continue @xmath258 in equation ( [ eq : continue ] ) , and using again the spectral decomposition ( [ eq : spectral ] ) , we derive an equation between retarded quantities : @xmath259 a calculation along the same lines for the fermionic self - energy @xmath260 leads to : @xmath261 the numerical implementation is then straightforward because ( [ eq : sxreal ] ) and ( [ eq : sfreal ] ) can each be expressed as the convolution product of two quantities , so that they can be calculated rapidly using fft . the algorithm is looped back using the dyson equations ( for real frequency ) : @xmath262 at each iteration , @xmath93 and @xmath54 are determined using a bisection on the equations ( [ eq : col3]-[eq : col4 ] ) , which can be properly expressed in terms of retarded green s functions : @xmath263 where @xmath234 , the average number of physical fermions , is : @xmath264 we note here that solving these real - time integral equations can be quite difficult deep in the kondo regime of the anderson model , or very close to the mott transition for the full dmft equations . the reason is that @xmath265 and @xmath266 develop low energy singularities ( this is analytically shown in the next appendix ) , that make the numerical resolution very unprecise if one uses fft . in that case , it is necessary to introduce a logarithmic mesh of frequency ( loosing the benefit of the fft speed , but increasing the accuracy ) , or to perform a pad extrapolation of the imaginary time solution . we present here for completeness the derivation of the slave rotor friedel s sum rule , equation ( [ eq : friedel ] ) , at half - filling . the idea , motivated by the numerical analysis as well as theoretical arguments @xcite , is that the pseudo - particles develop low frequency singularities at zero temperature : @xmath267 using the spectral representation @xmath268 we deduce the long time behavior of the green s functions ( we denote by @xmath269 the gamma function ) : @xmath270 we have similarly @xmath271 if one assumes a regular bath density of states at zero frequency . the previous expressions allow to extract the long time behavior of the pseudo - self - energies ( using the saddle - point equations ( [ eq : col1]-[eq : col2 ] ) ) : @xmath272 the next step is to use ( [ eq : repr ] ) the other way around to get from ( [ eq : sxw]-[eq : sfw ] ) the @xmath273-dependance of the self - energies : @xmath274 it is necessary at this point to calculate the real part of both self - energies . this can be done using the kramers - kronig relation , but analyticity provides a simpler route . indeed , by noticing that @xmath275 is an analytic function of @xmath276 and must be uni - valuated above the real axis , we find : @xmath277 } |z|^{1-\alpha_f}\\ \nonumber \sigma_f(z ) & = & \frac{1}{\pi } \frac{a_x \delta''(0)}{1-\alpha_x } \frac{e^{i\alpha_x\pi/2}}{\sin[\alpha_x\pi/2 ] } |z|^{1-\alpha_x}\end{aligned}\ ] ] we can therefore collect the previous expressions , using dyson s formula for complex argument : @xmath279 and this enables us to extract the leading exponents , as well as the product of the amplitudes : @xmath280 this proves ( [ eq : friedel ] ) . in principle , next leading order corrections can be computed by the same line of arguments , although this is much more involved @xcite . non fermi liquid correlations in the physical green s function would appear in this computation . for recent reviews , see : _ `` strong coulomb correlations in electronic structure calculations '' _ ( advances in condensed matter science ) , edited by v. anisimov . taylor and francis ( 2001 ) + k.held _ _ in _ `` quantum simulations of complex many - body systems '' _ edited by : j. grotendorst , d. marx and a. muramatsu ( nic - serie band 11 ) cond - mat/0112079 . zhang , m. j. rozenberg , and g. kotliar , phys . lett . * 70 * , 1666 , ( 1993 ) ; a. georges and w. krauth , phys . rev . b * 48 * , 7167 ( 1993 ) ; m.j . rozenberg , g. kotliar and x.y . zhang , phys . b * 49 * , 10181 ( 1994 ) . q. si and j. l. smith , _ phys . lett . _ * 77 * , 3391 ( 1996 ) ; h. kajueter _ phd thesis _ rutgers university , 1996 ; j.l . smith and q. si , phys . b * 61 * , 5184 ( 2000 ) ; a.m. sengupta and a. georges _ phys . rev . b _ * 52 * 10295 ( 1995 ) .

in this chapter , i present a summary of observational tests of the basic picture of disk accretion . an emphasis is placed on tests relevant to black holes , but many of the fundamental results are drawn from studies of other classes of systems . evidence is discussed for the basic structures of accretion flows . the cases of systems with and without accretion disks are discussed , as is the evidence that disks actually form . also discussed are the hot spots where accretion streams impact the disks , and the boundary layers in the inner parts of systems where the accretors are not black holes . the nature of slow , large amplitude variability is discussed . it is shown that some of the key predictions of the classical thermal - viscous ionization instability model for producing outbursts are in excellent agreement with observational results . it is also show that there are systems whose outbursts are extremely difficult to explain without invoking variations in the rate of mass transfer from the donor star into the outer accretion disk , or tidally induced variations in the mass transfer rates . finally , i briefly discuss recent quasar microlensing measurements which give truly independent constraints on the inner accretion geometry around black holes .

1312.1438

observations which , in hindsight , present evidence for mass transfer have been known about since antiquity ( e.g. the system algol ) . evidence for accretion power by black holes can be seen , again , in hindsight , as far back as the discovery of jets from active galactic nuclei ( curtis 1918 ) , and the bright , broad - lined nuclei of the seyfert galaxies ( 1943 ) . theoretical discussion of the possibility of accretion began in the 1940s ( kuiper 1941 ; bondi & hoyle 1944 ) , with the focus on mass transfer in contact binaries , and accretion from the interstellar medium , respectively . studies of the importance of accretion power in astrophysics first became prominent about 15 years later ( crawford & kraft 1956 ) , as it began to become clear that many of the classical , recurrent and dwarf novae were short orbital period spectroscopic binaries , often with photometric modulation at the same period ( e.g. walker 1954 ; joy 1954 ; joy 1956 ; greenstein & kraft 1959 ; kraft 1964 ) . the conclusion was thus drawn that mass transfer must , in some way , be responsible for the unusual phenomenology of the class of objects ( see also kopal 1959 for a synthetic discussion ) . the basic elements of accretion disk theory are reviewed in the chapter in this volume by omer blaes , while a review of the progress in connecting the `` microphysics '' of magnetohydrodynamics to the `` macrophysics '' of classical accretion disk theory is presented in the chapter by chris fragile . this article will focus on showing the observational tests that have verified that simple pictures of how accretion disks should work , largely developed in the 1970s , match many of the observational constraints at some broad level . it will also discuss the observational evidence for cases where the simple picture breaks down , and more complicated models must be invoked even though in some cases it is not clear what those more complicated models are . a variety of means of testing accretion disk theory can be made broadband spectroscopic measurements ( see e.g. juri poutanen s and jeff mcclintock s articles in this volume ) , and measurements of rapid variablity ( see e.g. the article by tomaso belloni & luigi stella in this volume ) are generally used to understand the processes in the inner accretion flows . in some cases , these methods also provide valuable information about the global accretion process . in this article , i will focus on discussing the observational constraints on accretion disk theory that come from other methodology chiefly , but not exclusively variability on much longer timescales . apart from some discussion of quasar microlensing , i will leave aside the discussion of topics such as the detailed structure of inner accretion flows around black holes , as this topic is covered elsewhere in this volume . accretion disk theory will be discussed in this article in broadbrush strokes , to set the stage for understanding which observations will be interesting , but will be discussed in detail in the other articles in this volume . in the standard picture of an accretion flow onto a compact object from a binary companion , we have the following structures : 1 . an accretion stream which begins at the donor star ( see e.g. albright & richards 1996 for a discussion of how evidence for streams can be found using doppler tomography ) 2 . an accretion disk a geometrically thin , optically thick flow in keplerian rotation , with a small inward drift due to either _ bona fide _ viscosity , or , as has become increasingly favored in recent years , a magnetic field effect that can be well modelled on a macroscopic scale as acting like viscosity . the accretion disk is generally assumed to extend outward to the `` circularization radius '' the radius where the specific angular momentum of the accreted material is the same as that for gas in a circular orbit around the accretor . if the stellar radius is larger than the circularization radius ( as can happen for wind - fed systems , and for accretion by `` normal '' stars , rather than by compact objects ) , a disk will fail to form , and the accretion stream will impact the accretor directly . a disk may also fail to form in the case of a rotating high magnetic field accretor , where the magnetic field of the central star becomes dynamically important for a circular disk outside the circularization radius ( e.g. ghosh & lamb 1979 ) . 3 . a `` hot spot '' where the accretion stream from the donor star collides with the outer accretion disk , and releases its excess kinetic energy . 4 . in systems without black hole accretors , a boundary layer should exist , where the excess rotational energy of the innermost part of the accretion disk is dissipated ( lynden - bell & pringle 1974 ) . in systems with black hole accretors , there may be excess rotational energy at the innermost stable circular orbit , but it should be transported across the event horizon . as the focus of this volume is black holes , we will not discuss the boundary layers further , except to say that evidence for them is found in both white dwarf ( e.g. pandel et al . 2003 ) and neutron star ( e.g. mitsuda et al . 1984 accretors , but that their spectra can be complicated , and , often , the emission does not show up as an extra blackbody component added to a disk model fit ( e.g. godon & sion 2005 ; piraino et al . 1999 ) . in active galactic nuclei , only the accretion disks are present . certain types of deviations from this picture are well - studied , and represent separate topics in their own right e.g. spectral state phenomenology ( in particular states in which the accretion disk becomes geometrically thick ) and the presence of jets . these topics are covered in detail in other chapters in this volume and will not be covered here in detail . a wide variety of classes of accreting objects exist in the universe . while a large fraction of the literature on accretion involves the study of mass - transferring binary stars , some accretion takes place onto isolated objects . in particular , active galactic nuclei are generally presumed to accrete from their local interstellar medium . in many protostars , the accreting object is a single star . since the focus of this volume is on accretion onto black holes , the focus of this article will be on the black hole accretors themselves . studies of stellar mass black holes are often plagued by small number statistics . studies of active galactic nuclei suffer from difficulty in making detailed measurements , as well as long timescales of variability . it is thus useful to supplement studies of black holes with studies of other classes of accreting objects . additionally , comparisons between black hole accretors and other classes of accretors can be excellent ways to determine which phenomenology is truly unique to black holes , rather than being generic to the process of accretion . the classes of mass transferring binaries seen in nature include : 1 . low mass x - ray binaries ( lmxbs ) . these are systems in which a neutron star or black hole accretes from a low mass main sequence star or a low mass subgiant star through roche lobe overflow . the mass transfer rates in these systems are determined by the rate at which either the donor star expands ( due to nuclear evolution ) or the orbit shrinks ( due to magnetic braking and/or gravitational radiation ) . high mass x - ray binaries ( hmxbs ) . these are systems in which a neutron star or black hole accretes from a massive star . many of these objects have be stars as the donor stars , and the accretion seems to take place from the equatorial wind . others have supergiant donors , and the accretion takes place through gravitational capture of the stellar wind of the supergiant . for these fast - wind systems , the accretion disk may begin to have a circular orbit very close to the compact object , or , in some cases , a disk may not form at all . finally , others have much shorter orbital periods , and their donor stars are either roche - lobe filling or nearly roche - lobe filling . in these systems the mass transfer takes place through either a focused wind , or actual roche lobe overflow . see corbet ( 1986 ) for a discussion of the different classes of high mass x - ray binaries . 3 . symbiotic stars . these are systems which , based on their initial definition , show evidence for both hot and cool components in their optical spectra . since their initial discovery , it has been realized that the cool components are red giants , and the hot components are various forms of accreting objects . most symbiotic stars have white dwarf accretors , but a small fraction have neutron star accretors ( hynes et al . 2013 and references within ) . the mass transfer is generally believed to take place through capture of the red giant star wind , although some symbiotic stars are at least very close to being in roche lobe contact . for a relatively recent review , see sokoloski ( 2003 ) . 4 . cataclysmic variables . these are systems in which mass is transferred from a low mass main sequence star or a low mass subgiant star to a white dwarf ; they are the analogs of low mass x - ray binaries , for systems where the accretors is a white dwarf . cataclysmic variables are broken into a large number of subclasses based on different observed phenomenology of variability . these sub - types are often named in terms of the prototype object , as is typical for nomenclature of variable stars . in this article , we will use nomenclature descriptive of the phenomenology , a practice which is thankfully becoming more common in the cataclysmic variable community as well . ultracompact binaries . these are binaries in which the donor star is degenerate either a white dwarf , or a low mass degenerate helium star . in the milky way , these have been seen only with white dwarf or neutron star companions . when the system is a double white dwarf binary with mass transfer , it is called an am cvn star , after the prototype object . when the accretor is a neutron star ( or , potentially , a black hole ) it is called an ultracompact x - ray binary . we refer the reader to nelemans & jonker ( 2010 ) for a review on the ultracompact x - ray binaries , and to maccarone et al . ( 2007 ) and zepf et al . ( 2008 ) for a discussion of the evidence for an ultracompact black hole x - ray binary in ngc 4472 . various classes of close binaries with two `` normal '' ( i.e. not compact remnant ) stars show evidence for accretion . these include , for example , algol systems ( in which mass transfer takes place from an evolved star to a main sequence star ) and w uma systems ( contact binaries , in which both stars overflow their roche lobes simultaneously ) . we refer the reader to thomas ( 1977 ) for a review of these systems . accretion disks form rarely in these systems ( although see olson 1991 for evidence of a disk in one algol , ku cyg ) . while active galactic nuclei represent one of the two major classes of accreting black holes , almost no fundamental tests of accretion theory have been put forth primarily from studies of active galactic nuclei . such tests would be exceedingly difficult tests based on spectroscopy would run into the problem that these systems are not fully ionized like the disks of x - ray binaries , making models of the inner accretion disks much more complicated ( e.g. done et al . 2012 ) trying to understand the inner accretion disks around black holes using active galactic nuclei is considerably more difficult than trying to understand the inner disks around stellar mass black holes . a further large part of our understanding of the outer parts of accretion disks comes from studies of variability on timescales long compared with the light crossing time at the schwarzschild radius . in this case , the problem for using active galactic nuclei as test cases stems from the fact that their viscous timescales , even at the inner edge of the accretion disk , are expected to be much longer than the durations of any light curves assembled by astronomers . as a result , nearly all of our understanding of the outer parts of accretion disks , as well , comes from studies of stellar mass accretors . it should be noted , of course , that many fundamental advances in the studies of outflows from accretion disks have been developed with essential contributions from observations of active galactic nuclei , and in recent years , studies of lensed quasars have started to give some distinctive observational tests of the accretion geometry in active galactic nuclei . it is thus the case that most of the fundamental constraints on accretion disk theory must come from studies of mass transferring binary systems , because they present accessible timescales , and have disks in an ionization state which is simpler than do active galactic nuclei . our goals should be to develop fundamental theories of accretion in general , and then to determine which phenomenology is specific to black holes . as a result , it often makes sense to incorporate observational constraints from other classes of accreting objects . the systems with neutron star accretors are the most similar to those with black hole accretors , given that the radiative efficiency for accretion onto a neutron star is very similar to that for a non - rotating black hole , at least in the context of a standard shakura - sunyaev accretion flow . the systems with white dwarf accretors ( and particularly the cataclysmic variable stars , rather than the symbiotic binaries ) are , however , the class of systems which have often provided the best constraints on accretion theory . like black hole and neutron star accretors , the cataclysmic variables have the emission from their primary stars dominated by the accretion process , rather than by core fusion , so that the radiation from the accretor can be taken as a tracer of the accretion disk s activity . that nuclear fusion contributes significantly to the emission only in low - duty cycle bursts is a fundamental difference between accreting compact objects and other kinds of accreting stars . it may seem a bit odd for a volume on black holes to include a substantial disucssion of the literature on cataclysmic variables , but in many cases , the cvs provide tighter constraints on the basic physics processes which can be expected to be generalizable to all of accretion physics . the chief advantage of cataclysmic variables relative to x - ray binaries for studying accretion is that the cvs are more numerous . about 10 times as many cvs as lmxbs are known , and the nearest cvs are about 10 times as close as the nearest low mass x - ray binaries . while in recent years , there have been a few geometric parallax measurements made of x - ray binaries in the radio ( e.g. miller - jones et al . 2009 ; reid et al . 2011 ) , such measurements have been made for much larger samples of cvs . it also turns out that the outbursting cataclysmic variable stars tend to have shorter recurrence times than the outbursts x - ray binaries , allowing for a large class of systems which have been studied repeatedly . the other advantages of the cvs being closer is that they are brighter optically , and are observed at lower extinction . furthermore by being more numerous , there are many more eclipsing cvs known than eclipsing neutron star , or , especially , black hole x - ray binaries . a variety of observational approaches has been used to establish the basic picture of disk accretion seen in binary systems . these include both spectroscopic observations , and variability studies . here , we first show that accretion disks really do form in many cases , and discuss the circumstances where accretion takes place without disks . we then discuss techniques that can be used to map out the accretion geometry in binary systems . among the first things worth testing are whether accretion really does take place in disks . in fact , the paper with the first association of stellar binarity with the production of classical novae ( walker 1954 ) presents a key piece of evidence that the accretion process , at least in those systems must take place via a disk . it shows the properties of the eclipse of the system relatively long ingresses and egresses and short minima for the primary eclipse , while having much weaker secondary eclipses ( constrained by walker & herbig 1956 to be less than 0.03 magnitudes , with primary eclipses of about 1 magnitude ) . in hindsight , these eclipse properties clearly indicate that the solid angle subtended by the accretion disk must be a small fraction of what would be subtended by a star of the same maximum radial extent . a second , more direct , but more recent piece of strong evidence that the objects believed to be accretion disks really are disks is that many of them show double - peaked emission lines ( see bailey & ward 1981 ; marsh 1988 for a discussion of cataclysmic variables ; eracleous & halpern 2003 for a discussion of active galactic nuclei ; and soria 2002 for a discussion of x - ray binaries ) . at the same time , there are clear examples of systems in which there is accretion which _ does not _ take place through a disk . these are systems which have circularization radii smaller than the radius at which a potential accretion disk would be disrupted . the most obvious such radius would be the radius of the accreting star , and indeed for non - compact stars , accretion disks are the exception rather than the rule . for compact stars , the relevant radius will usually be the magnetospheric radius of the accretor . the observational data on accreting objects support this theoretical scenario . there are classes of accreting white dwarfs and accreting neutron stars which lack the standard signatures of disk accretion . the accreting white dwarfs in this class are called polars . the name derives from the fact that their accretion light is often strongly polarized and that polarization stems from the fact that the accretion is channelled down the systems magnetic poles . these systems release large fractions of their emission in the x - rays relative to other cataclysmic variables , because the emission comes mostly from an accretion column rather than accretion disk . they also frequently show periodic emission , with the modulation taking place on the rotation period of the white dwarf . an analogous class of neutron stars are the accretion - powered x - ray pulsars . in both classes of objects , cyclotron lines have been seen ( e.g. reimers & hagen 2000 for polars ; hemphill et al . 2013 for x - ray pulsars ) . it is important to note that there are classes of systems which show magnetically dominated accretion and disk accretion at the same time the intermediate polars among cataclysmic variables ( warner 1983 ) , and both slow ( jonker & van der klis 2001 ; la barbera et al . 2001 ) and millisecond ( wijnands & van der klis 1998 ) x - ray pulsars among the neutron stars . at this point we have established that disks really do form in accreting objects . we also have sound theoretical reasoning , combined with empirical support , to show that the binary systems which do not have accretion disks have accretors which are fundamentally different from black holes they either have surfaces at large radii , or they have dynamically important magnetic fields . it is thus reasonable to assume that all accretion onto black holes takes place through accretion disks , and , in this volume about black holes , to worry only about disk accretion from this point on . now , we can determine whether certain specific predictions of the simplest accretion disk models are in agreement with the observational data . one of the most straightforward predictions is that any disk in a steady state should have @xmath0 . this result comes from equating the differential blackbody luminosity in an annulus , @xmath1 to the differential power released by gas falling through that annulus , @xmath2 , and solving for @xmath3 as a function of @xmath4 . in this framework @xmath5 , the accretion rate , and @xmath6 , the accretor mass , are constants in a steady state accretion disk , while @xmath3 is the temperature of the annulus , @xmath4 is the radial distance of the annulus from the compact object s center , @xmath7 is the luminosity of the annulus , and @xmath8 and @xmath9 are the gravitational constant and the stefan - boltzmann constant , respectively . the technique of choice for testing these models has been eclipse mapping of cataclysmic variables ( e.g. horne 1993 ; baptista 2001 ) . eclipse mapping is one of the few means of getting geometric information about continuum emission processes of accretion disks ( with quasar microlensing being the other major method ) . cataclysmic variables are the system class of choice for this work because there are optically bright eclipsing cataclysmic variables . no eclipsing low mass x - ray binaries with black holes are known , and the few eclipsing neutron star x - ray binaries are not as bright as the brightest eclipsing cvs . furthermore , bright accreting neutron stars are likely to have important effects from irradiation of the outer accretion disk by the inner accretion disk , meaning that the implications of a disagreement between the data and the standard theoretical @xmath0 law might be expected , and could be difficult to disentangle from other effects . it is most convenient to begin by attempting to model the `` novalike '' ( i.e. persistently bright ) cataclysmic variables , or the `` dwarf nova '' cataclysmic variables near the peaks of their outbursts . these are the systems that are expected to have nearly constant mass transfer rates on timescales of order the viscous propagation timescale through the accretion disks . the novalike systems have frequently failed to show @xmath10 temperature profiles ( wood et al . 1992a ; baptista et al . 1995 but see rutten et al . 1992 for an alternative result ) ; they often typically show much flatter temperature profiles . the quiescent dwarf novae almost universally show temperature profiles that are much flatter than @xmath10 ( e.g. wood et al . 1986,1992a ) . this can be interpreted as a build - up of mass in the outer accretion disk during quiescence , a loss of mass due to winds which take mass away from the inner disk relative to what is in the outer disk , the emission of that optically thin disk wind which emits substantial light , or some combination of the different effects ( e.g. wood et al . 1986 ; baptista et al . 1998 ) . knigge et al . ( 1998 ) showed that the integrated spectra found simply from summing optically thick blackbodies does not describe the integrated spectrum of ux uma , one of the prototype objects for eclipse mapping studies . baptista et al . ( 1998 ) and robinson et al . ( 1999 ) show that the results of eclipse mapping campaigns can be affected significantly by errors in what had previously been standard treatments the treatment of brightness temperatures as effective temperatures ( implictly assuming optically thick blackbody emission as the only source of light ) , and the failure to compute the effects of limb darkening properly . they find that even with careful treatment of limb darkening , the temperature profiles are flatter than @xmath10 . additionally , with a more careful treatment of the radiative transfer in accretion disks , it becomes possible to model the vertical structure of the disks . the standard shakura - sunyaev treatment yields a ratio of height @xmath11 to radius @xmath4 of @xmath12 . cataclysmic variable accretion disks typically span a range of a factor of only about 10 - 100 in radius between the surface of the white dwarf and the outer edge of the accretion disk , meaning that @xmath13 should change by , at most , a factor of 1.8 . only relatively large values of @xmath13 can be measured using eclipse mapping techniques , so generally , attempts are made only to estimate the scale height of the outer accretion disk . the numerical values of the theoretical scale heights for the shakura - sunyaev model indicate that @xmath14 should be typical for bright cvs . higher values have generally been found ( e.g. @xmath13 of about 0.06 for z cha ) , indicating that some additional process is puffing up the outer disks in these systems , or that some other geometric feature or radiative transfer process is not accounted for in the existing eclipse mapping analysis ( robinson et al . 1999 ) . a few eclipse mapping studies of x - ray binary accretion disks have been made as well . here , one expects irrations of the outer disk by the inner x - ray emitter to heat the outer disk , and cause it to have a larger scale height than expected in the context of the shakura - sunyaev disk model ( e.g. meyer & meyer - hofmeister 1982 ) . at least for the source x 1822 - 371 , the prediction of meyer & meyer - hofmeister ( 1982 ) is verified ( puchnariewicz et al . 1995 ; bayless et al . x - ray eclipse mapping has shown large spatial scale x - ray emission . this is sometimes interpreted in terms of the region in which the x - rays are produced being spatially very large ( church 2004 ) , but may be due to large scale optically thin disk winds which scatter a small fraction of the x - ray emission back into the observer s line of sight . it is generally true in x - ray binaries that the disk winds seem to be more important in soft states than in hard states ( neilsen & lee 2009 ) ; it is also true that the accretion disk corona sizes from eclipse mapping are larger in bright sources than in fainter sources ( church 2004 ) . to date , a single strong candidate eclipsing black hole x - ray binary is known ( see pietsch et al . 2006 for evidence of the eclipsing nature of the object and orosz et al . 2007 for the dynamical evidence that the object is a black hole x - ray binary ) . this system is a high mass x - ray binary with a luminous 70 @xmath15 donor star , ( orosz et al . the combination of the brightness of the donor star relative to the accretion disk , and the fact that the donor star should have a strong wind , and hence not act as a `` sharp edge '' for doing eclipsing mean that eclipse mapping of this accretion disk is not particularly promising . furthermore , the object is in m33 , at a distance of about 800 kpc . there are multiple lines of reasoning supporting the existence of hot spots where accretion streams impact the outer circular disks of accreting objects . in general , the hot spots in x - ray binaries can be quite a bit more difficult to detect than those in cataclysmic variables . this can be well understood in terms of the fraction of the total energy released as the material falls inwards . if one sets the expected luminosity of an accretion flow due to a fall through a potential from height @xmath16 to height @xmath17 , then @xmath18 . the hot spot luminosity can be obtained by setting @xmath16 to the orbital radius , and @xmath17 to the circularization radius , while the total luminosity can be obtained by setting @xmath17 to the radius of the compact star . for systems with orbital periods of a few hours , the few @xmath19 cm radii of white dwarfs will typically be of order 10% of the circularization radii , so a substantial fraction of the luminosity will be produced at the hot spot . for x - ray binaries , the fraction of the power produced at the hot spot will be a factor of order 1000 smaller . hot spots can thus be detected in black hole and neutron star accretors only if either the mass transfer rate is extremely low , and hence the radiative efficiency of the inner accretion flow is extremely small , or in the more common case , the systems are transients , and are being observed during a quiescent period , in which the accretion rate into the outer accretion disk far exceeds the accretion rate onto the central compact object ( mcclintock et al . 1995 ; froning et al . the hot spots tend to have a larger vertical height from the disk midplane than do the other parts of the outer disk , so there are also cases , for specific inclination angles , where the hot spot occults the inner disk when it is in the observers path to the compact object ( white & mason 1985 ) . in quiescent dwarf novae , the hot spot luminosity can be a very large fraction of the total luminosity from the system , leading to strong orbital modulations as the viewing angle of the hot spot changes ( e.g. wood et al . 1989 ) . the technique of doppler tomography ( marsh & horne 1988 ) allows a form of indirect imaging of accretion disks by looking at how line profiles change as a function of orbital phase . steeghs ( 2000 ) shows two - armed asymmetries in a several cataclysmic variables in outburst using doppler tomography . the evidence for such phenomena in x - ray binaries is much weaker , although it has been suggested that spiral density wave may help explain some of the large amplitude variability in grs 1915 + 105 ( tagger & pellat 1999 ) . one of the most important facets of the behavior of both cataclysmic variables and x - ray binaries is the presents of large amplitude , relatively smooth variations . these are often called dwarf nova outbursts in cataclysmic variables , and x - ray novae , or soft x - ray transient outbursts , in the x - ray binaries . in the x - ray binaries , these transients can lead to variations in the x - ray luminosities of the accreting systems of factors of @xmath20 or more . as it became clear that the dwarf novae and classical novae , often seen in the same objects , were fundamentally different phenomena , two models emerged for explaining the dwarf novae . both involved modulating the accretion rate onto the compact object . in one model , the mass transfer instability model , the rate at which mass _ enters _ the accretion disk is variable , while in the other model , the disk instability model , mass is supplied to the accretion disk at a constant rate , but there are instabilities in the way the mass flows through the disk . i will argue in this article that there is evidence supporting , if not demonstrating conclusively , the idea that both of these mechanisms apply at least some of the time . some mechanisms for producing mass transfer rate variations may be irradiation of the donor star by the accretor ( e.g. hameury et al . 1986 ; harpaz & rappaport 1991 ) , magnetic activity cycles in the donor stars ( e.g. bianchini 1990 ) , or changes in the eccentricity of the orbit of the inner mass transferring binary in a hierarchical triple system ( e.g. hut & paczynski 1984 ; maccarone 2005 ; zdziarski et al . mass transfer rates depend on the gas density at the inner lagrange point . the pressure scale height in a stellar atmosphere is typically of order @xmath21 of the orbital separation . therefore , changes in either the radius of the star or the orbital separation of the binary can lead to large changes in the mass accretion rate of order @xmath21 could lead to factors of a few changes in the mass transfer rate . thus , the lack of direct evidence that these changes should not be taken as proof that the mass transfer rate is not changing via these mechanisms . strong evidence for mass transfer variations can come in the form of finding variations in the accretion rate when one averages over timescales much longer than any reasonable timescale for mass to propagate through the accretion disk . this does not preclude variations in the mass transfer rate which happen faster than this timescale it is just that such faster variations are likely to be extremely difficult to disentangle from disk instabilities , and as we will show in this article , considerable evidence exists that disk instabilities do explain much of the outburst phenomenology in accretion disks . the thermal viscous instability is strongly favored as the primary source of disk instability . in this scenario , the viscosity parameter of the accretion disk is a function of the ionization state of the gas , so that when the disk is cold and neutral , the mass flow rate is smaller than when the disk is hot and ionized . given , also , a temperature - density relationship for the disk , a limit cycle instability will develop for accretion rates within a range commonly seen in binaries with accreting compact objects , so that many of these objects are expected to show outburst cycles as predicted by the limit cycles . the leading alternative , or perhaps complement , to the disk instability model is a model in which the mass transfer rate is varied . an extensive review of the features , successes , and shortcomings of the disk instability model is given by lasota ( 2001 ) . i will summarize some of the major points presented in that work , but will focus in this section on the observational developments since that time . the basic first order predictions of the disk instability model are seen to be followed pretty well by most transient accreting compact objects . by and large , the systems which have accretion rates high enough that one would expect their disks always to be fully ionized are persistent , and the systems which have accretion rates lower than that value are transient ( lasota 2001 ; see coriat et al . 2012 for a confirmation that the result still holds with a much larger sample of objects ) . a particular system near the threshold for stability is persistently accreting , but shows large amplitude variability ( maccarone et al . 2010 ) . an encouraging recent result came from the measurement of a precise geometric parallax distance for ss cyg . for quite some time it had simultaneously been held up as the prototype system for studying dwarf nova outbursts , and had been a system which appeared to have too high a mass accretion rate to allow the dwarf nova ionization instability to take place . the vlbi parallax distance found by miller - jones et al . ( 2013 ) indicates that the source is closer than was previously thought . the change of distance results in a mass transfer rate for ss cyg sufficiently low that the system is below the threshold for stable accretion in the ionization instability model . warner ( 1987 ) first found the relation between peak brightness and orbital period for cataclysmic variables , and that has recently been revisited with a larger sample by patterson ( 2011 ) who finds : @xmath22 a least - squares fit of the data from patterson ( 2011 ) to a power law relationship finds that the peak luminosity scales with @xmath23 . these data are plotted in figure 1 . given the relatively large scatter in the data , it is in fairly good agreement with the predictions for theoretical models which suggest that the whole accretion disk should be at a constant temperature in dwarf novae outbursts ( e.g. osaki 1996 ; cannizzo 1998 ; smak 2000 ) , yielding a @xmath24 relationship . some other indications that favor disk instabilities as a baseline model come from looking at the peak luminosities seen from x - ray binaries and from cataclysmic variables . in both cases , these are well - correlated with orbital period ( shahbaz et al . 1998 ; portegies zwart et al . p04 ; wu et al . find : @xmath25 for x - ray binaries , althought with a different treatment of the bolometric corrections , p04 found a steeper relationship for short orbital periods and a saturation at about the eddington luminosity for long orbital periods . these data are plotted in figure 2 . in any event , there is at least rough agreement with the finding of king & ritter ( 1998 ) that outburst peak luminosities should scale with the radius of the accretion disk . the durations of outbursts of many , but not all , systems , are relatively well in - line with the expected viscous timescales of the accretion disks for the black hole systems ( see e.g. chen , shrader & livio 1997 ; p04 ) . the outburst durations are shorter than the viscous timescales for the cv accretion disk , where irradiation is not important , and so cooling fronts can truncate the outbursts before the entire disks are accreted ( lasota 2001 ) . some significant amount of the data which represent exceptions to the basic ionization instability may be explained as the result of tidal effects ( see section 5.5.1 ) . while the ionization instability explains the phenomenology of x - ray binary and cv outbursts in broad brushstrokes , there are phenomena which are clearly not in agreement with that picture . in the cataclysmic variables , where the recurrence times between outbursts tend to be much shorter than in x - ray binaries , it can be clearly seen that there are variations from outburst to outburst in ways that have been fit , to date , only by adding in truncations of the inner accretion disk _ and _ variations of the mass transfer rate ( e.g. schreiber et al . 2003 ; lasota 2012 ) . there are a few other cases where strong evidence for mass transfer instabilities are expected . a prime recent example among black hole candidates is the ongoing outburst of swift j1753.5 - 0127 , which has been in outburst since 2005 , and has an orbital period of about 3.2 hours ( zurita et al . the outburst duration of 8 years ( and counting ) combined with the short orbital period is something that can not be explained in terms of an ionization instability model . while the observation of superhumps ( see the following section for a discussion of superhumps ) in this source ( zurita et al . 2008 ) should imply that this outburst is a `` super - outburst '' and hence should be longer than normal outbursts , the super - outbursts in well - studied systems are only a factor of a few longer than the normal outbursts . several neutron star accretors have also undergone outbursts that lasted far longer than the expected viscous timescales for the systems orbital periods ( e.g. wijnands et al . 2001 and references within ) . on the flip side , the 1999 transient episode of the accreting black hole xte j1819 - 254 showed strong evolution on timescales of a few hours , having reached a flux of about 12 crab , corresponding to a super - eddington luminosity for the source ( hjellming et al . 2000 ; orosz et al . the variations were clearly too fast to be the result of some global disk instability , and too strong to have been the result of the normal variability typically seen in x - ray binaries . an additional line of evidence for variations in the mass transfer rates perhaps the most direct such evidence comes from cantrell et al . they interpret the blue excess in the light from a 0620 - 00 in quiescence as coming from the accretion impact spot . in the context of that interpretation , the variability in the quiescent luminosity of the hot spot immediately implies that the rate at which matter is reaching the outer accretion disk is changing strongly as a function of time . the variation in the quiescent ultraviolet flux of the source also supports this interpretation , since the ultraviolet light can be demonstrated even more convincingly than the optical light to come from the hot spot ( mcclintock et al . 1995 ; froning et al . 2011 ) . in an x - ray binary , the presence of a donar star means that tidal forces on the accretion disk may be substantial and variable on the orbital period . there are a few observe phenomena which have very well motivated theoretical explanations as coming from tidal effects . there are also a few pheneomena which may more speculatively be associated with tidal effects . one phenomenon which is well - associated with tidal interactions is that of superhumping . superhumps are oscillations in the light curves of some outbursting dwarf novae and x - ray binaries . there is a critical mass ratio of 0.35 below which cvs in bright states show superhumps , and above which cvs never show superhumps ( patterson et al . the oscillations occur with a period very close to , but not exactly equal to , the orbital period of the binary . the oscillations are well - explain by a model in which an eccentric instability develops at the 3:1 resonance between the orbital period of the accretion disk and the orbital period of the binary system , and then this disk precesses due to the tidal forces ( whitehurst 1988 ) . the limit on the mass ratio is then given by the limit that the accretion disk not be tidally truncated inside the location of the 3:1 resonance region . the superhumping behavior is associated with `` super - outbursts '' of these systems . most of the cataclysmic variables which show superhumps show a sequence of outbursts , in which some small fraction of the outbursts are significantly brighter than the rest of them . for example , in v1504 cyg , which has been well - studied with kepler , the superoutbursts happen about once for every 14 normal outbursts ( osaki & kato 2013 ) . the prevailing view for these phenomena being so well coupled is that the thermal - tidal disk instability explains both effects ( e.g. osaki 1996 ) . in this model , the outbursts of the disk are caused by the standard thermal instability model , discussed above . the outer edge of the accretion disk moves outwards during each normal outburst . after a series of normal outbursts , the disk is outside the 3:1 resonance radius , and the next normal outburst triggers the tidal instability , which drives in material from far out in the disk , leading to an increased peak accretion rate and outburst duration . a modification of the tidal instability model has been proposed by truss et al . ( 2002 ) to explain the re - brightening in soft x - ray transients . in their model , they suggest that irradiation of the outer accretion disk is uneven . tidal forces cause the location of the irradiation region to change , allowing for the originally un - irradiated region to become irradiated later , allowing , at a relatively late time , a new portion of the disk to become hot and enter a high viscosity state . superhumps have also been seen in several short period x - ray binaries ( e.g. odonoghue & charles 1996 ; zurita et al . 2002 ; zurita et al . 2008 ; wang & chakrabarty 2010 ; see also wachter et al . 2002 for a more tentative result ) . the emission modulation mechanism in the cvs is thought to be tidal modulation of the viscosity , a mechanism which can not work in low mass x - ray binaries because such a small fraction of the energy is dissipated in the outer accretion disks of x - ray binaries ( e.g. haswell et al . 2001 ) . among the x - ray binaries which have shown some evidence for superhumps are short period black hole x - ray binaries ( e.g. nova mus 1991 , gro j0422 + 32 , xte j1118 + 480 and swift j1753.5 - 0127 ) , and a few recurrent transient neutron star x - ray binaries ( 4u 1608 - 52 wachter et al . , 2002 ; and aql x-1 r. jain , private communication and see also the discussion in wachter et al . interestingly , the two recurrent transient neutron stars are known to show outbursts with different amplitudes e.g. aql x-1 seems to show some kind of outburst roughly every 100 days , and a more major outburst every 300 or so days ( maitra & bailyn 2008 ) . there is also evidence for superhumping in the ultracompact x - ray binary 4u 1820 - 30 ( wang & chakrabarty 2010 ) this system is _ persistently _ bright , and at a short orbital period . the black holes which have been seen to show prominent superhumps in recent years have not shown outbursts of varied amplitude . it may be , however , that these systems have shown _ only _ super - outbursts . this idea has been put forth by maccarone & patruno ( 2013 ) as part of the reason why some short period black hole x - ray binaries seem to be brighter than one might expect given the wu et al . ( 2010 ) relation , and might then allow for the normal outbursts of the short period black hole x - ray binaries to manifest themselves as very faint x - ray transients ( see also knevitt et al . 2013 who discuss how short period transients may be absent from all - sky surveys ) . such a scenario is strongly bolstered by finding superhumps in the bright outbursts from short - period systems . it is worth further noting that the `` normal '' outbursts from such systems , because of radiative inefficiency for black holes in the low hard state , should be @xmath26 times fainter than their superoutbursts , rather than just the factor of a few difference seen for the cv superoutbursts and the candidate superoutbursts in the neutron star x - ray binaries . a much harder question to answer is whether active galactic nuclei are susceptible to the same type of accretion disk instabilities as x - ray binaries . in principle , they should be , as their accretion disks are much cooler than those of x - ray binaries , but theoretical calculations suggest that the outbursts may be quite a bit less dramatic than in x - ray binaries ( hameury et al . 2009 ) . determining observationally whether they do have outbursts is complicated by the fact that the typical one month timescales for outbursts of x - ray binaries , if scaled up to even the smallest , @xmath27 black holes in agn , would take place on timescales of several millenia . evidence for a rather sharp variation in the luminosities of the galactic center can be seen by looking at reflection spectra from molecular clouds , which indicate that it was several orders of magnitude brighter about 100 years ago than it is now ( e.g. ponti et al . 2010 ) still , there is no observational means to determine whether this variation was due to changes in the mass transfer rate into the agn s accretion disk , or changes in the rate of flow through the disk . koerding et al . ( 2006 ) found that the spectra of active galactic nuclei are consistent with following a hysteresis loop like that followed for black hole and neutron star x - ray binaries ( maccarone & coppi 2003 ) , which is suggestive of the idea that the agn pass through similar outburst curves , but this merely suggestive evidence is the strongest evidence to date that agn actually _ do _ have outbursts due to disk instabilities , rather than that they merely should have such outbursts . the evidence for disk winds in active galactic nuclei has been well - reviewed by ken pounds in this volume . in x - ray binaries , similar types of evidence for disk winds the discovery of x - ray absorption lines which appear to be dependent on inclination angle and on source spectral state ( e.g. diaz - trigo et al . 2006 ; neilsen & lee 2009 ; ponti et al . x - ray binaries provide an additional means of searching for evidence for disk winds . the mass transfer rates can be estimated from both the luminosities of the hot spots in quiescence , and from binary evolution modelling . these can then be compared with the long - term mean luminosities of the system , which provide an independent estimate of the mass accretion rate . if the mean mass transfer rate from the donor star is substantially larger than the mean mass accretion rate by the accretor , then there is additional evidence in support of winds being important . this methodology , too , shows that substantial mass loss is likely to be taking place from accretion disks around black holes ( e.g. froning et al . ideally , one would like to learn about the structures of accretion disks by imaging them , rather than by testing models of their spectral or variability properties . good prospects exist for making millimeter vlbi images of a very small number of very nearby galactic nuclei ( e.g. doeleman 2008 ) . relatively little hope exists in the near term for making direct images of a large sample of black hole accretion disks spanning a range of flux levels , and relatively little hope exists for doing small scale direct imaging at frequencies other than in the millimeter through sub - millimeter band . a technique has been exploited for making indirect imaging measurements of the accretion flows around active galactic nuclei namely using gravitationally lensed quasars ( see e.g. chen et al . 2012 for an extensive discussion of recent results ; chang & refsdahl 1979 for the first discussion of the possibility ) . when a quasar is behind a galaxy or group or cluster of galaxies , two types of lensing take place . the first is that the effect of the `` smooth mass '' of the lens ( i.e. the lens s dark matter halo plus the sum of the stars ) leads to the production of multiple images of the background object . the second is that individual stars in the lensing galaxies may microlens the background object . for cases where the lensed background source is much larger than the stars , the effects of the microlensing process are small , since then only a small part of the object is microlensed at a time . as a result , the magnitude of variability due to microlensing can be used to probe the size scale of the background object relative to the sizes of the stars doing the lensing . by observing the amplification factors due to microlensing at different wavelengths , one can map out the size scale of the accretion disks versus wavelength . the technique of choice for such work is time series analysis of microlensing . in the ideal case , the system will be well enough studied that the time delays due to the different path lengths light travels to form each of the observed multiple images are known . then one can correct for these time delays , and remove the variability instrinsic to the quasar , so that the variability due to microlensing can be isolated ( e.g. morgan et al . 2008 , 2010 ) . it can be difficult to arrange large numbers of epochs of monitoring data in the x - rays , particularly when arcsecond angular resolution is needed , and only chandra can provide the necessary data . the optical monitoring data are more readily obtained . as a result , methods which invoke less intensive x - ray coverage are desirable . pooley et al . ( 2006 ) show that optical data can be used to determine the magnitude of the optical microlensing anomalies , and then a single x - ray epoch can be used to estimate the variance in the x - ray magnifications due to microlensing in a statistical sense , from the variance in the x - ray images brightnesses . this can still often yield important information about the size scale of the x - ray emitting region , while using considerably less time on the most oversubscribed telescopes . a few key results come from morgan et al . ( 2008;2010 ) s studies : that the optical continuum comes from regions with spatial scales of @xmath28 , that the x - ray continuum comes from regions with spatial scales of @xmath29 , and that the fe k@xmath30 emission typically comes from regions even smaller than the x - ray continuum . microlensing of the broad line regions of quasars by the sheer fact that any microlensing is detected at all indicates that the broad line regions are not spherically symmetric ( sluse et al . 2012 ) . chen et al . ( 2013 ) have shown that the gravitational lensing by the quasar s own black hole can lead to factor of @xmath31 systematic errors on the estimates of the spatial scales . this result applies primarily to the small x - ray emission regions , since for much spatial scales of more than tens of schwarzschild radii , the effects of light bending by the black hole are very small . usually , the size of the emission region will be under - estimated , but the direction and magnitude of the effect depend on the inclination angle of the accretion disk , the spin of the black hole , and the emissivity profile of the accretion disk . these errors are of the same order as errors in black hole masses from most techniques used for active galactic nuclei . chen et al . ( 2013 ) also find that subtle differences in the time delays for different images should be detectable with excellent microlensing campaigns . with good enough data , the inclination angles and spins of black holes in quasars might be measureable using microlensing giving measurements independent of those which come e.g. from iron line measurements . it is important to note that low luminosity active galactic nuclei have not yet been well - studied using these techniques , and may have different accretion geometries the systems which have been analyzed are all bright quasars . the finding of very small x - ray emission regions thus does _ not _ have any clear implications for the controversy about whether the thin accretion disks around black holes in x - ray binaries are truncated , with an inner advection dominated region emitting most of the hard x - rays ( see e.g. rykoff et al . 2006 ; kolehmainen et al . these data thus _ do not _ help to resolve the controversies about whether the inner accretion disks in low / hard states are truncated , or extend in to the innermost stable circular orbits although the technique is , in principle , useful for resolving such controversies in low luminosity agn , if lensed llagn can be discovered which are bright enough to perform such studies . while there has been a recent discovery of a candidate low luminosity agn with short time delays between the different images and flux ratio anomalies ( anguita et al . 2009 ) , this particular object is extremely faint ( the brightest image is seen at magnitude 24.6 in the 606w and 814w filters with the hubble space telescope ) , and it is unlikely that it will be useful for understanding the x - ray geometry of low luminosity agn . several key pieces of accretion disk theory show good agreement between models and observations the basic structures of the accretion disks as geometrically thin , and optically thick in their outer parts , and the existence of hot spots and spiral arms are all well established . at the same time , there is much about accretion disks which is of great importance which is not fully understood particularly the development of a theory of why and how accretion disks vary . a great deal of the observed phenomenology agrees with the basic picture of the hydrogen ionization instability model . particularly , the accretion rate above which systems becomes stable is in good agreement with the predictions of that model , as is the relationship between outburst peak luminosities and system orbital periods . at the same time , there are observational results , such as the duration of the outbursts of several short period x - ray binaries , that probably require mass transfer variations as well , and the detailed shapes of the outbursts are not always well matched by models . the author thanks the conference organizers for having promoted a series of stimulating discussions . he also thanks the avett brothers , , whose sublime _ magpie and the dandelion _ made the process of finalizing this manuscript far more enjoyable than it would have been otherwise . albright g.e . , richards m.t . , 1996 , apj , 459 , l99 anguita t. , faure c. , kneib j .- , wambsganss j. , knobel c. , koekemoer a.m. , limousin m. , 2009 , a&a , 507 , 35 bailey j. , ward m. , 1981 , mnras , 194p , 17 baptista r. , horne k. , hilditch r. , mason k.o . , drew j.e . , 1995 , apj , 448 , 395 baptista r. , horne k. , wade r.a . , hubeny i. , long k. , rutten r.g.m . , 1998 , mnras , 298 , 1079 baptista r. , 2001 , lnp , 572 , 307 bianchini a. , 1990 , aj , 99 , 1941 bondi h. , hoyle f. , 1944 , mnras , 104 , 273 . cannizzo j.,k . , 1998 , apj , 493 , 426 cantrell a.g . , et al . , 2010 , apj , 710 , 1127 chang k. , refsdahl s. , 1979 , nature , 282 , 561 chen b. , dai x. , kochanek c.s . , chartas g. , blackburne j.a . , morgan c.w . , 2012 , apj , 755 , 24 chen b. , dai x. , baron e. , kantowski r. , 2013 , apj , 769 , 131 chen w. , shrader c.r . , livio m. , 1997 , apj , 491 , 312 church m.j . , 2004 , revmexaa , 20 , 143 church m.j . , baluciska - church m. , 2004 , mnras , 348 , 955 corbet r.h.d . , 1986 , mnras , 220 , 1047 coriat m. , fender r.p . , dubus g. , 2012 , mnras , 424 , 1991 crawford j.a . , kraft r.p . , 1956 , apj , 123 , 44 curtis h. , plico 13 , 55 ( 1918 ) . diaz - trigo m. , parmar a.n . , boirin l. , mendez m. , kaastra j.s . , 2006 , a&a , 445 , 179 doeleman s , 2008 , jphcs , 131 , 2055 done c. , davis s.w . , jin c. , blaes o. , ward m. , 2012 , mnras , 420 , 1848 eracleous m. , halpern j.p . , 2003 , apj , 599 , 886 froning c.s . , et al . , 2011 , apj , 743 , 26 ghosh p. , lamb f.k . , 1979 , apj , 234 , 296 godon p. , sion e.m . , 2005 , mnras , 361 , 809 greenstein j.l . , kraft r.p . , 1959 , apj , 130 , 99 hameury j .- , king a.r . , lasota j .- p . , 1986 , a&a , 162 , 71 hameury j .- m . , viallet m. , lasota j .- , 2009 , a&a , 496 , 413 harpaz a. , rappaport s. , 1991 , apj , 383 , 739 haswell c.a . , king a.r . , murray j.r . , charles p.a . , 2001 , mnras , 321 , 475 hemphill p.b . , rothschild r.e . , caballero i. , pottschmidt k. , khnel m.,frst f. , wilms j. , 2013 , apj , 777 , 61 hjellming r. et al . , 2000 , apj , 544 , 977 horne k. , 1993 , in accretion disks in compact stellar systems , ed j.c wheeler ( singapore : world scientific ) , 117 hut p. , paczynski b. , 1984 , apj , 284 , 675 hynes r.i . , et al . , 2013 , apj , in press jonker p. , van der klis m. , 2001 , apj , 553l , 43 joy a.h . , 1954 , apj , 120 , 377 joy a.h . , 1956 , apj , 124 , 317 krding e.g. , jester s. , fender r. , 2006 , mnras , 372 , 1366 king a.r . , ritter h. , 1998 , mnras , 293 , 42 knevitt g. , wynn g. , vaughan s. , watson m. , 2013 , mnras , in press knigge c. , long k. s.,wade r. a. , baptista r. , horne k. , hubeny i. , rutten r. g. m. , 1998a , apj , 499 , 414 kolehmainen m. , done c. , diaz trigo m. , 2013 , mnras , in press . kuiper g.p . , apj , 93 , 133 , ( 1941 ) . kraft r.p . , 1964 , apj , 139 , 4578 la barbera a. , burderi l. , di salvo t. , iaria r. , robba n. , 2001 , apj , 553 , 375 lasota j .- , 2001 , newar , 45 , 449 maccarone t.j . , coppi p.s . , 2003 , mnras , 338 , 189 maccarone t.j . , kundu a. , zepf s.e . , rhode k.l . , 2007 , nature , 445 , 183 maccarone t.j . , long k.s . , knigge c. , dieball a. , zurek d.r . , 2010 , mnras , 406 , 2087 maccarone t.j . , patruno a. , 2013 , mnras , 428 , 1335 macquart j .- p . , godfrey l.e.h . , bignall h.e . , hodgson j.a . , 2013 , apj , 2013 , apj , 765 , 142 maitra d. , bailyn c.d . , 2008 , apj , 688 , 537 marsh t.r . , 1988 , mnras , 231 , 1117 mcclintock j.e . , horne k. , remillard r.a . , 1995 , apj , 442 , 358 miller - jones j.c.a . , jonker p.g . , dhawan v. , brisken w. , rupen m.p . , nelemans g. , gallo e. , 2009 , apj , 706l , 230 miller - jones j.c.a . , sivakoff g.r . , knigge c. , krding e. , templeton m. , waagen e.o . , 2013 , science , 340 , 950 mitsuda k. , et al . , 1984 , pasj , 36 , 741 morgan c.w . , kochanek c.s . , dai x. , morgan n.d . , falco e.e . , 2008 , apj , 689 , 755 morgan c.w . , kochanek c.s . , morgan n.d . , falco e.e . , 2010 , apj , 712 , 1129 neilsen j. , lee j.c . , 2009 , nature , 458 , 481 odonoghue d. , charles p.a . , 1996 , mnras , 282 , 191 olson e.c . , 1991 , aj , 102 , 1423 orosz j.a . , et al . , 2001 , apj , 555 , 489 orosz j.a . , et al . , 2007 , nature , 449 , 872 osaki y. , 1996 , pasp , 108 , 39 osaki y. , kato t. , 2013 , pasj , in press pandel d. , cordova f.a . , howell s.b . , 2003 , mnras , 346 , 1231 patterson j. , 2011 , mnras , 411 , 2695 pietsch w. , et al . , 2006 , apj , 646 , 420 piraino s. , santangelo a. , ford e.c . , kaaret p. , 1999 , a&a , 349l , 77 pooley d. , blackburne j.a . , rappaport s. , schecheter p.l . , fong w - f . , 2006 , apj , 648 , 67 ponti g. , terrier r. , goldwurm a. , belanger g. , trap g. , 2010 , apj , 714 , 732 ponti g. , fender r.p . , begelman m.c . , dunn r.j.h . , neilsen j. , coriat m. , 2012 , mnras , 422 , l11 portegies zwart s.f . , dewi j. , maccarone t. , 2004 , mnras , 355 , 413 reid m.j . , mcclintock j.e . , narayan r. , gou l. , remillard r.a . , orosz j.a . , 2011 , apj , 742 , 83 reimers d. , hagen h .- j . , 2000 , a&a , 358l , 45 robinson e.l . , , wade r.a . , 1999 , apj , 514 , 952 rutten r.g.m.r . , van paradijs j. , tinbergen j. , 1992 , a&a , 260 , 213 rykoff e.s . , miller j.m . , steeghs d. , torres m.a.p . , 2007 , apj , 666 , 1129 schatzman e. , 1949 , anap , 12 , 281 schreiber m.r . , hemury j .- , lasota j .- , 2003 , a&a , 410 , 239 seyfert c.k . , apj , 97 , 28 ( 1943 ) . shahbaz t. , charles p.a . , king a.r . , 1998 , mnras , 301 , 382 shakura n.i . , sunyaev r.a , 1973 , a&a , 24 , 337 sluse d. , hutsemekers d. , courbin f. , meylan g. , wambsganss j. , 2012 , a&a , 544a , 62 smak j.i . , 2000 , aca , 50 , 399 sokoloski j. , 2003 , jaavso , 31 , 89 soria r. , 2002 , proceedings of the mgix mm meeting , eds . robert t jantzen & remo ruffini , world scientific publishing company steeghs d. , 23001 , lnp , 573 , 45 thomas h .- c . , 1977 , ara&a , 15 , 127 truss m.r . , wynn g.a . , murray j.a . , king a.r . , 2002 , mnras , 337 , 1329 wachter s. , hoard d.w . , bailyn c.d . , corbel s. , kaaret p. , 2002 , apj , 568 , 901 walker m.f . , 1954 , pasp , 66 , 230 walker m.f . , 1956 , apj , 123 , 68 wang z. , chakrabarty d. , 2010 , apj , 712 , 653 warner b. , 1983 , assl , 101 , 155 warner b. , 1987 , mnras , 227 , 23 white n.e . , mason k.o . , 1985 , ssrv , 40 , 167 white n.e . , peacock a. , hasinger g. , mason k.o . , manzo g. , taylor b.g , branduardi - raymont g. , 1986 , mnras , 218 , 129 whitehurst r. , 1988 , mnras , 232 , 35 wijnands r. , van der klis m. , 1998 , nature , 394 , 344 wijnands r. , miller j.m . , markwardt c. , lewin w.h.g . , van der klis m. , 2001 , apj , 560l , 159 wood j.h . , abbott t.m.c . , shafter a.w . , 1992a , apj , 393 , 729 wood j.h . , horne k. , vennes s. , 1992b , apj , 385 , 294 wood j.h . , horne k. , berriman g. , wade r.a . , 1989 , 341 , 974 wood j. , horne k. , berriman g. , wade r. , odonoghue d. , warner b. , 1986 , 219 , 629 wu y.x . , yu w. , li t.p . , maccarone t.j . , li x.d . , 2010 , apj , 718 , 620 zdziarski a.a . , wen l. , gierlinski m. , 2007 , mnras , 377 , 1006 zepf s.e . , stern d. , maccarone t.j . , kundu a. , kamionkowski m. , rhode k.l . , salzer j.j . , ciardullo r. , gronwall c. , 2008 , apjl , 683 , 139 zurita c. , et al . , 2002 , mnras , 333 , 791 zurita c. , durant m. , torres m.a.p . , shahbaz t. , casares j. , steeghs d. , 2008 , apj , 681 , 1458

the de branges spaces of entire functions generalise the classical paley - wiener space of square summable bandlimited functions . specifically , the square norm is computed on the real line with respect to weights given by the values of certain entire functions . for the paley - wiener space , this can be chosen to be an exponential function where the phase increases linearly . as our main result , we establish a natural geometric characterisation , in terms of densities , for real sampling and interpolating sequences in the case when the derivative of the phase function merely gives a doubling measure on the real line . moreover , a consequence of this doubling condition , is that the spaces we consider are one component model spaces . a novelty of our work is the application to de branges spaces of techniques developed by marco , massaneda and ortega - cerd for fock spaces satisfying a doubling condition analogue to ours .

1103.0566

an entire function @xmath0 is a hermite - biehler function if it has no real zeroes and @xmath1 for @xmath2 , where @xmath3 . for such functions , the corresponding de branges space is the hilbert space of functions @xmath4 where @xmath5 is the hardy space of the upper half - plane . the space @xmath6 is equipped with the norm @xmath7 the main example of a de branges space is the paley - wiener space , which is obtained by taking @xmath8 . inspired by this , it is usual to consider the polar decomposition @xmath9 the phase function @xmath10 is the increasing argument of @xmath0 . if @xmath11 is the zero set of @xmath0 , then for some @xmath12 we have @xmath13 this is easy to see , e.g. since @xmath14 is an inner function on the upper half - plane with singular support only at infinity . observe that when @xmath15 , which is an assumption we will make from now on , the measure @xmath16 is a sum of harmonic measures with respect to the complex conjugate of the points in @xmath11 . we say that the de branges space @xmath6 satisfies the doubling condition if this measure is a doubling measure . i.e. , if there exists a constant @xmath17 such that @xmath18 for all intervals @xmath19 , where @xmath20 is the interval co - centric to @xmath21 of double length . the measure @xmath22 induces a natural metric on @xmath23 , namely @xmath24 to motivate the significance of this metric , we consider the reproducing kernels of the space @xmath6 . these are functions @xmath25 , for @xmath26 , such that @xmath27 for every @xmath28 . it is not hard to see that in the inner product implied by the norm of @xmath6 , we have @xmath29 we denote by @xmath30 the normalised reproducing kernel . from this , it follows that for @xmath31 we have @xmath32 so , if we , for some fixed @xmath33 , let @xmath34 be a sequence of points such that @xmath35 , then the normalised reproducing kernels @xmath36 form an orthonormal sequence . it is a theorem of de branges @xcite , that for every such @xmath37 , except at most one , this sequence of reproducing kernels is an orthonormal basis for @xmath6 . the model spaces @xmath38 , where @xmath39 is an inner function , is the subspace of the classical hardy space @xmath5 defined by @xmath40 it is a simple observation that @xmath41 is a unitary homeomorphism between @xmath6 and the model subspace @xmath42 , where @xmath43 is the inner function . in this way , it is possible to show that the de branges spaces correspond exactly to the model subspaces with meromorphic inner functions @xcite . a model space @xmath38 satisfies the one component condition if there exists an @xmath44 such that the level set @xmath45 is connected . these spaces where introduced by cohn @xcite . by a characterization due to aleksandrov @xcite and lemma [ local doubling equivalences ] below , the meromorphic model spaces satisfying the one component condition essentially correspond to the class of de branges spaces satisfying a local doubling condition ( corollary [ aleksandrov corollary ] ) . sampling and interpolating sequences can be seen as dual concepts . we give the definitions for a general hilbert space @xmath46 of functions , analytic on a domain @xmath47 , with reproducing kernels @xmath48 . let @xmath49 be a sequence of points . then : * @xmath50 is called sampling for @xmath46 if there exists constants @xmath51 such that for all @xmath52 we have @xmath53 ( if only the upper inequality holds , @xmath50 is said to be a bessel sequence . ) * @xmath50 is called interpolating for @xmath46 if , for every sequence @xmath54 satisfying @xmath55 there exist @xmath52 such that @xmath56 . observe that a sequence @xmath50 is sampling if and only if the normalized reproducing kernels @xmath57 form a frame in @xmath46 , while @xmath50 is interpolating if and only if @xmath57 is a riesz sequence in @xmath46 ( for these , and related notions , see @xcite ) . we use the following notion of separation . a real sequence @xmath50 is called @xmath58separated when there exists an @xmath59 such that @xmath60 whenever @xmath61 and @xmath62 . observe that the points @xmath63 for which @xmath35 are at an integer distance from each other with respect to this metric . to study the _ real _ sampling and interpolating sequences for @xmath6 , we introduce the following beurling - type densities . recall that @xmath64 . given a @xmath58separated sequence @xmath65 , and a measure @xmath22 as above , the lower and upper uniform densities of @xmath50 are given by @xmath66 and @xmath67 where the @xmath21 are bounded intervals in @xmath23 and @xmath68 denotes the number of points in @xmath69 . our approach is motivated by work on sampling and interpolating sequences in the so - called weighted paley - wiener spaces @xcite , and in the fock space setting @xcite . i.e. , the space @xmath70 where @xmath71 is the planar lebesgue measure and @xmath72 is a subharmonic function on @xmath73 . in @xcite , the sampling and interpolating sequences are characterized in terms of beurling densities for the classical fock space @xmath74 ( for which @xmath75 ) . in subsequent work , the @xmath76 case for spaces with @xmath77 was studied @xcite . finally , in @xcite , the sampling and interpolating sequences are characterized in terms of beurling - type densities for spaces where @xmath78 gives a doubling measure on @xmath73 . the analogy to the de branges setting becomes clearer , perhaps , when we observe that the function @xmath79 has a laplacian which is supported on @xmath23 and satisfies @xmath80 there ( see lemma [ delta psi lemma ] ) . with this notation , the classical case ( paley - wiener space ) corresponds to @xmath75 when restricted to @xmath23 . since functions in @xmath6 have similar local properties as those in the corresponding fock space , we are able to use fock space techniques in our setting . recall that the density results for the paley - wiener case due to beurling , jaffard and seip @xcite extend essentially without changes to the so called weighted paley - wiener spaces , @xcite . in particular , they show that any such space is equal to a de branges space for which the zeroes of the hermite - biehler function do not approach the real axis , and for which @xmath81 . the family of spaces we study contains spaces which were not studied in that setting . however , the converse is also true , they give an example of a de branges space that is equal to the classical paley - wiener space , but for which the derivative of the phase of the hermite - biehler function is not doubling . in an example where the phase function is made to oscillate by placing zeroes exponentially sparse on a horizontal line , baranov @xcite proved that , in contrast to what is known for paley - wiener spaces , there are separated sequences that are not bessel sequences . this indicates that density results like the ones of theorems [ theorem_sampling ] [ theorem_interpolation ] , can not hold in this non - doubling situation . recently , this was studied in more generality by belov , mengestie and seip @xcite , who give a complete characterisation of bessel sequences for a class of `` small '' de branges spaces . ( see also the discussion following the statement of theorem [ theorem_interpolation ] . ) having said this , it appears natural to study the de branges spaces with doubling measure @xmath82 as an extension of paley - wiener spaces , using techniques adapted from the fock space setting . observe that this approach suggests the study of the de branges analogs to the `` large '' and `` small '' fock spaces introduced in @xcite and @xcite , among other problems . some of this work has been already done by mengestie in his forthcoming doctoral dissertation @xcite ( see also @xcite ) . we obtain the following description of sampling and interpolating sequences for the de branges spaces satisfying the doubling condition . [ theorem_sampling ] let @xmath0 be a hermite - biehler function with phase function @xmath10 , and suppose that @xmath83 is a doubling measure on @xmath23 . if @xmath50 is a real sampling sequence for @xmath6 , then it is a finite union of @xmath58separated subsequences and there exists a @xmath58separated subset @xmath84 which is sampling with @xmath85 . conversely , if a real sequence @xmath50 is a finite union of @xmath58separated sequences and there is a @xmath58separated subsequence @xmath84 such that @xmath86 , then @xmath50 is sampling . the interpolation part reads as follows . [ theorem_interpolation ] let @xmath0 be a hermite - biehler function with phase function @xmath10 , and suppose that @xmath83 is a doubling measure on @xmath23 . if @xmath50 is a real interpolating sequence for @xmath6 , then it is @xmath58separated with @xmath87 . conversely , if a real sequence @xmath50 is @xmath58separated and @xmath88 , then @xmath50 is interpolating . as we pointed out above , for some spaces @xmath6 , where the derivative of the phase function @xmath10 is highly oscillating , such results are no longer valid @xcite . an important motivation to study such de branges spaces , was to try to find a counterexample to feichtinger s conjecture , which in our setting says that a sampling sequence can be written as a finite union of interpolating sequences . however , it has been recently proved that feichtinger s conjecture for reproducing kernels holds true both in this case @xcite , and in the case of one - component model subspaces @xcite . ( observe that for de branges spaces with doubling phase , this also follows from theorems [ theorem_sampling ] and [ theorem_interpolation ] . ) this seems to suggest that counterexamples , if there are any , should be found when the phase goes from doubling to highly oscillatory . in section [ sec_preliminaires ] , we prove some preliminary results about sampling and interpolating sequences in de branges spaces with doubling and locally doubling measures . in section [ sec_basic_constructions ] , following beurling we construct a multiplier function . in section [ sec_main_results ] , we first prove the necessity of the density conditions in theorems [ theorem_sampling ] and [ theorem_interpolation ] using the techniques developed by ramanathan and steger . then , we prove the sufficiency for interpolation using a lagrange interpolating function and the sufficiency for sampling following the weak limit techniques developed by beurling . we begin by establishing some preliminary results needed in the rest of the paper , some of which are previously known . we only provide full proofs if the result is new , or there is some novelty in the argument . first some notation . for @xmath89 and @xmath90 , we set @xmath91 , and let @xmath92 denote the disk in @xmath73 with both diameter and intersection with @xmath23 equal to @xmath93 . for @xmath94 , we simply write @xmath95 and @xmath96 , respectively . by @xmath71 , we denote the lebesgue measure on @xmath73 , and we use the symbol @xmath97 to denote that there exists a constant @xmath17 such that @xmath98 for real valued functions @xmath99 and @xmath100 . the symbols @xmath101 and @xmath102 are defined analogously . for an entire function @xmath99 , we denote its zero set by @xmath103 . recall from the introduction that a positive borel measure @xmath22 is doubling if there exists a constant @xmath17 such that @xmath18 for every bounded interval @xmath21 . we say that @xmath22 is locally doubling if the same holds true whenever @xmath104 . ( observe that this notion of locally doubling measures is different than the one defined with the same name in @xcite . ) we state a standard result for doubling measures , see ( * ? ? ? * lemma 2.1 . ) or ( * ? ? ? * lemma 1 ) . [ lemma gamma doubling ] let @xmath22 be a positive ( borel ) doubling measure . then there exists @xmath105 such that for any pair of intervals @xmath106 with radius @xmath107 such that @xmath108 @xmath109 @xmath110 a version also holds for locally doubling measures . indeed , one gets the same conclusion ( [ doubling ] ) supposing that @xmath111 . for meromorphic inner functions @xmath39 , it was pointed out to us by a. baranov that there is a close connection between the one component condition and the doubling condition on the phase . in the following lemma , we essentially prove that for such functions , the local doubling condition on the phase is equivalent to the one component condition ( corollary [ aleksandrov corollary ] ) . [ local doubling equivalences ] let @xmath0 be a hermite - biehler function with phase @xmath10 . then the following are equivalent . * the measure @xmath83 is locally doubling . * there exist constants such that @xmath112 whenever @xmath113 . * the inequality @xmath114 holds . @xmath115 : suppose that @xmath22 is locally doubling , and let @xmath17 be the local doubling constant of @xmath22 for intervals @xmath21 such that @xmath116 . we first show that for @xmath117 in any interval @xmath21 of @xmath22-measure @xmath118 , then @xmath112 . observe that if @xmath119 is an interval , co - centric with @xmath21 , and such that @xmath120 then by local doubling @xmath121 . this implies that @xmath122 , and so @xmath123 . this implies @xmath124 . let @xmath125 denote the disk in @xmath73 such that @xmath126 . since @xmath127 is a sum of poisson kernels @xmath128 with respect to the zeros of @xmath129 , it follows that no such zero belongs to @xmath125 . indeed , if this was the case then the harmonic measure of @xmath20 with respect to this point , which is smaller than @xmath130 , would exceed @xmath131 . for each zero @xmath132 of @xmath129 , it now follows easily that @xmath133 , which yields @xmath112 . by a covering argument , we obtain @xmath134 . the converse is immediate . @xmath135 : if we denote the zeroes of @xmath129 by @xmath136 , then by the triangle inequality , @xmath137 where we have split the sum according to whether @xmath138 or @xmath139 , respectively . the first estimate is clear , @xmath140 to obtain the second estimate , we observe that for this case @xmath141 . indeed , first suppose that @xmath142 . letting @xmath143 be the point between @xmath144 and @xmath145 such that @xmath146 , it readily follows that @xmath147 on the other hand , if @xmath148 , then we know that @xmath149 . since @xmath150 ( always ! ) , the condition @xmath151 gives the required estimate . it now follows that @xmath152 the converse implication holds in general for any increasing function @xmath153 indeed , let @xmath154 and suppose that @xmath155 , then @xmath156 for inner functions @xmath39 on the unit disk @xmath157 with phase @xmath10 , it is proved by aleksandrov @xcite that condition @xmath158 above , under some mild additional restrictions that are automatically satisfied when @xmath39 is a meromorphic blaschke product , is equivalent to @xmath39 being one component . following the argument of aleksandrov , it is clear that the same result holds on the upper half - plane as long as one makes the additional hypothesis that @xmath159 has infinity in its spectrum ( cf . * theorem 3.4 ) ) . hence , we obtain the following consequence . [ aleksandrov corollary ] let @xmath0 be a hermite - biehler function with phase @xmath10 , and suppose that a subsequence of zeroes converges to infinity . then @xmath160 gives a locally doubling measure if and only if the inner function @xmath14 satisfies the one component condition . we point out that there exists a hermite - biehler function @xmath0 where the phase @xmath10 satisfies the local doubling condition , but not the doubling condition . this example is obtained by letting @xmath0 be the entire function with zeroes @xmath161 for @xmath162 . indeed , then @xmath163 . rewriting parts of the two previous lemmas in the notation of the metric @xmath164 we easily obtain the following . it is analogue to ( * ? ? ? * lemma 4 ) . [ technicallemma-2 ] let @xmath0 be a hermite - biehler function with phase @xmath10 . then there are constants depending on @xmath90 such that the following hold : * if the measure @xmath160 is locally doubling and @xmath165 , then @xmath166 . * if the measure @xmath167 is doubling and @xmath168 , then @xmath169 , for some @xmath170 depending only on the doubling constant of the measure . [ technicallemma ] when @xmath83 is a doubling measure , it follows by lemma [ technicallemma-2 ] that there exists some @xmath171 , such that for @xmath90 and @xmath168 , then @xmath172 with the constant depending on @xmath173 . [ miamadre ] observe that lemma [ technicallemma-2 ] still holds if we merely assume @xmath10 to be a smooth increasing function satisfying condition @xmath134 of lemma [ local doubling equivalences ] . the following lemma , while simple , explains the analogy between de branges spaces with a doubling phase function and fock spaces with a doubling laplacian . [ delta psi lemma ] let @xmath0 be a hermite - biehler function , and set @xmath174 then @xmath175 the function @xmath176 is analytic in a neighborhood of @xmath177 , so by the cauchy - riemann equations , @xmath178 . the result now follows by an argument using green s formula . we now show that for functions in a de branges space satisfying the local doubling property , a subharmonic - type estimate holds . this is a crucial tool , which in the context of fock spaces corresponds to ( * ? ? ? * lemma 19 ) . [ cmbsec : subharmonicity lemma ] let @xmath0 be a hermite - biehler function with phase function @xmath10 , and let @xmath72 be given by . if @xmath83 is a locally doubling measure on @xmath23 , then there are constants only depending on @xmath90 such that for all @xmath28 and @xmath89 we have @xmath179 and @xmath180 we first observe that an inequality of the type follows immediately by subharmonicity for a euclidean disk centered at @xmath181 , with radius @xmath182 , and which does not intersect @xmath23 . we now prove that such an inequality holds even if the disk intersects @xmath23 , and deduce the lemma from it . given @xmath183 , fix @xmath181 and set @xmath184 . suppose that the euclidean disc @xmath185 intersects @xmath23 . by green s formula , there exists a harmonic function @xmath186 on @xmath187 such that @xmath188 where @xmath189 is green s function for @xmath190 with pole at @xmath191 . observe that @xmath192 . we proceed to estimate @xmath193 for @xmath194 . set @xmath195 . by lemma [ delta psi lemma ] we have @xmath196 , and so @xmath197 it is clear that @xmath198 , so by lemma [ local doubling equivalences ] and the explicit form of the green function , we have @xmath199 applying this estimate to both @xmath200 and @xmath189 , we obtain @xmath201 . let @xmath202 be an analytic completion of @xmath186 . since @xmath46 is purely imaginary at the point @xmath132 , we can use the estimate on @xmath203 , and the subharmonicity of @xmath204 , to obtain @xmath205 where the implicit constant only depends on @xmath206 . the inequality now follows by observing that there exists an @xmath207 , depending only on @xmath173 , such that the ball @xmath208 is contained in @xmath92 . for , we use a simple contour argument involving the cauchy integral formula , and take the derivative to see that for any @xmath209 @xmath210 the conclusion now follows by choosing @xmath209 small enough , using the triangle inequality and applying . in this , and the next subsection , we use the fock space point of view to obtain the basic tools needed to study sampling and interpolating sequences for de branges spaces satisfying the doubling condition . most of these results have been proved by different methods in the context of one component model spaces . recall that a positive borel measure @xmath211 on @xmath212 is called carleson for the space @xmath5 if there exists a constant such that for all @xmath213 @xmath214 a well - known characterization , due to carleson @xcite and shapiro and shields @xcite , says that a positive borel measure @xmath211 is a carleson measure for @xmath5 if and only if @xmath215 for all bounded intervals @xmath19 , where @xmath216 is the square in @xmath212 with base @xmath21 . it is clear that the above holds true , with obvious modifications , for the space @xmath217 . carleson measures for model spaces were first considered by cohn @xcite . for positive measures supported on @xmath23 this notion carries over directly to de branges spaces : we say that a measure @xmath211 is a carleson measure for @xmath6 if @xmath218 we begin with a simple lemma . [ cmbsec : carleson lemma ] let @xmath0 be a hermite - biehler function with phase function @xmath10 , and set @xmath219 if @xmath83 is a locally doubling measure on @xmath23 , then the measure @xmath220 is carleson for @xmath221 . let @xmath19 be an interval with center @xmath222 , and denote by @xmath216 the carleson square with base @xmath21 . first , suppose that @xmath223 . by lemma [ local doubling equivalences ] , we have @xmath224 for all @xmath225 , and furthermore @xmath226 . hence , @xmath227 next , suppose that @xmath228 . let @xmath229 be a cover of @xmath21 by intervals with disjoint interiors such that @xmath230 . it follows from the doubling condition that @xmath231 with a uniform control of the implicit constants and @xmath232 . let @xmath233 denote the rectangle with base @xmath234 and height @xmath235 . since @xmath224 for @xmath236 , it now follows that @xmath237 for @xmath238 a similar argument works . [ cmbsec : discrete carleson remark ] a discrete version of this result also holds i.e. if @xmath239 is a @xmath58separated sequence of points in @xmath240 then @xmath241 is a carleson measure on @xmath242 . the following bernstein - type inequality is a special case of ( * ? ? ? * corollary 1.5 ) . we include a simple proof using our techniques . [ cmbsec : bernstein lemma ] let @xmath0 be a hermite - biehler function with phase function @xmath10 . if @xmath83 is a locally doubling measure on @xmath23 , then there exists a constant @xmath17 such that @xmath243 let @xmath244 . observe that @xmath245 implies @xmath246 . lemma [ cmbsec : subharmonicity lemma ] , followed by fubini s theorem yields @xmath247 the conclusion now follows by lemma [ cmbsec : carleson lemma ] . the above lemma also holds for any @xmath248 using the obvious generalizations . a well - known result by volberg and treil @xcite in the context of one component model spaces implies the following characterization . [ carleson_measures_lemma ] let @xmath0 be a hermite - biehler function with phase function @xmath10 . if @xmath160 is a locally doubling measure on @xmath23 , then a positive borel measure @xmath211 with support in @xmath23 is a carleson measure for the space @xmath6 if and only if @xmath249 for all intervals @xmath19 with @xmath250 . the following result is an immediate consequence of lemma [ carleson_measures_lemma ] . [ separation and sampling ] let @xmath0 be a hermite - biehler function with phase function @xmath10 . if @xmath83 is a locally doubling measure on @xmath23 , then a real sequence @xmath50 is bessel if and only if it is a finite union of uniformly @xmath58separated sequences . it follows that we may always assume that sampling sequences are @xmath58separated , which is a fact we need to prove our density results . we now establish some other basic results . the next lemma is a standard tool to prove perturbative results . indeed , combined with standard arguments , stability results for bessel sequences , sampling sequences and interpolating sequences follow ( see e.g. @xcite ) . while we include a simple proof of the lemma to illustrate the use of our techniques , we do not work out all of these consequences , instead we refer the interested reader to baranov @xcite , where such results are obtained for general one component model spaces using other tools . [ sasec : stability lemma ] let @xmath0 be a hermite - biehler function with phase function @xmath10 , and suppose that @xmath251 is a real @xmath58separated sequence . if @xmath83 is a locally doubling measure on @xmath23 , then for every @xmath171 and real sequence @xmath252 that satisfies @xmath253 , there exists a constant @xmath254 such that @xmath255 moreover , @xmath256 can be chosen to satisfy @xmath257 as @xmath258 . we begin by estimating @xmath259 to estimate @xmath260 we apply the cauchy - schwartz inequality to obtain @xmath261 in the case of @xmath262 , the cauchy - schwartz inequality followed by lemma [ local doubling equivalences ] yields @xmath263 let @xmath264 denote interval with end points @xmath265 . for @xmath171 small enough , the intervals @xmath264 do not overlap , and we simply sum the previous two estimates up to get @xmath266 the desired estimate for @xmath171 small , now follows from lemma [ cmbsec : bernstein lemma ] . to see that we get an inequality for any @xmath171 , with some constant , it is enough to note that in general there is a uniform bound , only depending on @xmath171 and @xmath0 , for how many times the intervals @xmath264 overlap . we end this section on preliminaries with two standard results on separation that we need in the proof of theorems [ theorem_sampling ] and [ theorem_interpolation ] . they do not seem to have appeared previously in our context . for a proof in the paley - wiener case see ( * ? ? ? * lemma 3.11 , 3.12 ) . we provide a proof of the first lemma to illustrate again the use of our techniques . let @xmath0 be a hermite - biehler function with phase function @xmath10 . if @xmath83 is a locally doubling measure on @xmath23 , then every interpolating sequence @xmath50 is @xmath58separated . for @xmath267 there exists a function such that for @xmath268 we have @xmath269 with @xmath270 . for @xmath271 , this implies @xmath272 for some @xmath145 in the interval with end points @xmath273 . by lemma [ cmbsec : subharmonicity lemma ] , this is less than @xmath274 by remark [ cmbsec : discrete carleson remark ] , the expression @xmath260 is bounded by some constant only depending on the norm of @xmath275 . hence , @xmath276 . but , by lemma [ technicallemma-2 ] , this implies that @xmath277 . let @xmath0 be a hermite - biehler function with phase function @xmath10 . if @xmath83 is a locally doubling measure on @xmath23 , then for every sampling sequence @xmath278 , there exists a @xmath10-separated sequence @xmath279 which is also a sampling sequence . this result can be deduced from lemma [ sasec : stability lemma ] . for a detailed proof in the bergman case see @xcite . in this section , we consider a larger class of spaces than the de branges spaces . in view of this , let @xmath280 denote a real valued subharmonic function with laplacian supported on the real line , and define the space @xmath281 note that if @xmath0 is a hermite - biehler function and , as above , we define @xmath282 then @xmath283 if and only if @xmath284 and @xmath285 are in @xmath286 , with the norm computed on @xmath23 . let @xmath22 denote the measure on @xmath23 given by @xmath287 . we make the additional assumptions that this measure is doubling , and that @xmath288 for some suitable function @xmath289 satisfying ( b ) of lemma [ local doubling equivalences ] , i.e. , there exist constants such that the following local doubling property holds : @xmath290 recall that by @xmath291 , @xmath292 and @xmath293 we denote the metric , open intervals and open balls induced by the measure @xmath294 , respectively , and that by @xmath103 we denote the zero set of the entire function @xmath99 . we first establish the existence of the so - called multiplier function . we follow the proof of theorem 41 in @xcite closely . [ multiplier theorem ] let @xmath295 and @xmath22 be as above . if the sequence @xmath296 of real numbers is separated in the metric @xmath291 and satisfies @xmath297 then we can construct a function @xmath298 such that @xmath299 , and , for some number @xmath300 , it holds that : * the set @xmath301 is separated from the other zeroes of @xmath99 in the sense that @xmath302 for all @xmath303 . * there exist constants such that @xmath304 for all @xmath303 and @xmath305 the idea in the first part of the proof , is to use to associate with @xmath288 a measure on @xmath73 that can be split up into pieces with good cancelation properties . to this end , we need the following lemma . [ ortega lemma ] let @xmath306 be a complex measure supported in an interval @xmath19 with @xmath307 and @xmath308 for @xmath309 . then there exists a set of points @xmath310 such that @xmath311 for @xmath312 . moreover , if @xmath313 , the points @xmath314 are in a ball @xmath315 for some constant @xmath316 depending only on @xmath317 and @xmath318 we sketch a proof due to ortega - cerd @xcite . let @xmath319 our objective is to solve the system of equations @xmath320 we first suppose that @xmath321 observe that @xmath322 the polynomials @xmath323 are called the power sum symmetric polynomials , and it is well - known that they generate the ring of symmetric polynomials with rational coefficients in @xmath317 variables . in particular , the coefficients of the polynomial @xmath324 which are the elementary symmetric polynomials , are in the ring of symmetric polynomials . in view of this , there exist polynomials @xmath325 such that @xmath326 as the elementary symmetric polynomials @xmath327 also generate the ring of symmetric polynomials in @xmath317 variables , there exist polynomials @xmath328 such that @xmath329 since the power sum symmetric polynomials are algebraically independent over @xmath330 @xcite , it now follows that @xmath331 let @xmath332 be the roots of the polynomial @xmath333 clearly , @xmath334 and therefore @xmath335 as it holds that @xmath336 the points @xmath337 are in @xmath338 for some constant @xmath17 depending only on @xmath317 and @xmath339 . in the general case , the interval @xmath340 can be sent to @xmath341 , by an affine map . if we apply the above procedure to the measure @xmath342 on @xmath341 induced by @xmath306 , we obtain a set of points @xmath343 . with this , it is not hard to see that the points @xmath344 have the desired properties . indeed , @xmath345 given @xmath177 , we set @xmath346 . observe that in this notation , @xmath347 define the measure @xmath348 fix @xmath162 so large that @xmath349 , where @xmath350 is the exponent appearing in lemma [ lemma gamma doubling ] for @xmath22 . by condition ( [ density condition ] ) , given any @xmath209 , if we choose @xmath351 large enough , then it holds for any interval @xmath19 , with @xmath352 , that @xmath353 choose @xmath354 so large that both @xmath355 and the above holds , and let @xmath356 be a partition of @xmath23 into intervals with @xmath357 . note that if @xmath354 is an integer , then so is @xmath358 hence , without loss of generality , we can add points to the set @xmath301 , while keeping the separation ( since no more than @xmath354 points have to be added to any single interval @xmath359 ) , so that @xmath360 let @xmath361 be the square , symmetric with respect to @xmath23 , such that @xmath362 , and consider the measure @xmath363 as a singularly supported measure on @xmath73 . let @xmath364 denote its restriction to @xmath361 . we apply lemma [ ortega lemma ] to each of the measures @xmath364 to find point sets @xmath365 for which the measures @xmath366 each have @xmath317 vanishing moments . here @xmath367 is the square co - centric with @xmath361 and with side - lengths scaled up by the factor @xmath368 the next step of the proof is to modify each @xmath369 to obtain sets @xmath370 such that for each @xmath371 , the measure @xmath372 also has @xmath317 vanishing moments , while keeping the set @xmath301 separated from @xmath373 in the sense that there exists an @xmath300 such that for all @xmath303 , the discs @xmath374 have empty intersection with @xmath375 . in this notation , the set @xmath370 may have some points of multiplicity @xmath317 . because of the separation of @xmath376 shrinking @xmath300 if necessary , we can suppose that the euclidean discs @xmath377 are pairwise disjoint for @xmath378 now , for each @xmath379 belonging to @xmath380 for some @xmath303 , set @xmath381 let @xmath382 be the union of all such sets for those @xmath383 . with this , define @xmath384 as mentioned above , in this definition we allow some points to have multiplicity @xmath317 . by construction , the set @xmath375 is clearly separated from @xmath301 in the sense given above ( see figure 1 ) . to see that the measure @xmath385 also has @xmath317 vanishing moments , it suffices to observe that for any @xmath386 and @xmath387 we have @xmath388 note that it may happen that the points in @xmath370 are now contained in some square larger than @xmath367 . however , since @xmath389 satisfies the local doubling property , it is easily seen that this may be remedied by increasing @xmath316 slightly , independent of @xmath390 . is obtained by replacing @xmath191 by points placed on a circle with center @xmath191 and radius depending on the proximity to the closest @xmath391 . the dotted circle indicates the placement of the @xmath317 new points , while the shaded area indicates the new separation between the new points and @xmath391 . ( note that we have normalized the distances . ) ] to define the desired multiplier function @xmath99 , let @xmath392 , and set @xmath393 in particular , @xmath394 . assume for a moment that @xmath395 is finite for all @xmath181 , and let @xmath100 denote any entire function with zeroes @xmath396 with appropriate multiplicities . then @xmath397 . i.e. , @xmath398 hence , by weyl s lemma , the function @xmath399 is harmonic on @xmath73 . therefore , there exists an entire function @xmath186 of which this is the real part , so @xmath400 . we set @xmath401 the final step of the proof is to obtain the required estimates for @xmath99 . indeed , to show that @xmath402 and that @xmath134 holds , it suffices to find positive constants @xmath403 for which @xmath404 and for small enough @xmath405 and all @xmath303 , @xmath406 for the estimates below , we let @xmath407 denote the center of @xmath408 and set @xmath409 . by the local doubling property , it follows readily that there is a radius @xmath410 such that for all @xmath371 we have @xmath411 . we first prove that holds . fix @xmath412 and let @xmath413 in this case , @xmath414 . by @xmath415 , denote the taylor polynomial of order @xmath416 of the function @xmath417 at the point @xmath418 . then , since @xmath419 , it holds that @xmath420 where @xmath421 is the segment in @xmath73 with endpoints @xmath418 and @xmath422 . so , by the @xmath317 vanishing moments of @xmath385 , @xmath423 let @xmath424 be such that @xmath425 . to see that the sum converges with a bound independent of @xmath132 , it suffices to establish that @xmath426 , for some constant @xmath427 only depending on @xmath428 . recall that @xmath349 , and @xmath350 is the exponent appearing in lemma [ lemma gamma doubling ] for @xmath22 . we leave it to the reader to find that such an estimate follows from the doubling property of @xmath22 . next , suppose that @xmath429 for some @xmath424 . by the doubling property , there exists a number @xmath430 , such that for @xmath431 , then @xmath432 . so , if we make the split @xmath433 then we can use the same method to get a bound @xmath434 as we did in the first case above . to estimate @xmath435 , we increase @xmath428 if necessary so that @xmath436 ( this is possible independently of @xmath390 ) . then we can use the first vanishing moment of @xmath385 to get @xmath437 since @xmath438 , and @xmath439 for @xmath440 , it follows by basic calculus that the integral on the right - hand side is comparable to @xmath428 . we turn to the demonstration of . consider @xmath303 and @xmath441 , where @xmath300 is a separation constant of @xmath301 from @xmath375 . to get the lower estimate , we make the same split @xmath442 as above , using @xmath391 in place of @xmath48 . the estimate of @xmath443 follows as before with the obvious modifications . in the lower estimate of @xmath435 , recall that above the expression @xmath444 was estimated simply as @xmath445 . the lower estimate of @xmath435 follows in the current case , if we replace this by @xmath446 as for the upper estimate of @xmath435 , since @xmath22 is a positive measure , @xmath447 it follows easily that the whole sum is bounded by some uniform constant , since every @xmath350 on the right - hand side is at least at a euclidean distance of @xmath448 from @xmath132 . this ends the proof . as an application of the above multiplier construction , we obtain a class of peak functions analogue to the ones used by beurling in his original work on sampling sequences . first , we record the following simple lemma . [ multiplier corollary ] let @xmath280 and @xmath389 be as above . then , for every @xmath449 there exist a function @xmath450 and a sequence @xmath451 that satisfy the following conditions , all with constants not depending on @xmath452 : * the set @xmath453 is separated . * the quantity @xmath454 is bounded below . * the map @xmath455 is bounded above . * we have @xmath456 for every @xmath89 . set @xmath457 . since @xmath458 we may use these @xmath278 to repeat the construction from theorem 3 we do this while keeping the sequence @xmath375 separated not only from @xmath278 but also from @xmath453 . this can be done in the same way as in the proof of theorem 3 , with a constant of separation not depending on @xmath452 . note that the constants @xmath459 and @xmath125 appearing in ( 10 ) and ( 11 ) depend only on the distribution of @xmath278 ( with respect to @xmath460 ) and the doubling constant of @xmath461 . hence , the bounds obtained in these constructions do not depend on @xmath452 . properties ( i ) , ( iii ) and ( iv ) follow immediately . to see that property ( ii ) is satisfied we make an estimate similar to the one on the right side of ( 11 ) , with @xmath462 , and use the fact that @xmath452 is uniformly separated from all points in @xmath463 . we are now ready to construct our peak functions . [ peak lemma ] let @xmath280 and @xmath389 be as above , and fix @xmath464 and @xmath465 . for every @xmath89 there is a function @xmath466 that satisfies @xmath467 and @xmath468 where @xmath316 is a constant not depending on @xmath145 . moreover , the following estimate holds whenever @xmath390 is big enough : @xmath469 for @xmath449 , let @xmath470 be the function from lemma [ multiplier corollary ] . property ( iii ) of the lemma implies that there exists @xmath471 , not depending on @xmath452 , such that the interval @xmath472 includes at least @xmath354 points from the corresponding sequence @xmath278 . choose such @xmath473 and set @xmath474 where @xmath475 is chosen to satisfy @xmath476 . it is clear that @xmath477 since outside of a big enough disc we have @xmath478 . the inequality is satisfied whenever @xmath354 is big enough , as in the proof of ( * ? ? ? * theorem 18 ) . to prove , we use to see that for @xmath479 we have @xmath480\frac{\psi'(y)dx}{1+d^{2k}_{\psi}(x , y)},\end{aligned}\ ] ] where the constant does not depend on @xmath481 . for the first integral on the right - hand side , the needed estimate is easily obtained . next , since @xmath482 is a doubling measure , by remark [ technicallemma ] , there exists a @xmath483 such that the second integral is dominated by @xmath484 by the change of variables @xmath485 , this is seen to converge for large @xmath390 . to get the necessary conditions for a real sequence of points to be interpolating or sampling , we use the technique developed by ramanathan and steger @xcite ( see also ( * ? ? ? * lemma 4 ) ) . since we have canonical orthonormal bases in de branges spaces , this theorem is easily applicable . let @xmath0 be a hermite - biehler function with phase function @xmath10 , and suppose that @xmath83 is a doubling measure on @xmath23 . if @xmath65 is a separated sampling sequence for @xmath6 and @xmath486 is an interpolating sequence for @xmath6 , then @xmath487 and @xmath488 . recall that @xmath489 is the reproducing kernel of @xmath6 at the point @xmath181 , while @xmath490 . we denote by @xmath491 and @xmath492 the dual frame of @xmath493 , and the biorthogonal basis of @xmath494 , respectively . for @xmath495 @xmath177 we define the sets @xmath496 and the corresponding orthogonal projections @xmath497 and @xmath498 onto @xmath499 and @xmath500 , respectively . the idea of the proof is the following observation . it is clearly sufficient to show that , given any @xmath59 , then for @xmath501 big enough , and all @xmath502 , the trace of the operator @xmath503 , as an operator on @xmath500 , satisfies @xmath504 on the finite dimensional space @xmath505 , the trace of the self - adjoint and positive operator @xmath506 is computed by @xmath507 the upper estimate follows easily from the fact that eigenvalues of @xmath506 are bounded by one in modulus : @xmath508 for the lower estimate , we use the cauchy - schwartz inequality and the biorthogonality property of @xmath509 to get @xmath510 the term @xmath511 is uniformly bounded , therefore it only remains to make @xmath512 small . since @xmath50 is sampling , a simple computation involving the projection @xmath497 yields @xmath513 now let @xmath514 @xmath515 and let @xmath171 be the separation constant of @xmath50 with respect to the metric @xmath516 . inserting the expression of our reproducing kernels , and making a simple estimate , this is bounded above by @xmath517 the left - most inequality holds as soon as @xmath518 , while the right - most holds for the @xmath519 given by lemma [ technicallemma-2 ] . with this , follows . for a de branges space @xmath6 with @xmath72 defined as in , we recall that @xmath520 . so with @xmath521 , given a @xmath10-separated sequence @xmath522 for which @xmath523 it follows from theorem [ multiplier theorem ] that for small enough @xmath405 there exists a function @xmath524 such that uniformly in @xmath303 we have @xmath525 and @xmath526 we now use this multiplier @xmath99 to solve the interpolation problem @xmath527 , where the data @xmath528 satisfies @xmath529 . our objective is to show that the solution is given by the following lagrange - type interpolation function : @xmath530 we proceed by duality . first , @xmath531 at each term of this sum , we multiply and divide by @xmath532 . the cauchy - schwarz inequality , in combination with the assumption on @xmath533 and , yields @xmath534 where @xmath46 denotes the hilbert transform on the real line . given @xmath303 , we consider the decomposition @xmath535 with this , the cauchy - schwarz inequality , lemma [ technicallemma-2 ] and , imply @xmath536 as for @xmath203 , we disconnect the @xmath391 appearing in the domain of integration and in the function , and define @xmath537 this function is analytic for @xmath538 . in a similar way we define the function @xmath539 , which is , however , only analytic for @xmath540 . by subharmonicity , and since the radius of @xmath541 is comparable to @xmath542 , we get the inequality @xmath543 since @xmath284 is bounded on @xmath23 , and the hilbert transform is bounded on @xmath544 , we use the fact that @xmath545 gives a carleson measure for both @xmath5 and @xmath217 , with constant only depending on @xmath546 , to see that @xmath547 plugging these estimates into , yields @xmath548 due to the @xmath10-separation of the points @xmath391 we get @xmath549 in addition , by lemma [ cmbsec : carleson lemma ] , @xmath550 this yields the desired conclusion as the cauchy transform is bounded on @xmath544 , and @xmath284 is bounded on @xmath23 . we follow the ideas of beurling in the case of the paley - wiener space @xcite , as adapted to the fock space setting by marco , massaneda and ortega - cerd in @xcite . for the space @xmath551 , defined in section [ sec_basic_constructions ] , we say that a sequence @xmath552 is a sampling sequence if there exists some constant @xmath17 such that for all @xmath553 we have @xmath554 the first lemma states that in order to prove that a sequence is sampling in a de branges space , it is enough to prove that it is sampling in such a uniform norm space . [ tango ] let @xmath0 be a hermite - biehler function with phase function @xmath10 , suppose that @xmath83 is a doubling measure on @xmath23 and let @xmath72 be defined by . moreover , suppose that @xmath278 is a @xmath10-separated sequence . if , for some @xmath555 , the sequence @xmath278 is a sampling sequence for @xmath556 , then it is a sampling sequence for @xmath6 . it is sufficient to show that if @xmath278 is a sampling sequence for @xmath556 , then every @xmath557 is pointwise equal to a series of the form @xmath558 with @xmath559 , where @xmath560 is the reproducing kernel in @xmath6 at @xmath561 . so , assume that for some @xmath59 , the sequence @xmath278 is a sampling sequence for @xmath556 . following a duality argument in @xcite , there exist functions @xmath562 , with @xmath563 uniformly in @xmath145 , so that for all @xmath564 such that @xmath565 as @xmath566 , we have @xmath567 let @xmath568 be the function from lemma [ peak lemma ] ( corresponding to the @xmath569 from above ) , and set @xmath570 . clearly , @xmath571 as @xmath572 , and so holds for this function . in particular , we have @xmath573 , so the formula gives pointwise @xmath574 moreover , by the estimate of lemma [ peak lemma ] , the functions @xmath575 belong to @xmath544 . now , to show that the series in ( [ the series ] ) converges in @xmath544 , it suffices to verify that the sequence @xmath576 is a bessel sequence in @xmath544 and that the sequence @xmath577 , for fixed @xmath578 , is in @xmath579 . the last claim follows from corollary [ separation and sampling ] , while the first holds as @xmath563 . given @xmath580 , we have @xmath581 . therefore , the equality ( [ the series ] ) yields @xmath582 . hence , the bessel property of @xmath583 implies that @xmath584 this concludes the proof . following the approach of beurling @xcite , we treat sampling sequences in @xmath585 by looking at certain translations . we need to consider translates , both of the sequence on which we sample and the function being sampled . since our spaces are not translation invariant , we modify the concept of translations accordingly . as in section [ sec_basic_constructions ] , we consider real valued subharmonic functions @xmath280 , for which @xmath288 is a doubling measure on @xmath23 induced by @xmath586 for some suitable function @xmath289 . moreover , we suppose that @xmath389 satisfies both ( b ) and ( c ) of lemma [ local doubling equivalences ] . i.e. , there exist constants such that @xmath587 and @xmath588 given a point @xmath449 we denote by @xmath589 the scaled translation @xmath590 with this , given a set @xmath301 , we make the following definitions , @xmath591 note that evaluating a function @xmath99 on the sequence @xmath278 is the same as evaluating @xmath592 on @xmath593 . also , @xmath594 and @xmath595 , and the radon - nikodym derivative of @xmath596 is @xmath597 . to translate @xmath280 , we consider the operator @xmath598(z)= \frac{1}{2\pi } \int_{{\mathbb{r } } } \left[\log { \left|1-\frac{z}{t } \right| } -\textrm{re}\big(p_q({z}/{t})\big)1\!\!1_{{\mathbb{r}}\setminus ( -1,1)}(t)\right ] f(t ) { \mathrm{d}}t,\ ] ] where @xmath599 is the taylor polynomial of degree @xmath430 of @xmath600 around @xmath601 . when it exists , the image @xmath602 $ ] is a subharmonic function which satisfies @xmath603= f(x){\mathrm{d}}x$ ] . to make the integral converge for @xmath604 , observe that since @xmath389 gives a doubling measure , by remark [ technicallemma ] , there exists some @xmath605 such that @xmath606 for all @xmath177 . a basic estimate ensures that the choice @xmath607 is sufficient . ( the number @xmath350 depends only on the doubling constant . ) given @xmath449 , we now define @xmath608 . \ ] ] recall , that the frchet distance between two closed sets @xmath609 is defined by @xmath610:=\max\{\sup_{x\in e } d(x , f),\sup_{y\in f } d(y , e)\}$ ] . we use beurling s definition of strong and compactwise limits of sets . a sequence @xmath611 of closed sets in @xmath23 , converges strongly to a set @xmath612 , denoted @xmath613 , if @xmath614\rightarrow 0 $ ] . the sequence @xmath615 converges compactwise to @xmath616 , denoted @xmath617 , if for every compact set @xmath618 in @xmath23 we have @xmath619 . given a set @xmath552 , we say that @xmath620 is a weak limit of @xmath278 , if there exists a sequence @xmath621 such that @xmath622 . [ w - l of l ] let @xmath389 be as above , and suppose that the sequence @xmath301 is @xmath389-separated . if @xmath623 is a sequence , then there exists a subsequence @xmath624 such that the following holds . * for some sequence @xmath625 , we have @xmath626 . * [ w - l of eta ] for a function @xmath627 , it holds that @xmath628 and @xmath629 uniformly on compacts . * let @xmath630 be as above . then it gives a doubling measure with doubling constant smaller than the one of @xmath631 . moreover , @xmath632 satisfies , with the same constants or better . * let @xmath630 be as above and set @xmath633 $ ] , then @xmath634 uniformly on compacts in @xmath73 . we write @xmath635 , and denote the set of all such week limits by @xmath636 . @xmath637 : fix some @xmath464 . for @xmath638 $ ] , by remark [ miamadre ] , it holds that @xmath639 for some @xmath170 . since in addition @xmath301 is @xmath389-separated , we get @xmath640)|}<\infty$ ] . passing to a subsequence we can assume that the number of points in each such set is constant . by a compactness argument we now find a set @xmath641 such that @xmath642\rightarrow \l_a$ ] . letting @xmath643 tend to infinity , we obtain @xmath625 from a usual diagonal argument . @xmath644 : for @xmath645 in an interval , there are constants , only depending on the euclidean length , such that @xmath646 . from this it follows that the functions @xmath647 and @xmath648 are uniformly bounded on the interval . by the arzel - ascoli theorem and cantor diagonalization , there exists a subsequence @xmath624 and a function @xmath627 such that @xmath629 uniformly on compacts . @xmath649 : by definition , the measure @xmath650 is doubling with a constant not bigger than the constant of @xmath22 , and taking limits the statement about @xmath651 follows . similarly , to show that holds for @xmath630 , it is sufficient to establish it for the @xmath652 . @xmath653 : to show that @xmath654-k[(\psi^*)']|}$ ] tends to zero on a given compact , split the domain of integration into a large interval @xmath655 and its complement . the latter part is made small by choosing @xmath173 and the degree of the taylor polynomial large enough , depending on the doubling constant of @xmath656 . the integral over @xmath655 is then handled by observing that the kernel of the operator is integrable and letting @xmath390 tend to infinity . in the next lemma , we list the properties preserved by weak limits . [ w - l of l2 ] let @xmath389 be as above , and suppose that the sequence @xmath301 is @xmath389-separated . if @xmath623 is a sequence such that @xmath657 uniformly on compacts for some function @xmath630 , and @xmath658 for some sequence @xmath625 , then the following holds : * @xmath625 is @xmath630-separated . * @xmath659 . @xmath637 : first , we note that , if @xmath278 is @xmath389-separated by some constant , then each @xmath660 is @xmath652-separated by the same constant . using the convergence of @xmath661 and @xmath662 to @xmath663 and @xmath625 , respectively , the conclusion follows . @xmath644 : by the definitions of @xmath650 and the density , it follows that for any @xmath664 we can choose @xmath90 sufficiently large to get @xmath665 observe that for big enough @xmath390 , we have @xmath666 ( this is because frchet limits allows an extra element on each endpoint of @xmath21 , and the sequences are uniformly separated ) . in addition , if @xmath667 then @xmath668 since @xmath669 uniformly on compacts , and so we obtain for big enough @xmath173 that @xmath670 by taking limits we get the result . given @xmath449 , let @xmath671 be an entire function with @xmath672 as its real part , and define an operator @xmath673 by @xmath674 the proof of the following lemma is straight - forward and omitted . [ w - l of f ] let @xmath280 and @xmath389 be as above . * @xmath675 is an isometry between @xmath676 and @xmath677 . moreover @xmath678 is a sequence of functions in @xmath679 , uniformly bounded in norm , and let @xmath623 . if @xmath680 uniformly on compacts , then there exists a subsequence @xmath624 for which @xmath681 converges uniformly on compacts to a function in @xmath682 . it is not difficult to see that if @xmath278 is a @xmath389-separated sampling sequence for @xmath551 and @xmath683 , then @xmath684 is a sampling sequence for @xmath685 and , in particular , a uniqueness set for the space . we will need the opposite direction of this claim , which is given in the following lemma . [ finaly using w - l thing ] let @xmath280 and @xmath389 be as above , and suppose that the sequence @xmath278 is @xmath389-separated . if the sequence @xmath684 is a uniqueness set for @xmath685 whenever @xmath683 , then @xmath278 is a sampling sequence for @xmath551 . assume that @xmath278 is not a sampling sequence for @xmath551 . this means that for every @xmath686 , there exists a function @xmath687 of unit norm and @xmath688 such that @xmath689 by lemma [ w - l of l ] , there exists a subsequence @xmath624 such that @xmath635 . by lemma [ w - l of f ] ( taking another subsequence if needed ) , @xmath681 converges uniformly on compacts to a function @xmath690 . we claim that on the one hand , @xmath691 and on the other hand , @xmath692 . this contradicts our assumption and therefore completes the proof . we omit the details . we can now use the previous lemma to connect between the density of a sequence and its sampling properties . [ finaly close to the end ] let @xmath280 and @xmath389 be as above , and suppose that the sequence @xmath278 is @xmath389-separated . if @xmath693 , then @xmath278 is a sampling sequence for @xmath679 . by lemma [ finaly using w - l thing ] , it is enough to show that if @xmath683 , and @xmath694 satisfies @xmath695 , then @xmath696 . assume to the contrary that there exists such @xmath99 which is non - zero . we may assume , without loss of generality , that @xmath697 ( otherwise we translate the setting as above ) . choose @xmath90 , by jensen s formula applied to @xmath99 , we get @xmath698|}}{t}dt\leq \frac{1}{2\pi}\int_0^{2\pi}\psi^*(re^{i\theta})d\theta + \textrm{o}(1).\ ] ] since @xmath699 , greens formula yields that the integral on the right - hand side of the last expression is equal to @xmath700)}{t}dt+\textrm{o}(1).\end{aligned}\ ] ] this contradicts our assumption about the density of @xmath278 , which , by lemma [ w - l of l2 ] , implies that there exists some @xmath59 , so that for any @xmath645 big enough , @xmath701|}}{\mu^*([-t , t])}\geq \frac{1+\epsilon}{2\pi}.\ ] ] this ends the proof . we are now ready to prove the main result . let @xmath72 be defined by , and recall that @xmath702 . since @xmath703 , there exists an @xmath59 small enough for @xmath704 to hold . set @xmath705 , and observe that @xmath706 . so , by lemma [ finaly close to the end ] , the sequence @xmath278 is sampling for @xmath679 . the result now follows from lemma [ tango ] . the authors would like to thank joaquim ortega - cerd , kristian seip , alexandru aleman and anton baranov for some helpful conversations . a. b. aleksandrov , _ a simple proof of the vol berg - treil theorem on the embedding of covariant subspaces of the shift operator _ , zap . nauchn . sem . otdel . mat . inst . steklov . ( pomi ) * 217 * ( 1994 ) , no . issled . po linein . 22 , 2635 , 218 . , _ embedding theorems for coinvariant subspaces of the shift operator . ii _ , zap . nauchn . sem . otdel . mat . inst . steklov . ( pomi ) * 262 * ( 1999 ) , no . issled . po linein . 27 , 548 , 231 . arne beurling , _ the collected works of arne beurling . vol . 2 _ , contemporary mathematicians , birkhuser boston inc . , boston , ma , 1989 , harmonic analysis , edited by l. carleson , p. malliavin , j. neuberger and j. wermer . i. g. macdonald , _ symmetric functions and hall polynomials _ , second ed . , oxford mathematical monographs , the clarendon press oxford university press , new york , 1995 , with contributions by a. zelevinsky , oxford science publications . a. l. volberg and s. r. treil , _ embedding theorems for invariant subspaces of the inverse shift operator _ , zap . nauchn . sem . otdel . mat . inst . steklov . ( lomi ) * 149 * ( 1986 ) , no . issled xv , 3851 , 186187 .

in 1985 , dunwoody showed that finitely presentable groups are accessible . dunwoody s result was used to show that context - free groups , groups quasi - isometric to trees or finitely presentable groups of asymptotic dimension 1 are virtually free . using another theorem of dunwoody of 1979 , we study when a group is virtually free in terms of its cayley graph , and we obtain new proofs of the mentioned results and others previously depending on these . _ 2000 mathematics subject classification . _ primary : 20f65 ; secondary : 05c25 _ key words . _ virtually free group , accessibility , cayley graph , asymptotic dimension .

0911.2177

let @xmath0 be a discrete , finitely generable , multiplicative group , fixed throughout the article . let @xmath1 be an oriented graph fixed throughout the article . recall that @xmath2 is a set with a specified non - empty subset @xmath3 @xmath4 is the complement of @xmath5 in @xmath2 , and @xmath6 are functions from @xmath4 to @xmath7 the elements of @xmath5 are called vertices , the elements of @xmath4 are called edges and the functions @xmath8 are called incidence functions . for two subsets @xmath9 @xmath10 of a set @xmath11 the complement of @xmath12 in @xmath13 will be denoted by @xmath14 ( and not by @xmath15 since we let @xmath16 denote the set of @xmath0-orbits of a left @xmath0-set @xmath17 ) . a _ sequence _ is a _ set _ endowed with a specified listing of its elements , usually represented as a vector in which the coordinates are the elements of ( the underlying set of ) the sequence . for two sequences @xmath9 @xmath18 their concatenation will be denoted @xmath19 we will find it useful to have a notation for intervals in @xmath20 that is different from the notation for intervals in @xmath21 let @xmath22 @xmath23 @xmath24 we write @xmath25 also , @xmath26 and @xmath27\coloneqq [ i{\uparrow}\infty [ \,\,\,\vee\,\ , \{\infty\}.$ ] we define @xmath28 $ ] to be the reverse of the sequence @xmath29,$ ] that is , @xmath30 we shall use sequence notation to define families of indexed symbols . let @xmath31 be a symbol . for each @xmath32 we let @xmath33 denote the ordered pair @xmath34 we let @xmath35}\coloneqq \begin{cases } ( v_i , v_{i+1 } , \cdots , v_{j-1 } , v_j ) & \text{if $ i \le j,$}\\ ( ) & \text{if $ i > j.$ } \end{cases}\ ] ] also , @xmath36 we define @xmath37}$ ] to be the reverse of the sequence @xmath38}.$ ] now suppose that @xmath38}$ ] is a sequence _ in _ the group @xmath39 that is , there is a specified map @xmath38 } \to g.$ ] we treat the elements of @xmath38}$ ] as elements of @xmath39 possibly with repetitions , and we define @xmath40}&\coloneqq \begin{cases } v_i v_{i+1 } \cdots v_{j-1 } v_j \in g & \text{if $ i \le j,$}\\ 1 \in g & \text{if $ i > j.$ } \end{cases}\\ \pi v_{[j{\downarrow}i ] } & \coloneqq \begin{cases } v_j v_{j-1 } \cdots v_{i+1 } v_i \in g & \text{if $ j \ge i,$}\\ 1 \in g & \text{if $ j < i .$ } \end{cases}\end{aligned}\ ] ] let @xmath41 be a set . suppose that @xmath42}$ ] is a sequence in the set of subsets of @xmath11 then we define @xmath43}\coloneqq \cap_{k\in [ i\uparrow j]}v_k$ ] and @xmath44}\coloneqq \cup_{k\in [ i\uparrow j]}v_k.$ ] let @xmath45 and for @xmath46 let @xmath47 denote @xmath48 @xmath49 denote @xmath50 and we extend @xmath6 to @xmath51 defining @xmath52 and @xmath53 a _ path _ @xmath54 _ in _ @xmath2 is a pair of sequences @xmath55},e_{[1\uparrow n]})$ ] , where @xmath56}$ ] is a sequence in @xmath3 @xmath57}$ ] is a sequence in @xmath51 such that for @xmath58,$ ] @xmath59 and @xmath60 the _ length _ of @xmath54 is @xmath61 the path is _ closed _ if @xmath62 the path @xmath54 is _ geodesic _ if for any other path @xmath63},f_{[1\uparrow m]})$ ] in @xmath2 with @xmath64 and @xmath65 @xmath66 a sequence @xmath56}$ ] in @xmath3 is _ a path in @xmath5 _ if there exists a sequence @xmath57}$ ] in @xmath51 such that @xmath55},e_{[1\uparrow n]})$ ] is a path in @xmath67 we use @xmath68 or simply @xmath69 to denote the path metric on @xmath7 that is , @xmath70 if there exists a geodesic path @xmath55},e_{[1\uparrow n]})$ ] in @xmath2 with @xmath71 and @xmath72 and @xmath73 if no such geodesic path exists . let @xmath74 be a finite generating set of @xmath75 the graph @xmath2 is the _ cayley graph _ of @xmath0 with respect to @xmath74 if @xmath76 @xmath77 and for @xmath78 @xmath79 and @xmath80 in this case , for @xmath81 , we put @xmath82 a group is _ virtually free _ if it has a free subgroup of finite index . by stalling s ends theorem , @xmath0 has _ more than one end _ if and only if there exists a @xmath0-tree with finite edge stabilizers such that no vertex is stabilized by @xmath0 . the group @xmath0 is _ accessible _ if there exists a @xmath0-tree with finite edge stabilizers , and each vertex stabilizer having at most one end . the group @xmath0 is _ context - free _ if for every finite generating set @xmath74 of @xmath39 the set @xmath83}\text { sequence in } x^{\pm 1 } : \pi x_{[1\uparrow n]}=1\}\ ] ] is a context - free language . see review [ rev : lang ] for a complete definition . in 1983 , muller and schupp ( * ? ? ? * theorem iii ) showed that @xmath0 is context - free and accessible if and only if @xmath0 is virtually free . by ( * theorem i ) , a group is context - free if every closed path in the cayley graph is @xmath84-triangulable . in particular it follows that being context - free can be expressed in terms of the cayley graph and that context - free groups are finitely presentable . in 1985 , dunwoody @xcite proved that finitely presented groups are accessible , and so the accessibility hypothesis of ( * ? ? ? * theorem iii ) can be removed . in this article we investigate how to characterize virtual freeness of a group in terms of its cayley graph . using an accessiblility result ( theorem [ t : dun ] ) which is implicit in an earlier paper of dunwoody @xcite of 1979 we obtain new proofs of several results which originally used the accessibility of finitely presented groups . let us list these results . let @xmath2 be the cayley graph of @xmath39 then @xmath0 is virtually free if and only if one of the following hold : 1 . there exists @xmath85 such that every closed path in @xmath2 is @xmath86-triangulable ( see ( * ? ? ? * theorems i and iii ) ) . @xmath2 is quasi - isometric to a cayley graph of a free group ( see @xcite ) . @xmath0 is finitely presentable and @xmath87 ( see ( * ? ? ? * cor 1.2 ) , ( * ? ? ? * thm 2 ) or ( * ? ? ? * cor g ) ) . 4 . @xmath2 admits a uniform spanning tree ( @xcite ) . the ends of @xmath2 have finite diameter ( @xcite ) . @xmath2 has finite strong tree - width ( @xcite ) . the original proofs of ( 1)-(3 ) directly invoke the accesibility of finitely presentable groups . the proofs of ( 4)-(6 ) invoke ( 1 ) . as ( 1 ) is equivalent to @xmath0 being context - free , we found two other characterizations of @xmath0 being virtually free : 1 . there exists a generating set of @xmath0 and @xmath88 such that every @xmath84-locally geodesic in @xmath0 is a geodesic ( @xcite ) 2 . there exists a generating set of @xmath0 such that @xmath0 admits a ( regular ) broomlike combing with respect to the generating set(@xcite ) . in section [ s : dun ] , we will review the classical geometric characterization of @xmath0 being virtually free in terms of actions on trees . we will use an accessibility result of dunwoody of 1979 , implicit in @xcite . in section [ s : graphs ] , we show the equivalence of many conditions for graphs . using results of section [ s : dun ] , we will show that the cayley graph of @xmath0 satisfies any of these conditions if and only if @xmath0 is virtually free . in particular , we give new proofs of ( 1),(2),(4)-(6 ) . at the end of the section , we recall the definition of a context - free group in review [ rev : lang ] , and we make some remarks about the connection of the @xmath86-triangulations with the characterizations ( 7 ) and ( 8) . in section [ s : adim ] , we review the concept of asymptotic dimension . we use ( 1 ) to give a proof of ( 3 ) . in this section , using a theorem of dunwoody , we show that @xmath0 is virtually free if and only if @xmath0 acts on a tree with finite stabilizers . let @xmath41 be a set and let @xmath89 denote the maps from @xmath0 to @xmath90 let @xmath91 and @xmath92 there is a natural right @xmath0-set structure on @xmath89 defined by @xmath93 for all @xmath94 the set @xmath95 is _ almost contained _ in @xmath11 written @xmath96 if @xmath97 is finite . if @xmath98 and @xmath99 then the sets @xmath95 and @xmath41 are _ almost equal _ , and it is denoted by @xmath100 let @xmath41 be a right @xmath0-set . a subset @xmath101 is _ almost - right-@xmath0-invariant _ if for all @xmath102 @xmath103 a function @xmath91 is _ almost - right-@xmath0-invariant _ if for all @xmath104 @xmath105 is almost - right-@xmath0-invariant . stallings ends theorem relates almost invariant sets of @xmath0 with graph of groups decompositions of @xmath75 in @xcite , using stallings theorem , karrass , pietrowski and solitar showed that finitely generable virtually free groups act on trees with finite stabilizers . we will follow the approach to the theory of dunwoody @xcite , using the following result . [ t : dun ] let @xmath0 be a finitely generable group and @xmath41 a finite set . let @xmath106 be almost - right-@xmath0-stable , then @xmath0 is the fundamental group of a graph of groups @xmath107 where @xmath17 is finite , the edge groups are finite and each vertex group is the right - stabilizer of some @xmath108 this result appears implicit in @xcite , as a combination of lemma 4.7 and lemma 4.6 with @xmath109 there is an explicit formulation in ( * ? ? ? * iii.3.4 ) . it can also be shown as a consequence of the almost stability theorem ( * ? ? ? * iii.8.5 ) . [ t : a ] let @xmath0 be a finitely generable group . the following are equivalent * @xmath0 acts on a tree with finite stabilizers . * @xmath0 is the fundamental group of a finite graph of finite groups . * @xmath0 is virtually free . * there exists a finite set @xmath41 and an almost - right-@xmath0-invariant function @xmath91 such that for all @xmath110 the right - stabilizer of @xmath111 is finite . ( a1 ) and ( a2 ) are equivalent by bass - serre theory , see ( * ? ? ? * i.4.1 ) and ( * ? ? ? * i.7.6 ) for a proof . ( a2 ) implies ( a3 ) by a classic result of karrass , pietrowski and solitar @xcite , see ( * ? ? ? * i.7.4 ) for a proof . ( a4 ) implies ( a2 ) by theorem [ t : dun ] . to complete the proof , it is enough to show that ( a3 ) implies ( a4 ) . let @xmath112 and suppose first that @xmath0 is free . let @xmath74 be a free generating set for @xmath39 and @xmath113 let @xmath91 be the function that assigns to each @xmath114 the first generator of the free reduced word in @xmath74 representing @xmath115 and @xmath116 let @xmath2 be the cayley graph of @xmath0 with respect to @xmath117 geometrically , @xmath118 is the function that collapses each connected component of @xmath119 to a vertex . suppose that @xmath120 then @xmath121 hence @xmath122 let @xmath123 @xmath124 and assume @xmath125 we will show that @xmath126 the @xmath0-stabilizer of @xmath127 if @xmath115 is cyclically reduced , @xmath128 then @xmath129 for infinitely many @xmath130 and thus @xmath131 for infinitely many @xmath132 . hence @xmath133 if @xmath115 is not cyclically reduced , then @xmath134 and there exist @xmath135 such that @xmath136 and @xmath137 by the previous discussion , @xmath138 and hence @xmath139 . we consider now the general case . let @xmath140 be a finite - index free subgroup of @xmath39 @xmath41 a finite set and @xmath141 an almost - right-@xmath140-invariant function such that for all @xmath142 @xmath143 is finite . let @xmath95 be a transversal for the left @xmath140-multiplication in @xmath75 for each @xmath81 , denote by @xmath144 and @xmath145 the unique elements @xmath146 and @xmath147 such that @xmath148 . let @xmath149 be defined by @xmath150 for @xmath104 @xmath151 hence @xmath152 is almost - right-@xmath0-stable . let @xmath153 for all but a finite number of @xmath154 @xmath155 let @xmath111 be the restriction of @xmath156 to @xmath140 . then for all but a finite number of @xmath154 @xmath157 as @xmath110 , it follows that @xmath158 is finite . throughout @xmath159 will be a fixed graph . we show the equivalence of many conditions for @xmath2 in a succession of lemmas . we summarize the results in theorem [ t : eqcond ] . furthermore , for cayley graphs , these conditions characterize virtual freeness . let @xmath160 and @xmath161 be two metric spaces . let @xmath162 be a map . the map @xmath163 is _ lipschitz continuous _ if there exists a constant @xmath164 such that for all @xmath165 @xmath166 the map @xmath163 is _ large - scale lipschitz _ if there exists a constant @xmath164 and a constant @xmath167 such that for all @xmath165 @xmath168 the map @xmath163 is _ bornologous _ if for every @xmath169 there exists @xmath170 such that for all @xmath171 , if @xmath172 then @xmath173 the map @xmath163 is _ coarse _ if it is bornologous and the inverse image , under @xmath174 of each bounded subset of @xmath17 , is a bounded set of @xmath117 the maps @xmath175 are _ close _ to each other if there exists @xmath176 such that for all @xmath177 @xmath178 let @xmath179 be a class of maps . we say that @xmath74 and @xmath17 are _ @xmath179-equivalent _ if there exist maps @xmath162 and @xmath180 in the class @xmath179 such that @xmath181 and @xmath182 are close to the identity . two metric spaces are @xmath74 and @xmath17 _ quasi - isometric _ if they are large - scale lipschitz equivalent . a _ uniform spanning tree _ of @xmath2 is a tree @xmath183 with a bijective map @xmath184 such that @xmath163 and @xmath185 are lipschitz continuous . the following result follows immediately from the definitions . [ l : mapclass ] @xmath163 lipschitz continuous @xmath186 @xmath163 large - scale lipschitz @xmath186 @xmath163 bornologous . [ r : finind ] recall that the cayley graphs of @xmath0 with respect to different generating sets are quasi - isometric . hence the quasi - isometric equivalence class of @xmath0 is well defined . recall also that a group @xmath0 is quasi - isometric to any finite index subgroup . hence cayley graphs of virtually free groups are quasi - isometric to trees . see ( * ? ? ? * proposition 3.19 ) . let @xmath187 an _ @xmath86-sequence _ of length @xmath132 in @xmath5 is a finite sequence @xmath56}$ ] in @xmath5 such that @xmath188 and @xmath189 for @xmath190.$ ] an @xmath86-sequence @xmath56}$ ] in @xmath5 is _ @xmath86-reducible _ if there exists @xmath191,$ ] such that @xmath192 in this event we can perform an _ @xmath86-reduction _ of @xmath193}$ ] and obtain a new @xmath86-sequence @xmath194}\vee v_{[i+1\uparrow n]}.$ ] an @xmath86-sequence is _ @xmath86-triangulable _ if it has length @xmath195 or if there exists an @xmath86-reduction that produces a sequence which is @xmath86-triangulable . [ ex:0red ] recall that a closed path @xmath55},e_{[1\uparrow n]})$ ] is _ reducible _ if @xmath196 or @xmath197 is a subsequence of @xmath57}.$ ] hence if a closed path @xmath55},e_{[1\uparrow n]})$ ] of length @xmath198 is reducible then the @xmath199-sequence @xmath56}$ ] is @xmath200-reducible . the term triangulation can be justified as follows . let @xmath56}$ ] be a 1-sequence which is @xmath86-triangulable . for each @xmath201,$ ] identify @xmath33 with the complex number @xmath202 and the @xmath199-sequence with the regular polygon @xmath203 of vertices @xmath204}.$ ] if @xmath205 then @xmath203 is a ( maybe very degenerate ) triangle . let @xmath206}\vee v_{[p+1\uparrow n]}$ ] be the @xmath86-sequence obtained by an @xmath86-reduction of the @xmath86-triangulation . add a diagonal from @xmath207 to @xmath208 this diagonal divides @xmath203 into a polygon with @xmath209 sides and a triangle . repeating this process , the @xmath86-triangulation of @xmath56}$ ] gives a triangulation of @xmath203 by diagonals . we still need some more definitions , however we state now the main theorem of this section . the proof is broken into a series of lemmas , and definitions will be introduced as needed . [ t : eqcond ] let @xmath2 be graph . then the following are equivalent : 1 . @xmath2 admits a uniform spanning tree . @xmath2 is quasi - isometric to a tree . @xmath2 is bornologous equivalent to a tree . there exists @xmath210 such that every @xmath199-sequence in @xmath5 is @xmath86-triangulable . there exists @xmath210 such that for every @xmath211 and every path @xmath54 from @xmath212 to @xmath31 the inequality @xmath213 holds . there exists @xmath210 such that for every @xmath214 and every @xmath215 the boundaries of @xmath216 have diameter less than or equal to @xmath217 7 . @xmath2 admits a strong tree decomposition of uniformly bounded diameter . moreover if @xmath2 is the cayley graph of @xmath0 with respect to a finite generating set , then ( b1)-(b7 ) and the following are equivalent : 1 . the ends of @xmath2 have uniformly bounded diameter . there exists @xmath218 such that for every @xmath219 every path @xmath54 from @xmath212 to @xmath31 and every @xmath220 in the geodesic from @xmath212 to @xmath31 the inequality @xmath221 holds . @xmath0 is virtually free . there exists a finite set @xmath41 and an almost - right-@xmath0-invariant function @xmath222 such that for all @xmath110 the right - stabilizer of @xmath111 is finite . notice that the value of the constant @xmath86 may be different from one condition to another . by theorem [ t : a ] ( a4)@xmath186(a3 ) , by remark [ r : finind ] ( a3)@xmath186(b2 ) . by lemma [ l : mapclass ] , ( b1 ) @xmath186 ( b2 ) @xmath186 ( b3 ) . the scheme of the rest of the proof is the following . first we close a cycle of implications from ( b1 ) to ( b7 ) and back to ( b1 ) . then we show ( b7)@xmath186 ( b8)@xmath186 ( b9)@xmath186 ( a4 ) . notice that ( b9 ) is a particular case of ( b5 ) , with @xmath223 . the reader interested in showing the muller and schupp theorem or that virtually free groups are quasi - isometric to trees can go from lemma [ l : triang->pathineq ] directly to lemma [ l : pathineq2->a ] . we start now the proof of the theorem . the following useful lemma is a classic result of nielsen @xcite . see also ( * ? ? * i proposition 2.2 ) . [ l : triangintree ] let @xmath183 be tree . for all @xmath224 every @xmath86-sequence in @xmath225 is @xmath86-triangulable . we will argue by induction on the length of the @xmath86-sequence . by definition all @xmath86-sequences of length @xmath195 are @xmath86-triangulable . let @xmath56}$ ] be an @xmath86-sequence in @xmath225 with @xmath226 and assume that any shorter @xmath86-sequence is @xmath86-triangulable . let @xmath227 be the smallest connected subgraph of @xmath183 containing @xmath56}.$ ] hence @xmath227 is a locally finite tree and it can be identified with a subgraph of the cayley graph of a free group @xmath228 with respect to a free generating set @xmath117 we may think of @xmath56}$ ] as a sequence in @xmath229 by the induction hypothesis , it is enough to show that @xmath56}$ ] is @xmath86-reducible for @xmath230,$ ] let @xmath231 hence @xmath232}=1.$ ] notice that for @xmath230,$ ] @xmath233 for @xmath234,$ ] let @xmath235 be the maximal common prefix of @xmath236 and @xmath237 let @xmath238 similarly , let @xmath239 be the common prefix of @xmath240 and @xmath241 if for some @xmath230,$ ] @xmath242 , then either @xmath243 or @xmath244 is greater than half @xmath245 . suppose that @xmath246 after cyclically permuting @xmath56},$ ] if necessary , we can assume that @xmath247 then @xmath248 and @xmath56}$ ] is @xmath86-reducible . we restrict then to the case @xmath249 for all @xmath250.$ ] let @xmath251 hence @xmath252}={a_1}^{-1}(\pi b_{[1\uparrow n]})c_n^{-1}.\ ] ] by construction , there are no reductions on the product @xmath253},$ ] @xmath254 for @xmath230.$ ] let @xmath255.$ ] if @xmath256 then @xmath257 and hence @xmath56}$ ] is @xmath86-reducible . if @xmath258 as @xmath259 either @xmath260 or @xmath261 suppose the former , the discussion for the latter is analogous . as @xmath262 @xmath263 also as @xmath264 @xmath265 hence @xmath266 and @xmath56}$ ] is @xmath86-reducible . in the case @xmath267 we conclude that @xmath268 and hence @xmath56}$ ] is @xmath86-reducible . [ c : born->trian ] let @xmath183 be a tree and suppose that @xmath2 is bornologous equivalent to @xmath183 . then for every @xmath269 there exists @xmath270 such that every @xmath84-sequence in @xmath5 is @xmath271-triangulable . let @xmath272 and @xmath273 be bornologous maps , such that @xmath274 is close to the identity . let @xmath275 such that for every @xmath276 @xmath277 . let @xmath278 there exists @xmath279 such that for all @xmath219 if @xmath280 then @xmath281 there exists @xmath282 such that if @xmath283 then @xmath284 and hence @xmath285 let @xmath56}$ ] be a @xmath84-sequence in @xmath5 . hence @xmath286})$ ] is an @xmath86-sequence in @xmath225 and , by lemma [ l : triangintree ] , it is @xmath86-triangulable . as each @xmath86-reduction of @xmath286})$ ] gives a @xmath287-reduction of @xmath56}$ ] , it follows that @xmath56}$ ] is @xmath287-triangulable . [ l : triang->pathineq ] let @xmath288 let @xmath2 be a graph such that every 1-sequence in @xmath5 is @xmath86-triangulable . then for every @xmath211 and any path @xmath289}$ ] in @xmath5 from @xmath212 to @xmath290 the following inequality holds : @xmath291})\leq \dfrac{1}{2}(\operatorname{d}_\gamma(z , u)+\operatorname{d}_\gamma(z , v)-\operatorname{d}_\gamma(u , v)+3m).\ ] ] let @xmath292}$ ] and @xmath293}$ ] be geodesic paths in @xmath5 from @xmath220 to @xmath212 and from @xmath31 to @xmath294 respectively . then @xmath292}\vee x_{[1\uparrow n-1]}\vee v_{[\operatorname{d}(z , v)\downarrow 0]}$ ] is a @xmath199-sequence and hence it is @xmath86-triangulable . notice that @xmath295 if @xmath296 as @xmath297 the inequality holds . thus , we may assume that @xmath298 there exists a sequence @xmath299}$ ] of @xmath86-sequences such that @xmath300}\vee x_{[1\uparrow n-1]}\vee v_{[\operatorname{d}(z , v)\downarrow 0]},$ ] for @xmath301 $ ] each @xmath302 is an @xmath86-reduction of @xmath303 and @xmath304 has length @xmath305 as @xmath306 are the first and last terms in all @xmath307 @xmath308,$ ] there exists @xmath309 $ ] such that @xmath310 only contains one term of @xmath311},$ ] say @xmath312 if @xmath310 has the form @xmath313 or @xmath314 then @xmath315})\leq m$ ] and the inequality hold . thus , it remains the case when @xmath310 has the form @xmath316 where @xmath3170\uparrow \operatorname{d}(z , u)],$ ] @xmath318 and @xmath319 is obtained from @xmath310 by suppressing @xmath312 [ figabc ] in this event @xmath320 is an @xmath86-sequence . then @xmath321})&\leq&\operatorname{d}(z , u_a)+m.\\ \operatorname{d}(z , x_{[0\uparrow n]})&\leq&\operatorname{d}(z , v_c)+m.\\\end{aligned}\ ] ] summing up the three inequalities we get @xmath322})\leq \operatorname{d}(z , u ) + \operatorname{d}(z , v ) + 3m\ ] ] which is equivalent to . recall that the _ gromov product _ of @xmath323 based at @xmath220 is @xmath324 hence can be reformulated as for all @xmath325 and all paths @xmath54 from @xmath326 to @xmath327 @xmath328 as we mentioned , readers not interested in ( b6)-(b8 ) can go directly to lemma [ l : pathineq2->a ] . let @xmath329 and @xmath330 the _ boundary _ of @xmath41 is @xmath331 the _ ball of center @xmath220 and radius _ @xmath332 written @xmath333 is the complete subgraph of @xmath334 having all the vertices @xmath335 at distance at most @xmath132 from @xmath220 as vertex set . [ l : pathineq->boundaries ] let @xmath288 suppose that for every @xmath211 and any path @xmath289}$ ] in @xmath5 from @xmath212 to @xmath31 the inequality hold . then for all @xmath214 and all @xmath336 all connected components @xmath337 of @xmath338 satisfy @xmath339 fix @xmath214 and @xmath340 and let @xmath341 be in the boundary of a connected component of @xmath342 let @xmath343}$ ] be a path in @xmath5 from @xmath212 to @xmath31 contained in @xmath342 then by , @xmath344})\leq 1/2(2n-\operatorname{d}(u , v)+3m)$ ] and hence @xmath345 [ def : rgraphpartition ] let @xmath346 be a metric space , @xmath347 a partition of @xmath74 and @xmath348 @xmath349-graph associated to @xmath347 _ is an unoriented graph @xmath350 with vertex set @xmath347 and an edge between @xmath351 if and only if @xmath352 . a _ strong tree decomposition _ of @xmath5 is a partition @xmath347 of @xmath5 such that the @xmath199-graph associated to @xmath347 is a tree . a strong tree decomposition @xmath347 has _ uniformly bounded diameter _ _ finite width _ ) if there exists a constant @xmath353 such that for all @xmath354 @xmath355 ( resp . @xmath356 ) . [ l : finwith ] if there is a strong tree decomposition @xmath347 of @xmath2 with uniformly bounded diameter and the degrees of the vertices of @xmath2 are uniformly bounded , the strong tree decomposition also has finite width . let @xmath357 such that for all @xmath354 @xmath358 let @xmath359 be a uniform bound of the degrees of the vertices of @xmath67 for @xmath354 let @xmath17 be a minimal subtree of @xmath2 containing all vertices of @xmath90 a ball of radius @xmath84 in a regular tree of degree @xmath359 has at most @xmath360 vertices , it follows that @xmath17 has at most @xmath360 vertices and hence , @xmath361 [ l : treedecom ] let @xmath2 be a connected graph and @xmath330 suppose that for all @xmath336 the boundaries of the connected components of @xmath338 have diameter @xmath362 then there exists a strong tree decomposition of @xmath2 with uniformly bounded diameter . moreover , if the degrees of the vertices of @xmath2 are uniformly bounded by @xmath363 the strong tree decomposition also has finite width . let @xmath364 hence , for all @xmath354 @xmath365 let @xmath350 be the @xmath199-graph associated to the partition @xmath366 to show that @xmath350 is a tree , it is enough to show that a closed path @xmath367},e_{[1\uparrow l]})$ ] in @xmath350 is reducible . without loss of generality , assume that @xmath368 by construction , there are no edges joining a vertex with itself , hence , for @xmath369,$ ] @xmath370 . there exists @xmath371 $ ] such that for all @xmath372,$ ] @xmath373 for some @xmath374 @xmath375 and @xmath376 are the boundaries of the connected component of @xmath377 containing @xmath378 hence @xmath379 and the path is reducible . [ figstree ] if the degrees of the vertices of @xmath2 are unifomly bounded , then , by lemma [ l : finwith ] , the strong tree decomposition also has finite width . [ l : std->usp ] suppose that @xmath2 admits a strong tree decomposition of uniformly bounded diameter . then @xmath2 admits a uniform spanning tree . let @xmath347 be a partition of @xmath5 into sets of diameter @xmath380 such that the @xmath199-graph associated to @xmath347 is a tree . for each @xmath354 take @xmath381 with @xmath382 let @xmath183 be the tree with vertex set @xmath3 such that for all @xmath383 and all @xmath384 there is an edge joining @xmath31 and @xmath385 and for all @xmath386 with @xmath387 there is an edge joining @xmath388 and @xmath389 let @xmath341 , @xmath390 @xmath391 @xmath392 and @xmath393 if @xmath394 then @xmath395 if @xmath396 then @xmath397 also @xmath398 hence for all @xmath219 @xmath399 notice that we have proved the equivalence between ( b1)-(b7 ) . [ r : treedecom ] there exists the more general concept of a _ tree decomposition _ of a graph , which we will not use here . this concept is important in the theory of graph minors . in @xcite , kuske and lohrey characterize virtual freeness in terms of the tree decompositions . by theorem 3.4 and theorem 3.7 of @xcite , a vertex - transitive locally - finite graph admits a strong tree decomposition of finite width and uniformly bounded diameter if and only if @xmath2 has a tree decomposition of finite width . hence , for cayley graphs , all these concepts turn out to be equivalent . let @xmath2 be a locally finite graph . let @xmath400 let @xmath401 be the set of infinite paths in @xmath5 which do not repeat a vertex . for all @xmath402 and @xmath403 as @xmath404 either @xmath405 or @xmath406 for each @xmath403 we define a map @xmath407\colon \mathfrak b \gamma \to \{0,1\}$ ] as follows : for @xmath402 we set @xmath407(b)=1 $ ] if @xmath405 and @xmath407(b)=0 $ ] otherwise . there is a natural equivalence relation @xmath408 in @xmath409 given by @xmath410 if and only if @xmath407=[p'].$ ] the space @xmath411 : p \in \mathfrak p \gamma \}$ ] is called _ the space of ends of @xmath2_. for @xmath412 let @xmath413\in \mathfrak e\gamma : [ p](b)=1\}\subseteq \mathfrak b \gamma \cup \mathfrak e \gamma\ ] ] and let @xmath414 then @xmath415 with the topology defined by @xmath416 is a totally disconnected compact hausdorff topological space . see ( * ? ? ? * iv.6 ) , for details . let @xmath417 an _ end @xmath407 $ ] of @xmath418 has diameter @xmath419 _ if there exists a sequence @xmath420 in @xmath421 such that for all @xmath215 @xmath422 @xmath423 and @xmath424.$ ] the sequence @xmath420 is _ a sequence of diameter @xmath419 converging to @xmath407$]_. the ends of @xmath2 have _ uniformly bounded diameter _ if there exists @xmath357 such that every @xmath407\in \mathfrak e \gamma$ ] has diameter @xmath425 [ l : std->ends ] if the vertices of @xmath2 have uniformly bounded degree and @xmath2 has a strong tree decomposition of uniformly bounded diameter , then the ends of @xmath2 have uniformly bounded diameter . by lemma [ l : treedecom ] , there exist @xmath84 and @xmath359 and a strong tree decomposition @xmath347 of @xmath2 such that for all @xmath354 @xmath426 and @xmath427 we may consider @xmath347 as the vertex set of a locally finite tree @xmath428 let @xmath429}\in \mathfrak p \gamma.$ ] the path @xmath430 can be viewed as an infinite sequence @xmath431 in @xmath432 with @xmath433 which repeats a vertex at most @xmath359 times . hence for @xmath434 either @xmath405 or @xmath435 we construct a sequence @xmath436}$ ] in @xmath437 such that @xmath438 and @xmath407(b_i)=1 $ ] for @xmath439.$ ] let @xmath440 be the connected component of @xmath441 such that @xmath407(b_0)=1.$ ] to construct @xmath442 from @xmath443 we consider the first @xmath444 such that @xmath445 for @xmath446 . as @xmath407(b_i)=1 $ ] such a @xmath447 exists . then there exists a connected component @xmath337 of @xmath448 strictly contained in @xmath449 such that @xmath407(c)=1 $ ] . let @xmath450 hence for @xmath451 @xmath452 @xmath453 and @xmath407\in \cap { \overline}{b}_{[1\uparrow \infty[}.$ ] let @xmath454 the set @xmath10 is almost equal to a subforest of @xmath183 with finitely many components . suppose that @xmath455 hence for a large enough @xmath332 @xmath456 is contained in a subforest of @xmath457 let @xmath458 such that @xmath459\in \cap { \overline}{b}_{[0\uparrow \infty[}.$ ] if @xmath459(b)=1,$ ] then @xmath460 and @xmath10 contains some @xmath461 therefore @xmath407(b)=1.$ ] similarly , if @xmath459(b)=0 $ ] then @xmath407(b)=0.$ ] hence @xmath407(b)=[q](b)$ ] for all @xmath412 thus @xmath462 and @xmath407=\cap { \overline}{b}_{[0\uparrow \infty[}.$ ] [ l : ends->pathineq2 ] let @xmath2 be a locally finite vertex - transitive graph . suppose that the ends of @xmath2 have uniformly bounded diameter . then there exists @xmath463 such that for all @xmath219 any @xmath464 geodesic from @xmath212 to @xmath290 any @xmath465 and any path @xmath54 from @xmath212 to @xmath290 @xmath466 let @xmath357 be a uniform bound of the diameter on the ends of @xmath67 let @xmath467 for @xmath407\in \mathfrak e\gamma,$ ] let @xmath468}_{[0\uparrow \infty[}$ ] be a sequence of diameter @xmath419 converging to @xmath407.$ ] the set @xmath469 } : [ p]\in\mathfrak e \gamma , n\in [ 0\uparrow \infty [ , \operatorname{d}(t , b_n^{[p]})>k \}\ ] ] is an open covering of @xmath470 as @xmath471 is compact , there exist @xmath472 such that @xmath473}.$ ] let @xmath474}.$ ] then @xmath41 is finite and @xmath475 let @xmath463 such that @xmath476\ ] ] and @xmath477 let @xmath478 and @xmath479 be as in the hypothesis . as @xmath2 is vertex transitive , we may assume that @xmath480 if @xmath481 or @xmath482 then @xmath466 if @xmath483 and @xmath484 then @xmath485 and @xmath486 for some @xmath487.$ ] if @xmath488 then there exist @xmath489 such that @xmath490 and @xmath491 as @xmath492 @xmath493 as @xmath220 is in the geodesic @xmath494 @xmath495 by construction of @xmath11 @xmath496 and @xmath497 hence @xmath498 combining and , @xmath499 but this is impossible since @xmath417 hence @xmath500 and then path @xmath54 goes through @xmath501 hence , by , @xmath502 now we connect the graph conditions using theorem [ t : dun ] . [ l : pathineq2->a ] let @xmath2 be a cayley graph of a group @xmath0 with respect to a finite generating set @xmath117 let @xmath463 such that for all @xmath219 any geodesic @xmath464 from @xmath212 to @xmath290 any @xmath465 and any path @xmath54 from @xmath212 to @xmath290 @xmath466 then there exists a finite set @xmath41 and an almost - right-@xmath0-invariant function @xmath91 such that for all @xmath503 the right - stabilizer of @xmath111 is finite . if @xmath0 is finite , the result is trivial . so , we assume that @xmath0 is infinite . let @xmath41 be the partition of @xmath0 formed by the ( vertex set of the ) connected components of @xmath504 and ( the vertex set of ) @xmath505 let @xmath506 be the function satisfying @xmath507 for all @xmath92 fix @xmath81 , and let @xmath508}$ ] be a sequence in @xmath509 such that @xmath510}=g$ ] . let @xmath511 such that @xmath512 then @xmath513}|>m$ ] for all @xmath514 $ ] and hence @xmath515 and @xmath516 lie in the same connected component of @xmath517 thus @xmath518 this shows that @xmath118 is almost - right-@xmath0-invariant . let @xmath108 there exists @xmath340 such that @xmath118 and @xmath111 agree on @xmath519 let @xmath81 such that @xmath520 and @xmath521}.$ ] let @xmath522}.$ ] hence , as the geodesic from @xmath523 to @xmath524 goes through @xmath525 any path from @xmath523 to @xmath524 goes through @xmath526 and hence @xmath523 and @xmath524 lie in different connected components of @xmath517 therefore @xmath527 and hence @xmath115 does not stabilize @xmath127 thus @xmath528 is finite . compare the proof of the lemma with the proof of ( a3 ) implies ( a4 ) in theorem [ t : a ] . this completes the proof of theorem [ t : eqcond ] the equivalence between ( b5 ) and ( a3 ) is due to muller and schupp @xcite . by ( * theorem i ) , a group is context - free if and only if ( b5 ) holds . the equivalence between ( b2 ) and ( a3 ) is attributed to gromov and appears explicitly in ( * ? ? ? * thm 7.19 ) . the equivalence between ( b1 ) , ( b4 ) and ( b8 ) appears in the article of woess @xcite . woess uses the result of muller and schupp to conclude the equivalence of these conditions with ( a3 ) . the equivalence between ( b7 ) and ( a3 ) is due to kuske and lohrey @xcite . see remark [ r : treedecom ] . [ rev : lang ] the original result of muller and schupp @xcite is about context - free groups . for completeness we recall the definition . by @xmath529 we denote the free monoid freely generated by a set @xmath117 a subset @xmath530 of @xmath529 is _ a language over x_. we identify @xmath529 with the set of sequences in @xmath74 with the operation given by concatenation . usually the languages are classified either by the type of grammars that generate the language or by the type of automatons that recognize the sequences in the language . we follow the second approach . a language @xmath530 is _ regular _ , if it is accepted by a finite - state automaton . a _ finite - state automaton _ @xmath13 over @xmath74 is a quadruple @xmath531 where @xmath532 is a finite set , called the set of states , @xmath533 is the initial state , @xmath534 is a function and @xmath41 is a subset of @xmath532 , called the set of accepting states . it is useful to think @xmath13 as an oriented labelled graph with vertex set @xmath532 and edges labelled by @xmath535 starting at @xmath212 and ending at @xmath536 for @xmath537 a sequence @xmath538}$ ] in @xmath74 is recognized by an automaton if there exists a sequence @xmath56}$ ] in @xmath532 such that @xmath539 @xmath540 for @xmath230 $ ] and @xmath541 a language @xmath530 is _ context - free _ , if it is accepted by a push - down automaton . it can also be defined as a language generated by a context - free grammar . roughly speaking , a push - down automaton is a finite - state automaton with a memory stack . now the transition function @xmath542 depends on the state and the top symbol of the stack , and outputs a new state and modify the top of the stack , by deleting , adding or keeping the top symbol . formally , a _ push - down automaton _ @xmath13 over @xmath74 is a 6-tuple @xmath543 where @xmath544 and @xmath41 are as for the finite - state automaton , @xmath545 is a finite set and @xmath546 and the function @xmath547 satisfies that if @xmath548 then @xmath111 is either equal to @xmath54 or obtained from @xmath54 by either adding a symbol at the end of the sequence , or deleting the last symbol of the sequence . a sequence @xmath538}$ ] in @xmath74 is recognized by the automaton if there exist sequences @xmath56}$ ] in @xmath532 and @xmath549}$ ] in @xmath550 such that @xmath551 @xmath552 for @xmath230 $ ] and @xmath541 given a group @xmath0 and a finite generating set @xmath74 of @xmath39 we say that @xmath553 is regular ( respectively context - free ) if the set @xmath554 of the sequences @xmath538}\in(x^{\pm 1})^*$ ] such that @xmath555}=1 $ ] in @xmath0 is a regular ( respectively context - free ) language . it can be shown that if @xmath556 is another finite generating set of @xmath0 and @xmath554 is regular ( respectively context - free ) , then @xmath557 is also regular ( respectively context - free ) . the regular groups are exactly the finite ones . this is a classical result of anisimov . the muller - shupp theorem states that context - free groups are exactly the virtually free ones . [ rem : otherchar ] there exist other characterizations of virtual - freeness that involve language conditions . for example , in @xcite , bridson and gilman show that it is possible to put more structure on the spanning tree of ( b1 ) of theorem [ t : eqcond ] . in particular they show that a group @xmath0 is virtually free if and only if there exists a finite generating set @xmath74 of @xmath0 such that @xmath0 admits a regular broomlike combing with respect to @xmath117 we remark that the result holds for any finite generating set . let @xmath558 be a surjective monoid morphism . a subset @xmath530 of @xmath529 is _ a combing _ with respect to @xmath74 if the restriction of @xmath559 to @xmath530 is a bijection with @xmath75 the combing is _ broomlike _ if there exists a constant @xmath84 such that for all @xmath560 where @xmath561 there exist @xmath562 such that @xmath563 @xmath564 and @xmath565 the subset @xmath530 is a _ regular broomlike combing _ with respect to @xmath74 for @xmath39 if @xmath530 is a regular language and a broomlike combing with respect to @xmath74 for @xmath75 the monoid @xmath529 has a natural tree structure , and hence @xmath530 inherits a tree structure . it is not hard to show that @xmath530 with the bijection @xmath566 is a uniform spanning tree . another result in the same spirit , can be found in @xcite , where gilman , hermiller , holt and rees proved the following : a group @xmath0 is virtually free , if and only if there exists a finite generating set @xmath74 and a constant @xmath567 such that every @xmath86-locally geodesic sequence in @xmath0 is geodesic . we remark that this result does not hold in general for every finite generating set . let @xmath538}$ ] be a sequence in @xmath568 it is is geodesic if @xmath569}|=n.$ ] the sequence is a @xmath86-locally geodesic sequence , if any subsequence of length @xmath86 is geodesic . again this characterization can be partially connected to the present work . using the ideas of ( * proposition 1 ) , it is not hard to show that if every @xmath86-locally geodesic sequence is geodesic , then every @xmath199-sequence is @xmath570-triangulable . in this section we give an elementary introduction to groups of asymptotic dimension 1 . a general reference about the topic is ( * ? ? ? * chapter 9 ) . the concept of asymptotic dimension was introduced by gromov in @xcite . he proposed two different definitions of what asymptotic dimension should be , which turn out to be equivalent . see ( * ? ? ? * theorem 9.9 ) . we will use the following . [ d : adim ] a metric space @xmath346 has _ asymptotic dimension _ at most @xmath332 if for every @xmath571 there exists a partition @xmath572 of @xmath74 into sets of uniformly bounded diameter such that the @xmath86-graph of @xmath572 is @xmath573-colorable , where the @xmath86-graph of @xmath572 has vertex set @xmath572 and two elements of @xmath572 are joined by an edge if they are at distance at most @xmath86 . the asymptotic dimension is invariant under coarse equivalence . hence , the asymptotic dimension of a finitely generable group is just the asymptotic dimension of its cayley graph with respect to a finite generating set . [ o : defadim ] the cayley graph @xmath2 of a finitely generated group @xmath0 has asymptotic dimension at most @xmath332 if for every @xmath463 there exist two functions @xmath577 such that @xmath578 @xmath579 is finite and there is a function @xmath580 $ ] such that for @xmath581 if @xmath582 then @xmath583 if @xmath604 then @xmath605 if @xmath606 then @xmath41 and @xmath607 lie in different connected components of @xmath608 and hence any path joining @xmath41 and @xmath607 must go through @xmath609 and hence @xmath610 as @xmath0 is finitely generable , so is @xmath140 . let @xmath13 be a finite generating set of @xmath140 and let @xmath10 be a finite subset of @xmath615 such that @xmath617 generates @xmath0 . the inclusion @xmath618 is an isometric embedding , and then @xmath619 let @xmath623 and @xmath624 a geodesic expression for @xmath625 has the form @xmath626 where @xmath627 are elements of @xmath628 and @xmath629 are words in @xmath630 as @xmath631 hence @xmath632 . thus @xmath633 fix @xmath463 . we are going to show that @xmath638 . by observation [ o : defadim ] , there exist @xmath639 such that @xmath640 for all @xmath641 @xmath642 is finite and there is a @xmath643-coloring of @xmath644 such that for @xmath645 if @xmath646 then @xmath647 let @xmath648 in particular @xmath183 is a finite subset of @xmath615 , and hence @xmath649 is a finite subgroup of @xmath615 . let @xmath650 be a transversal for the right @xmath228-multiplication action in @xmath651 thus the functions @xmath666 @xmath667 and @xmath668 are well defined and @xmath669 for all @xmath81 , @xmath670 is finite . let @xmath671 we have to show that for @xmath672 if @xmath581 and @xmath582 then @xmath583 let @xmath681 denote the cyclic group of order @xmath132 . the _ lamplighter group _ is the group @xmath682 the standard presentation of the lamplighter group arises from the wreath product structure @xmath683 , m , n \in \mathbb{z } \rangle.\ ] ] a _ diagram _ over @xmath2 for @xmath54 is a pair @xmath685 where @xmath686 is a compact , planar , simply connected 2-complex with a fixed embedding in @xmath687 and @xmath163 is a graph map from the underlying graph of @xmath686 to @xmath2 such that the boundary of @xmath686 maps to @xmath688 let @xmath689 be a presentation and @xmath2 be the cayley graph of @xmath690 with respect to @xmath74 . let @xmath691}$ ] be a sequence in @xmath509 such that @xmath692}=1.$ ] let @xmath559 be the closed path in @xmath2 given by @xmath693 a _ van kampen diagram _ over @xmath179 for @xmath694 is a diagram @xmath685 over @xmath2 for @xmath559 such that for each inner face @xmath695 of @xmath696 there exists a starting vertex and a direction for the closed path bounding @xmath697 say @xmath698},f_{[1\uparrow m]})$ ] , that satisfies that @xmath699 is a freely reduced word of @xmath95 . let @xmath689 be a presentation where each @xmath700 is a freely reduced word . let @xmath691}$ ] be a freely reduced word over @xmath74 with @xmath692}=1 $ ] in @xmath701 then there exists a van kampen diagram over @xmath179 for @xmath693 [ obs:1seqtopathseq ] let @xmath56}$ ] be an 1-sequence in @xmath5 . if @xmath56}$ ] is not a path sequence @xmath2 , then there exists @xmath190 $ ] such that @xmath702 it follows that @xmath703 and , we can perform a @xmath200-reduction . hence there exists a sequence of @xmath200-reductions of @xmath56}$ ] that produces a path sequence . the main result of this section is that @xmath0 is virtually free if and only if @xmath704 and @xmath0 is finitely presentable . this result has been proved independently in @xcite , @xcite and @xcite . the proof given here is more elementary , and it relies on the results of the previous section . let @xmath709 and let @xmath577 such that @xmath669 for all @xmath81 , @xmath579 is finite and there is a @xmath710-coloring for @xmath711 such that for @xmath672 if @xmath581 and @xmath582 then @xmath712 . let @xmath350 be the graph with vertex set @xmath711 and an edge connecting two vertices @xmath713 if and only if @xmath714 and @xmath715 it is straightforward to check that @xmath220 induces an ( unoriented ) graph quotient @xmath716 denoted again by @xmath717 let @xmath718}$ ] be a @xmath199-sequence in @xmath7 if we show that the 1-sequence in @xmath719 @xmath720})$ ] is @xmath199-triangulable , then it follows that the sequence @xmath718}$ ] is @xmath707-triangulable with @xmath721 by observation [ obs:1seqtopathseq ] , we can assume that @xmath718}$ ] is a path sequence in @xmath67 let @xmath684},e_{[1\uparrow n]})$ ] be a reduced closed path in @xmath2 . we may assume @xmath722 . by the van kampen lemma , there exists a van kampen diagram @xmath723 over @xmath708 for @xmath688 hence @xmath724 is a diagram over @xmath350 for @xmath725 as each inner face @xmath695 of @xmath726 is labeled by some @xmath727 @xmath728 we subdivide each inner face of @xmath726 into faces with at most 3 edges , following a triangulation of each @xmath729 denote this new complex by @xmath686 . we will show that there exists a triangular diagram @xmath742 over @xmath350 for @xmath743 with no inner vertex . note that if such diagram exists , then @xmath743 is @xmath199-triangulable in @xmath350 and hence @xmath54 is @xmath707-triangulable in @xmath67 * construction 1 : * if there exists an edge @xmath751 such that @xmath752 and @xmath745 is a weak edge , then @xmath753 is a vertex and @xmath163 is compatible with collapsing @xmath745 to a vertex in @xmath686 . collapsing the edge , we obtain a new triangular diagram @xmath742 over @xmath350 for @xmath743 with fewer inner vertices than @xmath750 since all the faces adjacent to @xmath31 have at most 3 edges , and if 2 edges of the face are strong , the other one should be weak , this implies that @xmath220 is constant on all the vertices of @xmath754 different from @xmath31 . then we can modify @xmath163 to be constant in @xmath754 as follows . let @xmath755 @xmath756 then we let @xmath757 and @xmath758 @xmath759 and @xmath760 @xmath761 the theorem [ t : adim1 ] can be reformulated for graphs as follows : let @xmath2 be a graph with @xmath764 and @xmath765 is generated by closed paths of length @xmath766 , then there exists @xmath357 such that every 1-sequence in @xmath2 is @xmath84-triangulable . * corollary 5 ) let @xmath767 be a short exact sequence with @xmath615 locally finite , @xmath0 finitely generable and @xmath140 free . then @xmath0 is finitely presentable if and only if @xmath615 is finite . if @xmath0 is finitely presentable then , by theorem [ t : adim1 ] , @xmath0 is virtually free . hence , there exists a short exact sequence @xmath769 with @xmath770 free and @xmath771 finite . as @xmath772 the restriction of @xmath773 to @xmath615 has trivial kernel , and hence @xmath615 is finite .

this is a survey on the state - of - the - art of the classification of finite - dimensional complex hopf algebras . this general question is addressed through the consideration of different classes of such hopf algebras . pointed hopf algebras constitute the class best understood ; the classification of those with abelian group is expected to be completed soon and there is substantial progress in the non - abelian case . _ dedicado a biblioco 34 _ 16t05 , 16t20 , 17b37 , 16t25 , 20g42 . hopf algebras , quantum groups , nichols algebras .

1403.7838

hopf algebras were introduced in the 1950 s from three different perspectives : algebraic groups in positive characteristic , cohomology rings of lie groups , and group objects in the category of von neumann algebras . the study of non - commutative non - cocommutative hopf algebras started in the 1960 s . the fundamental breakthrough is drinfeld s report @xcite . among many contributions and ideas , a systematic construction of solutions of the quantum yang - baxter equation ( qybe ) was presented . let @xmath0 be a vector space . the qybe is equivalent to the braid equation : @xmath1 if @xmath2 satisfies , then @xmath3 is called a braided vector space ; this is a down - to - the - earth version of a braided tensor category @xcite . drinfeld introduced the notion of quasi - triangular hopf algebra , meaning a pair @xmath4 where @xmath5 is a hopf algebra and @xmath6 is invertible and satisfies the approppriate conditions , so that every @xmath5-module @xmath0 becomes a braided vector space , with @xmath2 given by the action of @xmath7 composed with the usual flip . furthermore , every finite - dimensional hopf algebra @xmath5 gives rise to a quasi - triangular hopf algebra , namely the drinfeld double @xmath8 as vector space . if @xmath5 is not finite - dimensional , some precautions have to be taken to construct @xmath9 , or else one considers yetter - drinfeld modules , see [ subsec : repns ] . in conclusion , every hopf algebra is a source of solutions of the braid equation . essential examples of quasi - triangular hopf algebras are the quantum groups @xmath10 @xcite and the finite - dimensional variations @xmath11 @xcite . in the approach to the classification of hopf algebras exposed in this report , braided vector spaces and braided tensor categories play a decisive role ; and the finite quantum groups are the main actors in one of the classes that splits off . by space limitations , there is a selection of the topics and references included . particularly , we deal with finite - dimensional hopf algebras over an algebraically closed field of characteristic zero with special emphasis on description of examples and classifications . interesting results on hopf algebras either infinite - dimensional , or over other fields , unfortunately can not be reported . there is no account of the many deep results on tensor categories , see @xcite . various basic fundamental results are not explicitly cited , we refer to @xcite for them ; classifications of hopf algebras of fixed dimensions are not evoked , see @xcite . i thank all my coauthors for pleasant and fruitful collaborations . i am particularly in debt with hans - jrgen schneider , matas graa and ivn angiono for sharing their ideas on pointed hopf algebras with me . let @xmath12 and @xmath13 . the base field is @xmath14 . if @xmath15 is a set , then @xmath16 is its cardinal and @xmath17 is the vector space with basis @xmath18 . let @xmath19 be a group : we denote by @xmath20 the set of isomorphism classes of irreducible representations of @xmath19 and by @xmath21 the subset of those of dimension 1 ; by @xmath22 the centralizer of @xmath23 ; and by @xmath24 its conjugacy class . more generally we denote by @xmath25 the set of isomorphism classes of simple objects in an abelian category @xmath26 . the group of @xmath27-th roots of 1 in @xmath14 is denoted @xmath28 ; also @xmath29 . the group presented by @xmath30 with relations @xmath31 is denoted @xmath32 . the notation for hopf algebras is standard : @xmath33 , @xmath34 , @xmath35 , denote respectively the comultiplication , the counit , the antipode ( always assumed bijective , what happens in the finite - dimensional case ) . we use sweedler s notation : @xmath36 . similarly , if @xmath37 is a coalgebra and @xmath0 is a left comodule with structure map @xmath38 , then @xmath39 . if @xmath40 are subspaces of @xmath37 , then @xmath41 ; also @xmath42 and @xmath43 for @xmath44 . the first examples of finite - dimensional hopf algebras are the group algebra @xmath45 of a finite group @xmath19 and its dual , the algebra of functions @xmath46 . indeed , the dual of a finite - dimensional hopf algebra is again a hopf algebra by transposing operations . by analogy with groups , several authors explored the notion of extension of hopf algebras at various levels of generality ; in the finite - dimensional context , every extension @xmath47 can be described as @xmath37 with underlying vector space @xmath48 , via a heavy machinery of actions , coactions and non - abelian cocycles , but actual examples are rarely found in this way ( extensions from a different perspective are in @xcite ) . relevant exceptions are the so - called _ abelian extensions _ @xcite ( rediscovered by takeuchi and majid ) : here the input is a matched pair of groups @xmath49 with mutual actions @xmath50 , @xmath51 ( or equivalently , an exact factorization of a finite group ) . the actions give rise to a hopf algebra @xmath52 . the multiplication and comultiplication can be further modified by compatible cocycles @xmath53 , producing to the abelian extension @xmath54 . here @xmath53 turns out to be a 2-cocycle in the total complex associated to a double complex built from the matched pair ; the relevant @xmath55 is computed via the so - called kac exact sequence . it is natural to approach hopf algebras by considering algebra or coalgebra invariants . there is no preference in the finite - dimensional setting but coalgebras and comodules are locally finite , so we privilege the coalgebra ones to lay down general methods . the basic coalgebra invariants of a hopf algebra @xmath5 are : @xmath56 the group @xmath57 of group - like elements of @xmath5 . @xmath56 the space of skew - primitive elements @xmath58 , @xmath59 ; @xmath60 . @xmath56 the coradical @xmath61 , that is the sum of all simple subcoalgebras . @xmath56 the coradical filtration @xmath62 , where @xmath63 ; then @xmath64 . the category @xmath65 of left modules over a hopf algebra @xmath5 is monoidal with tensor product defined by the comultiplication ; ditto for the category @xmath66 of left comodules , with tensor product defined by the multiplication . here are two ways to deform hopf algebras without altering one of these categories . @xmath67 let @xmath68 be invertible such that @xmath69 and @xmath70 . then @xmath71 ( the same algebra with comultiplication @xmath72 ) is again a hopf algebra , named the _ twisting _ of @xmath5 by @xmath73 @xcite . the monoidal categories @xmath65 and @xmath74 are equivalent . if @xmath5 and @xmath75 are finite - dimensional hopf algebras with @xmath65 and @xmath76 equivalent as monoidal categories , then there exists @xmath73 with @xmath77 as hopf algebras ( schauenburg , etingof - gelaki ) . examples of twistings not mentioned elsewhere in this report are in @xcite . @xmath67 given a linear map @xmath78 with analogous conditions , there is a hopf algebra @xmath79 ( same coalgebra , multiplication twisted by @xmath80 ) such that the monoidal categories @xmath66 and @xmath81 are equivalent @xcite . a yetter - drinfeld module @xmath82 over @xmath5 is left @xmath5-module and left @xmath5-comodule with the compatibility @xmath83 , for all @xmath84 and @xmath85 . the category @xmath86 of yetter - drinfeld modules is braided monoidal.[page : yd ] that is , for every @xmath87 , there is a natural isomorphism @xmath88 given by @xmath89 , @xmath84 , @xmath90 . when @xmath5 is finite - dimensional , the category @xmath86 is equivalent , as a braided monoidal category , to @xmath91 . the definition of hopf algebra makes sense in any braided monoidal category . hopf algebras in @xmath86 are interesting because of the following facts discovered by radford and interpreted categorically by majid , see @xcite : @xmath92 if @xmath7 is a hopf algebra in @xmath86 , then @xmath93 with semidirect product and coproduct is a hopf algebra , named the _ bosonization _ of @xmath7 by @xmath5 . @xmath92 let @xmath94 be hopf algebra maps as in @xmath95 & \ar@/^-1pc/_{\iota}[l ] h}$ ] with @xmath96 . then @xmath97 is a hopf algebra in @xmath86 and @xmath98 . for instance , if @xmath99 , then the tensor algebra @xmath100 is a hopf algebra in @xmath86 , by requiring @xmath101 . if @xmath102 satisfies @xmath103 , @xmath104 the usual flip , then the exterior algebra @xmath105 is a hopf algebra in @xmath86 . there is a braided adjoint action of a hopf algebra @xmath7 in @xmath86 on itself , see e.g. ( * ? ? ? * ( 1.26 ) ) . if @xmath106 and @xmath107 , then @xmath108 . a quasitriangular hopf algebra @xmath109 is _ triangular _ if the braiding induced by @xmath7 is a symmetry : @xmath110 for all @xmath111 . a finite - dimensional triangular hopf algebra is a twisting of a bosonization @xmath112 , where @xmath19 is a finite group and @xmath113 has @xmath114 @xcite . this lead eventually to the classification of triangular finite - dimensional hopf algebras @xcite ; previous work on the semisimple case culminated in @xcite . the algebra of functions @xmath46 on a finite group @xmath19 admits a haar measure , i.e. , a linear function @xmath115 invariant under left and right translations , namely @xmath116 sum of all elements in the standard basis of @xmath45 . this is adapted as follows : a right integral _ on _ a hopf algebra @xmath5 is a linear function @xmath117 which is invariant under the left regular coaction : analogously there is the notion of left integral . the notion has various applications . assume that @xmath5 is finite - dimensional . then an integral _ in _ @xmath5 is an integral on @xmath118 ; the subspace of left integrals in @xmath5 has dimension one , and there is a generalization of maschke s theorem for finite groups : @xmath5 is semisimple if and only if @xmath119 for any integral @xmath120 . this characterization of semisimple hopf algebras , valid in any characteristic , is one of several , some valid only in characteristic 0 . see @xcite . semisimple hopf algebras can be obtained as follows : @xmath92 a finite - dimensional hopf algebra @xmath5 is semisimple if and only if it is cosemisimple ( that is , @xmath118 is semisimple ) . @xmath92 given an extension @xmath121 , @xmath5 is semisimple iff @xmath75 and @xmath122 are . notice that there are semisimple extensions that are not abelian @xcite . @xmath92 if @xmath5 is semisimple , then so are @xmath71 and @xmath79 , for any twist @xmath73 and cocycle @xmath80 . if @xmath19 is a finite simple group , then any twisting of @xmath45 is a simple hopf algebra ( i.e. , not a non - trivial extension ) @xcite , but the converse is not true @xcite . @xmath92 a bosonization @xmath123 is semisimple iff @xmath7 and @xmath5 are . to my knowledge , all examples of semisimple hopf algebras arise from group algebras by the preceding constructions ; this was proved in @xcite in low dimensions and in @xcite for dimensions @xmath124 , @xmath125 , where @xmath126 , @xmath127 and @xmath128 are primes . see ( * ? ? ? * question 2.6 ) . an analogous question in terms of fusion categories : is any semisimple hopf algebra weakly group - theoretical ? see ( * ? ? ? * question 2 ) . there are only finitely many isomorphism classes of _ semisimple _ hopf algebras in each dimension @xcite , but this fails in general @xcite . ( kaplansky ) . let @xmath5 be a semisimple hopf algebra . the dimension of every @xmath129 divides the dimension of @xmath5 . the answer is affirmative for iterated extensions of group algebras and duals of group algebras @xcite and notably for semisimple quasitriangular hopf algebras @xcite . nichols algebras are a special kind of hopf algebras in braided tensor categories . we are mainly interested in nichols algebras in the braided category @xmath86 , where @xmath5 is a hopf algebra , see page . in fact , there is a functor @xmath130 from @xmath86 to the category of hopf algebras in @xmath86 . their first appearence is in the precursor @xcite ; they were rediscovered in @xcite as part of a quantum differential calculus " , and in @xcite to present the positive part of @xmath10 . see also @xcite . there are several , unrelated at the first glance , alternative definitions . let @xmath99 . the first definition uses the representation of the braid group @xmath131 in @xmath27 strands on @xmath132 , given by @xmath133 , @xmath2 in @xmath134 tensorands ; here recall that @xmath135 . let @xmath136 be the matsumoto section and let @xmath137 be the quantum symmetrizer , @xmath138 . then define @xmath139 hence @xmath140 is a graded hopf algebra in @xmath86 with @xmath141 , @xmath142 ; by the algebra structure depends only on @xmath2 . to explain the second definition , let us observe that the tensor algebra @xmath100 is a hopf algebra in @xmath86 with comultiplication determined by @xmath143 for @xmath144 . then @xmath145 coincides with the largest homogeneous ideal of @xmath100 generated by elements of degree @xmath146 that is also a coideal . let now @xmath147 be a graded hopf algebra in @xmath86 with @xmath148 . consider the conditions @xmath149 these requirements are dual to each other : if @xmath150 has finite - dimensional homogeneous components and @xmath151 is the graded dual of @xmath150 , i.e. , @xmath152 , then @xmath150 satisfies if and only if @xmath7 satisfies . these conditions determine @xmath153 up to isomorphisms , as the unique graded connected hopf algebra @xmath150 in @xmath86 that satisfies @xmath154 , and . there are still other characterizations of @xmath145 , e.g. as the radical of a suitable homogeneous bilinear form on @xmath100 , or as the common kernel of some suitable skew - derivations . see @xcite for more details . despite all these different definitions , nichols algebras are extremely difficult to deal with , e.g. to present by generators and relations , or to determine when a nichols algebra has finite dimension or finite gelfand - kirillov dimension . it is not even known a priori whether the ideal @xmath145 is finitely generated , except in a few specific cases . for instance , if @xmath2 is a symmetry , that is @xmath155 , or satisfies a hecke condition with generic parameter , then @xmath153 is quadratic . by the efforts of various authors , we have some understanding of finite - dimensional nichols algebras of braided vector spaces either of diagonal or of rack type , see [ subsec : nichols - diagonal ] , [ subsec : nichols - rack ] . we now explain how nichols algebras enter into our approach to the classification of hopf algebras . recall that a hopf algebra has the _ dual chevalley property _ if the tensor product of two simple comodules is semisimple , or equivalently if its coradical is a ( cosemisimple ) hopf subalgebra . for instance , a _ pointed _ hopf algebra , one whose simple comodules have all dimension one , has the dual chevalley property and its coradical is a group algebra . also , a _ copointed _ hopf algebra ( one whose coradical is the algebra of functions on a finite group ) has the dual chevalley property . the lifting method is formulated in this context @xcite . let @xmath5 be a hopf algebra with the dual chevalley property and set @xmath156 . under this assumption , the graded coalgebra @xmath157 associated to the coradical filtration becomes a hopf algebra and considering the homogeneous projection @xmath158 as in @xmath159 & { \operatorname{gr}}h \ar@<-0.5ex>@{->}_{\quad \pi}[r ] & \ar@<-0.5ex>[l ] k } $ ] we see that @xmath160 . the subalgebra of coinvariants @xmath7 is a graded hopf algebra in @xmath161 that inherits the grading with @xmath162 ; it satisfies since the grading comes from the coradical filtration . let @xmath163 be the subalgebra of @xmath7 generated by @xmath164 ; then @xmath165 . the braided vector space @xmath0 is a basic invariant of @xmath5 called its _ infinitesimal braiding_. let us fix then a semisimple hopf algebra @xmath75 . to classify all finite - dimensional hopf algebras @xmath5 with @xmath166 as hopf algebras , we have to address the following questions . 1 . [ item : method - nichols ] determine those @xmath167 such that @xmath168 is finite - dimensional , and give an efficient defining set of relations of these . [ item : method - gen - deg - one ] investigate whether any _ finite - dimensional _ graded hopf algebra @xmath7 in @xmath161 satisfying @xmath162 and @xmath169 , is a nichols algebra . [ item : method - lifting ] compute all hopf algebras @xmath5 such that @xmath170 , @xmath0 as in [ item : method - nichols ] . since the nichols algebra @xmath153 depends as an algebra ( and as a coalgebra ) only on the braiding @xmath2 , it is convenient to restate question [ item : method - nichols ] as follows : 1 . determine those braided vector spaces @xmath171 in a suitable class such that @xmath172 , and give an efficient defining set of relations of these . 2 . for those @xmath0 as in ( a@xmath173 ) , find in how many ways , if any , they can be realized as yetter - drinfeld modules over @xmath75 . for instance , if @xmath174 , @xmath175 a finite abelian group , then the suitable class is that of braided vector spaces of diagonal type . in this context , question ( a@xmath176 ) amounts to solve systems of equations in @xmath175 . the answer to [ item : method - nichols ] is instrumental to attack [ item : method - gen - deg - one ] and [ item : method - lifting ] . question [ item : method - gen - deg - one ] can be rephrased in two equivalent statements : 1 . investigate whether any _ finite - dimensional _ graded hopf algebra @xmath150 in @xmath161 with @xmath148 and generated as algebra by @xmath177 , is a nichols algebra . 2 . investigate whether any _ finite - dimensional _ hopf algebra @xmath5 with @xmath178 is generated as algebra by @xmath179 . we believe that the answer to [ item : method - gen - deg - one ] is affirmative at least when @xmath75 is a group algebra . in other words , by the reformulation ( b@xmath176 ) : [ conj : gendegone ] _ @xcite _ every _ finite - dimensional _ pointed hopf algebra is generated by group - like and skew - primitive elements . as we shall see in [ subsec : gen - deg-1 ] , the complete answer to [ item : method - nichols ] is needed in the approach proposed in @xcite to attack conjecture [ conj : gendegone ] . it is plausible that the answer of ( b@xmath176 ) is affirmative for every semisimple hopf algebra @xmath75 . question [ item : method - lifting ] , known as _ lifting _ of the relations , also requires the knowledge of the generators of @xmath145 , see [ subsec : deformations ] . before starting with the analysis of the various questions in [ subsec : chevalley ] , we discuss a possible approach to more general hopf algebras @xcite . let @xmath5 be a hopf algebra ; we consider the following invariants of @xmath5 : @xmath56 the _ hopf coradical _ @xmath180}$ ] is the subalgebra generated by @xmath61 . @xmath56 the _ standard filtration _ @xmath180 } \subset h_{[1 ] } \subset \dots$ ] , @xmath181}=\wedge^{n+1 } h_{[0]}$ ] ; then @xmath182}$ ] . if @xmath5 has the dual chevalley property , then @xmath181 } = h_n$ ] for all @xmath183 . in general , @xmath180}$ ] is a hopf subalgebra of @xmath5 with coradical @xmath61 and we may consider the graded hopf algebra @xmath184}/h_{[n-1]}$ ] . as before , if @xmath185}$ ] is the homogeneous projection , then @xmath186 is a hopf algebra in @xmath187}}^{h_{[0]}}\mathcal{yd}}$ ] and @xmath188}$ ] . furthermore , @xmath189 with grading inherited from @xmath190 . this discussion raises the following questions . 1 . [ item : gen - method - coalgebras ] let @xmath37 be a finite - dimensional cosemisimple coalgebra and @xmath191 a bijective anti - coalgebra map . classify all finite - dimensional hopf algebras @xmath122 generated by @xmath37 , such that @xmath192 . [ item : gen - method - r ] given @xmath122 as in the previous item , classify all finite - dimensional connected graded hopf algebras @xmath7 in @xmath193 . [ item : gen - method - lifting ] given @xmath122 and @xmath7 as in previous items , classify all deformations or liftings , that is , classify all hopf algebras @xmath5 such that @xmath194 . question [ item : gen - method - coalgebras ] is largely open , except for the remarkable ( * ? ? ? * theorem 1.5 ) : if @xmath5 is a hopf algebra generated by an @xmath35-invariant 4-dimensional simple subcoalgebra @xmath37 , such that @xmath195 , then @xmath5 is a hopf algebra quotient of the quantized algebra of functions on @xmath196 at a root of unity @xmath197 . nichols algebras enter into the picture in question [ item : gen - method - r ] ; if @xmath198 , then @xmath153 is a subquotient of @xmath7 . question [ item : gen - method - lifting ] is completely open , as it depends on the previous questions . here we expose two important notions introduced in @xcite . let @xmath199 and @xmath200 . a _ basic datum _ of type @xmath201 is a pair @xmath202 , where @xmath203 is a set and @xmath204 is a map such that @xmath205 for all @xmath206 . let @xmath207 be the quiver @xmath208 over @xmath209 , with @xmath210 , @xmath211 ( here @xmath212 means target , @xmath213 means source ) . let @xmath214 be the free groupoid over @xmath207 ; in any quotient of @xmath214 , we denote @xmath215 i.e. , the implicit superscripts are those allowing compositions . coxeter datum _ is a triple @xmath216 , where @xmath202 is a basic datum of type @xmath201 and @xmath217 is a family of coxeter matrices @xmath218 with @xmath219 the _ coxeter groupoid _ @xmath220 associated to @xmath216 ( * ? ? ? * definition 1 ) is the groupoid presented by generators @xmath207 with relations @xmath221 a generalized root system ( grs for short ) is a collection @xmath222 , where @xmath223 is a family of generalized cartan matrices @xmath224 , cf . @xcite , and @xmath225 is a family of subsets @xmath226 . we need the following notation : let @xmath227 be the canonical basis of @xmath228 and define @xmath229 by @xmath230 , @xmath231 , @xmath232 . the collection should satisfy the following axioms : @xmath233 for all @xmath232 , @xmath234 . we call @xmath235 , respectively @xmath236 , the set of _ positive _ , respectively _ negative _ , roots . let @xmath237 , @xmath238 , @xmath239 , @xmath232 , and @xmath240 the subgroupoid of @xmath241 generated by all the @xmath242 , i.e. , by the image of the morphism of quivers @xmath243 , @xmath244 . there is a coxeter matrix @xmath218 , where @xmath245 is the smallest natural number such that @xmath246 . then @xmath217 fits into a coxeter datum @xmath216 , and there is an isomorphism of groupoids @xmath247 & \cw = \cw(\cx , \rho , \cc ) } $ ] @xcite ; this is called the _ weyl groupoid _ of @xmath248 . if @xmath249 , then @xmath250 , by . the sets of _ real _ roots at @xmath232 are @xmath251 ; correspondingly the _ imaginary _ roots are @xmath252 . assume that @xmath253 is connected . then the following conditions are equivalent ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * lemma 2.11 ) : * @xmath254 is finite for some @xmath255 , * @xmath254 is finite for all @xmath255 , * @xmath256 is finite for all @xmath255 , * @xmath253 is finite . if these hold , then all roots are real @xcite ; we say that @xmath248 is _ finite_. we now discuss two examples of grs , central for the subsequent discussion . [ exa : grs - lie - super ] @xcite let @xmath257 be a field of characteristic @xmath258 , @xmath259 , @xmath260 and @xmath261 . we assume @xmath262 for simplicity . let @xmath263 . let @xmath264 be the kac - moody lie superalgebra over @xmath257 defined as in @xcite ; it is generated by @xmath265 , @xmath266 and @xmath267 , @xmath206 , and the parity is given by @xmath268 , @xmath269 , @xmath270 , @xmath271 . let @xmath272 be the root system of @xmath264 . we make the following technical assumptions : @xmath273 the matrix @xmath274 is _ admissible _ if holds @xcite . let @xmath275 be given by @xmath276 we need the following elements of @xmath257 : @xmath277 with the help of these scalars , we define a reflection @xmath278 , where @xmath279 , with @xmath280 , and @xmath281 , with @xmath282 [ thm : isomorfismo ti ] there is an isomorphism @xmath283 of lie superalgebras given ( for an approppriate basis @xmath284 of @xmath265 ) by @xmath285 assume that @xmath286 ; then and hold . let @xmath287 then @xmath288 , where @xmath289 and @xmath290 , is a finite grs , an invariant of @xmath264 . [ exa : grs - nichols - ss - yd ] let @xmath5 be a hopf algebra , assumed semisimple for easiness . let @xmath291 be finite - dimensional , with a fixed decomposition @xmath292 , where @xmath293 . then @xmath294 and @xmath295 are @xmath296-graded , by @xmath297 for all @xmath298 , @xmath299 . recall that @xmath300 . [ th : grs - semisimple ] _ @xcite _ if @xmath301 , then @xmath82 has a finite grs . we discuss the main ideas of the proof . let @xmath302 . we define @xmath303 , @xmath304 then @xmath305 and @xmath306 is a generalized cartan matrix @xcite . also , @xmath307 is irreducible @xcite , @xcite . let @xmath209 be the set of objects in @xmath86 with fixed decomposition ( up to isomorphism ) of the form @xmath308 then @xmath309 , where @xmath310 , is a crystallographic datum . next we need : @xmath67 ( * ? ? ? * theorem 4.5 ) ; @xcite there exists a totally ordered index set @xmath311 and families @xmath312 in @xmath313 , @xmath314 such that @xmath315 as @xmath296-graded objects in @xmath86 , where @xmath316 for all @xmath317 , @xmath318 . let @xmath319 , @xmath320 , @xmath321 . then @xmath322 is a finite grs . [ th : classif - grs ] _ @xcite _ the classification of all finite grs is known . the proof is a combinatorial tour - de - force and requires computer calculations . it is possible to recover from this result the classification of the finite - dimensional contragredient lie superalgebras in arbitrary characteristic @xcite . however , the list of @xcite is substantially larger than the classifications of the alluded lie superalgebras or the braidings of diagonal type with finite - dimensional nichols algebra . let @xmath19 be a finite group . we denote @xmath323 for @xmath324 . so @xmath325 is a left @xmath19-module with a @xmath19-grading @xmath326 such that @xmath327 , for all @xmath328 . if @xmath329 , then the braiding @xmath88 is given by @xmath330 , @xmath331 , @xmath90 , @xmath332 now assume that @xmath333 is a finite abelian group . then every @xmath334 is a @xmath175-graded @xmath175-module , hence of the form @xmath335 , where @xmath336 is the @xmath337-isotypic component of @xmath338 . so @xmath339 is just the category of @xmath340-graded modules , with the braiding @xmath88 given by @xmath341 , @xmath342 , @xmath343 , @xmath344 , @xmath345 . let @xmath346 , @xmath200 . [ def : diagonal ] let @xmath347 be a matrix with entries in @xmath348 . a braided vector space @xmath3 is of _ diagonal type _ with matrix @xmath349 if @xmath0 has a basis @xmath350 with @xmath351 thus , every finite - dimensional @xmath352 is a braided vector space of diagonal type . question [ item : method - nichols ] , more precisely ( a@xmath173 ) , has a complete answer in this setting . first we can assume that @xmath353 for @xmath206 , as otherwise @xmath354 . also , let @xmath355 and @xmath356 a braided vector space with matrix @xmath357 . if @xmath358 and @xmath359 for all @xmath360 , then @xmath361 as braided vector spaces . [ th : classif - diagonal ] _ @xcite _ the classification of all braided vector spaces of diagonal type with finite - dimensional nichols algebra is known . the proof relies on the weyl groupoid introduced in @xcite , a particular case of theorem [ th : grs - semisimple ] . another fundamental ingredient is the following result , generalized at various levels in @xcite . [ th : pbw ] _ @xcite _ let @xmath0 be a braided vector space of diagonal type . every hopf algebra quotient of @xmath100 has a pbw basis . the classification in theorem [ th : classif - diagonal ] can be organized as follows : @xmath92 for most of the matrices @xmath362 in the list of @xcite there is a field @xmath257 and a pair @xmath363 as in example [ exa : grs - lie - super ] such that @xmath364 , and @xmath365 has the same grs as the nichols algebra corresponding to @xmath349 @xcite . @xmath92 besides these , there are 12 ( yet ) unidentified examples . we believe that theorem [ th : classif - diagonal ] can be proved from theorem [ th : classif - grs ] , via example [ exa : grs - lie - super ] . [ th : presentation - diagonal ] _ @xcite _ an efficient set of defining relations of each finite - dimensional nichols algebra of a braided vector space of diagonal type is known . the proof uses most technical tools available in the theory of nichols algebras ; of interest in its own is the introduction of the notion of convex order in weyl groupoids . as for other classifications above , it is not possible to state precisely the list of relations . we just mention different types of relations that appear . * quantum serre relations , i.e. , @xmath366 for suitable @xmath367 . * powers of root vectors , i.e. , @xmath368 , where the @xmath369 s are part of the pbw basis . * more exotic relations ; they involve 2 , 3 , or at most 4 @xmath370 s in @xmath201 . we now consider nichols algebras of objects in @xmath371 , where @xmath19 is a finite not necessarily abelian group . the category @xmath371 is semisimple and the simple objects are parametrized by pairs @xmath372 , where @xmath373 is a conjugacy class in @xmath19 and @xmath374 , for a fixed @xmath375 ; the corresponding simple yetter - drinfeld module @xmath376 is @xmath377 as a module . the braiding @xmath2 is described in terms of the conjugation in @xmath373 . to describe the related suitable class , we recall that a rack is a set @xmath378 with a map @xmath379 satisfying * @xmath380 is a bijection for every @xmath381 . * @xmath382 for all @xmath383 ( self - distributivity ) . for instance , a conjugacy class @xmath373 in @xmath19 with the operation @xmath384 , @xmath385 is a rack ; actually we only consider racks realizable as conjugacy classes . let @xmath15 be a rack and @xmath386 a decomposition of @xmath15 , i.e. , a disjoint family of subracks with @xmath387 for all @xmath388 . [ def : braiding - rack]@xcite a _ 2-cocycle of degree @xmath389 _ , associated to @xmath390 , is a family @xmath391 of maps @xmath392 such that @xmath393 given such @xmath349 , let @xmath394 and let @xmath395 be given by @xmath396 then @xmath397 is a braided vector space called of _ rack type _ ; its nichols algebra is denoted @xmath398 . if @xmath399 , then we say that @xmath349 is _ principal_. every finite - dimensional @xmath113 is a braided vector space of rack type ( * ? ? ? * theorem 4.14 ) . question ( a@xmath173 ) in this setting has partial answers in three different lines : computation of some finite - dimensional nichols algebras , nichols algebras of reducible yetter - drinfeld modules and collapsing of racks . the algorithm to compute a nichols algebra @xmath153 is as follows : compute the space @xmath400 of relations of degree @xmath370 , for @xmath401 ; then compute the @xmath402-th partial nichols algebra @xmath403 , say with a computer program . if lucky enough to get @xmath404 , then check whether it is a nichols algebra , e.g. via skew - derivations ; otherwise go to @xmath405 . the description of @xmath406 is not difficult @xcite but for higher degrees it turns out to be very complicated . we list all known examples of finite - dimensional nichols algebras @xmath398 with @xmath15 indecomposable and @xmath349 principal and abelian ( @xmath407 ) . [ exa : nichols - sim ] let @xmath408 be the conjugacy class of @xmath409-cycles in @xmath410 , @xmath411 . we start with the rack of transpositions in @xmath410 and the cocycles @xmath412 , @xmath337 that arise from the @xmath413 with @xmath414 , see ( * ? ? ? * ( 5.5 ) , ( 5.9 ) ) . let @xmath0 be a vector space with basis @xmath415 and consider the relations @xmath416 the quadratic algebras @xmath417 and @xmath418 were considered in @xcite , @xcite respectively ; @xmath419 are named the _ fomin - kirillov algebras_. it is known that @xmath56 the nichols algebras @xmath420 and @xmath421 are twist - equivalent , hence have the same hilbert series . ditto for the algebras @xmath422 and @xmath419 @xcite . @xmath56 if @xmath423 , then @xmath424 and @xmath425 are finite - dimensional @xcite ( for @xmath426 part of this was done by graa ) . in fact @xmath427 but for @xmath428 , it is not known whether the nichols algebras @xmath420 and @xmath421 have finite dimension or are quadratic . [ exa : nichols - cube ] @xcite the nichols algebra @xmath429 is quadratic , has the same hilbert series as @xmath430 and is generated by @xmath431 with defining relations @xmath432 [ exa : nichols - affines ] @xcite let @xmath274 be a finite abelian group and @xmath433 . the _ affine _ rack @xmath434 is the set @xmath274 with product @xmath435 , @xmath436 . let @xmath437 be a prime , @xmath438 a power of @xmath126 , @xmath439 and @xmath440 the multiplication by @xmath441 ; let @xmath442 . assume that @xmath443 , 4 , 5 , or 7 , with @xmath444 , @xmath445 , 2 or 3 , respectively . then @xmath446 , @xmath447 being the euler function , and @xmath448 , where @xmath449 is generated by @xmath450 with @xmath451 ; and @xmath452 is generated by @xmath453 and @xmath454 of course @xmath455 ; also @xmath456 . by duality , we get @xmath457 [ exa : nichols - tetra]@xcite there is another finite - dimensional nichols algebra associated to @xmath458 with a cocycle @xmath349 with values @xmath459 , where @xmath460 . concretely , @xmath461 and @xmath462 can be presented by generators @xmath463 with defining relations @xmath464 the ideas of example [ exa : grs - nichols - ss - yd ] in the context of decomposable yetter - drinfeld modules over groups were pushed further in a series of papers culminating with a remarkable classification result @xcite . consider the groups @xmath465 @xmath56 @xcite let @xmath19 be a quotient of @xmath466 . then there exist @xmath467 , @xmath468 such that @xmath469 and @xmath470 . @xmath56 @xcite let @xmath19 be a quotient of @xmath471 . then there exist @xmath472 , @xmath473 , @xmath474 , @xmath475 such that @xmath476 , @xmath477 , @xmath478 and @xmath479 , @xmath480 . @xmath56 @xcite let @xmath19 be a quotient of @xmath481 . then there exist @xmath482 , @xmath483 such that @xmath484 , @xmath485 and @xmath486 . @xmath56 @xcite let @xmath19 be a quotient of @xmath150 . then there exist @xmath487 , @xmath488 such that @xmath489 , @xmath490 and @xmath491 . _ let @xmath19 be a non - abelian group and @xmath492 such that @xmath19 is generated by the support of @xmath493 . assume that @xmath494 and that @xmath495 . then @xmath493 is one of @xmath496 , @xmath497 , above , and correspondingly @xmath19 is a quotient of either @xmath498 , @xmath499 , or @xmath150 . implicit in question ( a@xmath173 ) in the setting of racks is the need to compute all non - principal 2-cocycles for a fixed rack @xmath15 . notably , there exist criteria that dispense of this computation . to state them and explain their significance , we need some terminology . all racks below are finite . @xmath56 a rack @xmath15 is _ abelian _ when @xmath500 , for all @xmath501 . @xmath56 a rack is _ indecomposable _ when it is not a disjoint union of two proper subracks . @xmath56 a rack @xmath15 with @xmath502 is _ simple _ when for any projection of racks @xmath503 , either @xmath158 is an isomorphism or @xmath504 has only one element . _ @xcite , @xcite _ every simple rack is isomorphic to one of : 1 . affine racks @xmath505 , where @xmath126 is a prime , @xmath506 , and @xmath150 is the companion matrix of a monic irreducible polynomial @xmath507 $ ] of degree @xmath212 , @xmath508 . [ item : twisted - conjugacy ] non - trivial ( twisted ) conjugacy classes in simple groups . [ item : twisted - homogeneous ] twisted conjugacy classes of type @xmath509 , where @xmath510 , with @xmath122 a simple non - abelian group and @xmath511 ; and @xmath512 acts by @xmath513 , where @xmath514 . [ def : rack - typed ] @xcite we say that a finite rack @xmath15 is _ of type d _ when there are a decomposable subrack @xmath515 , @xmath516 and @xmath517 such that @xmath518 . also , @xmath15 is _ of type f _ @xcite if there are a disjoint family of subracks @xmath519 and a family @xmath520 with @xmath521 , such that @xmath522 , @xmath523 , for all @xmath524 . an indecomposable rack @xmath15 _ collapses _ when @xmath525 for every finite _ faithful _ 2-cocycle @xmath349 ( see @xcite for the definition of faithful ) . [ th : type4 ] _ @xcite ; @xcite _ if a rack is of type d or f , then it collapses . the proofs use results on nichols algebras from @xcite . if a rack projects onto a rack of type d ( or f ) , then it is also of type d ( or f ) , hence it collapses by theorem [ th : type4 ] . since every indecomposable rack @xmath15 , @xmath502 , projects onto a simple rack , it is natural to ask for the determination of all _ simple _ racks of type d or f. a rack is _ cthulhu _ if it is neither of type d nor f ; it is _ sober _ if every subrack is either abelian or indecomposable @xcite . sober implies cthulhu . @xmath56 let @xmath526 . let @xmath373 be either @xmath527 , if @xmath528 , or else @xmath529 if @xmath530 . the type of @xmath80 is formed by the lengths of the cycles in its decomposition . @xmath92 @xcite if the type of @xmath80 is @xmath531 , @xmath532 , @xmath533 , @xmath534 , @xmath535 , @xmath536 , @xmath537 , or @xmath538 , then @xmath373 is cthulhu . if the type of @xmath80 is @xmath539 , then @xmath373 is sober . @xmath92 @xcite let @xmath437 be a prime . assume the type of @xmath80 is @xmath540 . if @xmath541 or not of the form @xmath542 , @xmath128 a prime power , then @xmath373 is sober ; otherwise @xmath373 is of type d. assume the type of @xmath80 is @xmath543 . if @xmath544 or not of the form @xmath542 , @xmath128 a prime power , then @xmath373 is sober ; otherwise @xmath373 is of type d. @xmath92 @xcite for all other types , @xmath373 is of type d , hence it collapses . @xmath56 @xcite let @xmath545 and @xmath127 be a prime power . let @xmath546 not semisimple and @xmath547 . the type of a unipotent element are the sizes of its jordan blocks . @xmath92 assume @xmath548 is unipotent . if @xmath548 is either of type @xmath549 and @xmath127 is even or not a square , or of type @xmath550 and @xmath551 , then @xmath373 is sober . if @xmath548 is either of type @xmath552 and @xmath127 is even , or of type @xmath553 and @xmath554 then @xmath373 is cthulhu . if @xmath548 is of type @xmath555 and @xmath556 is even , then @xmath373 is not of type d , but it is open if it is of type f. @xmath92 otherwise , @xmath373 is either of type d or of type f , hence it collapses . @xmath56 @xcite let @xmath373 be a conjugacy class in a sporadic simple group @xmath19 . if @xmath373 appears in table [ tab : notd ] , then @xmath373 is not of type d. if @xmath557 is the monster and @xmath373 is one of , then it is open whether @xmath373 is of type d. otherwise , @xmath373 is of type d. @xmath150 & & @xmath558 & + @xmath559 & & @xmath560 & + @xmath561 & & @xmath562 & + @xmath563 & & @xmath564 & + @xmath565 & & @xmath566 & + @xmath567 & & @xmath568 & + @xmath569 & & @xmath570 & + @xmath571 & & @xmath572 & + @xmath573 & & @xmath574 & + @xmath575 & & @xmath576 & + @xmath577 & & @xmath578 & + @xmath579 & + here is the scheme of proof proposed in @xcite to attack conjecture [ conj : gendegone ] : let @xmath150 be a finite - dimensional graded hopf algebra in @xmath161 with @xmath148 and generated as algebra by @xmath177 . we have a commutative diagram of hopf algebra maps @xmath580 & & { { \mathfrak b}}(v ) \\ & t(v)\ar@{->}[ul]^p \ar@{->}[ur ] & } $ ] . to show that @xmath158 is injective , take a generator @xmath128 ( or a family of generators ) of @xmath145 such that @xmath581 and consider the yetter - drinfeld submodule @xmath582 of @xmath100 ; if @xmath583 , then @xmath584 . then @xmath126 factorizes through @xmath585 , where @xmath586 is the ideal generated by primitive generators of @xmath145 , and so on . the conjecture has been verified in all known examples in characteristic 0 ( it is false in positive characteristic or for infinite - dimensional hopf algebras ) . a finite - dimensional pointed hopf algebra @xmath5 is generated by group - like and skew - primitive elements if either of the following holds : @xmath92 _ @xcite _ the infinitesimal braiding is of diagonal type , e. g. @xmath587 is abelian . @xmath92 _ @xcite . _ the infinitesimal braiding of @xmath5 is any of @xmath588 , @xmath589 ( @xmath590 ) , @xmath591 , @xmath592 , @xmath593 , @xmath594 , @xmath595 . we address here question [ item : method - lifting ] in [ subsec : chevalley ] . let @xmath15 be a finite rack and @xmath596 a 2-cocycle . a hopf algebra @xmath5 is a _ lifting _ of @xmath597 if @xmath61 is a hopf subalgebra , @xmath5 is generated by @xmath179 and its infinitesimal braiding is a realization of @xmath598 . see @xcite for liftings in the setting of copointed hopf algebras . we start discussing realizations of braided vector spaces as yetter - drinfeld modules . let @xmath346 and @xmath599 . first , a _ yd - datum of diagonal type _ is a collection @xmath600 where @xmath601 , @xmath602 , @xmath603 ; @xmath19 is a finite group ; @xmath604 ; @xmath605 , @xmath206 ; such that @xmath606 , @xmath603 . let @xmath3 be the braided vector space of diagonal type with matrix @xmath607 in the basis @xmath608 . then @xmath113 by declaring @xmath609 , @xmath206 . more generally , a _ yd - datum of rack type _ @xcite is a collection @xmath610 where @xmath15 is a finite rack ; @xmath596 is a @xmath611-cocycle ; @xmath19 is a finite group ; @xmath612 is an action of @xmath19 on @xmath15 ; @xmath613 is equivariant with respect to the conjugation in @xmath19 ; and @xmath614 is a family of 1-cocycles @xmath615 ( that is , @xmath616 , for all @xmath617 , @xmath618 ) such that @xmath619 and @xmath620 for all @xmath621 . let @xmath622 be the associated braided vector space . then @xmath0 becomes an object in @xmath371 by @xmath623 and @xmath624 , @xmath625 , @xmath617 . second , let @xmath626 be a yd - datum of either diagonal or rack type and @xmath627 as above ; let @xmath628 . the desired liftings are quotients of @xmath629 ; write @xmath630 in these quotients instead of @xmath631 to distinguish them from the elements in @xmath632 . let @xmath241 be a minimal set of generators of @xmath145 , assumed homogeneous both for the @xmath633- and the @xmath19-grading . roughly speaking , the deformations will be defined by replacing the relations @xmath634 by @xmath635 , @xmath636 , where @xmath637 belongs to a lower term of the coradical filtration , and the ideal @xmath638 generated by @xmath639 , @xmath636 , is a hopf ideal . the problem is to describe the @xmath639 s and to check that @xmath640 has the right dimension . if @xmath581 has @xmath19-degree @xmath440 , then @xmath641 for some @xmath642 ; depending on the action of @xmath19 on @xmath128 , it may happen that @xmath643 should be 0 . in some cases , all @xmath636 are primitive , so all deformations can be described ; see @xcite for quantum linear spaces ( their liftings can also be presented as ore extensions @xcite ) and the examples [ exa : lifting - transpositions - s3 ] and [ exa : lifting - transpositions - s4 ] . but in most cases , not all @xmath636 are primitive and some recursive construction of the deformations is needed . this was achieved in @xcite for diagonal braidings of cartan type @xmath644 , with explicit formulae , and in @xcite for diagonal braidings of finite cartan type , with recursive formulae . later it was observed that the so obtained liftings are cocycle deformations of @xmath632 , see e.g. @xcite . this led to the strategy in @xcite : pick an adapted stratification @xmath645 @xcite ; then construct recursively the deformations of @xmath646 by determining the cleft extensions of the deformations in the previous step and applying the theory of hopf bi - galois extensions @xcite . in the examples below , @xmath647 for all @xmath617 by ( * ( d ) ) . [ exa : lifting - transpositions - s3 ] @xcite let @xmath648 be a yd - datum . let @xmath649 be such that @xmath650 let @xmath651 be the quotient of @xmath629 by the relations @xmath652 then @xmath653 is a pointed hopf algebra , a cocycle deformation of @xmath654 and @xmath655 ; @xmath656 iff @xmath657 for some @xmath658 . conversely , any lifting of @xmath659 is isomorphic to @xmath660 for some yd - datum @xmath648 and @xmath649 satisfying , . [ exa : lifting - transpositions - s4 ] @xcite let @xmath661 be a yd - datum . let @xmath662 be such that @xmath663 let @xmath651 be the quotient of @xmath629 by the relations @xmath664 then @xmath653 is a pointed hopf algebra , a cocycle deformation of @xmath665 and @xmath666 ; @xmath656 iff @xmath657 for some @xmath658 . conversely , any lifting of @xmath667 is isomorphic to @xmath660 for some yd - datum @xmath661 and @xmath662 satisfying , . [ exa : lifting - affine - f4 ] @xcite let @xmath668 be a yd - datum . let @xmath662 be such that @xmath669 let @xmath651 be the quotient of @xmath629 by the relations @xmath670 @xmath671 then @xmath653 is a pointed hopf algebra , a cocycle deformation of @xmath654 and @xmath672 ; @xmath656 iff @xmath657 for some @xmath658 . conversely , any lifting of @xmath673 is isomorphic to @xmath660 for some yd - datum @xmath668 and @xmath662 satisfying , . [ exa : lifting - affine - f5 ] @xcite let @xmath674 be a yd - datum . let @xmath662 be such that @xmath675 let @xmath651 be the quotient of @xmath629 by the relations @xmath676 @xmath677 . then @xmath653 is a pointed hopf algebra , a cocycle deformation of @xmath654 and @xmath678 ; @xmath656 iff @xmath657 for some @xmath658 . conversely , any lifting of @xmath679 is isomorphic to @xmath660 for some yd - datum @xmath674 and @xmath662 satisfying , . [ exa : lifting - affine - f5 - 3 ] @xcite let @xmath680 be a yd - datum . let @xmath662 be such that @xmath681 let @xmath651 be the quotient of @xmath629 by the relations @xmath682 where @xmath683 . then @xmath653 is a pointed hopf algebra , a cocycle deformation of @xmath654 and @xmath678 ; @xmath656 iff @xmath657 for some @xmath658 . conversely , any lifting of @xmath684 is isomorphic to @xmath660 for some yd - datum @xmath680 and @xmath662 satisfying , . here is a classification from @xcite . let @xmath685 be a yd - datum of diagonal type as in with @xmath175 a finite abelian group and let @xmath352 be the corresponding realization . we say that @xmath626 is a _ cartan datum _ if there is a cartan matrix ( of finite type ) @xmath686 such that @xmath687 , @xmath688 . let @xmath689 be the root system associated to @xmath690 , @xmath691 a choice of simple roots , @xmath692 the set of connected components of the dynkin diagram of @xmath689 and set @xmath693 whenever @xmath694 belong to the same @xmath695 . we consider two classes of parameters : * @xmath696 is a family in @xmath697 with @xmath698 when @xmath699 or @xmath700 . * @xmath701 is a family in @xmath14 with @xmath702 when @xmath703 , @xmath704 , and @xmath705 or @xmath706 . here @xmath707 for an arbitrary @xmath708 . we attach a family ( @xmath709 in @xmath710 to the parameter @xmath711 , defined recursively on the length of @xmath712 , starting by @xmath713 . from all these data we define a hopf algebra @xmath714 as the quotient of @xmath715 by the relations @xmath716 the hopf algebra @xmath717 is pointed , @xmath718 and @xmath719 . let @xmath5 be a pointed finite - dimensional hopf algebra and set @xmath720 . assume that the prime divisors of @xmath721 are @xmath722 . then there exists a cartan datum @xmath626 and parameters @xmath643 and @xmath711 such that @xmath723 . it is known when two hopf algebras @xmath717 and @xmath724 are isomorphic . the proof offered in @xcite relies on @xcite . some comments : the hypothesis on @xmath725 forces the infinitesimal braiding @xmath0 of @xmath5 to be of cartan type , and the relations of @xmath153 to be just quantum serre and powers of root vectors . the quantum serre relations are not deformed in the liftings , except those linking different components of the dynkin diagram ; the powers of the root vectors are deformed to the @xmath726 that belong to the coradical . all this can fail without the hypothesis , see @xcite for examples in rank 2 . we present some classification results of pointed hopf algebras with non - abelian group . we say that a finite group @xmath19 _ collapses _ whenever any finite - dimensional pointed hopf algebra @xmath5 with @xmath727 is isomorphic to @xmath45 . @xmath67 @xcite let @xmath729 and let @xmath626 be the corresponding yd - datum . let @xmath5 be a finite - dimensional pointed hopf algebra with @xmath730 . then @xmath5 is isomorphic either to @xmath731 , or to @xmath732 , or to @xmath733 , cf . example [ exa : lifting - transpositions - s3 ] . @xmath67 @xcite let @xmath734 be a finite - dimensional pointed hopf algebra with @xmath735 . let @xmath736 , @xmath737 , @xmath738 , with corresponding data @xmath739 , @xmath740 and @xmath741 . then @xmath5 is isomorphic to one of @xmath742 here @xmath743 is as in example [ exa : lifting - transpositions - s4 ] ; @xmath744 is the quotient of @xmath745 by the relations @xmath746 , @xmath747 , @xmath748 and @xmath749 is the quotient of @xmath750 by the relations @xmath751 clearly @xmath752 , @xmath753 , @xmath754 . also @xmath755 iff @xmath756 for some @xmath757 , and @xmath758 iff @xmath759 for some @xmath757 . @xmath67 @xcite let @xmath5 be a finite - dimensional pointed hopf algebra with @xmath760 , but @xmath761 . it is not known whether @xmath762 . let @xmath739 , @xmath740 be the data corresponding to @xmath763 , @xmath764 . if the infinitesimal braiding of @xmath5 is not @xmath765 , then @xmath5 is isomorphic to one of @xmath743 , @xmath766 ( defined as in example [ exa : lifting - transpositions - s4 ] ) , or @xmath767 , or @xmath768 ( defined as above ) . @xmath67 @xcite let @xmath769 . let @xmath770 be a finite - dimensional pointed hopf algebra with @xmath771 . then the infinitesimal braiding of @xmath5 is @xmath772 , where the type of @xmath80 is @xmath773 and @xmath774 , @xmath775 or @xmath776 ; it is an open question whether @xmath777 , see example [ exa : nichols - sim ] . if @xmath778 , there are two more nichols algebras with unknown dimension corresponding to the class of type @xmath537 , but they are conjugated to those of type @xmath779 by the outer automorphism of @xmath780 . @xmath67 @xcite let @xmath781 , @xmath782 with @xmath783 . let @xmath784 be the dihedral group of order @xmath785 . then there are infinitely many finite - dimensional nichols algebras in @xmath371 ; all of them are exterior algebras as braided hopf algebras . let @xmath5 be a finite - dimensional pointed hopf algebra with @xmath786 , but @xmath787 . then @xmath5 is a lifting of an exterior algebra , and there infinitely many such liftings . we say that a semisimple hopf algebra @xmath75 _ collapses _ if any finite - dimensional hopf algebra @xmath5 with @xmath788 is isomorphic to @xmath75 . thus , if @xmath19 collapses , then @xmath46 and @xmath789 collapse , for any twist @xmath73 . next we state the classification of the finite - dimensional copointed hopf algebras over @xmath790 @xcite . let @xmath729 as a yetter - drinfeld module over @xmath791 . let @xmath792 be such that @xmath793 . let @xmath794 be the quotient of @xmath795 by the relations @xmath796 , @xmath797 , @xmath798 , for @xmath799 . then @xmath800 is a hopf algebra of dimension 72 and @xmath801 . any finite - dimensional copointed hopf algebra @xmath5 with @xmath802 is isomorphic to @xmath803 for some @xmath643 as above ; @xmath804 iff @xmath643 and @xmath805 are conjugated under @xmath806 .

we include an energy term based on dark matter ( dm ) self - annihilation during the cooling and subsequent collapse of the metal - free gas , in halos hosting the formation of the first stars in the universe . we have found that the feedback induced on the chemistry of the cloud _ does _ modify the properties of the gas throughout the collapse . however , the modifications are not dramatic , and the typical jeans mass within the halo is conserved throughout the collapse , for all the dm parameters we have considered . this result implies that the presence of dark matter annihilations does not substantially modify the initial mass function of the first stars , with respect to the standard case in which such additional energy term is not taken into account . we have also found that when the rate of energy produced by the dm annihilations and absorbed by the gas equals the chemical cooling ( at densities yet far from the actual formation of a proto stellar core ) the structure does _ not _ halt its collapse , although that proceeds more slowly by a factor smaller than few per cent of the total collapse time . stars : formation stars : populationiii dark ages , reionization , first stars dark matter

1003.0676

in the currently favoured @xmath0cdm cosmological model , the bulk of the matter component is believed to be made of ( so far ) electromagnetically undetected particles , commonly dubbed dark matter ( dm ) . although the evidence for the existence of dm is compelling on different scales , yet its nature is unknown , and many particle models beyond the standard one have been proposed in the literature as dm candidates . we address the reader to a recent review of observational evidence and particle candidates for dm ( e.g. bertone , hooper & silk 2005 ) , and will concentrate in this paper on a particular class of candidates , i.e. weakly interacting massive particles ( wimps ) . many wimp dm models are stable ( under the conservation of the suitable symmetry , for each model ) and hence do never decay into standard model particles . however , in many of these very same models the wimps are majorana particles , thus carrying the remarkable property of being self - annihilating ; the value of the self annihilation cross section , arising naturally in many wimp models , reproduces the dark matter relic abundance required by the @xmath0cdm cosmology , if the mass scale of wimps is within the gev / tev scale and they are to be thermally produced in the early universe . we adopt this as a benchmark scenario for our paper , and will often refer to it as a `` vanilla wimp '' . the actual dm distribution in the local universe is such that even in the densest regions ( e.g. galactic nuclei and black hole surroundings ) from which the annihilation signal could be in principle detected , the energy released by wimp dm annihilations ( hereafter , dmas ) is only a negligible fraction of the one associated with standard gas processes . this implies that , locally , dm affects the host system almost uniquely through its gravitational effects , perhaps with the only possible exception of peculiar locations , such as the central parsec of the milky way ( fairbairn , scott & edsjo 2008 , scott , edsjo & fairbairn 2009 , casanellas & lopes 2009 ) . the effects of annihilating ( or decaying , a scenario we do not consider here ) dm upon the evolution of the intergalactic medium ( igm ) at high redshift have been thoroughly studied ( e.g. chen & kamionkowski 2004 ; padmanabhan & finkbeiner 2005 ; mapelli & ferrara 2005 ; mapelli , ferrara & pierpaoli 2006 ; furlanetto , oh & pierpaoli 2006 ; zhang et al . 2006 ; ripamonti , mapelli & ferrara 2007a ; valds et al . 2007 ; shchekinov & vasiliev 2007 ) , and are now believed to be small , except perhaps in the case of an extremely high clumping factor ( myers & nusser 2008 ; chuzhoy 2008 ; natarajan & schwarz 2008 , lattanzi & silk 2009 ) , if one takes into consideration a standard , vanilla wimp scenario . the effects of dmas upon primordial star - formation might be more significant . as the igm could be heated by the energy deposition from dmas , its temperature might in principle exceed the virial temperature of the smallest halos with the result of quenching gas accretion onto them . this effect has been shown to be unimportant by ripamonti , mapelli & ferrara ( 2007b , hereafter rmf07 ) . however , dmas are expected to become more important as the collapse proceeds to protostellar scales ( ascasibar 2007 ) . spolyar , freese & gondolo ( 2008 ; hereafter sfg08 ) found that during the protostellar collapse of the first ( pop iii ) stars , the energy released by dmas and absorbed by the gas could compensate ( or even overcome ) the radiative cooling of the gas . the increasing importance of such process arises from the combined enhancement during the collapse of dm density ( due to gravitational dragging ) and gas optical depth , implying a higher annihilation luminosity and absorption by the gas . the final phases of the collapse , after the formation of a hydrostatic core for gas central densities @xmath1 ( where @xmath2 is the central baryonic density , and @xmath3 is the proton mass ) , have been investigated by iocco et al . ( 2008 , hereafter i08 ) , freese et al . ( 2008b , 2009 ) and spolyar et al . ( 2009 ) . initially , the dm pile - up is purely driven by gravitational interactions , but as the protostar approaches the zero age main sequence , dm accretion becomes dominated by the capture of wimps located in the star host halo after they scatter stellar baryons . as a consequence of the peculiar formation process of pop iii , following the smooth collapse of the gas cloud at the very center of the dm halos hosting them , iocco ( 2008 ) and freese , spolyar & aguirre ( 2008 ) suggested that dm capture is relevant for primordial stars ; however , it can be safely neglected once local star formation is concerned , as the latter takes place anywhere in galactic discs , and it does nt follow from a single , centered gas collapse episode . further studies ( i08 , yoon , iocco & akiyama 2008 ; taoso et al . 2008 ) have concluded that wimp dm capture s most remarkable effect is the possible increase of the stellar lifetime . quite surprisingly , the early phases of the collapse have received so far less attention with respect to the more advanced ones , i.e. after hydrostatic core formation . for example , it is still unclear if the energy injection following annihilations results in a nett heating or cooling of the gas . in fact , high energy photons and electrons heat the gas through ionizations ; however , this heat input could be overwhelmed by the increased production of cooling species ( as for example molecular hydrogen ) stimulated by the larger abundances of free electrons , thus resulting in a net gas cooling . this , among others , is one of the aspects of the collapse of first stars in presence of wimp annihilation that we would like to address here . we plan to do so by a set of sophisticated numerical simulations including all the relevant chemical reactions and cooling processes . a first attempt to model the effects of dma energy input was presented in ripamonti et al . ( 2009 ) ; this paper represents a substantial extension and improvement of that study . throughout the paper we assume the following set of cosmological parameters : @xmath4 , @xmath5 , @xmath6 , @xmath7 , and @xmath8 . we base our investigation on a 1-d spherically symmetric code described by ripamonti et al . ( 2002 ; hereafter r02 ) . the original code , which includes the treatment of gravitation , hydrodynamics , and especially the chemistry and cooling of primordial gas , was originally conceived for the study of the last phases of the collapse of a primordial protostar ( see also omukai & nishi 1998 ) ; later , it was extended in several ways ( see ripamonti 2007 , ripamonti et al . 2007a , and rmf07 ) . our simulations are based on those described in rmf07 ; here we list their most important properties , especially when they differ from rmf07 : a single typical halo with mass @xmath9 virializing at @xmath10 ( virial radius @xmath11 ) is considered ; the baryon fraction inside such halo is assumed to be equal to the cosmological value ( @xmath12 ) ; the simulations are started at @xmath13 and involve a comoving volume 1000 times larger than that of simulated halo ; initial baryonic density and temperature are constant in all the simulation shells , and equal to the cosmological values ; before virialization , the gravitational effects of dm are treated as in the nfw case of rmf07 : a predetermined ( but time - dependent until virialization ) dm potential is added to the gas self gravity . such potential mimics the evolution of a halo in the top - hat approximation : as the dm potential becomes steeper , the ( initially uniform ) gas falls towards the centre of the halo , similarly to what is predicted by theory and consistent with the results of simulations ( see e.g. fig . 2 of abel et al . 2002 ) ; after virialization , the evolution of the previously described artificial dm potential is stopped : its state at virialization is set to a nfw profile ( navarro , frenk & white 1996 ) with @xmath14 and @xmath15 ; this is reasonably close to the results shown in fig . 2 of abel et al . the evolution of the dm profile is also followed in response to the baryon contraction ( see later ) in order to compute the dma rate ; rmf07 investigated _ whether _ stellar formation might occur in a halo , whereas here we investigate _ how _ it starts ; for this reason , simulations are stopped only when their computational costs become very large ( usually at number densities @xmath16 ) ; the dm energy input is computed only after halo virialization , and only in regions with high baryon density ( @xmath17 , i.e. @xmath18 ) , rather than at all times and everywhere : this is because rmf07 already showed that before virialization and at low baryon densities the effects of dmas are small ; for the purpose of evaluating the dma rate ( see below ) , the dm density @xmath19 is evaluated by assuming the conservation of the so called `` adiabatic '' invariant ( see blumenthal et al . we implement the algorithm described by gnedin et al . ( 2004 ) , following the details in i08 , and using the nfw and gas profiles described above as initial conditions ; the specific luminosity due to dmas is @xmath20 where @xmath21 is the thermally - averaged annihilation rate , and @xmath22 is the wimp mass ; in the following we adopt @xmath23 , whereas we consider @xmath22 as a free parameter in the range @xmath24 ( i.e. @xmath25 ) ; the energy @xmath26 that each baryon actually absorbs from dm annihilations ( per unit time ) is calculated through a detailed radiative transfer calculation , formally identical to the one performed for gray continuum radiation . such calculation is based on a constant gas opacity @xmath27 ( roughly similar to the values used by sfg08 ) . moreover , since it is believed that the energy from annihilations splits roughly equally into electrons , photons , and neutrinos , we assume that only @xmath28 of @xmath29 ( i.e. the fraction not going into neutrinos ) can be absorbed ; similar to rmf07 , @xmath26 can go into ionization , heating and excitation of atoms and molecules ; we employ the results of valds & ferrara ( 2008 ) ( see also shull & van steenberg 1985 , and furlanetto & johnson stoever 2010 ) to estimate how to split the energy input into these three . also note that the `` ionization '' component is split into ionization of h , d , he , he@xmath30 , and dissociation of , hd and h@xmath31 . in our `` standard '' treatment each species receives a fraction of the ionization energy which is proportional to its total baryonic content ( in number , see rmf07 for details ) ; r02 switched to equilibrium chemistry ( e.g. allowing the use of saha equations instead of the detailed balance ones ) for shells with number densities @xmath32 . here we drop this simplification since ( i ) dmas effects change the chemical evolution of the gas , and usually delay the approach to equilibrium , and ( ii ) we never venture to densities @xmath33 . given the standing ignorance on the precise detail of feedback effects on the ionization and dissociation of atoms and molecules ( especially h@xmath34 ) , in addition to the standard , fiducial set of runs we performed runs with either enhanced or reduced feedback , in order to bracket the possible impact of such process . in the same way , since the opacity @xmath35 we employ represents only a very rough estimate , we performed runs with either higher or lower values of @xmath35 . [ tab : runlist ] lccl name & @xmath36[gev ] & fbk & notes + m1000 & 1000 & std . & minimal + m100 & 100 & std . & fiducial + m10 & 10 & std . & sub - maximal + m1 & 1 & std . & maximal + n1000 & 1000 & off & + n100 & 100 & off & + n10 & 10 & off & + n1 & 1 & off & + e1000 & 1000 & enh . & + e100 & 100 & enh . & + e10 & 10 & enh . & + e1 & 1 & enh . & + l100 & 100 & std . & @xmath37 + l10 & 10 & std . & @xmath37 + l1 & 1 & std . & @xmath37 + h1000 & 1000 & std . & @xmath38 + h100 & 100 & std . & @xmath38 + h10 & 10 & std . & @xmath38 + nodm & & & control + we test the effects of dmas varying different sets of parameters : _ ( i ) _ the normalization of the dma rate , which is regulated by the ratio @xmath21/@xmath22 ; _ ( ii ) _ the feedback strength on chemistry ; and _ ( iii ) _ the `` gray '' gas opacity @xmath35 . we anticipate that the strength of feedback on chemistry has little impact on the overall results , and that the effects of a variation in @xmath35 are somewhat similar to a variation of the same amount in @xmath21/@xmath22 . we will discuss the dependence on these two latter parameters in section [ sec : opacfeed ] . here we start by introducing in more details the physics of first star formation in presence of dmas for our fiducial models ( `` m '' labeled runs , see table [ tab : runlist ] for details ) . it is worth anticipating that the effects of dm become more relevant ( and they become efficient earlier during the collapse ) for higher self - annihilation cross sections / lower dm masses ( see eq . [ eq : defdmlum ] ) . in what follows , we always adopt @xmath21=3@xmath3910@xmath40 @xmath41/s , and vary the particle mass @xmath42 . given the degeneracy in the dma energy injection term , the results can be interpreted at fixed mass and correspondingly varying the self - annihilation rate . of the central shell of our simulation ( whose mass is @xmath43 ) as a function of its baryon number density @xmath44 , starting slightly before virialization . the thick solid line shows the evolution in the control run with no dm energy input ( nodm ) . the thin solid ( blue ) corresponds to the fiducial model ( run m100 ) ; the dashed ( red ) to the maximal model ( m1 ) ; the dot - dashed ( green ) to the sub - maximal model ( m10 ) ; and the thin dotted ( magenta ) to the minimal model ( m1000).,scaledwidth=48.0% ] figure [ fig : plot_n_t ] shows the evolution of the temperature @xmath45 of the innermost shell as a function of the density of the same central shell of the simulated objects ( @xmath44 ) , for the five most representative models , m100 , m10 , m1 , m1000 , and nodm ( see table [ tab : runlist ] ) . as expected , dma effects , which can be quantified by the deviations from the thick solid line , are most prominent in runs with low mass particles @xmath46 ( m1 run , corresponding to the maximal dma energy injection rate ) , and become very limited in the case of dm particles with high mass @xmath47 ( m1000 run , minimal dma energy injection rate ) ; we will comment extensively on such dependence later in this section . in the following we will refer to the `` fiducial '' run ( m100 ; where the choice of parameters is quite standard ) , to the `` minimal '' run ( m1000 ; where the high value of @xmath22 reduces the energy injection from dmas , providing a check on the level where its effects become negligible ) , to the `` maximal '' run ( m1 ; where the parameters are chosen in a way which maximizes the energy injection from dmas and their effects upon the gas evolution ) , and to the `` sub - maximal '' run , m10 . the maximal model , m1 , is likely to represent a sort of upper limit on the effects of dmas on the formation of primordial stars . in fact , the dm parameters corresponding to this run ( @xmath21=3@xmath3910@xmath48 , @xmath36=1 gev ) are currently severely disfavored by multimessenger constraints on dm , in particular by ( astrophysical ) parameter - independent ones ( e.g. galli et al . 2009 ; cirelli , iocco and panci 2009 ) . of what is shown here , since the total dma energy input also includes energy actually going into excitations and ionizations ( see text for details).,scaledwidth=48.0% ] figure [ fig : plot_n_lambdah2_dminput ] compares the total energy input from dmas into the gas ( which needs to be splitted into ionizations , heating and excitations ) to the cooling rate due to . this comparison largely oversimplifies the thermal balance ( see next subsection ) ; however , it makes clear that , even in the maximal case , the heating from dmas remains a relatively small fraction of the cooling until @xmath44 becomes @xmath49 . at earlier stages ( i.e. lower densities ) the effects of dmas are more subtle than what could be simply inferred by the direct heating , which would not be able to significantly change the temperature of the gas . in fact , it is quite remarkable that in the @xmath50 range , the _ additional _ energy injection from dmas results in a _ decrease _ of the central temperature ( see also fig . [ fig : plot_n_t ] ) . such counter - intuitive temperature decrease is due to indirect effects of dmas , and in particular to their feedback on the chemistry ; this remarkable property clearly emerges upon inspection of the curves in fig . [ fig : plot_n_lambdah2_dminput ] showing that cooling ( thin solid line ) is stronger when dmas are included ( at least for @xmath51 ) . in fact , as it can be inferred from fig . [ fig : plot_n_f ] , dma _ ionization _ effects keep the gas more ionized than in the nodm case ( i.e. @xmath52 abundance @xmath53@xmath54 , rather than @xmath55 ) . in turn , this relatively high ionization level favors the formation of through the reaction chain @xmath56 which is the main formation route for in moderately dense , dust - free gas , and where the electrons act as catalysts ( see e.g. galli & palla 1998 ; glover & jappsen 2007 ; stasielak , biermann & kusenko 2007 ) . fractions in the central shell of our simulation as a function of its baryon number density . in each panel the fraction ( thin solid line ) and @xmath52 fraction ( thin dashed line ) from one of our runs ( from top to bottom : fiducial , sub - maximal , maximal runs ) are compared to the same quantities from the control nodm run ( thick solid line for and thick dashed line for @xmath52).,scaledwidth=48.0% ] as ripamonti et al . ( 2009 ) remarked , the increase in the fraction can amount to 2 - 3 orders of magnitude , which apparently might imply same order - of - magnitude differences in the cooling rate , and large differences also in the gas temperature . however , while the increase in the cooling rate is quite large ( though smaller than what is naively expected from the difference in abundance ) , the corresponding reduction of the gas temperature is much smaller , of the order of 30 per cent ( see fig.[fig : plot_n_t ] ) . this is due to the very strong temperature dependence of cooling , especially below @xmath57 ( the temperature corresponding to the transition from the fundamental state to the lowest rotationally excited level ) : in fact , in this regime a moderate reduction in temperature results in a much larger decrease in the cooling rate , which in turn slows down the temperature decrease . , whereas the bottom panel compares the various models by showing the ratio of @xmath58 in a given model to @xmath59 in the nodm run.,scaledwidth=48.0% ] as a result , since the jeans mass sensitivity to the temperature is quite mild , @xmath60 dmas might reduce the jeans mass scale but only by a factor @xmath61 ( see fig . [ fig : plot_n_mjeans ] ) . [ tab : criticalpoints ] ccc @xmath36 & @xmath62[@xmath63 & @xmath64[@xmath65 + 1 gev@xmath66 & @xmath67 ( @xmath68)@xmath69 & 780 ( @xmath70800)@xmath69 + 10 gev@xmath66 & @xmath71 ( @xmath72)@xmath69 & 1070 ( @xmath73)@xmath69 + 100 gev@xmath66 & @xmath74 ( @xmath75)@xmath69 & 1580 ( @xmath701300)@xmath69 + 1000 gev@xmath66 & @xmath76 ( @xmath77)@xmath69 & 1740 ( @xmath78)@xmath69 + figure [ fig : plot_n_lambdah2_dminput ] shows that when the density increases above @xmath79 ( @xmath80 ) for the fiducial ( maximal ) run , the energy input from dmas finally overcomes the cooling ( see also table [ tab : criticalpoints ] for details and a comparison with previous literature ) . this marks the beginning of what we call the direct feedback phase of the collapse , since the dmas _ direct _ effects ( especially the heating ) finally start to dominate both cooling and the more subtle dmas indirect effects discussed above . we will refer to this condition ( equality of dma heating and cooling terms ) as to the `` critical point '' . the study of this phase is particularly interesting because when the dma heating starts to compensate the radiative cooling , the protostar becomes unable to shed away its gravitational energy ; it had been previously proposed ( see sfg08 ) that this could induce the stop of the collapse , and the formation of a new type of stable astrophysical object powered by dmas . the evolution of all the quantities beyond the critical point , shown in figures [ fig : plot_n_t ] and [ fig : plot_n_lambdah2_dminput ] , imply the existence of a dynamical evolution of the system also after the onset of the direct feedback phase -and hence no stalling of the object- even if there are important differences among the various runs . in the fiducial run , the evolution of the central shell proceeds relatively undisturbed . for @xmath81 the central temperature remains below the one of the control run , even if the gap tends to close ( see fig . [ fig : plot_n_t ] ) . conversely , in the maximal run there is a substantial steepening of the evolution in the @xmath82 plane , and the central temperature overcomes the value of the control run at @xmath83 . the temperature increase becomes particularly dramatic , with a sudden rise by a factor @xmath84 when the central density reaches @xmath85 . in order to understand this phase , we need to describe the heating and cooling processes in detail . [ fig : plot_n_heatcool ] shows the contributions from the main heating and cooling mechanisms during the contraction of the central shell , for our reference runs . . the panels refer to four different runs : from top to bottom , we show results for the control ( without dmas ) , fiducial , sub - maximal , maximal runs . each panel shows line cooling ( thin solid line ) , continuum cooling ( thick solid line ) , dma heating ( dashed line ) , adiabatic heating ( dotted line ) , and chemical heating or cooling ( dot - dashed line ) . we note that i ) the continuum cooling line can be plotted only when @xmath86 ( corresponding to @xmath87 ) , however it is negligible at lower temperatures ; ii ) the dma heating is different from the dma energy input shown in fig . [ fig : plot_n_lambdah2_dminput ] , since it does not include the ionization and excitation components ; iii ) the chemical term does not have a constant sign : shaded regions indicate the density ranges where it corresponds to a cooling term , whereas the ranges where it corresponds to a heating term are not shaded.,scaledwidth=48.0% ] figure [ fig : plot_n_heatcool ] shows the large importance of chemical heating / cooling , i.e. of the energy which is released / absorbed as a consequence of chemical reactions , and in particular ( in this regime , at least ) of the formation / dissociation of molecules ) ; at the densities we are considering here , most of this energy is eventually converted into thermal energy of the gas ( see e.g. hollenbach & mckee 1979 ) ; on the other hand , the dissociation of through collisional reactions absorbs the same amount of thermal energy from the gas . ] : in every run there is a phase where formation is the main heating mechanism . in the control run , chemical heating dominates over the adiabatic heating in the @xmath88 density range , which coincides with the regime where 3-body reactions ( see palla , salpeter & stahler 1983 ) turn most of the hydrogen into molecular form . in the runs with dmas such a phase is even more extended , as the dma feedback on the chemistry anticipates the epoch of rapid formation : in the maximal run , chemical heating becomes dominant already at @xmath89 . it is remarkable that the chemical term can contribute both to heating and to cooling . it was already known ( see e.g. omukai & nishi 1998 ; r02 ) that because of this property , it tends to act as a `` thermostat '' which stabilizes the evolution of the collapse , even if it can not always prevent sudden transitions . the evolution of the maximal and sub - maximal models are particularly significant in this respect . in these models the fraction peaks ( at a level @xmath90 ; see fig . [ fig : plot_n_f ] ) when @xmath91@xmath92 : in such conditions 3-body formation is slowed down by the relative scarcity of atomic h left , and must compete with the increasingly large number of dissociations due to the ionization component of the dma energy input . when the molecular fraction starts to go down , the cooling rate must follow , because it is dominated by and the number of molecules is being reduced ; but so does the heating , because its dominant component ( chemical heating ) is reduced by the energy lost in molecular dissociations . the fact that both the heating and the cooling mechanisms follow the same trend prevents a significant steepening of the temperature evolution and the consequent stop in the protostellar collapse . this is because the `` excess '' energy which is not radiated away is mainly used to dissociate . however , this can not last indefinitely , as the reservoir of is finite , and the dma energy injection increases with increasing density . in the case of the maximal run , where the dma energy injection is quite high , is rapidly exhausted . at @xmath93 there is essentially no left , which also implies almost no radiative cooling . then , the temperature goes up very fast ( see fig . [ fig : plot_n_t ] ) , developing a pressure gradient which is able to momentarily stop the collapse of the innermost shell ( actually , there is even a very brief re - expansion episode- more details are discussed in section [ sec : evolspatprof ] ) ; however , the increase in temperature also results in a very strong increase in the continuum radiation which is produced by collision - induced emission , h@xmath94 , and atomic h : continuum radiation increases by some 6 orders of magnitude ( cfr . lenzuni , chernoff & salpeter 1991 ; marigo & aringer 2009 ) and replaces line radiation as the main cooling mechanism for the gas . the collapse is then able to resume , even if further investigations are needed in order to understand its details beyond @xmath95 . at present in fact , computational costs become almost unaffordable beyond @xmath96 because of the strong sensitivity of cooling rate on temperature . in spite of these limitations we consider the conclusions based on these runs very robust as we have pushed the evolution to densities exceeding @xmath97 times the critical " density at which the cooling becomes comparable to the dma energy transferred to baryons . things are quite different in the case of the sub - maximal run : here a moderate dissociation rate can compensate both the heating and the ionization parts of the ( relatively small ) dma energy input for a significant span . in fact , at @xmath98 ( i.e. at a density @xmath99 times larger than the density at the onset of dissociation , and @xmath73 times larger than the density at the `` critical point '' ) the fraction is still @xmath100 : it is still possible that exhaustion ( which we expect to happen at @xmath101 ) is accompanied by a transition similar to the one we discussed for the maximal case ; however , such transition is likely to involve a smaller adjustment in temperature ( say , an increase by a factor @xmath102 rather than @xmath84 ) because the required increase in the continuum cooling is much smaller ( somewhat less than 3 orders of magnitude rather than 6 , as can be inferred from fig . [ fig : plot_n_heatcool ] ) than what was required in the maximal case . the fiducial and minimal runs exhibit trends which are somewhat similar to the sub - maximal run : the peak in the abundance at @xmath103 is followed by an extremely slow decline , during which the adiabatic heating is often as important as the dma heating : then , it is quite likely that the transition from line cooling to continuum cooling is as smooth as in the scenario without dmas . in fact , it should be noted that at @xmath98 the continuum cooling amounts already to about 10 per cent of the total heating , and it is increasing faster than any heating mechanism ( see fig . [ fig : plot_n_heatcool ] ) . for run coding.,scaledwidth=48.0% ] in order to check the effect of our assumptions about feedback ( in particular , destruction of molecules by the energy injected by dmas into the gas ) we have run two extra sets of simulations : the first ( label : e ) in which feedback is enhanced , and the second ( label : n ) with dm - induced dissociations switched off . in the e - models , we have assumed that _ all _ dma energy input which goes into excitations is in the form of photons in the lyman - werner band ( @xmath104 ) , and that each of these photons dissociates one molecule , where @xmath105 is the baryon number density , @xmath26 is the energy input per baryon due to dmas , @xmath106 is the fraction of this energy which goes into excitations ( i.e. into photons with @xmath107 ) , and @xmath108 is the average energy of a lyman - werner photon . such rate is added to the one produced by the fraction the dma energy input which goes into ionizations . ] . this clearly represents an upper limit on the feedback upon , as in reality about half of the photons going into dissociations has energy below the lower limit of the lyman - werner band ; furthermore , in low density regions a significant fraction of lyman - werner photons can escape . as it can easily be noticed from the upper panel of figure [ fig : plot_t_fbk_kappa ] , ( showing the evolution of models m1 , e1 and l1 in the @xmath82 plane ) switching off the feedback results in a sensible change of the temperature only in the final stages of collapse ( because the absence of dma - induced dissociations strongly reduces the heating ) ; on the other hand , strengthening the feedback results in appreciable modification of the temperature of the cloud in earlier stages ( because it reduces the abundance ) , but the difference reduces drastically in later phases ( when the direct effects of dmas dominate the ones on chemistry ) . we also note that in models with @xmath109 the differences induced by the treatment of feedback are much smaller than in the models shown in fig . [ fig : plot_t_fbk_kappa ] ( with @xmath46 ) . we have also studied the effects of varying the gray opacity @xmath35 , by running sets of simulations in which we have increased / decreased the fiducial value @xmath35=0.01@xmath110 by one order of magnitude . at low and intermediate densities the effects on the evolution of the cloud are very similar to those of a decrease / increase of the particle mass @xmath22 ; whereas models with the same @xmath22 tend to converge at high densities . in fact , an opacity variation affects sensibly the properties of the gas only when it is of `` optically thin '' to the energy produced by dmas : in later stages of the collapse , ( i.e. in regions of the halo where the gas is `` optically thick '' to the dma energy ) , the energy absorption does not depend on @xmath35 . this can be appreciated from the lower panel of figure [ fig : plot_t_fbk_kappa ] , where it is clear that the modified opacity models ( h , l labeled ) overplot to the corresponding standard ( m labeled ) models at higher central densities . we can definitely infer that `` astrophysical '' ( and `` numerical '' ) parameter dependence is mostly degenerate with dm parameters , and in any case not drastically affecting the qualitative picture we have drawn so far . = 10@xmath111 ( black ) , 10@xmath112 ( blue ) , 10@xmath113 ( green ) , 10@xmath114 ( red ) ; the different stages can be easily distinguished through the position of their innermost point , as the lines referring to high - density stages extend to lower radii . in the upper left panel , the dotted lines refer to dm densities in the m100 case ( note that the dm density profile for the @xmath115 stage is very similar to the nfw profile which is assumed to form immediately after virialization ) . the geometrical markers indicate the position enclosing baryonic masses of 10@xmath116m@xmath117 ( hexagons ) , 10@xmath118m@xmath117 ( pentagons ) , 1m@xmath117 ( squares ) , 10@xmath119m@xmath117 ( triangles ) , 10@xmath120m@xmath117 ( circles ) , and 10@xmath121m@xmath117 ( diamonds ) ; see text for comments.,scaledwidth=48.0% ] figure [ fig : profilo_r ] shows the spatial profile of gas properties at four different stages ( corresponding to central shell gas densities @xmath122 , @xmath123 , @xmath124 , and @xmath125 ) in the m100 fiducial case and , for comparison , in the nodm control run . the top - left panel shows also the dm density in the case of m100 run at the same stages ( dm density stays constant and equal to the initial profile in the nodm control run ) . the first important remark to file is that the gas density profiles are virtually unchanged by the presence of dmas : such profiles are indistinguishable until central gas densities of @xmath12610@xmath127 are reached , and even then the discrepancies are of minor entity ( e.g. , the slope outside the core becomes @xmath128 rather than @xmath129 ) . we can conclude that dmas do not alter the self - similar phase of the collapse ( larson 1969 , penston 1969 ) , at least for run m100 and for @xmath130 . the evolution of the dm density profile is in good agreement with previous studies , see for instance fig . 1 in sfg08 , and fig . 1 in i08 . it is also relevant to notice that ( as could be expected ) the differences in infall velocity , temperature profile and fraction affect only the central @xmath131 of gas ( i.e. the gas within the hexagonal marks ) . actually , relevant ( although yet not dramatic ) differences are limited to the inner @xmath132 ( region within the pentagonal marks ) : the overall properties of the @xmath9 halo are conserved in presence of dmas . the lower panels visualize the changes in temperature ( lower left ) following the modification of fraction ( lower right ) which we have extensively commented in the previous sections . here we only note that the molecular core is significantly more extended than in the control run : the region where is the dominant chemical species encloses @xmath133 , rather than @xmath134 ; furthermore , the fall in abundance is less steep than in the case without dmas . the upper right panel is somewhat intriguing , as it gives a glimpse of the modification of the dynamical properties of the collapse : while there are very little changes in the central region ( say , within the square marks , corresponding to a @xmath135 enclosed mass ) and in the outskirts of the halo ( at enclosed masses @xmath136 , i.e. outside a point roughly mid - way between the pentagonal and hexagonal marks ) , the infall velocity of the intermediate - mass gas shells is slowed down . the reduced temperature of the cloud ( due to the more efficient cooling which we discussed in the previous sections ; see also the bottom - right panel of fig . [ fig : profilo_r ] ) is likely responsible for this change . in fact , it is well known ( see e.g. stahler , palla & salpeter 1986 , and references therein ) that during the self - similar evolution , the mass infall rate @xmath137 in the region where the density falls as a power law with index @xmath84 should roughly scale with @xmath138 , where @xmath139 is the isothermal sound speed . since @xmath140 is practically the same in the m100 and the nodm run , the reduced temperature should result in a reduced infall velocity , similar to what we obtain . it is worth noting that on the contrary , the collapse of the innermost gas shell is _ accelerated _ , albeit very slightly , by the presence of dm , until central densities of 10@xmath141(/10@xmath142 ) for the m100(/m1 ) case , see table [ tab : collapsetime ] . however , such infall time alteration are of very small entity , and in any case negligible with respect to the total collapse time of the order of @xmath143 . [ tab : collapsetime ] lccc log ( @xmath44[@xmath63 ) & nodm & m100 & m1 + 23 & 2.117@xmath3910@xmath144 & 2.116@xmath3910@xmath144 & 2.107@xmath3910@xmath144 + 34 & 8.382@xmath3910@xmath111 & 8.326@xmath3910@xmath111 & 8.077@xmath3910@xmath111 + 45 & 3.204@xmath3910@xmath111 & 3.200@xmath3910@xmath111 & 3.401@xmath3910@xmath111 + 56 & 1.158@xmath3910@xmath111 & 1.177@xmath3910@xmath111 & 1.190@xmath3910@xmath111 + 67 & 3.930@xmath3910@xmath116 & 4.026@xmath3910@xmath116 & 3.895@xmath3910@xmath116 + 78 & 1.261@xmath3910@xmath116 & 1.250@xmath3910@xmath116 & 1.221@xmath3910@xmath116 + 89 & 3.927@xmath3910@xmath145 & 3.757@xmath3910@xmath145 & 3.544@xmath3910@xmath145 + 910 & 1.210@xmath3910@xmath145 & 1.118@xmath3910@xmath145 & 1.072@xmath3910@xmath145 + 1011 & 366.4 & 333.3 & 333.7 + 1112 & 103.2 & 102.0 & 111.3 + 1213 & 29.42 & 31.93 & 31.05 + 1314 & 9.210 & 9.480 & + but comparing runs m1 ( maximal ) and nodm ( control ) . solid lines refer to gas quantities in the m1 run , dashed lines to gas quantities in the nodm control run . all the symbols and lines are the same as in fig . [ fig : profilo_r ] , apart from the fact that the red lines refer to a central gas density @xmath44=10@xmath146 ( rather than 10@xmath114 ) ; see text for comments.,scaledwidth=48.0% ] figure [ fig : profilo_rmax ] , analogue to fig . [ fig : profilo_r ] , compares the control run to the maximal run , rather than to the fiducial run ; in this case the highest central density we plot is @xmath147 ( rather than @xmath125 ) , because the maximal run was stopped before reaching @xmath98 . the top - left panel shows that even in this case dmas do not alter the self - similar part of the collapse ; however , it should be remarked that the highest density stage is starting to deviate from the expected self - similar profile : in fact , the density profile within the core is not flat , but there is a maximum at the edge of the core . this is more clear when we examine the other panels : the lines referring to the three stages with @xmath148 are qualitatively similar to their counterparts in fig . [ fig : profilo_r ] ; but the lines representing the highest density stage are clearly different , and we will focus on them . in the lower panels we can see a huge drop in the abundance near the centre of the protostar , and a corresponding discontinuity in the gas temperature ; both could be expected by our discussion in the previous sections , even if here we get a further piece of information about the size of the region where the was dissociated ( @xmath149 , enclosed mass @xmath150 ) . also the velocity panel is affected : this is the only case where the infall velocities of the run with dmas significantly exceed those in the control run in an extended region ; furthermore , we can see that there are secondary peaks in velocity at low radii ( in particular one is coincident with the edge of the region where was dissociated ) , which are likely to result from the propagation of the density wave which was generated by the re - expansion of the core which occurred when @xmath93 . all these properties indicate that the core of this object is approaching hydrostatical equilibrium , even if it is not quite there . it is interesting to notice that the innermost regions , of enclosed mass between 1m@xmath117 and 10m@xmath117 , are infalling with higher velocity than in the nodm case , whereas between 10m@xmath117 and 10@xmath145m@xmath117 the collapse is slowed , when central density is @xmath44=10@xmath151 . again , this is equivalent to what we have observed for the fiducial run , and shown in table [ tab : collapsetime ] , and the feature only becomes more pronounced in the case of the maximal run . we have studied the effects of self - annihilating dark matter on the collapse of the gas structures harboring the formation of the first stars , known as pop iii . for the first time in the literature we follow self - consistently the evolution of the dark matter profile as a consequence of the gravitational drag of the collapsing gas and include the feedback of energy injection by dm annihilations on the chemistry of the gas , in the yet unexplored regime between the virialization of the halo and the formation of a hydrostatic core . we have explored the effects of dmas by spanning a range of masses and annihilation cross sections around the values of the vanilla wimp scenario , namely those able to reproduce the relic abundance of dm with a thermal decoupling , finding similar results but with variations in the details and onset times of the different phases that we are to summarize . in the following , quoted numbers refer to our fiducial case ( @xmath152 , @xmath153 , @xmath154 ) . as expected , heavier wimps ( or smaller self - annihilation rates , or lower opacity since the energy injection term is proportional to @xmath21/m@xmath155 and depends also on @xmath35 ) effects become relevant at later times during the collapse , the same dma rate being achieved at higher dm densities . lighter particles ( or higher @xmath21 , or higher @xmath35 ) produce the same effects at earlier times ( and therefore smaller gas densities ) . this general scenario is robust with respect to parameter variations within physically acceptable ranges . a few key points of our study are worth some emphasis : _ ( i ) _ as known from existing literature ( rmf07 ) , before the virialization of the halo , at gas density @xmath156 ( and therefore extremely small gas opacity ) dmas do not sensibly affect any gas process ; _ ( ii ) _ between halo virialization and a gas density @xmath157 dmas contribute mainly through indirect feedback effects : the free electron floor created by the ionizations induced by dmas catalyzes formation . in turn , molecular provides more cooling to the cloud than in the standard case ( without dmas ) and the temperature of the cloud _ decreases _ as a consequence of dmas ; _ ( iii ) _ finally , at @xmath158 the dma heating rate becomes equal to the gas cooling rate . to a first approximation this leads to a balance between losses and gain , which we dubbed as the `` critical point '' . our results generally agree with previous ones : as the semi - analytical estimates of sfg08 and the analytical study ( based on simulation data ) from natarajan , tan and o shea ( 2009 ) , we confirm the existence of a `` critical point '' . we also find that the equality of cooling vs dma heating terms takes place ( in the innermost shell ) at central densities of approximately @xmath85 , details depending on parameters ( dm mass , opacity coefficient etc . see previous section , and following discussion ) , in agreement with the above mentioned studies . this is particularly relevant as our analysis is the first fully numerical , self - consistent study including a reasonably accurate treatment of radiative transfer allowing to reach the critical point and beyond it in at least three cases . it had been previously suggested ( sfg08 ) that , after reaching of the `` critical point '' , the collapse would stop and the whole structure stall , thus generating a new type of celestial object . with our numerical simulations we have accessed this stage of the collapse , finding that the system does _ not _ halt its collapse in three cases out of four ; in the only case where the collapse stops , this happens far after the critical point was reached , and its duration is very short ( @xmath159 ) after which the gas restarts contracting . moreover , the dm parameters for which the astrophysical system finds important changes after the critical point are actually strongly disfavored by dm constraints based on local and primordial universe observations . by changing the dm mass or self - annihilation rate , the scenario we have described does not change qualitatively , within the range of values studied and the physical regime we have accessed with our simulations . while stressing again that when our objects reach the critical point the structure does _ not _ halt the collapse , and instead it continues its evolution toward the formation of a protostellar core , it is also useful to comment on the alteration of the dynamics of the collapse induced by dmas . the duration of the gas collapse from halo virialization down to densities of @xmath160 , in presence of dmas appears to be _ shorter _ ( especially in the first phases ) than in the standard case ( by about 1 per cent ) of the total collapse time . however we point out that the change is a small fraction of the total , and that the effects of dmas on collapse time are very difficult to quantify . in fact , the dmas effect can indeed accelerate the collapse in some shells , and decelerate it in others , thus making it difficult to obtain a homogeneous definition of time delay . for example , the reduction of the infall velocity of shells enclosing @xmath132 implies that the collapse time of these shells is longer ( by 1020 per cent ) than in the control ( nodm ) case . we feel confident in stressing , however , that the shells which are mostly affected by the infall time decrease ( @xmath7020% of the infall velocity ) are placed between 10@xmath118m@xmath117 and 10@xmath145m@xmath117 , and would become part of a hydrostatic protostar only in the final stages of the collapse . it is not clear whether these shells will eventually end up in the hydrostatic core ( omukai & palla 2003 ) , however it is reasonable to expect that such modification will not alter the _ total _ time for the formation of the star by more than @xmath7020 per cent . even if our runs could not reach the regime when a hydrostatic core forms ( with the partial exception of the maximal run , where we probably got close ) , it is worth to examine the likely consequences of our results for the further evolution of the protostar in presence of dmas . we start reminding that in the standard case ( without dmas ) the simulations of r02 show that when the protostar becomes opaque to continuum radiation ( at @xmath161 ) , the formation of the hydrostatic core is delayed by the thermostatic effects of dissociations , so that the core actually forms when @xmath162 . however , at the end of our runs the central abundance in the maximal and sub - maximal runs is @xmath163 ; even in the fiducial run , the abundance , while still high , is decreasing much earlier than in the standard case ; this is very different from the results of r02 , where @xmath164 up to @xmath165 . because of the lower amount of h@xmath34 , it is reasonable to expect that in the cases with dmas the delaying effects of dissociation are absent , or smaller than in the standard case . then , the hydrostatic core would form at lower densities ( e.g. , @xmath166 , if the protostar becomes optically thick to continuum radiation at the same density as in the r02 runs ) , and its initial mass would be larger ( probably in the 0.010.1 @xmath167 range , rather than @xmath168 ) . however , we point out that r02 found that in the case with no dmas the mass of the hydrostatic core grows very fast , reaching @xmath169 in less than 3 years ( see e.g. their fig . 6 ) : then , the difference in the initial size is relatively unimportant . it is probably more relevant to note that the lower temperatures and infall velocities of the layers outside the hydrostatic core imply that in the cases with dmas the accretion rate might be slower than what was found by r02 . we have studied the effects of wimp dark matter annihilations ( dmas ) on the evolution of primordial gas clouds hosting the first stars . we have followed the collapse of gas and dm within a @xmath170 halo virializing at redshift @xmath171 , from @xmath13 to slightly before the formation of a hydrostatic core , properly including gas heating / cooling and chemistry processes induced by dmas , and exploring the dependency of the results on different parameters ( dm particle mass , self - annihilation cross section , gas opacity , feedback strength ) . independently of such parameters , when the central baryon density , @xmath44 , is lower than the critical density , @xmath172 , corresponding to a model - dependent balance between dma energy input and gas cooling rate , dma ionizations catalyze an increase in the abundance by a factor @xmath99 . the increased cooling moderately reduces the temperature ( by @xmath173 per cent ) but does not significantly reduce the fragmentation mass scale . for @xmath174 , the dma energy injection exceeds the cooling , with the excess heat mainly going into dissociations . in the presence of dma the transition to the continuum dominated cooling regime occurs earlier and generally is not associated with abrupt temperature variations . in conclusion , no significant differences are found with respect to the case without dmas ; in particular , and contrary to previous claims , the collapse does not stall and the cloud keeps contracting even when @xmath175 . our simulations stop at central densities @xmath176 , and can not follow the hydrostatic core formation , nor its accretion . at the final simulation stage , the lower temperature / infall velocity of the layers enclosing a mass of @xmath177 suggest that dmas might lead to slightly longer stellar formation timescales , with a possible @xmath178 per cent increase over models without dmas . the latter finding strengthens our conclusions , although the final answer will come from numerical simulations ( hopefully also three - dimensional ) able to address this very same problem in the yet unexplored density regime @xmath179 . f. i. acknowledges support from miur through grant prin-2006 during the early stages of this work , and from the european community research program fp7/2007/2013 within the framework of convention # 235878 . a. b. and p. m. acknowledge financial contribution from contract asi i/016/07/0 . we thank m. mapelli for stimulating discussions . 99 abel t. , bryan g. l. , norman m. l. , 2002 , science , 295 , 93 ascasibar y. , 2007 , a&a , 462 , l65 bertone g. , hooper d. , silk j. , 2005 , phys . , 405 , 279 blumenthal g. r. , faber s. m. , flores r. , primack j. r. , 1986 , apj , 301 , 27 casanellas j. , lopes , i. , 2009 , apj , 705 , 305 chen x. , kamionkowski m. , 2004 , phys . d , 70 , 043502 chuzhoy l. , 2008 , apj , 679 , l65 cirelli m. , iocco f. , panci p. , 2009 , jcap , 009 , 0910 fairbairn m. , scott p. , edsjo j. , 2008 , phys . d , 77 , 047301 freese k. , spolyar d. , aguirre a. , 2008 , jcap , 0811 , 014 [ freese et al . ( 2008a ) ] freese k. , bodenheimer p. a. , spolyar d. , gondolo p. , 2008 , apj , 685 , l101 [ freese et al . ( 2008b ) ] freese k. , spolyar d. , bodenheimer p. a. , gondolo p. , 2009 , new j. phys . , 105014 furlanetto s. r. , johnson stoever s. , 2010 , mnras , 404 , 1869 furlanetto s. r. , oh s. p. , pierpaoli e. , 2006 , phys . rev . d. , 74 , 103502 galli d. , palla f. , 1998 , a.&a , 335 , 403 galli s. , iocco f. , bertone g. , melchiorri a. , 2009 , phys . d. , 80 , 023505 glover s. c. o , jappsen a .- k . , 2007 , apj , 666 , 1 gnedin o. y. , kravtsov a. v. , klypin a. a. , nagai d. , 2004 , apj , 616 , 16 hollenbach d. , mckee c. f. , 1979 , apjs , 41 , 555 iocco f. , 2008 , apj , 677 , l1 iocco f. , bressan a. , ripamonti e. , schneider r. , ferrara a. , marigo p. , 2008 , mnras , 390 , 1655 [ i08 ] larson r. b. , 1969 , mnras , 145 , 271 lattanzi m. , silk j. , 2009 , phys . rev . d , 79 , 083523 lenzuni p. , chernoff d. f. , salpeter e. e. , 1991 , apjs , 76 , 759 mapelli m. , ferrara a. , 2005 , mnras , 364 , 2 mapelli m. , ferrara a. , pierpaoli e. , 2006 , mnras , 369 , 1719 marigo p. , aringer b. , 2009 , a&a , 508 , 1539 myers z. , nusser a. , 2008 , mnras , 384 , 727 natarajan a. , schwarz d. j. , 2008 , phys . d. , 78 , 103524 natarajan a. , tan j. c. , oshea b. w. , 2009 , apj , 692 , 574 navarro j. f. , frenk c. s. , white s. d. m. , 1996 , apj , 462 , 563 omukai k. , nishi r. 1998 , apj , 508 , 141 omukai k. , palla f. , 2003 , apj , 589 , 677 padmanabhan n. , finkbeiner d. p. , 2005 d , 72 , 023508 palla f. , salpeter e. e. , stahler s. w. , 1983 , apj , 271 , 632 penston m. v. , 1969 , mnras , 144 , 425 ripamonti e. , haardt f. , ferrara a. , colpi m. , 2002 , mnras , 334 , 401 [ r02 ] ripamonti e. , 2007 , mnras , 376 , 709 ripamonti e. , mapelli m. , ferrara a. , 2007a , mnras , 374 , 1067 ripamonti e. , mapelli m. , ferrara a. , 2007b , mnras , 375 , 1399 [ rmf07 ] ripamonti e. , iocco f. , bressan a. , schneider r. , ferrara a. , marigo p. , 2009 , proceedings of science idm08 , 75 ( arxiv:0903.0346 ) scott p. , edsjo j. , fairbairn m. , 2009 , mnras , 394 , 82 shchekinov y. a. , vasiliev e. o. , 2007 , mnras , 379 , 1003 stahler s. w. , palla f. , salpeter e. e. , 1986 , apj , 302 , 590 shull j. m. , van steenberg m. e. , 1985 , apj , 298 , 268 spolyar d. , freese k. , gondolo p. , 2008 lett . , 100 , 051101 [ sfg08 ] spolyar d. , bodenheimer p. a. , freese k. , gondolo p. , 2009 , apj , 705 , 1031 stasielak j. , biermann p. l. , kusenko a. , 2007 , apj , 654 , 290 steigman g. , sarazin c. l. , quintana h. , faulkner j. , 1978 , aj , 83 , 1050 taoso m , bertone g. , meynet g. , ekstrom s. , 2008 , phys . d , 78 , 123510 valds m. , ferrara a. , 2008 , mnras , 387 , l8 valds m. , ferrara a. , mapelli m. , ripamonti e. , 2007 , mnras , 377 , 245 yoon s. c. , iocco f. , akiyama s. , 2008 , apj , 688 , l1 yoshida n. , omukai k. , hernquist l. , abel t. , 2006 , apj , 652 , 6 zhang l. , chen x. , lei y .- a . , si z .- g . , 2006 , phys rev . d , 74 , 103519 one of the hypothesis underlying the ac formalism ( as derived by blumenthal et al . 1986 ) is that the orbital time of dm particles should be short when compared to the collapse time of the baryons . in the case we are studying , this is a problem : in fact , the baryonic collapse happens essentially on a free - fall time - scale , and the dm orbital times can not be shorter than that . the case of a `` fast collapse '' ( where the baryon collapse time is similar to the free - fall time ) was studied by steigman et al . ( 1978 ; see their sec . iiib ) , in the assumptions that i ) the baryons stand still at @xmath180 ( so that they will collapse in free - fall ) , ii ) the dm contribution to the total mass @xmath181 is small , and iii ) dm particles have a dispersion velocity @xmath182 . the case can be treated analytically , and steigman et al . ( 1978 ) obtain that if the radius of the sphere enclosing all the baryons goes from @xmath183 ( at time @xmath180 ) to @xmath184 ( at time @xmath185 ) , the ratio of the number @xmath186 of dm particles within @xmath184 ( at @xmath187 ) to the number @xmath188 of dm particles within @xmath183 ( at @xmath180 ) is roughly proportional to @xmath189 ( cfr . 31 and 32 of steigman et al . 1978 rather than @xmath190 , since this is the value that can be inferred from the more general eq . 31 . ] ) , at least in the limit of @xmath191 . then , the density @xmath192 of dm particles should grow asymptotically as @xmath193 , or as @xmath194 . we can compare this behaviour to the dependence of the central dm density ( @xmath195 ) upon the central baryonic density ( @xmath196 ) which can be obtained from the ac formalism : spg08 find that @xmath197 , and our calculations essentially confirm their result . then , it is clear that the `` fast collapse '' slope is significantly flatter than what can be inferred from the application ot the ac formalism . however , there are some extra facts that need to be kept into account . * for central baryonic number densities @xmath198 the dm density is larger than the baryonic one , and the application of the ac does not modify the dm density profile ; * the `` fast collapse '' slope was obtained with the assumption of an uniform density profile of both the baryons and the dm , whereas the nfw profile of our initial dm profile is strongly peaked : then , the calculations of @xmath199 by steigman et al . ( 1978 ) likely underestimate the value appropriate in our case ; as a consequence , it is likely that the slope of the @xmath200 relation is higher than 0.5 ; * the dm profile seen in the simulations of abel et al . 2002 appears to be more consistent with the ac predictions than with those of the `` fast collapse '' model : in fact , the innermost dm density shown in their fig . 2 is a factor @xmath201 lower than the baryonic density @xmath202 : this is somewhat consistent with the difference by a factor @xmath203^{1 - 0.81}\sim 6 $ ] predicted by ac , but quite smaller than the difference by a factor @xmath203^{1 - 0.81}\sim 100 $ ] predicted using the `` fast collapse '' . then , it appears that the ac formalism can be safely used at least for densities up to @xmath204 . it is quite possible ( and even likely ) that at higher density it will significantly overestimate the dm density . however , such overestimate will be large ( a maximum @xmath205 for @xmath206 ) but still reasonable in the range of densities where we perform our simulations . furthermore , we remark that the ac approach allows a better comparison with previous studies , and , given our conclusions , is the _ conservative _ choice .

we propose the autocorrelator of conductance peak heights as a signature of the underlying chaotic dynamics in quantum dots in the coulomb blockade regime . this correlation function is directly accessible to experiments and its decay width contains interesting information about the underlying electron dynamics . analytical results are derived in the framework of random matrix theory in the regime of broken time - reversal symmetry . the final expression , upon rescaling , becomes independent of the details of the system . for the situation when the external parameter is a variable magnetic field , the system - dependent , nonuniversal field scaling factor is obtained by a semiclassical approach . the validity of our findings is confirmed by a comparison with results of an exact numerical diagonalization of the conformal billiard threaded by a magnetic flux line .

cond-mat9510118

during the last five years the fabrication of ballistic two - dimensional electron gas microstructures in gaas / ga@xmath0al@xmath1as heterostructures has opened up a new field : the study of quantum manifestations of classical chaos in condensed matter systems.@xcite the first theoretical discussion@xcite of the so - called quantum chaos in microstructures dealt with open systems , i.e. , cavities on the micrometer scale or smaller connected to the external world by leads that admit a continuous flow of electrons . several successful experiments followed.@xcite nowadays the theoretical study of quantum chaos is also well developed for closed microstructures , known as quantum dots.@xcite these are cavities coupled to the external leads by tunnel barriers that prevent a continuous flow of electrons and make the electric charge inside the dot quantized . the striking characteristic of quantum dots is the appearance of very sharp peaks in the conductance as a function of the gate voltage . these peaks are roughly equidistant due to the strong charging effect , but their heights vary randomly by an order of magnitude or more . the general interest in this phenomenon , combined with the existence of interesting experimental data on quantum chaos for closed systems,@xcite motivated several theoretical studies.@xcite these works proposed universal conductance peak distributions for quantum dots in the coulomb blockade regime as fingerprints of the underlying chaotic dynamics . similar predictions for open systems@xcite have been difficult to observe experimentally@xcite because of the strong influence of dephasing.@xcite the initial experiments on coulomb blockade peak heights in quantum dots@xcite have paved the way for the first convincing measurements of statistical distributions of peak heights.@xcite however , the full distribution of the conductance peaks is not a trivial quantity to measure . since each peak sequence is not very long , good statistics demands great experimental effort . one possible solution is a true ensemble averaging , which requires many different samples . in ref . it was suggested that one could obtain such an average from a single device by changing its shape _ in situ _ using external electrostatic potentials . this difficult procedure has now been tested experimentally in open@xcite and closed systems.@xcite another seemingly simpler solution is to generate more statistics by varying an applied magnetic field ( taking care to measure at @xmath2-field steps larger than the conductance autocorrelation width @xmath3).@xcite in order to test more thoroughly quantum manifestations of classical chaos in quantum dots in the tunneling regime we propose a different quantity to be investigated : the parametric conductance peak autocorrelation function . because it is not difficult to cover a parametric range which extends over several autocorrelation lengths , correlation functions need a small number of peaks and should be easy to obtain experimentally . moreover , correlators of a fixed eigenstate are also of theoretical interest . analytical calculations using random matrix theory beyond a perturbative approach seem to be very challenging . in addition , the nonuniversal scaling contains interesting information about the underlying classical dynamics . to put the parametric dependence in contact with physical quantities and develop a satisfactory theoretical treatment of nonuniversal scaling factors , one has to invoke semiclassical arguments . in this sense the study of the autocorrelator of conductance peaks is an example where semiclassical and random matrix theories complement each other in the understanding of mesoscopic phenomena . the first step towards the description of the main features of chaos in quantum dots is to consider the problem of electrons moving in a cavity . we assume that the coulomb interaction within the cavity can be taken into account in a self - consistent way , yielding noninteracting quasiparticles . as a result of extensive studies of two - dimensional ( 2d ) dynamical systems , one has learned that when the classical motion in the cavity is hyperbolic , the quantum spectrum and the wave functions exhibit certain universal features.@xcite throughout this paper we shall consider a statistical approach which is valid for chaotic systems and also works fairly well for systems whose phase space is predominantly chaotic.@xcite the paper is organized as follows . in sec . [ sec : stochastic ] we present our analytical and numerical treatment based on random matrix theory ( rmt ) . we concentrate on the case of broken time - reversal symmetry ( tr ) , but also show numerical results for the tr preserved case ( spin - orbit interactions are always neglected ) . in sec . [ sec : dynamical ] we study a dynamical model to illustrate our findings . there we conjecture that the semiclassical periodic orbit theory can be used to estimate the typical magnetic field scale of any correlator of a quantum dot in a realistic situation . we conclude in sec . [ sec : conclusions ] with a discussion on how the presented results are robust with respect to dephasing effects and on possible experimental realizations . several more technical details are left to the appendices . one of the basic tools to investigate the conductance in mesoscopic devices is the landauer - bttiker formula.@xcite for a two - lead geometry , where leads are denoted by @xmath4 ( left ) and @xmath5 ( right ) , the conductance @xmath6 is given by @xmath7 where the sum runs over the channels @xmath8 on the lead @xmath4 and the channels @xmath9 on the lead @xmath5 . the scattering matrix @xmath10 connects right and left channels and is evaluated at the energy @xmath11 and magnetic flux @xmath12 . the factor of 2 is due to spin degeneracy . in the case of quantum dots in the tunneling regime this formula becomes particularly simple , since the @xmath13 matrix can be written as @xmath14 where the matrix elements @xmath15 give the probability amplitude of a state in channel @xmath16 to couple to the resonance @xmath17 in the dot . the total decay width is given by @xmath18 . we shall assume that the contribution of direct processes is very small and neglect the off - resonance term @xmath19 . equation ( [ eq : smatrix ] ) is a very good approximation to the @xmath13 matrix when the average decay width @xmath20 is much smaller than the mean spacing between resonant states @xmath21 . this is indeed the case for the isolated resonances observed in quantum dot experiments . ( hereafter we use @xmath21 to denote the level spacing caused solely by single - particle states within the cavity , as opposed to the spacing between conductance peaks observed in the experiments . ) when the condition @xmath22 is not met , unitarity corrections to eq . ( [ eq : smatrix ] ) become important and they demand a different parametrization of the @xmath13 matrix ( see , for example , ref . ) . the matrix elements in eq . ( [ eq : smatrix ] ) can be obtained by invoking the @xmath5-matrix formalism.@xcite such an approach was originally proposed in the study of nuclear resonance scattering and has more recently been successfully applied in the context of mesoscopic physics.@xcite the partial decay amplitude @xmath23 is given by @xmath24 where @xmath25 is the eigenfunction of the resonance @xmath17 _ inside _ the dot with the appropriate boundary condition and @xmath26 is the wave function in the channel @xmath27 at energy @xmath28 . the integration is done over the contact region between lead and cavity . a useful quantity for the forthcoming discussion is the partial decay width , defined as latexmath:[$\gamma_{c\nu } = not contain barrier penetration factors . throughout this paper we will assume that penetration factors are channel independent and therefore will influence only the average decay width @xmath20 , but not its distribution . this approximation is certainly valid in the case of a single open channel in each lead , the case for which we specialize our analytical and numerical results . when the constrictions connecting leads to the cavity are smaller than one electron wavelength but the cross section in the leads is not , strong correlations among partial decay amplitudes at a fixed resonance exist.@xcite for this general situation it is still possible to incorporate different penetration factors and the effect of correlations into our results , although we do not believe that the final expressions could be cast in a simple analytical form . nevertheless , one would need to obtain all parameters either from the experiment , or from a satisfactory dynamical model that comprises both cavity and leads . finite - temperature averaging caused by the rounding of the fermi distribution can be done straightforwardly . in the limit @xmath30 the conductance peak corresponding to an _ on - resonance _ measurement is given by@xcite @xmath31 where @xmath32 is the decay width for the resonance @xmath17 into open channels in the @xmath33 lead . in other words , @xmath34 . equation ( [ eq : peakcond ] ) does not take into account thermal activation effects . however , one should have in mind that in the tunneling regime very small values of the partial decay widths may occur relatively close in energy to large ones . in addition to that , the single - particle spacing fluctuates and can reach values smaller than @xmath21 . as a result , even at low temperatures ( less than 100 mk ) a very short peak may be substantially influenced by neighboring resonances . this effect causes the conductance peak to cross over to a different regime , where the height becomes proportional to @xmath35 rather than @xmath36 ( even though the inequality @xmath37 is still approximately satisfied).@xcite we will avoid further consideration of this or any other similar effect by restricting our analysis to peaks whose heights have exclusively the characteristic @xmath36 behavior . in what follows we are going to evaluate the autocorrelator of conductance peak heights in terms of an external parameter @xmath38 which acts on the system hamiltonian . ( at this point we do not need to specify @xmath38 , which could be , for instance , a shape variable or a magnetic flux . ) in order to obtain the autocorrelator we use a statistical approach that shall be described in detail in the next subsection . unfortunately , a complete analytical solution for this problem is not yet available and presently seems to be a formidable task . therefore our analytical treatment is based on a perturbative expansion for small values of @xmath38 . we resort to numerical diagonalizations of large random matrices to conjecture the complete form of the autocorrelator for all other values of @xmath38 . we also derive an exact analytical expression for the joint probability distribution of conductance peak heights and their first parametric derivative . this distribution , besides being of theoretical interest by itself , can also be employed alternatively in the calculation of the small-@xmath38 asymptotic limit of the conductance peak correlator . before proceeding , we first present some basic quantities entering the calculation . since we are only considering two - point functions , the knowledge of the partial decay widths at two parameter space points , @xmath39 and @xmath40 , is required . we will need to specify @xmath41 only up to first - order terms in an expansion around @xmath42 , namely , @xmath43 where @xmath44 . the coefficient @xmath45 is given by first - order perturbation theory in terms of eigenfunctions , eigenvalues , and matrix elements of the perturbation . assuming that the hamiltonian may be written in the form @xmath46 and using eq . ( [ eq : gamma ] ) , we find that @xmath47 with @xmath48 and all partial decay amplitudes , eigenvalues , and matrix elements evaluated at @xmath39 . so far no statistical assumption has been made . the starting point of our discussion of the conductance peak autocorrelation function is eq . ( [ eq : peakcond ] ) . in this subsection we shall mainly treat the situation of tr broken by an external magnetic field . we identify the magnetic flux with the previously introduced variable @xmath38 . in the spirit of rmt,@xcite the analysis that follows will lead to a universal curve for the conductance peak autocorrelation , which is expressed in terms of a scaled @xmath38 and the coupling of the dot to the external leads , denoted by the parameters @xmath49 . both @xmath49 and the typical scale of @xmath38 are system - specific quantities , depending on the dot shape , the quality of the dot - lead couplings , and the fermi energy . ideally , the scaling of @xmath38 should be extracted from the experimental resonance ( level ) velocity correlator.@xcite when this information is not available , it is not straightforward to put @xmath38 in correspondence with experimental parameters . to overcome this problem we invoke a dynamical system and shall discuss it at length in sec . similarly , to obtain @xmath49 one needs to model specific features of the tunneling barriers . we avoid this procedure by expressing our results in terms of the mean decay width @xmath20 , which then becomes a fitting parameter in comparisons with experiments . the autocorrelator of peak heights is defined as @xmath50 where @xmath51 denotes averages over the resonances @xmath17 and over different values of @xmath52 , but will later be interpreted as an average over the gaussian unitary ensemble ( gue ) of random hamiltonians.@xcite the difficulties involved in the exact calculation of @xmath53 become clear once we expand the conductance peak height in powers of @xmath38 and assume translation invariance ( independence of @xmath52 ) , @xmath54 ^ 2 \right\rangle - \langle g_\nu(\bar{x } ) \rangle^2 . \label{eq : expcx}\ ] ] ( notice that odd powers disappear since the autocorrelator is by construction an even function of @xmath38 . ) as presented in eq . ( [ eq : expcx ] ) , the task of finding the complete functional form of @xmath53 requires not just the knowledge of the second moment of all derivatives of @xmath55 , but also the summation of the taylor series . presently we do not know of any other alternative method that could lead to the _ complete _ analytical form of @xmath53 . an analogous situation is encountered in the calculation of the more commonly studied level velocity autocorrelator@xcite ( see the detailed discussion of ref . ) . the small @xmath38 asymptotic limit can , however , be evaluated exactly through standard random matrix methods . moreover , an approximate curve for all values of @xmath38 can be found through numerical simulations . we start with the small-@xmath38 asymptotics . for this purpose we keep in eq . ( [ eq : expcx ] ) only the two lowest order terms , @xmath56 with @xmath57 and @xmath58 now we introduce a statistical model : we assume that due to shape irregularities and the presence of a magnetic field , the hamiltonian can be modeled as a member of the gue.@xcite under this assumption , @xmath59 has already been obtained.@xcite for example , in the simplest case , where one has two equivalent uncorrelated open channels ( one in each lead ) , @xmath60 . the second - order coefficient requires a more complex treatment , since eq . ( [ eq : c0pp ] ) involves not simply eigenfunctions but eigenvalues as well . for gaussian ensembles , a great simplification is possible due to the statistical independence of matrix elements of the hamiltonian . this will allow us to divide the ensemble average into averages over eigenfunctions and eigenvalues . for the sake of simplicity , we will take the matrix elements of @xmath61 as gaussian distributed,@xcite with @xmath62 where @xmath63 is a measure of the perturbation strength and @xmath64 is the number of eigenstates ( resonances ) considered . it immediately follows that @xmath65 and @xmath66 with @xmath67 . up to this point the formulation of the problem is very generic and no restriction has been made either on the number of open channels or on their transmission coefficients . a quick inspection of eqs . ( [ eq : averlambda1 ] ) and ( [ eq : averlambda2 ] ) indicates that the solution of the most generic case ( multichannel leads ) is possible , but the algebra is very tedious and intricate . recent experimental realizations of quantum dots in the tunneling regime @xcite indicate that one typically has _ only one open channel _ in @xmath4 and @xmath5 . owing to that , we limit our calculations to the one - channel case . carrying out the average over @xmath61 by using eqs . ( [ eq : averlambda1 ] ) and ( [ eq : averlambda2 ] ) to simplify eq . ( [ eq : c0pp ] ) , we get @xmath68 \right\rangle_{\{\gamma,\varepsilon_\mu\}}. \label{eq : c0ppbis}\ ] ] as pointed out before , the independence of eigenvalues from eigenfunctions allows the factorization of the right - hand side of eq . ( [ eq : c0ppbis ] ) into two decoupled averages , namely , @xmath69 the first factor is given by @xmath70 \right\rangle_{\{\gamma_r,\gamma_l\ } } , \label{eq : bamp}\end{aligned}\ ] ] where the average @xmath71 is taken over both leads @xmath4 and @xmath5 and resonances @xmath17 and @xmath72 with @xmath73 ( terms with an odd number of distinct amplitudes drop out ) . the second factor is @xmath74 and it involves only an average over the eigenvalues . since the lead contacts are typically farther than one wavelength @xmath75 apart , the decay widths at @xmath5 and @xmath4 are assumed to fluctuate independently . hence , the calculation of @xmath76 requires the convolution of four @xmath77 distributions with two degrees of freedom,@xcite @xmath78 ( notice that we are considering a single open channel per contact and identical leads ) , and we obtain @xmath79 . the eigenvalue average in eq . ( [ eq : beig ] ) is more complicated to carry out and needs some additional assumption . because we are going to take the thermodynamic limit ( @xmath80 ) later on , effects due to variations in the density of states are negligible . consequently , we may place the reference eigenstate @xmath28 at the center of the spectrum and write @xmath81 where @xmath82 is the average density of states . before proceeding to evaluate this quantity in the rmt framework , let us first discuss an illustrative and extreme situation , the picket fence spectrum with spacing @xmath21 . in this limit the averaged sum simplifies to @xmath83 , which readily yields @xmath84 . because the gaussian ensembles hypothesis implies fluctuations in the spectrum , the actual value of @xmath85 for the gue should be larger but of the same order of magnitude . the average over eigenvalues in eq . ( [ eq : beig1 ] ) can be restated as an average over a spectral determinant , @xmath86 where @xmath87 is an @xmath88 gue matrix and the normalization constant is given by @xcite @xmath89 [ notice that @xmath90 is the mean level spacing at the center of the spectrum . ] in the limit of @xmath91 eq . ( [ eq : beig2 ] ) can be evaluated by the fermionic method @xcite ( see appendix [ sec : appendixa ] for details ) and one finds that @xmath92 as expected , the actual value of @xmath85 is larger , but still in fair agreement with the picket fence estimate . the next step is to rescale the perturbation @xmath38 to a dimensionless form , in such a way that all system - dependent parameters are eliminated . one possibility is to use @xmath93 as the rescaling parameter . as will become clear later on , we do not find this procedure very interesting from the physics viewpoint because this quantity can not be easily calculated given the underlying dynamical system . we rather follow an idea originally proposed by szafer , simons , and altshuler : recall that the perturbation strength @xmath63 ( the nonuniversal scale in the above calculations ) also appears in the level velocity correlator @xmath94,@xcite @xmath95 when evaluated at @xmath96 , namely , @xmath97 ^ 2\rangle_u}{\delta^2 } = \frac{\sigma^2}{n\delta^2 } \;. \label{eq : vrms}\ ] ] the statement implicit in the original works of refs . and is that the quantity @xmath98 sets the scale for _ any _ averaged parametric function @xmath99 , provided that the system dynamics is chaotic in the classical limit . in sec . [ sec : dynamical ] we will show that @xmath100 can be obtained by semiclassical arguments , once details of the confining geometry of the dot are known . therefore , in analogy to their analysis of level velocity correlator , we apply the rescalings @xmath101 and @xmath102 to arrive at the following universal ( dimensionless ) form : @xmath103 valid for @xmath104 . to extract the nonperturbative part of @xmath105 , as well as its @xmath106 asymptotic limit , we relied on a numerical simulation . we performed a series of exact diagonalizations of random matrices of the form @xmath107 , with @xmath108 and @xmath109 denoting two @xmath110 matrices drawn from the gue . this model for the parametric dependence is rather convenient for the simulations and later data analysis because it does not make the level density depend on @xmath38 , nor does it cause the eigenvalues of @xmath111 to drift with @xmath38 . it is helpful to think of @xmath112_{kl}$ ] as the matrix element of the hamiltonian in a discrete space representation . as a result , for the one - channel lead case we may simply equal decay widths to the renormalized ( @xmath113 ) wave function intensities at a given point , @xmath114 where @xmath115 is the site number and @xmath17 the eigenstate label . in this way , we were able to generate more than @xmath116 different configurations of the two - lead geometry , out of which only @xmath117 were used ( we stress that wave functions taken at different @xmath17 or @xmath115 are statistically independent when the size of the matrix is large enough ) . for each realization of @xmath108 and @xmath109 we varied @xmath38 in the interval @xmath118 $ ] and considered only the 100 central eigenstates in order to avoid having to unfold the spectrum . in total we have run 50 realizations . the final result is presented in fig . [ fig : guesimul ] . for comparison , we have also shown in the inset the result obtained for the decay width correlator , @xmath119 we remark that this correlator is not directly accessible to experiments in quantum dots ( recall that conductance peak widths are dominated by the thermal rounding of the fermi surface).@xcite here we have introduced it with the unique purpose of checking the reliability of the numerical simulations in the @xmath120 range . contrary to the situation for @xmath105 , both @xmath121 _ and _ @xmath120 asymptotic limits of the rescaled @xmath122 can be evaluated analytically . one finds that ( see appendix [ sec : appendixb ] ) @xmath123 the data obtained from the random matrix simulations indicate that the asymptotic tail of @xmath105 is well described by an @xmath124 law . in the light of eq . ( [ eq : corwwasymp ] ) , this is not very surprising : if one lead were more strongly coupled to the cavity than the other , say , @xmath125 , we would have that @xmath126 and therefore @xmath127 in leading order . we can not rigorously prove , though , that this asymptotic form is also exact when right and left leads are identical . to conclude this subsection , we briefly discuss the universality class of preserved tr without spin - orbit coupling ( the case when spin - orbit coupling is present , the symplectic ensemble , will not be discussed since it is not relevant to semiconductor heterostructures ) . the simplest experimental realization is a measurement of the evolution of conductance peak heights as a function of shape deformation in the absence of a magnetic field . the general approach is the same as above and we assume that the system hamiltonian can be modeled as a member of the gaussian orthogonal ensemble ( goe).@xcite however , the @xmath104 asymptotics of the conductance peak correlator is now more difficult to calculate . this is not a daunting problem because , as seen above , the numerical results reliably recover the correct behavior of the correlator for small values of @xmath128 . consequently , for the tr - preserved case we relied entirely on numerical simulations and did not attempt any analytical calculation . we used the same parametric dependence of @xmath111 , but this time drew @xmath108 and @xmath109 from the goe . all other steps were identical to the gue simulation . the resulting correlation functions ( after the proper rescalings ) are shown in fig . [ fig : goesimul ] . notice that the large-@xmath128 asymptotics of both @xmath105 and @xmath129 are well described by an @xmath124 law , as for the gue simulations . here , analogously to the gue , we do not know how to prove analytically that this decay is rigorously true for @xmath105 ; on the other hand , we do know that this power law decay is indeed exact for @xmath129.@xcite we emphasize that the most important characteristic of the tr - preserved , universal @xmath105 as compared to the tr - broken one is the larger decay width . the widths at half maximum height differ by approximately @xmath130 . the power spectra of conductance peak height oscillations [ the fourier transform of @xmath105 ] for both gue and goe are shown in fig . [ fig : power ] . notice that the behavior is exponential only over a small range of @xmath115 values . one then concludes that a lorentzian can only be used as a _ rough _ approximation to the exact curves . besides , a lorentzian can not accommodate simultaneously the small and large asymptotic limits of @xmath105 presented above in either tr - preserved or tr - broken cases . in this subsection we evaluate the joint probability distribution @xmath132 where @xmath133 and the hamiltonian belongs to the gue . this distribution , although not suitable for direct experimental investigations , allows one to easily evaluate higher moments of the conductance peak height and its first parametric derivative ; in particular , one can use @xmath134 to obtain the two first coefficients in the expansion of eq . ( [ eq : secndcx ] ) , @xmath135 and @xmath136 we begin by recalling eq . ( [ eq : lambda ] ) and introducing the coefficients @xmath137 into eq . ( [ eq : jointf1 ] ) , @xmath138 where @xmath139 \label{eq : mexp}\ ] ] taking the fourier transform of eq . ( [ eq : jointf2 ] ) with respect to @xmath140 , we get @xmath141 following the same assumptions of the previous subsection , we break up the ensemble average into four partial averages . recalling eqs . ( [ eq : averlambda1 ] ) and ( [ eq : averlambda2 ] ) , we first carry out the average over the perturbation @xmath61 : @xmath142 next , we evaluate the average over the decay widths @xmath143 for @xmath144 only , @xmath145^{-1}\;. \label{eq : jointf3}\ ] ] at this point , instead of also taking the average over the remaining decay widths , we consider first the average over the eigenvalues @xmath146 , @xmath147 . here we use the following relation , valid in the large-@xmath64 limit and for @xmath28 at the center of the spectrum:@xcite @xmath148^{-1}\right\rangle_{\{\varepsilon_\nu\ } } = { \widetilde b}(k)\ ; , \label{eq : eigaver}\ ] ] where @xmath149 is the fourier transform of @xmath150 [ the evaluation of the average in eq . ( [ eq : eigaver ] ) requires a generalization@xcite of the approach used to derive eq . ( [ eq : beig3 ] ) . a brief description is given in appendix [ sec : appendixc ] . ] after inserting ( [ eq : eigaver ] ) into eq . ( [ eq : jointf3 ] ) and inverse fourier transforming the result we get @xmath151 which is more conveniently expressed in the form @xmath152 the remaining average over the reduced widths @xmath153 and @xmath154 appears only in the evaluation of the bell - shaped function @xmath155 in fact , one can easily carry out one of the above integrations and arrive at @xmath156 it is possible to represent @xmath157 in terms of special functions , but we did not find it particularly clarifying and therefore do not do it here . the asymptotic limits follow directly from the integral representation shown above . for @xmath158 , we obtain @xmath159 whereas for a fixed @xmath160 , we have that @xmath161 notice that eqs . ( [ eq : jointf5 ] ) and ( [ eq : bellcurve2 ] ) together immediately lead to the same @xmath104 asymptotics for the peak height correlator shown in eq . ( [ eq : cxsmall ] ) . from eq . ( [ eq : hasympt2 ] ) it is obvious that @xmath162 diverges for @xmath163 . this can be ultimately related to the level repulsion present in the spectrum : strong anticrossings of levels can cause anomalously large variations of the conductance peak height as a function of the external parameter @xmath38 . when this happens , one finds that @xmath164 , where @xmath165 . since the probability of this event goes as @xmath166 for the unitary ensemble,@xcite one finds that @xmath167 for @xmath168 , in agreement with our exact calculation . the aim of this section is to compare the results of the previous section with exact numerical diagonalizations of a dynamical model . the essential characteristic of a dynamical model for this type of study is a fair resemblance to the actual experimental conditions , combined with its adequacy to numerical computations . for this reason we chose the ( two - dimensional ) conformal billiard penetrated by an infinitely thin aharonov - bohm flux line carrying a flux of @xmath12 . this model was originally introduced in ref . and later adopted in the study of statistical features of conductance in quantum dots.@xcite using complex coordinates , the shape of the billiard in the @xmath169 plane is given by @xmath170 in the following area preserving conformal mapping : @xmath171 where @xmath172 , @xmath27 , and @xmath173 are real parameters chosen in such a manner that @xmath174 for @xmath175 . the classical dynamics of a particle bouncing inside this billiard is predominantly stochastic and is unaffected by the presence of the flux line . to describe the flux line we chose the following gauge for the vector potential @xmath176 : @xmath177 where @xmath178 $ ] . this particular gauge , combined with neumann boundary conditions , permits a separation of the schrdinger equation into polar coordinates @xmath179 of the complex parameter @xmath180 . the eigenstates @xmath181 thus obtained correspond to the resonant wave function appearing in eq . ( [ eq : gamma ] ) . @xcite in our numerical treatment the wave function @xmath26 in the lead is equal to a transverse sine function multiplied by a longitudinal plane wave . in practical calculations,@xcite one fixes the value of the flux @xmath12 and uses as a truncated basis the lowest ( in our case 1000 ) eigenstates of the circular billiard ( @xmath182 ) . these have the form @xmath183 , where @xmath184 is the bessel function of fractional order @xmath17 and @xmath185 is the @xmath186th root of its derivative , @xmath187 . the dependence on the magnetic flux enters through @xmath188 . to solve the schrdinger equation one has to calculate several thousand matrix elements of the jacobian @xmath189 in this @xmath12-dependent basis this operation is very time consuming . changes in shape are not a major obstacle , since @xmath172 , @xmath27 , and @xmath173 act as prefactors to the matrix elements and no further calculation is necessary . ( for instance , in ref . only two different values of @xmath12 needed to be used . ) in the present work , however , we wanted to change the flux to simulate the simplest experimental setup and this required the use of 76 different values of @xmath12 . the spectrum is seen to be not only @xmath190 periodic in @xmath12 , but also symmetric around @xmath191 . for this reason and , furthermore , to avoid the special points 0 and 1/2 we let @xmath192 vary in the interval @xmath193 $ ] . to circumvent the overwhelming problem of too large amounts of computing time we employed the following strategy . we calculated the jacobian matrix @xmath194 for only seven different values of @xmath195 , namely , from 0.10 to 0.40 in steps of 0.04 . then , for any other value of @xmath196 , the matrix element @xmath197 was found by polynomial interpolation . this was checked to give a relative error of at most @xmath198 . to improve the statistics we also calculated the eigenstates for five different shapes by keeping @xmath199 and letting @xmath200 , with @xmath201 . the spectra corresponding to these shapes are statistically uncorrelated.@xcite due to the truncation of the basis only the lowest 300 of the calculated 1000 eigenstates were accurate enough to be used in the analysis . we also discarded the lowest 100 eigenstates because of their markedly nonuniversal behavior . we should emphasize that it is not a trivial task to increase the number of usable states . for the asymmetric conformal billiard no symmetry reduction of the resulting eigenvalue problem is possible . the presence of the flux line constraints the method of analysis to the diagonalization of large hermitian matrices , limiting the number of eigenstates that can be treated efficiently . in the following , we present the numerical results for the correlators of level velocity , decay widths , and conductance peak heights . ( recall that at present only the last can be directly measured in real experiments . ) figure [ fig : cvbilliard ] shows the level velocity correlator defined by eq . ( [ cvxdef ] ) rescaled according to @xmath202 , with @xmath203 . the plotted data were obtained by averaging over 76 equidistant values of @xmath196 in the interval @xmath204 $ ] , over the eigenstates between 201 and 300 , and over the five different shapes mentioned above . we observe a good agreement with the analytical results@xcite for small values of @xmath128 and with random matrix simulations in general . a thorough discussion of the scaling factor @xmath98 dependence on energy and billiard shape is postponed to the next subsection . next we employed the billiard model to obtain the decay width autocorrelation function @xmath205 , eq . ( [ eq : corww ] ) . the decay widths @xmath206 for the billiard were calculated as described in sec . [ sec : stochastic ] . the position of the leads and their widths were specified in the following way . based on autocorrelation calculations of the eigenfunction @xmath181 along the perimeter of the billiard , we found that the spatial decorrelation for the relevant levels ( levels 201 to 300 ) takes place over a distance around 1/60 of the perimeter . we therefore decided to take the width of the leads to be 1/24 of the perimeter , i.e. , 2.5 times the decorrelation length , yielding 24 adjacent leads . to improve the statistics we used all 24 lead position for each @xmath181 . due to the relatively large width of the leads , adjacent leads are not correlated , as was verified by obtaining the same result ( with larger fluctuations ) using only every second or every third lead . the result of the calculation is shown in fig . [ fig : cwbilliard ] . a fair agreement with eq . ( [ eq : corwwasymp ] ) is noted . finally , we calculated the conductance peak correlation @xmath105 , eq . ( [ eq : defcx ] ) , for the conformal billiard using the above mentioned decay widths . we chose all possible pair configurations among the 24 lead positions . the result is shown in fig . [ fig : cgbilliard ] and , again , a fair agreement with the predictions of sec . [ sec : subcondpeakcor ] is observed . notice that the data for the conformal billiard are not fully consistent with a lorentzian if we fix the @xmath104 asymptotics of the curve to be identical to eq . ( [ eq : cxsmall ] ) . a squared lorentzian does not seem to provide a better approximation either , although in a recent experiment@xcite @xmath105 was measured and such a curve was fitted to the data . it would be interesting to check how well our result for @xmath105 based on random matrix calculations matches the available experimental data _ without _ any fitting parameter ( in sec . [ sec : energydep ] we will present a way to predict the typical field correlation scale ) . for all three correlators we have noticed large statistical fluctuations between data taken at different billiard shapes . we found that most levels around the 300th one ( the upper limit of reliability in our calculation ) still do not show more than one full oscillation within the range of flux allowed by symmetry . the averaging over shape deformation was thus crucial to get rid of the remaining nonuniversal features . after averaging over the five values of @xmath173 mentioned previously ( see also the following subsection ) , we found that @xmath100 near the @xmath64th level obeys the law @xmath207 . we ascribe the small mismatch between theory and numerics , particularly at @xmath208 , to poor statistics . as mentioned before , the only way to circumvent this problem is to compute higher eigenstates . fortunately , this difficulty does not appear in real experiments where the magnetic flux is extended over the whole area of the cavity and the dependence of @xmath12 is not periodic . one of the important features of a dynamical model is that its quantum fluctuations display a marked dependence on energy . a _ quantitative _ understanding of the field scale of the fluctuations and its dependence on energy is important to put any random matrix result in contact with measurements in quantum dots . for this purpose , the semiclassical approach can be used in a relatively simple form . the aim of this subsection is to discuss the energy dependence of the level velocity correlator @xmath100 , which is a measure of the quantum fluctuations , and show that one can successfully estimate this quantity using exclusively classical quantities . it was already properly noticed by berry and robnik @xcite that the typical flux necessary to induce a crossover from goe- to gue - like spectral fluctuations in a chaotic billiard threaded by an aharonov - bohm flux line scales with the energy as @xmath209 . the origin of this dependence has a simple semiclassical explanation which was nicely worked out by ozrio de almeida and coworkers.@xcite the nonuniversal scaling @xmath100 can be obtained in an analogous manner . our point of departure is a recent work by berry and keating@xcite based on the gutzwiller trace formula . most of our semiclassical considerations follow their findings . however , our interpretation and the method we use to quantify @xmath100 are different . to make the exposition self - contained , we shall briefly describe points of their work which are relevant to our discussion and comment when necessary . the initial step is to approximate @xmath210 by a two - point correlator . ( the nature of the approximation is evident , since to track down the parametric evolution of a single eigenvalue for _ any _ @xmath12 is a task that can not be exactly achieved by considering only green s functions with a finite number of points . ) as in ref . , we write @xmath211 where @xmath212 is a smoothed cumulative level density which counts the number of single - particle states up to an energy @xmath11 at a magnetic flux @xmath12 . for the smoothing it is convenient to adopt the form @xmath213 , where the parameter @xmath214 is chosen to be much smaller than @xmath21 . ( we would like to point out a change in notation : throughout this subsection angular brackets @xmath51 will denote energy averages in distinction to the previous ensemble averages . ) the energy average in eq . ( [ ssacorr ] ) is taken over a range @xmath215 around @xmath11 . ordinarily , the average over the magnetic flux is taken over a window @xmath216 in flux which corresponds to little change in the classical dynamics but is semiclassically large , i.e. , it corresponds to sizable fluctuations in the spectrum . for billiards threaded by aharonov - bohm flux lines this is not an actual constraint , since flux variations have no effect on classical trajectories . nonetheless , when considering the correlator of eq . ( [ ssacorr ] ) one has , in principle , to avoid @xmath217 pertaining to the tr - breaking crossover regime . when levels are much farther apart than @xmath214 , it is straightforward to show that @xmath218 can be approximated by @xmath219 . for @xmath12 larger than a certain @xmath220 such that the correlation between level velocities of different states @xmath221 is much weaker than for @xmath222 , @xmath223 is equivalent to @xmath210 and independent of @xmath214 , provided that the energy levels are taken to be within @xmath215 . this equivalence also holds in the limit of @xmath224 when the spectrum is nondegenerate . in summary , @xmath225 is a good approximation to the level velocity correlator @xmath210 only for @xmath224 and @xmath226 . in fact , @xmath225 is also the quantity evaluated analytically by szafer and altshuler @xcite in the context of disordered metallic rings , using a diagrammatic perturbation theory based on impurity averaging . following ref . , we now turn to the semiclassical treatment . in particular , we specialize the results to billiards threaded by a flux line . this simplifies the problem enormously since if @xmath227 is the action of a periodic orbit @xmath186 , upon applying a magnetic flux @xmath12 we find that @xmath228 , where @xmath229 is the number of times the orbit @xmath186 winds around the flux line . the cumulative level density is expressed semiclassically using the gutzwiller trace formula and one writes the correlator as @xmath230 where the amplitudes @xmath231 contain information on the stability of the orbit @xmath186 as well as its maslov index . the smoothing of the staircase function gives rise to the exponential damping factor @xmath214 times @xmath232 , the period of the closed orbit @xmath186 . in the semiclassical limit , @xmath233 , the flux range of the tr - breaking crossover is much smaller than @xmath234 . therefore , without affecting considerably the calculation for the pure tr - broken case , we take the limit of @xmath235 . now one arrives at one of the delicate points of the semiclassical approach . the correlator @xmath223 is expressed as a sum of diagonal and off - diagonal contributions . in contrast to the fact that it is a settled matter how to compute the diagonal part , the evaluation of the off - diagonal term is still an unsatisfactorily solved problem . it seems that for @xmath236 it is reasonable@xcite to neglect the off - diagonal contribution and we will do so hereafter . the diagonal part of @xmath223 in eq . ( [ sccorr ] ) is still difficult to evaluate since , in principle , it requires the knowledge of the full set of periodic orbits up to the cutoff @xmath237 . this can be simplified by the following considerations . for chaotic systems , the number of periodic orbits grows exponentially as a function of their length @xmath4 and ergodicity ensures that as the orbits become longer they tend to explore the phase space more uniformly . thus , one can define a critical length @xmath238 corresponding to an uniform coverage of the phase space by the periodic trajectories . for a fixed energy @xmath11 , @xmath238 determines @xmath239 , the time when the hannay and ozrio de almeida sum rule @xcite is applicable , allowing the sum over periodic orbits to be calculated as an integral over orbital times . applying this sum rule to the diagonal part of eq . ( [ sccorr ] ) one obtains @xmath240 here the overbar stands for an average over the phase space , or , in practice , over an ensemble of trajectories . the replacement of the summation over periodic orbits by an average over trajectories makes the problem amenable for a computational evaluation of @xmath241 . such a procedure has already been successfully used to estimate the goe to gue transition parameter in the stadium billiard.@xcite by writing the winding number as an integral over the angular velocity , @xmath242 , it is simple to see that for an ergodic system @xmath243 . the winding number variance is given by @xmath244 where @xmath245 . for chaotic systems in general the correlator @xmath246 decays sufficiently fast in @xmath35 to assure the convergence of the integral in eq . ( [ wsimple]).@xcite the knowledge of the winding number distribution @xmath247 allowed us to evaluate the phase space average in eq . ( [ fcomput ] ) . to determine @xmath247 for the conformal billiard we have randomly chosen @xmath248 initial conditions and computed trajectories up to 250 bounces for the particular deformation @xmath199 and different @xmath173 s . we have generated a histogram recording @xmath169 and @xmath35 every time a trajectory winds around the flux line located at the origin.@xcite the results displayed in fig . [ fig : woft ] confirm eq . ( [ wsimple ] ) . the variance of @xmath249 is better written as@xcite @xmath250 where @xmath251 is the billiard area and @xmath252 is a _ system - dependent _ quantity computed for the scaled billiard with unit area and trajectories with unit velocity . moreover , our numerical results give us confidence that for any given time @xmath247 is a gaussian distribution ( see fig . [ fig : pofw ] ) . this is in agreement with the conjecture of ref . for periodic orbits . substituting this result into ( [ fcomput ] ) and baring in mind that in the semiclassical limit @xmath253 , for @xmath224 we obtain @xmath254 using the leading term in the weyl formula , @xmath255 , we can write @xmath256^{1/4 } } \;.\ ] ] notice that @xmath257 . although @xmath258 fails as an accurate approximation for @xmath210 over the entire range of @xmath12 , we expect it to work for @xmath224 and @xmath259 . therefore , the semiclassical quantity @xmath220 given above should yield a good approximation to the exact inverse field scale @xmath98 . we used the same procedure described to obtain @xmath260 to compute @xmath252 as a function of @xmath173 . the results are show in fig . [ fig : fig9 ] . figure [ fig : scc0 ] shows @xmath100 as a function of @xmath64 from quantum mechanical ( sec . [ sec : numerics ] ) and semiclassical ( @xmath261 ) calculations . from the numerical diagonalizations we found that the proportionality factor between @xmath100 and @xmath262 varies between 0.94 ( @xmath263 ) and 1.54 ( @xmath264 ) , while the semiclassical estimate gives 0.90 for @xmath265 , for instance . clearly , both quantum and semiclassical calculations indicate that @xmath100 depends on the billiard shape . when comparing the results of these two calculations , one should also note the size of the large fluctuations in the data presented in fig . [ fig : scc0 ] . as already remarked in the previous subsection , these large statistical fluctuations are due to the limited data set used in the simulations . to put the semiclassical result in direct contact with experiments , one needs some system - specific information to compute @xmath252 . in the above discussion , @xmath252 gives a measure of how fast the winding number variance increases with time . for the more physical situation of extended @xmath2 fields , @xmath252 measures the rate of increase in the variance of accumulated areas as a function of time.@xcite it is interesting to notice that the semiclassical interpretation of the scales of the conductance autocorrelation function for quantum dots and open cavities is very similar . since quantum dots have @xmath266 , the escape time @xmath267 is always semiclassically large . in this regime , the physics is dominated by the classical decorrelation time @xmath268 implied in eq . ( [ wsimple ] ) . the situation is very different for open cavities , where the escape time plays an important role , since it is comparable with @xmath268 . for the quarter stadium and extended @xmath2 fields , ref . gives @xmath269 . in this study we observed that @xmath252 is a relatively robust number , since very different shapes of the conformal billiard give values of @xmath252 that differ at most by @xmath270 . therefore we believe that with help of eq . ( [ eqscale ] ) and taking @xmath252 to be of order unity , one can estimate the magnitude of @xmath220 for other chaotic billiards . in this paper , we have proposed that the universal form of the parametric correlator of conductance peak heights indicates the chaotic nature of the electron dynamics in quantum dots in the coulomb blockade regime . in experiments , the simplest parameter to vary is an external magnetic field . whereas random matrix theory provides the universal form of the correlation function , the nonuniversal field scale can be understood in simple semiclassical terms : it is related to the average winding number per unit of time of periodic orbits bouncing between the confining walls of the quantum dot . this field scale is rather sensitive to the geometry of the dot and the fermi energy . to compare our analytical and numerical predictions against the experimental result , the magnetic flux through the dot has to be larger than one quantum unit of flux @xmath271 , but such that the cyclotron radius is much larger than the dot diameter . the former condition assures that time - reversal symmetry is broken ; the latter implies that the bending of classical trajectories is mainly due to scattering by the boundaries . we point out that electron - electron correlations can be taken into account by assuming that the single - particle spectrum results from a self - consistent ( hartree - fock ) treatment.@xcite our theoretical prediction for the correlator of peak heights is based on the hypothesis that the statistical properties of the system hamiltonian can be described by random matrix theory . although we could only derive analytically expressions for the limit of small field variations , the complete form of the correlation function was obtained by numerical simulations of large gaussian matrices . we have compared the random matrix results with the exact correlator obtained from the conformal billiard after averaging over energy and shape deformation . the agreement found was good , given the limitations imposed by the size of the data set . in addition to that , we found that the result of the classical calculation for the magnitude of the field scale and its dependence on energy matches moderately well the quantum result . for experimental tests of our theory , it is important to note that dephasing in the small quantum dot has to be kept low enough , namely , the dephasing length has to be larger than the system size . also , the parametric results presented for the gue case were illustrated in a dynamical model in which only the magnetic field was varied . this was very convenient , since it allowed us to the relate the field scaling parameter to the underlying classical dynamics using a well - developed formalism . alternatively , the external parameter could be taken as the shape variation . for @xmath272 , the conductance autocorrelation function should then follow our numerical results for the goe case . or , by varying the shape with @xmath273 , the conductance peak correlator should be given by our gue results . ( while finishing this work we learned that such an experiment has already been performed.@xcite ) unfortunately , the semiclassical analysis for these situations is more difficult than the one presented here and still remains an open problem . some billiard geometries have well - pronounced short periodic orbits which , for insufficient averaging over energy and magnetic field , can lead to strong nonuniversal features to the curves presented in this work . we believe that overcoming this problem will be one of the most serious challenges for the experiments . in particular , to verify experimentally the asymptotic behavior of @xmath105 should be very difficult . since @xmath105 involves the subtraction of two numbers and for @xmath120 these numbers become very close , statistical ( nonuniversal ) fluctuations can easily drive the experimental @xmath105 below zero . another cause of deviations from the predicted universal behavior , presumably weaker , is the existence of wave function correlations which extend over the dot . in other words , our findings assume that the channels at different leads are decorrelated , which may not be completely true if the dot size is not much larger than the electron wavelength . the agreement of our random matrix results with the numerical calculations using the conformal billiard supports the assumption of independent channels for @xmath274 . smaller systems should be more influenced by short or direct orbit effects . we should also mention that if there were strong correlations between heights of neighboring peaks of a given sequence , they would influence mostly the @xmath120 region of the correlator . this is because averaging over a finite range of magnetic field usually yields less statistics for large field differences and , consequently , more pronounced data correlation effects . lastly , we point out the fact that , independently of previous considerations , another very interesting experiment is a direct measurement of @xmath275 . despite the very large coulomb energy , which makes the conductance peak spacing very regular at first sight very , a careful experiment should be able to observe the small fluctuations of the peak position as a function of an applied magnetic field . this will give direct information about the single - particle level dispersion or , equivalently , @xmath275 . it will also provide a direct test of our estimate of the flux correlation scale @xmath220 . _ note added . _ after the submission of this manuscript we learned of similar work by alhassid _ et al._@xcite h.b . was supported by the european commission under grant no.erbchbgct 930511 . c.h.l . acknowledges the financial support of the national science foundation . both h.b . and c.h.l thank nordita for its hospitality . we are grateful to b. altshuler , o. klein , c. marcus , b. simons , n. taniguchi , and s. tomsovic for helpful discussions and to r. jalabert for suggestions concerning the manuscript . we also thank h. baranger and c. marcus for providing us with refs . and , respectively , prior to publication . this appendix is devoted to a rather detailed evaluation of the average over the determinant shown in eq . ( [ eq : beig2 ] ) . let us call @xmath277 where @xmath87 is now an @xmath278 gue matrix . first , we notice that there is a more convenient way to express this average , namely , @xmath279 where the generating function @xmath280 can be evaluated by the fermionic method:@xcite @xmath281 \exp[-\chi^\dagger(h\otimes\openone_2 + \alpha \openone_n\otimes l)\chi ] \right\rangle_h , \label{eq : apc3}\ ] ] where @xmath282 , @xmath283 , and @xmath284 , with @xmath285 and @xmath286 representing @xmath64-component fermionic vectors . averaging over the gue matrix @xmath87 we find that @xmath287 \ ; , \label{eq : apc4}\ ] ] where @xmath288 is the following @xmath289 matrix @xmath290 the quartic term can be decoupled by a hubbard - stratonovich transformation , namely , @xmath291 \exp\left [ -\frac{n}{2}\mbox{tr}(q^2)- i\lambda\mbox{tr}(qu)\right ] = \left(\frac{2\pi}{n}\right)^2 \exp\left[-\frac{\lambda^2}{2n}\mbox{tr}(u^2 ) \right ] , \label{eq : apc6}\ ] ] where @xmath292 is a @xmath289 hermitian matrix and @xmath293=dq_{11}dq_{22}dq_{12}dq_{21}$ ] . as a result , we have @xmath294 \exp\left [ -\frac{n}{2}\mbox{tr}(q^2)\right ] \int\ ! d[\chi ] \exp [ -\chi^\dagger(\alpha \openone_n\otimes l+i\lambda \openone_n\otimes q)\chi ] . \label{eq : apc7}\ ] ] the gaussian integral over the fermionic variables can be easily carried out , yielding @xmath295 \exp\left\ { -\frac{n}{2}\mbox{tr}(q^2 ) + n\mbox{tr}[\ln(\alpha l+i\lambda q)]\right\}. \label{eq : apc8}\ ] ] when @xmath296 the above integral over @xmath292 can be evaluated by the saddle - point approximation ( which becomes exact in the limit @xmath80 ) . for this purpose we first separate angular and radial components of @xmath292 , namely , @xmath297 , where @xmath298 and @xmath35 is an su(2 ) matrix . the differential breaks up into @xmath293=d\mu(t)j(q)d[q]$ ] , where @xmath299 is the jacobian of the transformation , @xmath300=dq_1dq_2 $ ] , and @xmath301 is the group measure normalized to unity . this yields @xmath302 j(q ) \exp(-nf[q ] ) \int d\mu(t ) \exp\left[-\frac{i\alpha n}{\lambda}\mbox{tr}(q^{-1}tlt^\dagger)\right ] , \label{eq : apc9}\end{aligned}\ ] ] where @xmath303 = ( 1/2)\mbox{tr}(q^2)-\mbox{tr}[\ln(i\lambda q)]$ ] and we have only kept terms to lowest order in @xmath196 . the saddle - point expansion now involves only the radial part of the action : @xmath304 & = & 1 - 2\ln(\lambda ) - \frac{i\pi}{2}(\sigma_1+\sigma_2 ) + \delta q_1 ^ 2 + \delta q_2 ^ 2 + o(\delta q^3 ) , \label{eq : apc10}\end{aligned}\ ] ] where @xmath305 and @xmath306 . the relevant saddle points correspond to @xmath307 , resulting in @xmath308 . \label{eq : apc11}\end{aligned}\ ] ] the integral over the su(2 ) manifold can be evaluated through the well - known itzykson - zuber formula,@xcite which in a simplified form reads @xmath309 = \frac{\det[\exp(\beta l_{i}l_{j})]}{\beta(l_1-l_2)^2 } , \label{eq : apc12}\ ] ] with @xmath310 denoting the eigenvalues of @xmath4 . hence , @xmath311 ( the factor of 2 takes into account the double saddle point ) . finally , we obtain @xmath312 the small-@xmath38 asymptotics of the correlator of decay widths can be determined by the same method used in sec . [ sec : subcondpeakcor ] for the conductance peak height correlator . beginning with the definition presented in eq . ( [ eq : corww ] ) , we expand @xmath313 up to first order in @xmath38 [ see eq . ( [ eq : expgamma ] ) ] . the zeroth - order term of @xmath314 is then given by @xmath315 for the unitary ensemble . an expression analogous to eq . ( [ eq : expcx ] ) is used to write the second - order coefficient of @xmath314 in terms of the amplitudes @xmath137 , namely , @xmath316 carrying out the average over the matrix elements of the external perturbation @xmath61 [ see eq . ( [ eq : averlambda1 ] ) ] , we find that @xmath317 we now average separately over the eigenvalues and partial widths and obtain @xmath318 upon rescaling both @xmath319 and @xmath320 , we arrive at @xmath321 the large-@xmath38 asymptotics of @xmath314 can be inferred from the asymptotics of another correlator , namely , @xmath322 where @xmath323 is the system volume , @xmath324 , and @xmath325 . recall that the wave function intensities are proportional to the decay widths @xmath326 and @xmath327 for pointlike contacts . at @xmath328 and large @xmath38 , the interlevel correlations are secondary to intralevel ones ; as a result , the @xmath173 function in eq . ( [ eq : cgtilde ] ) acts as a kronecker @xmath173 , causing @xmath329 and @xmath314 to coincide ( up to a prefactor equal to @xmath330 ) to leading order in @xmath331 . let us for convenience assume a finite size space basis to represent the system hamiltonian . we can then reduce eq . ( [ eq : cgtilde ] ) to @xmath332_{kk } \mbox{im}\left[g(e_2-i\epsilon;x_2)\right]_{kk } \right\rangle - \frac{1}{\delta^2 } \label{eq : apd2}\ ] ] with @xmath333^{-1}$ ] and @xmath334 . the above expression can be rewritten in the more convenient form @xmath335_{kk } \left[g(e_2-i\epsilon;x_2)\right]_{kk } \right\rangle - \frac{1}{\delta^2 } \;. \label{eq : apd3}\ ] ] in general , an expression like eq . ( [ eq : apd3 ] ) requires the evaluation of the following quantity : @xmath336_{kl } [ g(e_2-i\epsilon;x_2)]_{mn } \right\rangle . \label{eq : apd4}\ ] ] the correlator @xmath337 can be calculated exactly in the zero - mode approximation of the supersymmetric nonlinear @xmath63 model @xcite ( or , equivalently , in the rmt framework ) . this calculation is standard nowadays ( for a recent review , see ref . ) and has already been presented in the literature.@xcite here we will only mention the resulting expression for the unitary ensemble , which is @xmath338 \ ; , \label{eq : apd21}\end{aligned}\ ] ] where @xmath339 \label{eq : apd22}\end{aligned}\ ] ] and @xmath340 . \label{eq : apd23}\ ] ] when writing these equations we have rescaled the variables to @xmath341 , @xmath342 , and @xmath343 . going back to eq . ( [ eq : apd3 ] ) , we arrive at @xmath344 . \label{eq : apd24}\ ] ] since @xmath345 and @xmath346 as @xmath347 , we have that @xmath348 in the same limit . therefore , we expect that @xmath349 finally , we remark that the @xmath350 universal asymptotics of both @xmath351 and @xmath352 can also be obtained by the diagrammatic perturbation theory of disordered metals expressed in terms of diffuson modes . in this appendix we give a schematic description of the calculation of @xmath354 . the starting point is eq . ( [ eq : eigaver ] ) . here we go through the same steps of sec . [ sec : subcondpeakcor ] to evaluate @xmath355 [ see eq . ( [ eq : beig ] ) ] . first we fix the reference eigenvalue @xmath28 to the center of the spectrum , obtaining @xmath356^{-1 } \right\rangle_{\{\varepsilon_\nu\}}\;. \label{eq : appc1}\ ] ] next , we rephrase this expression in terms of an average over a spectral determinant , namely , @xmath357 } \right\rangle_h , \label{eq : appc2}\ ] ] where @xmath358 is a constant [ such that @xmath359 and the average is performed over a @xmath88 gue matrix @xmath87 . the appearance of determinants in both numerator and denominator in eq . ( [ eq : appc2 ] ) makes its evaluation technically more difficult than eq . ( [ eq : apc1 ] ) . it is necessary to introduce not only four anticommuting auxiliary variables , but also two commuting ( complex ) ones . the resulting symmetry group is u@xmath360 ( the pseudo - unitarity is due to the structure of the denominator ) . fortunately , a general solution for such graded symmetry problems has been recently worked out.@xcite the derivation is a nontrivial generalization of the method of appendix [ sec : appendixa ] . for an expression with the structure of eq . ( [ eq : appc2 ] ) , one arrives at the following formula:@xcite @xmath361 , \label{eq : apcc3}\end{aligned}\ ] ] where the sum runs over all six nonequivalent combinations of pairs of @xmath362 , @xmath363 , and @xmath364 is an unspecified constant . to get @xmath149 we need to take the limit @xmath365 for all @xmath366 at a given order . after some algebra , one finds that @xmath367 finally , inverse fourier transforming the above expression , we arrive at @xmath368 reviews of experimental and theoretical aspects of quantum chaos in nanostructures can be found in chaos * 3 * , ( 1993 ) . see also h. u. baranger , in _ nanotechnology _ , edited by g. timp ( aip , new york , in press ) ; r. m. westervelt , _ ibid . _ for classical hyperbolic systems one expects quantum universal behavior in the semiclassical region of the spectrum , i.e. , where the electron wavelength at the fermi surface is small enough to resolve the typical irregularities " of the cavity . this is usually the case for two - dimensional systems with more than @xmath369 electrons . the averaging is done explicitly here to simplify the calculation . in principle @xmath61 is not required to be a random matrix . however , once we choose for basis the eigenstates of @xmath370 , the matrix elements @xmath371 will be gaussian distributed , regardless of the nature of @xmath61 . the decay of classical autocorrelators for ergodic dynamical systems is presently a matter of intensive investigation . in the absence of a rigorous proof we base our statement about the convergence of @xmath373 on numerical evidence . the underlying idea is again the uniform coverage of the phase space . for an ergodic system the typical winding number as a function of length should not depend on whether the trajectory is periodic or not , provided that @xmath374 . an example of this procedure for the case of ballistic quantum dots in the presence of high magnetic fields is given by p. l. mceuen _ et al . _ , b * 45 * , 11 419 ( 1992 ) . for heavily doped , diffusive dots , a single - particle description of the spectrum is probably insufficient ( see ref . ) .

over the last three years , the galactic center region has been monitored with the _ chandra x - ray observatory_. besides the x - ray emission from the target object , sgr a * , diffuse x - ray emission was detected throughout most of the 17 17 field of view . with 11 _ chandra _ observations through 2002 june , the total effective exposure reaches ks , providing significant photon statistics on much of the detailed structure within the faint , often filamentary , diffuse x - ray emission . the _ true - color _ x - ray image and the _ equivalent width _ images for the detected elemental species of the galactic center region demonstrate that the diffuse x - ray features have a broad range of spatio - spectral properties . enhancements of the low - ionization - state fe line emission ( 6.4 kev ) to the northeast of sgr a * can be interpreted as fluorescence within the dense interstellar medium resulting from irradiation by hard , external x - ray sources . they may also be explained by emission induced by the bombardments by high energy particles on the ambient medium , such as may accompany unresolved supernova ejecta intruding into dense interstellar medium . the detection of molecular cloud counterparts to the 6.4 kev fe line emitting features indicates that these fe line features are associated with dense galactic center clouds and/or active star - forming regions , which strongly supports the proposed origins of the x - ray reflection and/or supernova ejecta for the fe line emission . we detect highly ionized s and si lines which are generally coincident with the low - ionization - state fe line emission and the dense molecular clouds in the northeast of sgr a*. these hot plasmas are then most likely produced by massive star - forming activities and/or supernova remnants in the galactic center . in contrast , we find that highly ionized he - like fe line emission ( 6.7 kev ) is primarily distributed along the plane instead of being concentrated in the northeast of sgr a*. the implied high temperature and the relatively uniform , but , strong alignment along the plane are consistent with the magnetic confinement model , suggesting this hot gas component has been re - heated by the strong interstellar magnetic fields in the galactic center to produce the observed he - like fe line emission .

astro-ph0311460

the central @xmath2100 pc around the galactic center is an extremely complex region containing a variety of astrophysical activities : i.e. , cold and warm molecular clouds , star clusters / formation , supernova remnants ( snrs ) , and hii regions , to name a few . earlier galactic plane surveys with the _ tenma _ and _ ginga _ satellites have discovered the highly ionized fe emission line at @xmath3 @xmath2 6.7 kev , which is strongly enhanced toward the galactic center region @xcite . _ asca _ and _ bepposax _ observations have confirmed the presence of the ionized fe emission line as well as detecting emission lines from other highly ionized elemental species of si , s , and ar @xcite . these results suggested that the difuse x - ray emission in the plane and the galactic center regions arises from thermal plasmas with multiple temperatures and ionization states , although the temperatures of the individual plasma components are inconclusive ( e.g. , koyama et al . 1996 ; kaneda et al . 1997 ; valinia & marshall 1998 ; park & ebisawa 2001 ) . the origins of these hot plasmas have been suggested to be multiple supernova explosions and magnetic confinement of the interstellar medium ( ism ) ( e.g. , kaneda et al . 1997 and references therein ) . besides the apparent hot thermal components , the presence of giant molecular cloud complexes ( e.g. , sanders et al . 1984 ) , as well as a number of hard x - ray emitting point sources near the galactic center ( e.g. , skinner et al . 1987 ) has led to the prediction that fluorescent x - ray emission should arise where the cold , dense ism is illuminated by external x - ray sources @xcite . this proposed new class of x - ray sources has been discovered with _ asca _ by detecting low - ionization - state fe line emission ( fei @xmath4 fexvii ) in the galactic center region emerging at a characteristic energy of @xmath3 @xmath2 6.4 kev @xcite . a detailed analysis of this low ionization , or `` neutral '' 6.4 kev fe line emission toward the galactic giant molecular cloud sgr b2 has successfully described this fluorescent x - ray emission from the molecular cloud in terms of an x - ray reflection nebula ( xrn ) model @xcite . strong past activity of the supermassive galactic black hole candidate sgr a * was proposed to be the primary external illuminating source to produce the observed neutral fe line emission from sgr b2 . a similar xrn feature has also been detected in another galactic center molecular cloud , sgr c @xcite . alternatively , such neutral fe line emission may also be explained by the bombardment of energetic electrons on metal - rich sn ejecta fragments @xcite . massive stars appear to be forming at several sites in the central few hundred parsecs @xcite , so the proposed unresolved sn ejecta scenario for the origin of the observed neutral fe line emission is plausible as well . the x - ray line emission from the hot / cold ism is therefore a useful tool to probe the origins of the complex structure of the ism in the galactic center . the poor angular resolutions of previous detectors were , however , insufficient for such studies in complex regions of the galactic center . with the superb sub - arcsecond angular resolution and good sensitivity in the 0.3 @xmath4 8 kev band , the _ chandra x - ray observatory _ provides an excellent opportunity to investigate the nature of the diffuse x - ray emission in the galactic center regions . preliminary results from earlier _ chandra _ observations showed that the diffuse x - ray emission features in the galactic center appear to have a variety of spectral and morphological characteristics @xcite . the interpretations of these diffuse x - ray features have however been highly speculative primarily due to the limited photon statistics with previous data . recently , we have performed deep observations of the galactic center with _ the exposure has reached @xmath2600 ks , which is an order of magnitude deeper than previous observations , so that these deep _ chandra _ observations represent an unprecedented wealth of x - ray data on the galactic center . in fact , these _ chandra _ data have resolved over @xmath22300 point sources within @xmath220 pc of the galactic center @xcite . besides these point sources , including the target object sgr a * , the _ chandra _ data show a complex structure of the diffuse x - ray emission within the 17@xmath0 @xmath1 17@xmath0 field of view ( fov ) . we here present the latest deep _ observations of the galactic center . the high angular resolution _ chandra _ data with decent photon statistics from the deep exposure allow us extensive imaging and spectral analysis of detected bright filamentary features . the observation is described in [ sec : obs ] , and the x - ray image analysis is presented in [ sec : image ] and [ sec : ewi ] . we present the results from spectral analysis for some selected emission features in [ sec : spec ] , and compare these results with molecular line emission features in [ sec : mc ] . some implications are then discussed in [ sec : disc ] . a summary is presented in [ sec : summary ] . for the last three years , the galactic supermassive black hole candidate sgr a * has been monitored , as part of guaranteed time observation ( gto ) and guest observer ( go ) programs , with the advanced ccd imaging spectrometer ( acis ) @xcite on board the _ chandra x - ray observatory _ ( table [ tbl : tab1 ] ) . as of 2002 june , combining 11 _ chandra_/acis observations ( except for obsid 1561a , which was severely contaminated by a bright transient source within the fov ) , the total exposure has reached @xmath2590 ks , which is the deepest ever observation of the galactic center region in x - rays . we first applied the algorithm developed by townsley et al . ( 2002a ) to correct the spatial and spectral degradation of the acis data caused by radiation damage , known as charge transfer inefficiency ( cti ; townsley et al . [ 2000 ] ) . the expected effects of the cti correction include an increase in the number of detected events and improved event energies and energy resolution @xcite . we then screened the data by status , grade , and the flight timeline filter . we have also removed observation time intervals of strong flaring in the background . all individual event files were then reprojected to the tangent plane at the radio position of sgr a * ( ra[j2000 ] = 17@xmath5 45@xmath6 [email protected] , dec[j2000 ] = @xmath429@xmath8 00@xmath0 28@xmath9118 ) in order to generate the composite data . the detailed descriptions of these data reduction processes and the resulting broadband raw image from the composite data may be found in muno et al . ( 2003a ) . during the monitoring observations of sgr a * ( ra[j2000 ] = 17@xmath5 45@xmath6 [email protected] , dec[j2000 ] = @xmath429@xmath8 00@xmath0 28@xmath90 ) , the 17@xmath0 @xmath1 17@xmath0 fov of the acis - i array also imaged the interstellar surroundings of sgr a * , which reveals the complex environment within @xmath220 pc of the galactic center with unprecedented detail in x - rays ( hereafter , we assume a distance of @xmath10 = 8 kpc to the galactic center [ reid 1993 ] ) . a `` true - color '' x - ray image of the galactic center region is presented in figure [ fig : fig1 ] . each subband image has been exposure - corrected utilizing the exposure map produced by muno et al . ( 2003a ) , and adaptively smoothed to achieve signal - to - noise ( s / n ) ratio of 4 by using the _ ciao _ tool _ csmooth_. while over @xmath22000 point - like sources prevail over the fov , the point source contribution to the observed x - ray flux is @xmath210% and the bulk of the x - ray emission is _ truly _ diffuse @xcite . figure [ fig : fig1 ] indeed demonstrates the presence of diffuse extended x - ray emission features on a variety of angular scales . the target object sgr a * is located near the center of the image where the bright x - ray emission from a massive star cluster within the inner parsec of the galactic center dominates . the bright diffuse x - ray emission to the immediate east of sgr a * is the snr sgr a east @xcite . symmetric lobes of soft x - ray emission extended from sgr a * , perpendicular to the galactic plane , are indicated with the yellow dotted line in figure [ fig : fig1 ] . these features may represent a bipolar outflow from sgr a * , which are extensively discussed by baganoff et al . ( 2003 ) and morris et al . ( 2003 , in preparation ) . outside of these complex central regions are larger angular scale ( @xmath2a few arcmin ) x - ray features , consisting of a few x - ray `` loops '' around sgr a * ( e.g. , the white dotted curves in figure [ fig : fig1 ] ) . the soft nature of these loop - like features is typical of thermal shocks interacting with the galactic ism . considering significant star - forming activities in the galactic center regions , these red , soft x - ray features are perhaps caused by supernova blast waves or stellar winds / bubbles . we also detect a few hard x - ray knots and filaments . the bright knot to the south of sgr a * ( region a in figure [ fig : fig1 ] ) is the x - ray counterpart of the bright , nonthermal radio knot in the boundary shell of the snr [email protected] @xcite . we have identified at least one more continuum dominated arc - like feature in the northeast of sgr a * ( region b in figure [ fig : fig1 ] ) , which shows a hard photon index . the detailed analysis of these continuum x - ray sources are presented elsewhere ( baganoff et al . 2003 ; sakano et al . 2003 ; morris et al . 2003 , in preparation ; yusef - zadeh et al . 2003 , in preparation ) . in contrast , some other blue , hard filaments to the northeast of sgr a * ( regions 1 , 2 , and 3 in figure [ fig : fig1 ] ) exhibit strong neutral fe k@xmath11 line emission @xcite . the origin of these neutral fe line sources has been elusive @xcite , and we will discuss the detailed spectral characteristics of these neutral fe line features in [ sec : spec ] . in order to examine the overall distributions of the diffuse emission line features around the galactic center region , we construct _ equivalent width _ ( ew ) images for the detected atomic emission lines , following the method described by park et al . ( 2002 ) . after removing all detected point sources from the broadband image ( see muno et al . [ 2003a ] for the details of the point source detection ) , subband images for the line and continuum bandpasses were extracted for each spectral line of interest ( table [ tbl : tab2 ] ) . these subband images were adaptively smoothed to achieve a s / n ratio of 3 @xmath4 4 . the underlying continuum was calculated by logarithmically interpolating between images made from the higher and lower energy `` shoulders '' of each broad line . the estimated continuum flux was integrated over the selected line width and subtracted from the line emission . the continuum - subtracted line intensity was then divided by the estimated continuum on a pixel - by - pixel basis to generate the ew images for each element . in order to avoid noise in the ew maps caused by poor photon statistics near the ccd chip boundaries , we have set the ew values to zero where the estimated continuum flux is low . we also set the ew to zero where the integrated continuum flux is greater than the line flux . we present the ew images of the fe he@xmath12 ( @xmath3 @xmath2 6.7 kev ) , the neutral fe k@xmath11 ( @xmath3 @xmath2 6.4 kev ) , s he@xmath12 + ly@xmath12 ( @xmath3 @xmath2 2.5 kev ) and si he@xmath12 ( @xmath3 @xmath2 1.8 kev ) line emission ( figure [ fig : fig2 ] ) . we note that cosmic x - ray background ( cxb ) and the particle background ( pb ) , both of which have not been subtracted in the ew generation , could have contaminated our ew images to some extent . we found that the cxb contribution to the observed fluxes is negligible ( @xmath131% ; e.g. , muno et al . the pb contamination appears to be more significant than the cxb and may contribute up to @xmath220%@xmath430% of the estimated continuum emission where the surface brightness is low . the ew estimations are thus affected by the pb contamination by up to @xmath220% . this embedded contamination is not however critical for our purposes of the ew images : i.e. , we intended to utilize these ew images in order to qualitatively investigate overall variation and distribution of the bright x - ray line emission over the galactic center region . we do not use the estimated absolute ew values unless the actual regional spectal analysis has been performed . the adaptive smoothing also introduces some systematic uncertainties for the small - scale , faint ew features which we do not rely on for the analysis . the large - scale ( @xmath14arcmin ) , relatively bright ew features are , on the other hand , consistent with the actual spectral analysis and are reliable in the context with previous observations ( see the following discussions and sections ) . despite the embedded uncertainties , these ew images are therefore useful as first - order information in order to investigate the overall variation of the x - ray line emission within the fov and to qualitatively discuss some noticeable characteristics . we also note that the line flux maps as well as the ew maps might also be useful to understand the line emission distributions . since the line flux maps show the _ same _ overall intensity variations as the ew maps , we do not explicitly present the line flux maps although our discussion on the line features is in parts based on the line flux distribution in addition to the ew variations . the fe he@xmath12 line ew is generally enhanced along the plane . the enhancements in the fe he@xmath12 line ew of this galactic plane component are marginal ( roughly around @xmath2400 ev ) , which is consistent with the _ asca _ measurements of the hard galactic ridge x - ray emission ( grxe ) component ( @xmath3 @xmath2 8 kev ) @xcite . this marginal ew enhancements are also consistent with those from the detected point sources in the galactic center @xcite . a fraction of this fe he@xmath12 ew should thus originate from unresolved point sources . the substantial ew variations on arcmin scales , however , suggest that a bulk of this hot fe line emission has a diffuse origin arising from a hot ( @xmath3 @xmath2 5 @xmath4 10 kev ) plasma . sgr a east is a strong fe he@xmath12 line source @xcite , and is immediately recognized as an extremely bright source with an @xmath21@xmath153 diameter in the center of the fe he@xmath12 ew image ( figure [ fig : fig2]a ) . the estimated fe he@xmath12 line ew is @xmath22500 ev @xmath4 5000 ev within the central @xmath230@xmath16 diameter region of sgr a east , which is consistent with the results from the spectral analysis by maeda et al . this strong fe ew appears to be from hot fe - rich ejecta heated by the reverse shock of the snr sgr a east as posited by maeda et al . ( 2002 ) . the neutral fe ew is largely distributed along the plane , and , in contrast to the he - like fe ew , is strongly enhanced in the northeast of sgr a * ( figure [ fig : fig2]b ) . the bright , hard fe knots in the northeast of sgr a * ( regions 1 , 2 , and 3 in figure [ fig : fig1 ] ) are featureless in the fe he@xmath12 ew image ( figure [ fig : fig2]a ) , but remarkably bright in the neutral fe ew image ( figure [ fig : fig2]b ) . the regions to the northeast of sgr a * are also bright in the si and s ew images ( figure [ fig : fig2]c and figure [ fig : fig2]d ) . the si and s maps represent high ionization states in hot thermal plasmas which are typically produced by snr and/or star - forming activity . the extinction toward the hot s and si emission in the northeast of sgr a * is consistent with that of typical column density toward the galactic center ( see [ sec : spec ] , also muno et al . 2003b ) , which indicates the galactic center origin of the hot gas rather than a nearby foreground component . it is interesting that these high - ionization - state hot s , si material and the low - ionization - state cold fe sources are in general coincident in the northeast of sgr a*. on the other hand , albeit that the angular resolution and the positional accuracy of the ew images are significantly degraded due to the adaptive smoothing of faint diffuse emission , the one - to - one correlations between the bright s / si ew and the neutral fe ew enhancements appear to be weak : i.e. , the locations of the strong enhancements of hot and cold gas are probably not `` exactly '' coincident . these cold , neutral fe line emitting gas and the hot , ionized s / si line emitting plasma also appear to have galactic center molecular cloud counterparts ( see [ sec : mc ] ) . these features suggest that the high- and low - ionization - state line emission might be associated with the star - forming / sn activities nearby to the dense molecular cloud complexes in the northeast of sgr a*. both of s / si and neutral fe ews are weak in the southwest of sgr a * , which appears to be caused by intrinscally less activities there ( for more discussion , see [ sec : mc ] and [ sec : disc ] ) . we perform a spectral analysis of a few diffuse emission features , particularly those with strong neutral fe ews . the individual acis observations comprising our images utilized different roll - angles : i.e. , the first three observations have @xmath2270@xmath8 roll - angles and the remainder @xmath280@xmath8 . the fe knots have thus been detected on either acis - i0 or acis - i3 chips depending on the roll - angle . about @xmath285% of the total exposure was taken with roll - angles of @xmath280@xmath8 , however , so the bulk of the photons from these fe knots have been detected on the acis - i0 chip . we thus utilize the response matrices appropriate for acis - i0 . the fe knots 1 , 2 , and 3 contain @xmath212000 , @xmath224000 , and @xmath211000 photons , respectively , and we bin the data to contain a minimum of 60 @xmath4 200 counts per channel for the spectral fitting . these emission features are fairly bright and have well - confined structures , and so we estimate the background contamination by extracting spectra from nearby source - free regions after removing all detected point - like sources from the data . for the spectral analysis of our cti - corrected data , we utilize the response matrices appropriate for the spectral redistribution of the ccd , as generated by townsley et al . ( 2002b ) . the low energy ( @xmath3 @xmath13 1 kev ) quantum efficiency ( qe ) of the acis has degraded because of the molecular contamination on the optical blocking filter . although effects from this qe degradation on our heavily absorbed spectra are most likely negligible , we correct this time - dependent qe degradation by modifying the ancillary response function ( arf ) for each extracted spectrum by utilizing the idl acisabs software . the x - ray spectral characteristics of knots 1 and 2 are similar to each other : i.e. , a strong 6.4 kev fe emission line with underlying power - law continuum of photon index @xmath17 @xmath2 2 @xmath4 3 ( figure [ fig : fig3]a and figure [ fig : fig3]b ) . both spectra are heavily absorbed ( @xmath18 @xmath2 3.5 @xmath1 10@xmath19 @xmath20 ) and the photon fluxes are insignificant below @xmath3 @xmath13 4 kev . we thus present the best - fit parameters based on spectral fittings in the 4 @xmath4 8 kev band ( table [ tbl : tab3 ] ) . the measured ew of the fe line is @xmath21 kev for both features , which is consistent with that of the sgr c cloud @xcite . we note that the knot 1 spectrum may also suggest the presence of an fe k@xmath21 line at @xmath3 @xmath2 7 kev ( figure [ fig : fig3]a ) . we have added another gaussian line component in order to fit this weak feature , assuming a narrow line width . the best - fit k@xmath21 line center energy is 7.02@xmath22 kev with an ew of @xmath2200 ev . the measured k@xmath21 to k@xmath11 line flux ratio is [email protected] . this k@xmath21 line feature is not apparent in the knot 2 spectrum , at least with the current data . we place an upper limit ( 2@xmath24 ) on the k@xmath21 line ew of 120 ev for knot 2 . the fe knot 3 spectrum shows the same characteristics as the spectrum of the other two knots in the hard x - ray band ( @xmath3 @xmath25 4 kev ) : i.e. , a strong 6.4 kev fe k@xmath11 line with underlying continuum of @xmath17 @xmath2 1.8 ( figure [ fig : fig3]c ) . the measured 6.4 kev fe line ew is @xmath21.3 kev . this spectrum also appears to show a weak fe k@xmath21 line feature as seen in the knot 1 spectrum . the best - fit line center energy of this weak feature , the measured ew , and the line flux ratio between k@xmath21 to k@xmath11 lines are consistent with those of knot 1 . the knot 3 spectrum , on the other hand , is less absorbed ( @xmath18 @xmath2 1.5 @xmath1 10@xmath19 @xmath20 ) than knots 1 and 2 , revealing remarkably different features than knots 1 and 2 in the soft band ( @xmath3 @xmath26 4 kev ; figure [ fig : fig3]c ) . the soft band x - ray spectrum of knot 3 is dominated by thermal emission with a strong s he@xmath12 line . for comparison , a single power law model fit is relatively poor ( @xmath27 = 1.4 ) and the inclusion of an additional thermal plasma model improves the overall fit significantly based on an f - test . the strong s line emission from the soft component of the knot 3 spectrum indicates x - ray emission from a thermal plasma having an electron temperature of @xmath28 @xmath2 0.3 kev . the metal abundance for the soft component appears to be high and is best - fitted with a few times solar , although unconstrained with the current data ( e.g. , a 2@xmath24 lower limit is @xmath20.5 solar ) . this soft thermal component appears to be associated with the galactic center molecular clouds ( [ sec : mc ] ) , suggesting its location in the galactic center instead of a nearby foreground origin . the spectral fit with a two - component absorption for knot 3 indeed indicates a significant foreground column ( n@xmath29 @xmath2 1 @xmath1 10@xmath19 @xmath20 ) even only for the soft thermal component . the nearby foreground origin for this soft thermal emission is therefore unlikely . the best - fit parameters for the knot 3 spectrum are presented based on the spectral fitting in the 2 @xmath4 8 kev band ( table [ tbl : tab3 ] ) . the spatial distribution and spectral x - ray characteristics revealed by the data suggest that some of the diffuse x - ray emission features in the galactic center be associated with the cold , dense component of the ism and/or active star - forming regions . we thus examine galactic center molecular cloud maps in order to investigate correspondences between dense clouds and the detected x - ray emission features . considering the high molecular densities in the galactic center region , the cs molecule , with its relatively high critical density , @xmath30(h@xmath31 ) @xmath2 10@xmath32 @xmath33 , is an appropriate tracer of dense molecular clouds there . the cs line maps are , unlike those of co , expected to be nearly free of strong contamination by the galactic disk and are more representative of the molecular column density rather than the surface temperatures of clouds ( tsuboi et al . 1999 and references therein ) . the cs data also provide an angular resolution of @xmath230@xmath16 ( tsuboi et al . 1999 ) , allowing a meaningful cross - comparison with the high resolution _ chandra _ image . we have thus searched for potential molecular cloud counterparts to x - ray features by investigating various velocity components of the 49 ghz cs ( j = 1@xmath40 ) line emission @xcite . the overall distribution of dense molecular clouds in the velocity range @xmath34 = @xmath425 to @xmath3550 km s@xmath36 is along the galactic plane , and we compare their distributions with our ew images of neutral fe and highly ionized s line emission ( figure [ fig : fig4 ] ) . the molecular cloud contours in the presented velocity ranges display a general correspondence with the x - ray emission - line ew distributions , particularly to the northeast of sgr a*. the detailed comparisons indeed suggest an intimate relationship , but higher spatial resolution in the molecular line maps will be required to unambiguously demonstrate the nature and geometry of the physical relationship . the + 40 to + 50 km s@xmath36 clouds , which , according to vollmer et al . ( 2003 ) , are located close to , but slightly in front ( @xmath25 pc ) of sgr a * , are well correlated spatially with the s and the neutral fe ew images throughout the region to the northeast of sgr a * ( figure [ fig : fig4]e and figure [ fig : fig4]f ) . the most prominent cloud at + 10 to + 30 km s@xmath36 the `` 20 km s@xmath36 molecular cloud '' located to the south of sgr a * is located substantially in front ( @xmath220 @xmath4 50 pc ) of sgr a * @xcite , and does not show an x - ray emission line counterpart . in this same velocity range , there is a smaller molecular cloud in the northeast corner of the fov ( figure [ fig : fig4]c and figure [ fig : fig4]d ) which is adjacent to the radio arc and may be interacting with the arc @xcite . this cloud also appears to be spatially correlated with the most prominent emission in the ew images , and with knot 3 in particular . this correlation may be best demonstrated by the high angular resolution 98 ghz cs ( j=2@xmath41 ) data @xcite overlaid with the neutral fe line ew image ( see the lower - left corner inset in figure [ fig : fig4]c ) . the line - of - sight locations of the negative velocity clouds remains undetermined . except for knot 1 ( see below ) , the molecular emission in the @xmath425 to @xmath45 km s@xmath36 range appears to be anti - correlated with the s and neutral fe ew maps in the northeast corner of the fov ( figure [ fig : fig4]a and [ fig : fig4]b ) . some of the bright , hard x - ray filaments have interesting spatial correspondences with cs clouds . knot 1 falls upon a strong cs line peak at negative velocities ( figure [ fig : fig4]a ) , although it also coincides with a prominent ridge in the + 50 km s@xmath36 cloud ( figure [ fig : fig4]e ) . knot 2 is superimposed on the edge of those same molecular clouds , so its relationship with those clouds , if any , is unclear . and , as mentioned above , knot 3 has plausible molecular counterparts in the velocity ranges + 10 to + 30 km s@xmath36 and + 40 to + 50 km s@xmath36 ( figure [ fig : fig4]c and figure [ fig : fig4]e ) . toward the south and the southwest of sgr a * , the overall broadband surface brightness , as well as the si and s line fluxes , are relatively low , and little correspondence with molecular gas is evident . the foreground `` 20 km s@xmath36 '' cloud , as represented by the emission integrated from + 10 to + 30 km @xmath36 in figures [ fig : fig4]c and [ fig : fig4]d , is prominent in this region . this dense molecular cloud complex ( @xmath37 @xmath2 2 @xmath1 10@xmath32 @xmath33 ) implies a significant column density ( @xmath18 @xmath2 4 @xmath1 10@xmath19 @xmath20 ) and a correspondingly large visual extinction , @xmath38 @xmath14 400 mag @xcite , so this cloud should be optically thick in the soft x - ray band . this possibly accounts for the observed low intensities in the broadband surface brightness and for the soft x - ray line fluxes to the south of sgr a*. the relative lack of detected x - ray point sources toward this 20 km s@xmath36 cloud @xcite is consistent with a strong absorption effect . we also note that the si , s , and neutral fe line fluxes and the ews to the southwest of sgr a * where the negative velocity cloud is present , but where the 20 km s@xmath36 cloud is absent , are weak . this stands in contrast to the strong line emission in si , s , and neutral fe toward dense clouds in the northeast portion of the fov . the low line fluxes in this southwest region can not be caused by absorption in the 20 km s@xmath36 cloud simply because of the absence of this cloud in this direction . it is also unlikely to be caused by absorption in the negative velocity cloud because there is no evidence for substantial obscuration of the x - ray point sources in this southwest region toward the negative velocity cloud @xcite . thus , the weak x - ray line features to the southwest of sgr a * , where the 20 km s@xmath36 cloud is absent , appear to be intrinsic : that is , this region is lack of supernova activity , or perhaps of energetic star - forming activity compared to the northeast regions . based on the spectral analysis presented in [ sec : spec ] , we realize that the characteristics of the knot 3 spectrum are consistent with those of knot 1 , except for the low absorbing column density which reveals the soft thermal component . the column density toward knot 3 ( @xmath18 @xmath2 1.5 @xmath1 10@xmath19 @xmath20 ) is consistent with typical columns to the galactic center @xcite . knot 1 ( and 2 ) , on the other hand , has significantly higher column ( @xmath18 @xmath2 3.5 @xmath1 10@xmath19 @xmath20 ) . we speculate that the larger column for knot 1 ( and 2 ) might be attributable to the negative velocity molecular cloud in this direction . consistent with this hypothesis is the fact that sgra east , which has the same column as knot 3 @xcite , is not coincident with emission in the @xmath425 to @xmath45 km s@xmath36 cs map ( figure [ fig : fig4 ] ) . the apparent angular size of the @xmath425 to @xmath45 km s@xmath36 cloud to the northeast of sgr a * is @xmath26@xmath0 , so the physical extent is @xmath215 pc . if the bulk of the excess column toward knot 1 ( @xmath18 @xmath2 2 @xmath1 10@xmath19 @xmath20 ) is attributed to this cloud , a cloud density of @xmath30 @xmath2 4 @xmath1 10@xmath39 @xmath33 is implied . assuming a simple spherical geometry for the cloud , the total mass of the cloud is @xmath40 @xmath2 2 @xmath1 10@xmath41 @xmath42 . albeit crude estimates , the derived density and mass of the cloud are not unreasonable compared with other giant molecular cloud complexes in the galactic center . based on these x - ray absorption features , we therefore propose that this cloud is in the `` foreground '' of fe knot 1 ( and 2 ) , although some portions appear to be associated with knot 1 itself . the soft x - ray emission to the northeast of sgr a * , as presented by the s ew images ( figure [ fig : fig4]b ) , is apparently anti - correlated with the negative velocity cs emission , which can be accounted for by the presence of this absorbing foreground cloud . the bright , hard x - ray filaments to the northeast of sgr a * stand out for having strong neutral fe emission . the spectral characteristics of knot 1 , for example , are consistent with those of the sgr b2 and sgr c clouds , both of which have been plausibly described by an xrn model illuminated by past activity of sgr a * @xcite . knot 1 has a cometary morphology , with the `` head '' in the north and extended south by @xmath21@xmath0 , along a minor ridge of the negative velocity cs cloud . the spectral features of the strong 6.4 kev fe k@xmath11 line , a weak 7.02 kev fe k@xmath21 line , and the underlying continuum with @xmath17 @xmath2 2 are typical for the proposed xrn model @xcite . while the fe k@xmath11 ew of @xmath141 kev is weaker than that from the sgr b2 xrn @xcite , it nonetheless suggests that the fe line emission is dominated by irradiation from external sources ( sunyaev & churazov 1998 and references therein ) . since the thomson cross - section is dependent on the scattering angle of the incident photon , while the fluorescent line emission is isotropic , the observed ew the ratio of fluorescent line intensity to the scattered continuum is inversely proportional to 1 + cos@xmath43@xmath44 , where @xmath44 is the angle subtended at the scatterer between the observer and the illuminating source . if we assume that sgr a * was the primary external illuminating source for knot 1 , the lower fe ew for the knot 1 spectrum , relative to that of sgr b2 , therefore suggests that @xmath44 is near 0 or @xmath45 and thus that the primary reflecting cloud toward knot 1 is located in front of or behind the plane between sgr a * and sgr b2 . ( the ew may as well depend on other cloud parameters , such as the optical depth , abundances , and the ionization state . extensive studies of the determinants of the ew characteristics may thus require detailed xrn simulations , which are beyond the scope of the current work . ) the fe k@xmath21 to k@xmath11 line flux ratio , @xmath20.15 , is consistent with the laboratory prediction for the xrn model @xcite . the sharp absorption edge at @xmath3 @xmath2 7.1 kev is also in good agreement with previously reported xrn interpretations @xcite . the strong neutral fe line emission from knot 1 is thus well described by the xrn model . knot 2 shares the overall spectral characteristics of knot 1 . however , the molecular counterpart is weak or absent for knot 2 , and the apparent absence of the fe k@xmath21 line in the knot 2 spectrum is problematical for the xrn model . the identification of knot 2 as an xrn might thus be inconclusive . the knot 3 spectrum shows similar features to the knot 1 spectrum in the hard band ( @xmath3 @xmath25 4 kev ) , but reveals a substantial contribution from a thermal component in the soft band ( @xmath3 @xmath26 4 kev ) . the implied electron temperature ( @xmath28 @xmath2 0.3 kev ) and likely high abundances might be suggestive of emission from sn ejecta heated by a blast wave reverse shock . this suggested contribution from snr(s ) to the observed knot 3 spectrum is not surprising because there are dense molecular cloud complexes in these regions of the galactic center , which may contain active star - forming regions . in fact , the cs maps indicate associations of dense clouds with the x - ray emitting material at the position of knot 3 ( figures [ fig : fig4]c,[fig : fig4]d,[fig : fig4]e , and [ fig : fig4]f ) . based on the xrn model , a past state of dramatically enhanced activity of sgr a * , with an x - ray luminosity of @xmath46 @xmath14 10@xmath47 ergs s@xmath36 , has been proposed as the primary external illuminating source for the sgr b2 and sgr c clouds @xcite . the overall large - scale geometry of the neutral fe ew enhancements to the northeast of sgr a * is consistent with the fe emission lying predominantly within a conical region with an apex toward sgr a * ( figure [ fig : fig2]b ) . we thus consider that sgr a * might have also been the external irradiating source for the detected neutral fe knots . in the reflection models , the luminosity of the irradiating external source can be derived in terms of the observed 6.4 kev photon flux , the distance to the reflecting cloud from the source , and the apparent angular size of the cloud as viewed from the source position @xcite . assuming a solar abundance for fe and a thomson scattering depth @xmath48@xmath49 @xmath2 0.1 ( @xmath18 @xmath2 10@xmath19 @xmath20 ) , we derive an irradiating luminosity from sgr a * , @xmath46 @xmath2 10@xmath50 ergs s@xmath36 for knot 1 and @xmath46 @xmath2 10@xmath51 ergs s@xmath36 for knot 3 . these luminosities are @xmath22 @xmath4 3 orders of magnitude higher than that of the quiescent state flux of sgr a * @xcite , but only @xmath260.1% of those which might have produced the xrne in sgr b2 and sgr c clouds . the x - ray luminosity of sgr a * from the recently detected @xmath23 hr flaring event ( @xmath46 @xmath2 10@xmath50 ergs s@xmath36 ; baganoff et al . [ 2001 ] ) is comparable to the derived irradiating luminosity . considering the physical sizes of these fe knots ( @xmath22 pc for the angular sizes of @xmath21@xmath0 at @xmath10 = 8 kpc ) , however , the duration of the irradiating event must have been a few years or longer instead of on the order of @xmath2hours . the inferred sgr a * activity , if it was the irradiating source , should thus have been persistent much longer than the observed short flares . the inferred irradiating luminosities might have also been generated by galactic transient sources such as low mass x - ray binaries . there are likely a number of transient sources in active regions such as the galactic center . in fact , several transient x - ray sources have been detected near the galactic center with x - ray luminosities of @xmath46 @xmath2 10@xmath50 @xmath4 10@xmath52 ergs s@xmath36 ( e.g. , branduardi et al . 1976 ; maeda et al . 1996 ; pavlinsky et al . although durations of x - ray transient outbursts can be as long as a decade , typical burst periods are relatively short on the order of months . ( e.g. , wijnands 2001 ) . the nearby x - ray transient sources may thus not be completely ruled out as contributors to the observed neutral fe line emission , but unlikely . in terms of continuous irradiation sources , it is noteworthy that the derived irradiating fluxes are consistent with the x - ray luminosity of sgr a east @xcite . the sky position of sgr a east is between sgr a * and the fe knots , and so the steady emission from sgr a east might have also contributed the reflected neutral fe line emission at the positions of the observed fe knots . the measured temperatures for sgr a east , @xmath28 @xmath2 2 kev with _ chandra _ @xcite and @xmath28 @xmath2 4 kev with _ xmm - newton _ @xcite , however , appear to be too soft to produce a strong fe line at @xmath3 = 6.4 kev . moreover , if sgr a east is a steady illuminating source capable of producing the observed xrn features , a strong 6.4 kev fe line feature would also have been expected from the nearby neutral ism , for example , the circumnuclear disk of sgr a * ( e.g. , latkoski et al . 1999 and references therein ) , which is apparently not the case . steady emission from sgr a east is thus ruled out as a primary irradiating source . alternatively , the characteristic 6.4 kev fe line emission may originate from unresolved sn ejecta . fast nonthermal particles accelerated by the fermi mechanism in collisionless magnetohydrodynamic shocks diffuse through cold , metal - rich ejecta knots , and produce low - ionization - state k@xmath11 line x - ray emission @xcite . this is most effective when the ejecta knot is propagating through dense clouds ( @xmath37 @xmath2 1000 @xmath33 ) @xcite . the model predicts an underlying , non - thermal bremsstrahlung continuum which is harder ( @xmath17 @xmath2 1.5 ) than that expected from the xrn model ( @xmath17 @xmath2 2 ) . x - ray line broadenings caused by the fast - moving unresolved sn ejecta are also expected . the presence of the soft thermal component , particularly in knot 3 , could be apparent evidence for such a contribution from unresolved sn ejecta . although the observed continuum of knot 3 appears to be softer than the typical prediction from the sn ejecta model , the large uncertainties on the best - fit photon index of the observed underlying continuum still make the ejecta model a plausible candidate to describe the spectum . although a soft thermal component is not detected from knot 1 and knot 2 , the neutral fe line characteristics are nearly identical for all of these knots . the absence of soft thermal component from knot 1 and knot 2 may thus be merely due to absorption , and then the supernova ejecta model can not be ruled out as a source of knot 1 and knot 2 as well . the estimated fe line luminosity ( @xmath46 @xmath2 10@xmath53 ergs s@xmath36 ) can be produced by fast - moving ejecta from a single sn explosion interacting with a dense molecular cloud @xcite . we note that large - scale correlations between the neutral fe line and the sio ( j=1@xmath40 ) emission in the galactic center regions have been observed @xcite . this gas phase sio emission due to grain processing by strong shocks , being spatially coincident with the 6.4 kev fe line features , may support the sn ejecta origin of the neutral fe line emission in the galactic center . one way to distinguish between the xrn and sn ejecta models is to make conclusive measurements of the fe line broadenings . the detector position of knot 3 is close to the ccd readout ( rows @xmath26 256 ) , so that the energy resolution is achieved down to the instrumental limit ( @xmath2150 ev fwhm ) @xcite . the measured line width of the 6.4 kev line is then @xmath24 = 10@xmath54 ev for the fitted gaussian line profile . this is only a weak detection of line width , yet suggests that the fe line width might have been resolved , thus indicating a line broadening in this region . if confirmed ( by follow - up observations ) , this line broadening would provide support for the contribution from unresolved sn ejecta to the observed neutral fe line emission . observations with instruments of good spectral resolution such as _ astro - e2 _ should be helpful for this purpose . better constraints on the photon index of the underlying continuum may also be helpful in order to understand the detailed nature of these features . follow - up deep observations with _ xmm - newton _ , by utilizing the high sensitivity compared to the _ chandra_/acis , would be useful for this purpose . the sn ejecta model also predicts strong infrared emission from the radiative shocks that accompany ejecta fragments @xcite , and so infrared line observations would help to discriminate models as well . although the sn ejecta contribution to the observed neutral fe line emission appears to be plausible , we note a caveat that non - thermal radio emission features have not been reported in the northeast regions of sgr a * within the acis fov . this absence of the non - thermal radio features is thus admittedly a difficult problem in terms of the snr interpretation . the presence of highly ionized elemental species , si , s , and fe indicates that thermal diffuse x - ray emission , as well as the nonthermal component , also prevails in the galactic center . these emission lines from the he - like species over the wide range of photon energies ( @xmath3 @xmath2 2 kev for si / s and @xmath3 @xmath2 7 kev for fe ) confirm the multiple - temperature origin for the diffuse thermal emission in the galactic center / plane as indicated by the _ asca _ data ( e.g. , kaneda et al . 1997 ) . the thermal components are , with some exceptions of bright features near sgr a * , relatively faint and large - scale features . the direct spectral analysis of this diffuse thermal emission thus requires substantially more careful treatment than the bright , well - confined , nonthermal neutral fe line features . we will present the extensive spectral analysis of the galactic center thermal diffuse x - ray emission in subsequent papers ( muno et al . 2003b ; morris et al . 2003 , in preparation ) . we here qualitatively discuss the overall characteristics of the thermal emission based primarily on the image analysis . the si and s line ews trace enhancements of the soft component of the thermal emission , which is enhanced to the northeast of sgr a * ( figure [ fig : fig2]c and figure [ fig : fig2]d ) . these regions are close to the radio arc and the arches cluster , where the formation of massive stars has taken place in at least two episodes over the past 5 million years ( figer et al . 1999 ; yusef - zadeh et al . massive galactic center molecular clouds are present in these regions ( c.f . , [ sec : mc ] ) , so the observed si and s line emission most likely arises from the hot ism heated by powerful stellar winds and/or supernova blast waves . as we have discussed in previous sections , the overall broadband x - ray surface brightness and the s / si ews are low in the southwest regions , which is uncorrelated with the absorption effect . this suggests that the active star forming processes are present _ only _ in the northeast of sgr a * and that such activity is not manifested in the southwest . knot 3 provides a sample spectrum of this soft thermal component . as presented in the previous section , this thermal spectrum shows characteristics typical of hot plasma produced by snrs . the _ asca _ observations of the grxe have successfully demonstrated that the soft component emission in the plane can be furnished by snrs with a sn rate of @xmath2one per century @xcite . for the case of knot 3 , the angular size of the emission region ( @xmath21@xmath0 ) corresponds to @xmath22 pc at a distance of 8 kpc , which is typical for young galactic snrs , but might also be of a relatively older age in a dense environment such as the galactic center . the hard thermal component is representd by the he - like fe line ew ( figure [ fig : fig2]a ) . except for the brightest central feature of sgr a east , the highly ionized fe line emission is primarily distributed along the plane . in contrast to the other ews in figure [ fig : fig2 ] , this hot fe emission shows no particular enhancements in the northeast of sgr a*. the _ asca _ data suggested that this hot component might not be adequately described by the snrs , because such a hot plasma may not be confined in the plane by the galactic gravity and/or by typical interstellar pressure @xcite . in order to maintain this hot plasma in the plane , a high sn rate and a low ionization state of the plasma are required @xcite . the electron temperature of the plasma required to produce the observed strong 6.7 kev fe line emission ( @xmath28 = 5 @xmath4 10 kev ) is also unusually high for typical snrs , although we do detect the 6.7 kev fe line from metal - rich ejecta in young snrs ( e.g. , hughes et al . 2000 ; lewis et al . 2003 ) . this hot gas with @xmath28 @xmath2 5@xmath410 kev might rather have been produced when cooler plasma is confined and compressed by strong galactic magnetic fields @xcite . the required magnetic field strength to achieve this is @xmath55 = 20@xmath430 @xmath56 g . the interstellar magnetic field in the galactic center ( @xmath55 @xmath21 mg , e.g. , yusef - zadeh & morris 1987 ; lang et al . 1999 ) is sufficiently high for this model . this magnetic confinement model might not require recent massive star - forming activity or sn explosions in order to produce the observed hot plasma . the relatively uniform distribution of the he - like fe ew along the plane may also indicate this hot plasma s association with the strong magnetic field in the galactic center along the plane rather than with star - forming regions . deep _ chandra_/acis observations of the galactic center reveal a complex spatial and spectral structure of diffuse x - ray emission with significant photon statistics . the thermal ( e.g. , hot si and s ) and nonthermal ( e.g. , neutral fe ) x - ray line emission is generally enhanced in a region lying to the northeast of sgr a * , a region in which massive star - formation has been active over the past 5 million years ( e.g. , the arches cluster ) , and in which dense galactic center molecular clouds are currently present . the highly ionized line emission from si and s in these regions is therefore most likely associated with the hot ism heated by the shocks from sne and/or massive stellar winds , while shocks interacting with dense , cold ism may produce emission in the neutral fe line . curiously , we detect little or no enhancement of the emission lines to the south and southwest of sgr a*. this weak line emission to the south is apparently the result of an absence of activity there , in contrast to the northeast regions , and it is owed in part to the substantial absorption by the foreground 20 km s@xmath36 cloud . in contrast , the highly ionized he - like fe line emission shows a more uniform distribution along the plane , with no apparent enhancements to the northeast of sgr a*. this feature suggests that the young star - forming and/or sne activities have less effect on the he - like fe distribution than they do on the neutral fe line emission . the hot fe line emission might rather be caused by the hot ism trapped and then heated by the strong galactic center magnetic fields . the spectral characteristics of the strong neutral fe line emitting features to the northeast of sgr a * are plausibly described by the xrn model . we identify dense molecular counterparts for these fe knots , which is strong support for the hypothesized xrn origin . if past activity of sgr a * is the primary external illuminating source , the required luminosities are @xmath46 @xmath2 10@xmath50@xmath410@xmath51 ergs s@xmath36 with durations of at least a few years . alternatively , nonthermally - accelerated electrons by sn shocks penetrating back into dense , cold supernova ejecta may also explain the observed neutral fe line features . the presence of active star - forming regions and the dense molecular clouds near the neutral fe features are supportive of this scenario . unambiguous discriminations between these models are however difficult with the current data . follow - up observations with the current and upcoming observatories may be helpful for a detailed understanding of the origins for these features . baganoff , f. k. , bautz , m. w. , brandt , w. n. , chartas , g. , feigelson , e. d. , garmire , g. p. , maeda , y. , morris , m. , ricker , g. r. , townsley , l. k. , & walter , f. 2001 , nature , 413 , 45 baganoff , f. k. , maeda , y. , morris , m. , bautz , m. w. , brandt , w. n. , cui , w. , doty , j. p. , feigelson , e. d. , garmire , g. p. , pravdo , s. h. , ricker , g. r. , & townsley , l. k. 2003 , apj , 591 , 891 bamba , a. , murakami , h. , senda , a. , takaki , s. , yokogawa , j. , & koyama , k. 2002 , proceedings of the symposium `` new visions of the x - ray universe in the xmm - newton and chandra era '' , noordwijk - nl , nov . 2001 ( astro - ph/0202010 ) branduardi , g. , ives , j. c. , sanford , p. w. , brinkman , a. c. , & maraschi , l. 1976 , mnras , 175 , 47 bykov , a. m. 2002 , a&a , 390 , 327 figer , d. f. , kim , s. s. , morris , m. , serabyn , e. , rich , r. m. & mclean , i. s. 1999 , apj , 525 , 750 garmire , g. p. , bautz , m. w. , ford , p. g. , nousek , j. a. , & ricker , jr . , g. r. 2003 , `` x - ray and gamma - ray telescopes and instruments for astronomy '' , proceedings of spie , ed . , j. e. trmper & h. d. tananbaum , vol . 4851 , 28 ho , p. t. , jackson , j. m. , barrett , a. h. , & armstrong , t. 1985 , apj , 288 , 575 hughes , j. p. , rakowski , c. e. , burrows , d. n. , & slane , p. o. 2000 , apj , 528 , l109 kaneda , h. , makishima , k. , yamauchi , s. , koyama , k. , matsuzaki , k. , & yamasaki , n. y. 1997 , apj , 491 , 638 koyama , k. , makishima , k. , tanaka , y. , & tsunemi , h. 1986 , pasj , 38 , 121 koyama , k. , ikeuchi , s. , & tomisaka , k. 1986b , pasj , 38 , 503 koyama , k. , awaki , h. , kunieda , h. , takano , s. , tawara , y. 1989 , nature , 339 , 603 koyama , k. , maeda , y. , sonobe , t. , takeshima , t. , tanaka , y. , & yamauchi , s. 1996 , pasj , 48 , 249 lang , c. , morris , m. , & echevarria , l. 1999 , apj , 526 , 727 latvakoski , h. m. , stacey , g. j. , gull , g. e. , & hayward , t. l. 1999 , apj , 511 , 761 lewis , k. t. , burrows , d. n. , hughes , j. p. , slane , p. o. , garmire , g. p. , & nousek , j. a. 2003 , apj , 582 , 770 maeda , y. , koyama , k. , sakano , m. , takeshima , t. , & yamauchi , s. 1996 , pasj , 48 , 417 maeda , y. , baganoff , f. k. , feigelson , e. d. , morris , m. , bautz , m. w. , brandt , w. m. , burrows , d. n. , doty , j. p. , garmire , g. p. , pravdo , s. h. , ricker , g. r. , & townsley , l. k. 2002 , apj , 570 , 671 makishima , k. 1997 , in x - ray imaging and spectroscopy of cosmic hot plasmas , ed . , f. makino ( tokyo : universal academy press ) , 137 martin - pintado , j. , de vicente , p. , rodriguez - fernndez , n. j. , fuente , a. , & planesas , p. 2000 , a&a , 356 , l5 morris , m. 1993 , apj , 408 , 496 muno , m. p. , baganoff , f. k. , bautz , m. w. , brandt , w. n. , broos , p. s. , feigelson , e. d. , garmire , g. p. , morris , m. r. , ricker , g. r. , & townsley , l. k. 2003a , apj , 589 , 225 muno , m. p. , baganoff , f. k. , bautz , m. , broos , p. s. , feigelson , e. d. , garmire , g. p. , morris , m. r. , park , s. , ricker , g. r. , & townsley , l. k. 2003b , apj , submitted murakami , h. , koyama , k. , sakano , m. , & tsujimoto , m. 2000 , apj , 534 , 283 murakami , h. , koyama , k. , tsujimoto , m. , maeda , y. , & sakano , m. 2001a , apj , 550 , 297 murakami , h. , koyama , k. , & maeda , y. 2001b , apj , 558 , 687 park , s. , & ebisawa , k. , 2001 , apj , 547 , 777 park , s. , roming , p. w. a. , hughes , j. p. , slane , p. o. , burrows , d. n. , garmire , g. p. , & nousek , j. a. , 2002 , apj , 564 , l39 pavlinsky , m. n. , grebenev , s. a. , & sunyaev , r. a. 1994 , apj , 425 , 110 reid , m. j. 1993 , ara&a , 31 , 345 sakano , m. , mishiuchi , m. , maeda , y. , koyama , k. , & yokogawa , j. 1998 , iaus , `` the central regions of the galaxy and galaxies '' , ed . , y. sofue ( london : kluwer ) , 184 , 443 sakano , m. , warwick , r. s. , & decourchelle , a. 2002 , the proceedings of `` japan - germany workshop on galaxies and clusters of galaxies '' , shuzenji , japan , ( astro - ph/0212464 ) sakano , m. , warwick , r. s. , decourchelle , a. , & predehl , p. 2003 , mnras , 340 , 747 sanders , d. v. , solomon , p. m. , & scoville , n. z. 1984 , apj , 276 , 182 sidoli , l. , & mereghetti , s. 1999 , , 349 , l49 skinner , g. k. , willmore , a. p. , eyles , c. j. , bertram , d. , & church , m. j. 1987 , nature , 330 , 544 sunyaev , r. , markevitch , m. , & pavlinsky , m. 1993 , apj , 407 , 606 sunyaev , r. , & churazov , e. 1998 , mnras , 297 , 1279 townsley , l. k. , broos , p. s. , garmire , g. p. , & nousek , j. a. 2000 , apj , 534 , l139 townsley , l. k. , broos , p. s. , nousek , j. a. , & garmire , g. p. 2002a , nuclear instruments & methods in physics research section a , 486 , 751 townsley , l. k. , broos , p. s. , chartas , g. , moskalenko , e. , nousek , j. a. , & pavlov , g. g. 2002b , nuclear instruments & methods in physics research section a , 486 , 716 tsuboi , m. , ukita , n. & handa , t. 1997 , apj 481 , 263 tsuboi , m. , handa , t. & ukita , n. 1999 , apjs , 120 , 1 valinia , a. , & marshall , f. e. , 1998 , apj , 505 , 134 vollmer , b. , zylka , r. , & duschl , w. j. 2003 , a&a , in press ( astro - ph/0306200 ) wijnands , r. 2001 , in proceedings of the 113th meeting of the astronomical society of the pacific , `` the high energy universe at sharp focus : chandra science '' ( astro - ph/0107600 ) yamauchi , s. , & koyama , k. 1995 , pasj , 47 , 439 yusef - zadeh , f. , & morris , m. 1987 , apj , 320 , 545 yusef - zadeh , f. , law , c. , wardle , m. , wang , q. d. , fruscione , a. , lang , c. c. , & cotera , a. 2002 , apj , 570 , 665 zylka , r. , mezger , p. g. , & wink , j. e. 1990 , a&a , 234 , 133 ccc 0242 & 1999 - 09 - 21 & 41 + 1561a & 2000 - 10 - 26 & 36 + 1561b & 2001 - 07 - 14 & 14 + 2943 & 2002 - 05 - 22 & 35 + 2951 & 2002 - 02 - 19 & 12 + 2952 & 2002 - 03 - 23 & 12 + 2953 & 2002 - 04 - 19 & 12 + 2954 & 2002 - 05 - 07 & 12 + 3392 & 2002 - 05 - 25 & 167 + 3393 & 2002 - 05 - 28 & 158 + 3663 & 2002 - 05 - 24 & 38 + 3665 & 2002 - 06 - 03 & 90 + cccc si ( he@xmath12 ) & 1700 @xmath4 1900 & 1550 @xmath4 1640 & 2100 @xmath4 2150 + s ( he@xmath12 + ly@xmath12 ) & 2350 @xmath4 2700 & 2100 @xmath4 2150 & 2730 @xmath4 2810 + fe ( neutral ) & 6250 @xmath4 6500 & 5000 @xmath4 6100 & 7150 @xmath4 7300 + fe ( he@xmath12 ) & 6550 @xmath4 6800 & 5000 @xmath4 6100 & 7150 @xmath4 7300 + cccccccc 1 & 1.99@xmath57/ - & 32.9@xmath58 & 6.40@xmath59 & 1.19@xmath60 & [email protected] & 1.3 & 52.9/51 + 2 & 3.12@xmath61/ - & 36.8@xmath62 & 6.40@xmath59 & 1.03@xmath63 & [email protected] & 2.4 & 33.7/44 + 3 & 1.77@xmath64/0.31@xmath65 & 15.8@xmath66 & 6.39@xmath67 & 1.29@xmath60 & [email protected] & 1.5 & 134.7/121 + +

this paper presents the generalization of weighted distances to modules and their computation through the chamfer algorithm on general point - lattices . a first part is devoted to the definitions and properties ( distance , metric , norm ) of weighted distances on modules , with a presentation of weight optimization in the general case , to get rotation invariant distances . a general formula for the weighted distance on any module is presented . the second part of this paper proves that , for any point - lattice , the sequential 2-scan chamfer algorithm produces correct distance maps . finally , the definitions and computation of weighted distances are applied to the face - centered cubic ( fcc ) and body - centered cubic ( bcc ) grids . weighted distance , distance transform , chamfer algorithm , non - standard grids

0808.0665

given a binary image consisting of object and background grid points , the distance transform assigns to each object grid point its distance value to the closest background grid point . the distance transform provides valuable information for shape analysis and is widely used in several applications @xcite , such as , for example , skeleton extraction @xcite , template matching @xcite , shape based interpolation @xcite or image registration @xcite . computing the distance from each object grid point to each background grid point would lead to a far too high computational cost . numerous authors have thus investigated alternative ways of computing distance maps . to do so , they use the spatial consistency of a distance map , which allows propagation of local information . in this paper , we focus on one such algorithm , the chamfer algorithm , which computes the _ weighted distance transform _ ( wdt ) . when using the chamfer algorithm , only a small neighborhood of each grid point is considered . a weight , a _ local distance _ , is assigned to each grid point in the neighborhood . by propagating the local distances in the two - scan algorithm , the correct distance map is obtained . for example , the well - known two dimensional city - block ( @xmath0 ) and chessboard ( @xmath1 ) distances can be obtained in this way by using unit weights for the neighbors . it has been shown that in two dimensions , the hexagonal grid is in many ways to prefer over the usual square grid , @xcite . for example , since the hexagonal grid constitutes the closest sphere packing @xcite , only @xmath2 of the number of samples needed in the square grid can be used without information loss , @xcite . extending this to three dimensions , the face - centered cubic ( fcc ) grid is the lattice with the highest packing density , resulting in that less samples are needed for the fcc grid without loosing information , compared to the cubic grid , @xcite . the reciprocal grid of the fcc grid is the body - centered cubic ( bcc ) grid . this means that if the fcc grid is used in the spatial / frequency domain , then the bcc grid is used to represent the image in the frequency / spatial domain . the fcc grid has the highest packing density , so this grid is preferably used in frequency domain resulting in that only @xmath3 of the samples are needed for an image on the bcc grid compared to the cubic grid . if the fcc grid is used in spatial domain instead , @xmath4 of the samples can be used without loosing information , @xcite . this reasoning requires that it is possible to acquire images directly to the grids with non - cubic voxels . in theory , any tomography - based technique for acquiring volume images ( i.e. , that gives a sequence of 2d slices ) such as magnetic resonance imaging ( mri ) , computed tomography ( ct ) , single photon emission computed tomography ( spect ) , or positron emission tomography ( pet ) can be adjusted to work on the fcc or bcc grids . for spect and ct , using the filtered back - projection method or direct fourier methods , the fact that the fourier transform for the fcc and bcc grids ( and for two - dimensional planes that correspond to the 2d slices ) exists is enough . images on the fcc and bcc grids acquired using the algebraic reconstruction technique for spect or ct is found in @xcite . they use spherically - symmetric volume elements ( blobs ) , @xcite . with the pet - technique , the origin of gamma - rays ( produced by annihilation of an emitted positron and an electron ) are computed . the technique does not depend on the underlying grid except in the digitization step , which just as well can be performed on the fcc and bcc grids . dealing with digital images means dealing with discrete grids . indeed , digital images are discrete representation of the continuous world . two kind of approaches can be adopted to deal with discrete grids . the first one applies continuous methods to the discrete grid . this is the case of the euclidean distance transform ( edt ) which aims at computing exact euclidean distance on discrete images . the second approach , called discrete geometry , consists in developing tools directly devoted to digital images and discrete grids . this is the case of the wdt which can , for example , be computed using a two - scan algorithm , a _ chamfer algorithm _ in the usual square grid @xcite . since computing the wdt is less memory demanding and faster than computing the edt , the wdt is preferable in situations where the data set is too big to fit into the central memory ( such as in @xcite ) or the computational time should be minimized ( recent contributions involve speedup of level - set methods , @xcite , and real - time tele - operated force - feedback , @xcite ) . it should be mentioned that there are linear - time ( or almost linear ) algorithms for computing the edt obtained by sequentially propagating vectors , @xcite , and dimensionality reduction , @xcite . these algorithms are , however , not as fast as the chamfer algorithm since the edt requires a larger number of arithmetic operations . also , the vector propagation algorithm is never guaranteed to be error - free , @xcite . the wdt has other properties which makes it useful in applications . it consists of only integers without loosing the metric property which is not the case for the edt ( real - valued distances ) or the squared edt ( does not satisfy the triangular inequality ) . this leads to nice properties , e.g. , when extracting the centers of maximal balls for computing skeletons which is fast and simple for the most common wdts , @xcite , in contrast to the case for edt , @xcite . observe that , when using large neighborhoods in the wdt , then the situation can be complex also for wdts , @xcite . in general , the weighted distances are well - suited for morphological operations , @xcite . to allow image processing on images on the non - standard grids , algorithms must be designed to work directly on these grids . weighted distances and the chamfer algorithm has been applied to non - standard grids such as the two - dimensional hexagonal grid , @xcite and the fcc and bcc grids , @xcite . in this paper , the weighted distance and the chamfer algorithm are considered on _ modules_. this gives a very general framework in which @xmath5 , the hexagonal , fcc , and bcc grids are all considered in parallel . the general framework implies that the theory also applies to the higher - dimensional generalizations of these grids . this paper is organized as follows : we first summarize weighted distance definitions and properties that can be found in the literature and generalize them to sub - modules of @xmath6 . then we exhibit conditions for the chamfer algorithm to work on our framework . finally , we propose examples of chamfer masks for the body - centered cubic ( bcc ) grid and the face - centered cubic ( fcc ) grid . this section generalizes definitions and weighted distance properties found in literature to our general framework of modules . we denote the set of real numbers , the set of integers , and the set of natural numbers , with @xmath7 , @xmath8 , and @xmath9 , respectively . let @xmath10 be an abelian group ( @xmath11 could be , for example , @xmath6 or @xmath12 for @xmath13 ) . to define weighted distances , we need not only an internal operator ( here @xmath14 ) , but also an external operation : @xmath15 with @xmath16 and @xmath17 . vector spaces ( as for example @xmath18 ) handle external operations , but are too restrictive ( @xmath19 , for example , is not a vector space ) . to be general enough , we use modules . let @xmath20 be a commutative ring ( for example @xmath21 ) with two neutral elements @xmath22 and @xmath23 . a set @xmath11 is called a _ module on @xmath20 _ ( or @xmath20-module ) if @xmath11 has a commutative group operation @xmath14 , an external law @xmath24 , and satisfies the following properties : @xmath25 we now consider sub - rings @xmath26 of @xmath27 . given an abelian group @xmath10 we consider the @xmath20-module @xmath28 . [ rem : vspace_module ] the main difference between a module and a vector space is the non - invertibility ( with respect to the external law ) of the elements of its associated ring @xmath20 ( e.g. @xmath29 but @xmath30 ) . a basis of a module @xmath11 of dimension @xmath31 is a family of @xmath31 independent vectors @xmath32 ( @xmath33 \textrm { , } \alpha_{i } = 0 $ ] ) . but a linearly independent family of @xmath31 vectors may not be a basis of @xmath11 . for example , @xmath34 is a family of two independent vectors of @xmath35 but is not a basis of @xmath35 ( @xmath36 can not be reached by a linear combination of @xmath37 and @xmath38 with coefficients taken in the ring @xmath8 ) . a _ distance _ on a group @xmath11 , having values in @xmath20 , called @xmath39 is a function @xmath40 which satisfies the following properties : @xmath41 given a distance @xmath39 , @xmath42 is called a _ metric _ if it also satisfies the following property : @xmath43 given a metric @xmath39 , @xmath42 is called a _ norm _ on the module @xmath28 if it also satisfies the following property : @xmath44 an _ image _ is a function @xmath45 , where @xmath46 is a finite subset of @xmath11 . [ def : dist - map ] given a binary image @xmath47 , i.e. , @xmath48,let @xmath49 be the foreground and @xmath50 be the background . given a distance @xmath51 , the _ distance map _ , @xmath52 , of @xmath47 is a grey level image , @xmath53 , where the value of each point of the foreground corresponds to its shortest distance to the background , i.e. @xmath54 given a module ( i.e. a set of vectors ) @xmath55 and an affine space ( i.e , a set of points in @xmath47 ) @xmath56 , we say that @xmath56 is equivalent to @xmath11 in the the following sense : @xmath57 this vector @xmath58 is denoted @xmath59 . for every module @xmath11 , there exists an equivalent affine space @xmath56 and vice versa . these spaces always have the same dimension . given a point @xmath60 of @xmath56 , @xmath61 is an affine space with an origin . an origin is required to build a basis of @xmath56 and allows to define the following operations . @xmath62 in the following , we will consider @xmath56 and @xmath11 as the same set and denote @xmath63 . we now consider more particularly sub - modules of @xmath64 , ( for example @xmath65 and @xmath66 ) equivalent to the affine space @xmath6 , with the origin @xmath67 and the canonical basis @xmath68 . for any vector or point @xmath69 , we denote @xmath70 the decomposition of @xmath71 in this canonical basis . among numerous other methods , distance maps can be computed by propagation of local distances using a chamfer mask . the latter is defined as follows : [ def : chamfer - mask ] a _ chamfer mask _ @xmath72 is a finite set of weighted vectors @xmath73 } \in \mathcal{g } \times \mathcal{r } \}$ ] which contains a basis of @xmath55 and satisfies the following properties : @xmath74 intuitively , we speak about a distance between two points as the length of the shortest path between these points . in the case of weighted distances , we restrict the possible paths to those allowed by the chamfer mask . in this case , a path is an ordered sequence of chamfer mask vectors @xmath75 with @xmath76 having its origin at @xmath77 , @xmath78 having its end at @xmath79 , and @xmath80 $ ] . in other words , @xmath81 { { \overrightarrow{\mathbf { v}}}}_{i_{j } } \in \mathcal{c}$ ] and @xmath82 . as @xmath11 is an abelian group and a module on @xmath20 , the order of the vectors does not matter , and we can bring together the different vectors @xmath83}$ ] of the path equal to the same mask vector @xmath84 $ ] . observe that @xmath85 is the number of vectors within the path and @xmath86 is the size of the chamfer mask . moreover , as for each vector @xmath87 , @xmath88 , we can consider only positive coefficients . we obtain the following definition : [ def : chamfer - path ] given a chamfer mask + @xmath89 } \in \mathcal{g } \times \mathcal{r}\}$ ] and two points @xmath90 , a path @xmath91 from @xmath77 to @xmath79 is a sequence of vectors @xmath92 of the mask @xmath72 such that : @xmath93 \textrm { , } \alpha_{k } \in \mathcal{r}^{+}\ ] ] [ def : chamfer - cost ] the cost @xmath94 of a such a path @xmath91 is defined by : @xmath95 since a mask @xmath72 contains a basis of @xmath11 , and is symmetric , such a path always exists for any couple of points @xmath96 with positive coefficients . [ def : weighted - distance ] a weighted distance @xmath97 associated with a chamfer mask @xmath72 between two points @xmath77 and @xmath79 in @xmath11 is the minimum of the costs associated with paths @xmath91 linking @xmath77 to @xmath79 . @xmath98 a weighted distance is invariant under translation . this proof can be found in @xcite . given any three points @xmath99 in @xmath56 , @xmath100 . indeed , @xmath101 and any path @xmath102 from @xmath103 to @xmath104 is also a path from @xmath77 to @xmath79 , with the same cost : @xmath105 . a property of a weighted distance between any two points @xmath106 of @xmath11 can be expressed as a property of a weighted distance between the origin @xmath60 and a point @xmath77 . taking @xmath107 leads to @xmath108 . in the following , we will express the properties of weighted distances from the origin to any point @xmath77 of @xmath11 and denote @xmath109 . to be able to forecast the weighted distance map , from a single point , from the chamfer mask @xmath72 , we divide @xmath72 into several sectors spanning the points of @xmath11 such that the distance to the origin of a point lying within a sector will only depend on a restricted number ( actually @xmath31 the dimension ) of mask weights . we define the following objects : given a family of @xmath31 independent vectors of @xmath11 , @xmath110}$ ] , the _ @xmath7-sector _ @xmath111 is the region of @xmath6 spanned by the vectors @xmath112 i.e. : @xmath113 the _ @xmath11-sector _ @xmath114 is the set of points belonging to @xmath11 which are included in the @xmath7-sector @xmath115 . example of @xmath7-sector and @xmath11-sector in @xmath116 . ] it is possible that a @xmath11-sector does not correspond to the set of points @xmath77 such that @xmath117 . for example , as illustrated in figure [ fig : r - g - sectors ] , if we take the module @xmath118 , the point @xmath119 lies in the @xmath7-sector @xmath120 @xmath121 and as it is a point of @xmath122 , it also lies in the @xmath11-sector @xmath123 . but it can not be written as @xmath124 with @xmath125 . [ remark : r - g - sectors ] we call a _ wedge _ of a chamfer mask @xmath72 , a @xmath11-sector formed by a family @xmath126 of @xmath31 vectors of @xmath72 which does not contain any other vector of @xmath72 . to avoid the situation illustrated in remark 2.2 we consider only @xmath11-sectors based on a basis of @xmath11 . we call a @xmath11-sector @xmath127 where + @xmath128 is a basis of @xmath129 a @xmath11-basis - sector . by definition of a basis , a @xmath11-basis - sector corresponds exactly to the the set of points @xmath77 such that @xmath130 . given a family @xmath131}$ ] of @xmath31 independent vectors , we denote @xmath132 @xmath133 and @xmath134 $ ] , we consider the function @xmath135 such that : @xmath136 [ lemma : value ] a family @xmath137}$ ] is a basis of @xmath138 , iff @xmath139 \textrm { } \frac{1}{\delta_{\mathcal{f}}^{0 } } \times \delta_{\mathcal{f}}^{k}({{\overrightarrow{\mathbf { x } } } } ) \in \mathcal{r}.\ ] ] the proof in @xmath140 can be found in @xcite . + by definition of a basis , a family @xmath141}$ ] is a basis of @xmath142 iff @xmath143 , @xmath144 such that @xmath145 . + as we consider sub - modules of @xmath6 , @xmath146 and @xmath134 $ ] , @xmath147 . if the vectors of @xmath148 are not independent , @xmath148 is not a basis of @xmath11 , and as @xmath149 , @xmath150 . if @xmath148 is an independent family of vectors , as @xmath6 is a vector space , @xmath148 is a basis of @xmath151 and @xmath152 such that @xmath145 . this can be written : @xmath153 as @xmath148 is an independent family , @xmath154 and the matrix @xmath155 can be inverted such that @xmath156 inverting @xmath155 using cramer s rule leads to @xmath157 \textrm { , } \alpha_{k } = \frac{1}{\delta_{\mathcal{f}}^{0 } } \times \delta_{\mathcal{f}}^{k},\ ] ] and @xmath148 is a basis of @xmath158 iff @xmath134 $ ] , @xmath159 . when @xmath160 we obtain the condition @xmath161 for cone regularity defined in @xcite . we can always organize @xmath72 as a set of wedges ( taking @xmath31-tuples of vectors that do not contain any other vectors ) . to avoid the situation in remark 2.2 and to be able to forecast the final weighted distance from the chamfer mask , we choose masks whose wedges are all @xmath11-basis . when @xmath162 , taking every couple @xmath163 of adjacent vectors of @xmath72 in clockwise ( or counter - clockwise ) order gives such an organization . if , for each couple @xmath164 , then , they are a @xmath11-basis ( cf . @xcite ) . for example , in figure [ fig : shortestpath ] of remark [ rem : shortestpath ] , @xmath165 , @xmath166 , @xmath167 , @xmath168 , @xmath169 , @xmath170 is an organization of @xmath72 in @xmath171-basis wedges . for @xmath172 , this organization may be more complicated . indeed several ways of organizing @xmath173 independent vectors into two wedges may exist : for example , if we take the vectors @xmath174 , @xmath175 , @xmath176 , @xmath177 and their symmetric vectors , the wedges @xmath178 and @xmath179 are a @xmath11-basis , but we can also consider the wedges @xmath180 and @xmath181 . in @xcite , an automatic recursive method is given ; it is based on farey series to organize chamfer masks of @xmath182 with @xmath11-basis wedges . in the general case , considering a mask @xmath72 containing only a @xmath11-basis @xmath183 and their symmetric wedges @xmath184 , leads to two @xmath11-basis wedges ( if @xmath185 , \frac{1}{\delta_{\mathcal{f}}^{0 } } \delta_{\mathcal{f}}^{k}({{\overrightarrow{\mathbf { x } } } } ) \in \mathcal{r}$ ] , @xmath186 as @xmath20 is a group ) . moreover , the other wedges are @xmath187 for @xmath188 $ ] . they all are @xmath11-basis wedges as @xmath189 , \frac{1}{\delta_{\mathcal{f}_{l}}^{0 } } \delta_{\mathcal{f}_{l}}^{k}(x ) = \pm \frac{1}{\delta_{\mathcal{f}}^{0 } } \delta_{\mathcal{f}}^{k}(x)\in \mathcal{r}$ ] as @xmath20 is a group . we can then add vectors to this mask , taking care of keeping an organization in @xmath11-basis wedges . [ rem : shortestpath ] given a chamfer mask @xmath190 } \in \mathcal{g } \times \mathcal{r}\}$ ] which is organized in @xmath11-basis wedges , and given any point @xmath77 lying in a wedge @xmath191 of @xmath72 , there exist a path @xmath192 with @xmath134 $ ] ( i.e. the point @xmath77 can by reached by a linear combination of only @xmath31 vectors among the @xmath86 vectors of the mask @xmath72 with coefficients in @xmath193 ) . however , the final weighted distance at this point may not be a linear combination of the @xmath31 weights corresponding to _ these _ @xmath31 vectors in the chamfer mask . for example , in @xmath194 , let us consider the mask @xmath72 containing the weighted vectors @xmath195 , @xmath196 , @xmath197 and their symmetric vectors . the families @xmath198 and @xmath199 , generating the wedges @xmath200 and @xmath201 respectively , are a basis of @xmath202 . indeed , @xmath203 and @xmath204 , @xmath205 , @xmath206 and : @xmath207 @xmath208 in the same way for @xmath209 , we have : @xmath210 and @xmath211 , @xmath205 , @xmath206 and : @xmath212 @xmath213 example where the weighted distance does not depend on @xmath11-basis sector vectors . ] however , as illustrated in figure [ fig : shortestpath ] the point @xmath214 lying in the @xmath11-basis - sector @xmath215 can be reached only by the path @xmath216 containing only vectors of @xmath217 and @xmath218 . but @xmath77 can also be reached by the following path @xmath219 belonging neither to the sector @xmath215 nor to the sector @xmath220 with a cost @xmath221 . as the weighted distance of a point ( with respect to the origin ) is the minimum cost of all paths allowed by the mask , we have @xmath222 and thus @xmath223 . to avoid the situation mentioned in remark [ rem : shortestpath ] , we add restrictions to the mask weights . these restrictions rely on the fact that the polytope formed by the chamfer mask vectors normalized by their weights is convex . we call the polytope of @xmath6 whose faces are the @xmath224-dimensional pyramids formed by the @xmath31 vectors of each wedge of a chamfer mask @xmath225 } \in \mathcal{g } \times \mathcal{r}\}$ ] normalized by their weights , i.e. @xmath226 $ ] for each wedge @xmath227 of @xmath72 the _ normalized chamfer mask polytope _ , denoted @xmath228 . note that as @xmath28 is a module but not a vector space , these points may not be in @xmath11 . [ lemma : wedge : convex ] if the normalized polytope @xmath228 is convex , the weighted distance of any point @xmath77 lying inside a wedge @xmath229 of a chamfer mask @xmath230 } \in \mathcal{g } \times \mathcal{r } \}$ ] depends only on the weights @xmath231 . this proof can be found in @xcite for @xmath202 . + if @xmath232 , the proof is obvious . let a point @xmath233 , @xmath234 be given . as @xmath235 is a @xmath11-basis , there exists a path @xmath236 from @xmath237 to @xmath77 containing only vectors of @xmath148 having positive coefficients ( @xmath238 , \alpha_{k } \geq 0 $ ] ) . the cost of this path is @xmath239 . we can write @xmath240 as @xmath241 with @xmath242 since @xmath243 is a convex combination of the @xmath31 normalized vectors of @xmath244 $ ] , @xmath245 and @xmath246 , @xmath243 lies on @xmath228 . moreover , as the faces of @xmath228 are convex ( @xmath224-dimensional polytope with @xmath31 vertices ) @xmath243 also lies on the face formed by the family @xmath148 of @xmath228 , i.e. on the boundary of @xmath228 . + let us now consider another path @xmath247 from @xmath237 to @xmath77 containing arbitrary vectors of @xmath72 . as @xmath72 is symmetric , we can take @xmath134 , \beta_{k } \geq 0 $ ] without loss of generality . then @xmath248 . indeed , @xmath249 with @xmath250 @xmath251 is a convex combination of @xmath86 normalized vectors of @xmath72 , and lies thus within the convex polytope @xmath228 . moreover , we have @xmath252 with @xmath243 and @xmath251 having the same direction ( @xmath253 ) with positive coefficients . @xmath254 . as @xmath251 lies within @xmath228 and @xmath243 lies on the boundary of @xmath228 , @xmath255 ( as @xmath228 is centered in @xmath237 , if @xmath256 , @xmath257 would be farther from the origin than a point of the boundary , and thus outside @xmath228 ) thus @xmath258 and @xmath259 . figure [ fig : convex : proof ] shows an example of vectors @xmath243 and @xmath251 of a mask @xmath260 . example where @xmath228 is convex . we consider the vectors @xmath261 and the mask @xmath262 , @xmath263 , @xmath264 , @xmath265 , @xmath266 , @xmath267 . we consider the path @xmath268 and @xmath269 from @xmath237 to @xmath270 . we then have @xmath271 . ] [ corollary : convexhull ] if the vertices of each face of the convex hull @xmath272 of @xmath228 are normalized vectors corresponding to @xmath11-basis sectors of @xmath72 , then the vectors of @xmath72 whose corresponding normalized vectors do not lie on the convex hull of @xmath228 are not used to compute the final weighted distance . the proof can be found in @xcite for vectorial spaces . suppose there exists a vector @xmath273 such that @xmath274 does not lie on the convex hull of @xmath228 ( i.e. @xmath275 lies within @xmath228 ) . let us consider the @xmath11-basis sector @xmath276 in which @xmath277 lies and whose corresponding normalized vectors form a face @xmath148 of @xmath278 . there exists a path @xmath279 from @xmath237 to @xmath280 and with @xmath281 . indeed , @xmath282 with @xmath283 a convex combination of @xmath31 vectors of @xmath148 lying within @xmath148 . we thus have @xmath284 and @xmath285 . finally , as a weighted distance is the minimum of the costs of all possible path , in any path containing @xmath277 , @xmath277 will be replaced by @xmath240 whose cost is lower . note that this is what happens in remark [ rem : shortestpath ] . [ rem : convexhull ] if there exist faces of @xmath278 formed by normalized vectors whose corresponding mask vectors are not @xmath11-basis , the vectors of @xmath72 which does not lie on @xmath278 may be used , but this leads to a final weighted distance which may not be homogeneous along some directions , and we also may not be able to forecast the weighted distance inside a wedge with the only vectors generating the wedge . for example , if we consider the mask @xmath72 with the weighted vectors : @xmath286 , @xmath287 , @xmath288 , @xmath289 and their symmetric vectors , as shown in figure [ fig : nonconvex ] , @xmath290 . moreover , @xmath291 does not only depend on @xmath292 and @xmath293 or only on @xmath293 and @xmath294 . note that the vectors @xmath292 and @xmath294 generating the corresponding face of @xmath278 do not form a @xmath122 basis ( @xmath295 ) . when @xmath296 the weighted distance may not be homogeneous . ] in the following , we consider only chamfer masks whose normalized polytope is convex . indeed , if this is not the case , the mask may be redundant ( corollary [ corollary : convexhull ] ) , or even if not , we may not be able to forecast the final weighted distance , ( remark [ rem : convexhull ] ) . we can note that this condition implies that collinear vectors ( which are note opposite vectors ) are removed from the mask . considering previous lemmas and remarks , we can re - define a chamfer mask with stronger conditions as follows : [ def : chamfer - mask : restricted ] a _ chamfer mask _ @xmath297 is a finite set of weighted vectors @xmath73 } \in \mathcal{g } \times \mathcal{r } \}$ ] which satisfies the following properties : @xmath298 [ theorem : value ] given a chamfer mask @xmath299 } \in \mathcal{g } \times \mathcal{r } \right\}$ ] , defined as in definition [ def : chamfer - mask : restricted ] , the weighted distance of any point @xmath77 lying in a wedge @xmath229 can be expressed by : @xmath300 this formula was given for @xmath182 in @xcite without the entire proof . let @xmath77 be a point of @xmath11 lying in the wedge @xmath301 of @xmath297 . as @xmath302 is a basis of @xmath55 ( condition [ maskcond : wedges ] of definition [ def : chamfer - mask : restricted ] ) there exists a path @xmath240 from @xmath237 to @xmath77 and @xmath303 such that @xmath304 . the proof of lemma [ lemma : value ] gives @xmath134 , \alpha_{k } = \frac{1}{\delta_{\mathcal{f}}^{0 } } \times \delta_{\mathcal{f}}^{k}({{\mathbf { p}}})$ ] and the proof of lemma [ lemma : wedge : convex ] gives that the weight of @xmath240 is minimal . thus @xmath305 . a weighted distance computed with a chamfer mask + @xmath306 } \in \mathcal{g } \times \mathcal{r}\}$ ] as defined in definition [ def : chamfer - mask : restricted ] is a _ norm _ on @xmath307 . here , we have to show that @xmath308 is definite , positive and symmetric and satisfies the triangular inequality and the positive homogeneity properties . by definition of a weighted distance ( definition [ def : weighted - distance ] ) , given any points @xmath309 , there exist @xmath310 , @xmath311 , ... , @xmath312 @xmath313 such that @xmath314 and @xmath315 ( cf . definitions [ def : chamfer - path ] , [ def : chamfer - cost ] and [ def : weighted - distance ] ) . 1 . _ positivity _ ( needs conditions [ maskcond : positiveweights ] and [ maskcond : symmetry ] of definition [ def : chamfer - mask : restricted ] ) + by condition [ maskcond : symmetry ] of definition [ def : chamfer - mask : restricted ] we can choose @xmath316 for all @xmath317 $ ] . by condition [ maskcond : positiveweights ] we also have @xmath318 for all @xmath317 $ ] . we thus have @xmath319 . _ definitivity _ ( needs conditions [ maskcond : positiveweights ] and [ maskcond : symmetry ] of definition [ def : chamfer - mask : restricted ] ) + if @xmath232 , then the path @xmath320 is a path from @xmath237 to @xmath77 and its cost is @xmath321 . thus , @xmath322 . by positivity , we also have @xmath323 . it follows that @xmath324 . + conversely , if @xmath324 , as @xmath325 $ ] , @xmath326 , we have @xmath327 $ ] , @xmath328 and @xmath329 . thus @xmath232 . symmetry _ ( needs the condition [ maskcond : symmetry ] of definition [ def : chamfer - mask : restricted ] ) + by condition [ maskcond : symmetry ] of definition [ def : chamfer - mask : restricted ] , we have that @xmath330 is a path from @xmath77 to @xmath237 and @xmath331 . moreover , this cost is minimal . indeed , let us consider a path @xmath332 such that @xmath333 . then @xmath334 is a path from @xmath237 to @xmath77 and @xmath335 which is impossible by definition of a weighted distance ( definition [ def : weighted - distance ] ) . thus @xmath336 . 4 . _ triangular inequality _ ( needs conditions [ maskcond : positiveweights ] and [ maskcond : symmetry ] of definition [ def : chamfer - mask : restricted ] ) + the proof can be found in @xcite for @xmath337 and in @xcite ( corollary 3.4 ) for a general module . we want @xmath338 , @xmath339 . let @xmath340 be the minimum cost path between @xmath237 and @xmath79 ( i.e. @xmath341 ) , and by eq . ( [ alpha - form ] ) , @xmath342 . as the mask is symmetric , @xmath343 can be written @xmath344 for some @xmath345 and @xmath346 . let us suppose @xmath347 . the path @xmath348 is a path from @xmath77 to @xmath79 . as @xmath349 is an abelian group , @xmath350 which is impossible by definition of a weighted distance ( definition [ def : weighted - distance ] ) . by contradiction , we have @xmath351 . + + note that a decomposition in @xmath11-basis wedges is not needed for this condition . the only condition needed is to be able to extract a basis among all mask vectors . this is the case , for example , for masks containing only vectors corresponding to knight displacements . indeed , each wedge of this mask is not a @xmath8-basis ( see figure [ fig : r - g - sectors ] and remark 2.2 . however , this mask induces a metric , @xcite . + the triangular inequality does not depend on the choice of the weights . 5 . _ positive homogeneity _ ( needs conditions [ maskcond : positiveweights ] , [ maskcond : symmetry ] , [ maskcond : wedges ] and [ maskcond : convex ] of definition [ def : chamfer - mask : restricted ] ) + the proof can be found in @xcite for @xmath352 and in @xcite for a general module . let be @xmath353 . let @xmath354 be the wedges of @xmath297 in which @xmath77 lies . by theorem [ theorem : value ] , we have @xmath355 . + if @xmath356 , the point @xmath357 also lies in the wedge @xmath302 ( it is a point of @xmath11 , and @xmath358 with @xmath134 , \lambda \times \alpha_{i_{k } } \in \mathcal{r}^{+}$ ] ) . by theorem [ theorem : value ] we have : @xmath359 + if @xmath360 , @xmath361 with @xmath362 ( symmetry property ) and thus , @xmath363 . @xmath364 in addition to their metric and norm properties , weighted distances can be made more invariant to rotation . as a weighted distance is obtained by computing the smallest weight of several paths between two points , the first improvement to obtain a weighted distance with high rotational invariance is to allow a larger number of allowed directions for the paths . this means increasing precision by increasing the number of weighted vectors of the chamfer mask . to increase accuracy , another way is to choose suitable weights for mask vectors . this more challenging issue as been often addressed in the literature . the first optimal chamfer weights computation was performed for a 2-d @xmath365 mask in a square grid @xcite . then authors computed optimal weights with different optimality criteria @xcite , for larger masks @xcite and for anisotropic grid @xcite . authors of @xcite proposed an automatic computation of optimal chamfer weights for rectangular grids . observe that computing the distance transform and finding optimal weights is not directly related to the problem of estimating the length of straight lines in a discrete image @xcite . for optimal weights for such estimations , see @xcite . in all of the previous papers , the computation of optimal chamfer weights is performed the same way : 1 . first , a chamfer mask is built and decomposed in wedges . then , the final weighted distance from the origin to an arbitrary point of the grid is expressed . the variables corresponding to the mask weights are unknown , but variables corresponding to vector coordinates are known . in the general case , given a chamfer mask defined as definition [ def : chamfer - mask : restricted ] , and a point @xmath17 lying in the wedge @xmath366 the value of the weighted distance at this point is @xmath367 3 . in the same way , the euclidean distance from the origin to this point @xmath77 is expressed . in the general case , given a grid which is a sub - module of @xmath6 with an elongation of @xmath368 in each canonical direction , the euclidean distance between the origin and the point @xmath77 can be expressed in the following way : @xmath369 4 . after these three steps , the error between the weighed distance and the euclidean one can be expressed for any point @xmath17 . this error can be either absolute ( the difference between these two values ) @xcite or relative ( the difference is divided by the euclidean distance ) @xcite . the error can be expressed as follows : @xmath370 the general relative error can be expressed as follows : @xmath371 5 . the previous errors are @xmath31-dimensional functions of the coordinates @xmath372 , ... @xmath373 of @xmath77 . to reduce the number of these variables and be able to find extrema , the maximal error is computed either on a hyperplane or on a sphere . for example , in three dimensions , the error can be computed on a plane @xmath374 @xcite or on the sphere of radius @xmath375 @xcite . this error function is continuous on a compact set ( the @xmath224 pyramid formed by the @xmath31 vectors of each wedge ) . thus it is bounded and attains its bounds . these bounds can be located either at the vertices of the pyramid or inside ( including other bounds such as edges ) the pyramid . computing the derivatives of the error function gives the point @xmath376 at which the error is maximum ( due to the sign of the derivatives ) inside the wedge . in this way , the maximum error within the wedge @xmath377 can be obtained ( if @xmath376 lies within the wedge ) . 7 . the other extrema ( minima ) @xmath378 are obtained for the @xmath31 vectors delimiting the wedge ( the vertices of the @xmath224 pyramid ) . when computing the error on a sphere of radius @xmath379 , in the general case , the extrema can be expressed in the following way ( @xmath380 $ ] ) : @xmath381 with @xmath382 being the euclidean norm of the vector @xmath277 expressed in the world coordinate . 8 . minimizing the maximum error leads to computing optimal real weights with the following equation : @xmath383 . 9 . a depth - first search in an integer weights tree taking the error into account can lead to optimal integer weights set for a given mask @xcite . we have generalized weighted distance properties found in the literature to modules . these properties are true for the well - known cubic grid , but can also be applied to other grids such as the fcc and bcc grids as we will see in section [ sec : bccfcc ] . in applications , an efficient algorithm for computing distance transforms is is needed . in the following section , we discuss such an algorithm : the two - scan chamfer algorithm . we prove that the algorithm produces correct result for images on general point lattices . there are basically three families of algorithms for computing weighted distance transforms bucket - sort ( also known as wave - front propagation ) , parallel , and sequential algorithms . initially in the bucket - sort algorithm @xcite , the border points of the object are stored in a list . these points are updated with the distance to the background . the distances are propagated by removing the updated points from the list and adding the neighbors of these points to the list . this is iterated until the list is empty . the parallel algorithm @xcite is the most intuitive one . given the original image , a new image is obtained by applying the chamfer mask simultaneously to each point of the image giving the minimum distance value at each of these points . this process is applied to the new image . by applying the procedure iteratively until stability , the distance map is obtained ( a proof using infimal convolution is found in @xcite ) . rosenfeld & pfaltz @xcite showed that this is equivalent to two sequential scans of the image for grids in @xmath35 and a @xmath384 mask . sequential means that we use previously updated values of the same image to obtain new updated values . two advantages are that a two - scan algorithm is enough to construct a distance map in any dimension and that the complexity is known ( @xmath385 , where @xmath386 is the number of points in the image ) . the worst case complexity of the parallel algorithm is @xmath387 @xmath386 iterations of operations on @xmath386 points . this result is now generalized and proved to be correct on any grid for which all grid point coordinates can be expressed by a basis . in this section , @xmath46 is a _ finite _ subset of @xmath11 and @xmath388 . the cardinality of @xmath46 is denoted @xmath389 . observe that the chamfer weights are in @xmath390 ( definition [ def : chamfer - mask : restricted ] , condition [ maskcond : positiveweights ] ) and that the weighted distance have values in @xmath391 ( definition [ def : weighted - distance ] ) . given @xmath392 , where @xmath393 and @xmath394 . let @xmath395 where @xmath396 is one of the relations @xmath397 or @xmath398 . for example , @xmath399 is a hyperplane and @xmath400 , @xmath401 , @xmath402 , and @xmath403 are half - spaces separated by the hyperplane @xmath399 . [ def : scanningmasks ] let a chamfer mask @xmath72 be given . let also @xmath404 defining the hyperplane @xmath405 such that @xmath406 be given . the _ scanning masks with respect to @xmath72 _ are defined as @xmath407 @xmath408 in the following , the notation @xmath409 and @xmath410 , @xmath411 , means that the pair @xmath412 will be used . a _ scanning order _ is an ordering of the @xmath389 points in @xmath46 , denoted @xmath413 . for a scanning mask to propagate distances efficiently , it is important that , in each step of the propagation , the values at the points in @xmath46 from which the mask can propagate distances have already been updated . this is guaranteed if each point that can be reached by the scanning mask either has already been visited or is outside the image . let @xmath413 be a scanning order and @xmath414 a scanning mask . the scanning mask @xmath414 _ supports the scanning order _ if @xmath415 given a scanning mask @xmath414 and an image @xmath46 , there is a scanning order such that @xmath414 supports the scanning order . let @xmath405 such that @xmath406 be the hyperplane defining @xmath416 and @xmath417 ( we consider @xmath416 here , the proof for @xmath417 is similar ) . now we consider two sets , @xmath418 and @xmath419 , where @xmath420 is the complement of @xmath46 and the elements in @xmath421 are the already visited points in @xmath46 . let @xmath422 , @xmath423 . for every @xmath424 , there is a @xmath425 ( @xmath426 ) such that @xmath427 . since @xmath418 is finite , there is a point @xmath428 with corresponding @xmath429 such that @xmath430 , @xmath431 , and @xmath432 . thus , @xmath433 and @xmath434 , so @xmath435 and thus @xmath436 which has an empty intersection with @xmath418 , so @xmath437 . let this @xmath77 be @xmath438 , assign it to @xmath421 and repeat until @xmath439 . [ chamfer_alg ] let the scanning masks @xmath416 and @xmath417 and scanning orders @xmath413 and @xmath440 such that the masks support these scanning orders be given . * initially , @xmath441 and @xmath442 . * the image is scanned two times using the two scanning orders . * for each visited point @xmath443 in scan @xmath85 , @xmath444 usually , in the first scan @xmath445 is omitted , as @xmath445 is only computed for @xmath446 , so we know @xmath447 . now we prove that , in a distance map @xmath448 , for any point @xmath449 on a path defining the distance between @xmath450 and its closest background point @xmath451 , the value in the distance map at point @xmath449 is given by @xmath452 . in other words , the closest point in the background from @xmath449 is @xmath79 . [ chamfer_lemma ] let @xmath450 and @xmath453 be such that @xmath454 and @xmath455 . there is a set @xmath456 such that the point @xmath77 can be written @xmath457 and @xmath458 . for any set @xmath459 , if @xmath460 , then @xmath461 . let @xmath462 . since @xmath463 is the length of a path between @xmath464 and @xmath79 , the weighted distance can not be larger than this . assume that @xmath465 . then there is a @xmath466 such that @xmath467 now , @xmath468 since @xmath469 is the length of one path ( we can not assume that it is a shortest path ) between the points @xmath464 and @xmath470 , it follows that @xmath471 and thus @xmath472 which contradicts @xmath473 . to assure that the propagation does not depend on points outside the image , the definition of border points _ or _ of wedge preserving images defined below can be used . [ def : borderpoint ] given a chamfer mask @xmath72 , a point @xmath424 is _ a border point _ if @xmath474 [ border_lemma ] if for all border points @xmath475 , then for all @xmath476 such that @xmath477 , all points in any shortest path between @xmath77 and @xmath79 are in @xmath46 . assume that a point @xmath449 in a shortest path between @xmath77 and @xmath79 is not in @xmath46 . then , since all border points are in the background , there is a background grid point in all paths between @xmath77 and @xmath449 . since @xmath478 , it follows that there must be a border point @xmath479 such that @xmath480 which contradicts the assumption @xmath481 . [ def : wedge_pres ] given a chamfer mask @xmath72 and @xmath394 , @xmath482 such that @xmath483 . the half - spaces @xmath403 and @xmath401 are _ wedge - preserving _ if @xmath484 where @xmath302 denotes wedges in @xmath72 . [ wedge_pres_halfspace ] given a chamfer mask @xmath72 , the image @xmath46 is _ wedge - preserving _ if it is the intersection of wedge - preserving half - spaces . [ lem_wedge_equival ] if an image @xmath46 is wedge - preserving , then for all @xmath485 , all points in any shortest path between @xmath77 and @xmath79 are in @xmath46 . let @xmath46 be wedge - preserving and let @xmath486 such that @xmath487 with @xmath488 , where @xmath302 is a wedge . assume that there is a fixed @xmath489 and a set @xmath490 such that @xmath491 but @xmath492 . in other words , if there is a point not in @xmath46 in the shortest path between @xmath77 and @xmath79 , then some point @xmath449 in @xmath46 has a neighbor @xmath493 that is not in @xmath46 . since @xmath46 is wedge - preserving , there is a plane defined by @xmath494 such that @xmath495 and @xmath496 . since @xmath497 @xmath498 @xmath499 taking any @xmath500 , @xmath501 , which by definition [ wedge_pres_halfspace ] implies that @xmath502 . this implies that @xmath503 now , @xmath504 , but @xmath505 @xmath506 @xmath507 @xmath508 since @xmath496 , @xmath509 if @xmath510 and @xmath511 , and using eq . [ geqvect ] . we have @xmath512 , i.e. @xmath513 . contradiction . as the image generally is given first , a mask can often be built with preserved wedges and then cut into two scanning masks . then the points in the image can be ordered such that the masks support the scanning directions and then the distance map can be computed using our algorithm . [ chamf - alg - conditions ] [ chamf - alg - theorem ] if either * for all border points @xmath514 or * @xmath46 is wedge - preserving , then the chamfer algorithm in definition [ chamfer_alg ] produces distance maps as defined in definition [ def : dist - map ] . let @xmath450 and @xmath453 be such that @xmath454 and @xmath455 . there is a set @xmath456 such that the point @xmath77 can be written @xmath457 and @xmath458 . let @xmath515 be the image after the first scan . by lemma [ chamfer_lemma ] , using that the masks support the scanning order , the local distances are propagated in the following way . also , by lemma [ border_lemma ] and lemma [ lem_wedge_equival ] , all points below are in @xmath46 . @xmath516 let @xmath517 be the image after the second scan . using the notation @xmath518 , we get @xmath519 this proves that all steps of a shortest path between @xmath77 and @xmath79 have been processed . indeed , after the second scan , the value of the point @xmath77 is @xmath520 . in figure [ chamf - alg - conditions ] , the need for the conditions in theorem [ chamf - alg - theorem ] are shown in example images . the bcc grid ( @xmath521 ) and the fcc grid ( @xmath522 ) are defined as follows : @xmath523 @xmath524 a voxel is defined as the voronoi region of a point @xmath77 in a grid . figure [ neighborhoods ] ( a ) shows a voxel of a bcc grid . a bcc voxel has two kinds of face - neighbors ( but no edge- or vertex - neighbors ) , which results in the 8-neighborhood ( figure [ neighborhoods ] ( b ) ) and the 14-neighborhood ( figure [ neighborhoods ] ( c ) ) . on the fcc grid , each voxel ( see figure [ neighborhoods ] ( d ) ) has 12 face - neighbors and 6 vertex - neighbors . the resulting 12- and 18-neighborhoods are shown in ( figure [ neighborhoods ] ( e ) ) and ( figure [ neighborhoods ] ( f ) ) , respectively . note that the high number of face neighbors ( 12 for the fcc grid and 14 for the bcc grid ) implies these grids are more compact than the cubic grid , which has only @xmath525 face neighbors . to apply the results of sections [ sec : wdp ] and [ sec : chamferalgo ] to the bcc grid , we should check that : * @xmath521 is a sub - module of @xmath526 * masks decomposed in @xmath521-basis wedges can be created * a scanning order can be defined on the bcc lattice . [ lemma : bccsubmodule ] let @xmath527 , then @xmath528 is a sub - module of @xmath6 . let us prove first that @xmath529 is an abelian group . given @xmath77 , @xmath530 , there exist @xmath531 , and @xmath532 such that @xmath533\ p^{k } = p^{1 } + 2 \alpha_{k}$ ] and @xmath534 . the point @xmath535 also belongs to @xmath529 . indeed , @xmath533 $ ] , @xmath536 with @xmath537 . moreover , @xmath538 as @xmath539 with @xmath540 . we now have to prove that @xmath541 , @xmath542 , @xmath543 . indeed , @xmath134 , \lambda p^{k } \in \mathbb{z}$ ] and @xmath533 \lambda p^{k } = \lambda p^{1 } + 2 ( \lambda \alpha_{k } ) \textrm { with } ( \lambda \alpha_{k } ) \in \mathbb{z}$ ] . this result is true for all @xmath544 and obviously for @xmath545 . [ lemma : bccbasis ] a family @xmath546 is a basis of @xmath521 iff @xmath547 first , we prove that any determinant of 3 vectors @xmath548 of @xmath521 is a multiple of @xmath549 . indeed , given a vector @xmath550 , @xmath551 iff there exists @xmath552 such that @xmath553 and @xmath554 . then the determinant of @xmath555 and @xmath293 is @xmath556 with the notations of paragraph [ sec : wdp ] we obtain for any vector @xmath557 of @xmath521 : @xmath558 and @xmath559 iff @xmath560 which means @xmath561 . the same result applies for @xmath562 and @xmath563 . by lemma [ lemma : value ] we obtain @xmath148 is a basis of @xmath521 iff @xmath561 . to compute a weighted distance map on an image stored on a bcc grid , one can consider a mask built using 8-neighbors : @xmath564 this mask contains only one type of weight as the euclidean distance between each face - sharing neighbor to the central voxel is the same . this mask can be decomposed into @xmath521-basis sectors . indeed , let us consider the wedge @xmath565 , and its symmetric wedges . @xmath566 and from lemma [ lemma : bccbasis ] , @xmath567 is a @xmath521-basis sector . if we want larger masks , we can split the wedge @xmath567 and its symmetric wedges according to the farey triangulation technique @xcite . to do so , given a @xmath521-basis wedge @xmath568 , choose an edge to be split , say the edge between @xmath569 and @xmath292 . we create @xmath570 such that @xmath571 defined by @xmath572 , @xmath573 and @xmath574 . the two obtained wedges @xmath575 and @xmath576 are @xmath521-basis sectors . indeed , @xmath577 in the same way , @xmath578 . if we apply the first steps of this method , we can retrieve the chamfer mask using the 14-neighborhood proposed by @xcite . but we can also build larger masks by splitting the new wedges . to compute optimal integer chamfer weights for bcc grid the depth - first search method of @xcitetc18/code - data - set ] can be used . this method computes weights sets leading to weighted distance maps with small relative error with respect to the corresponding euclidean map . moreover , it keeps a weights set only if the normalized polytope of the corresponding mask is convex , which allows application of the results from section [ sec : wdp ] . the following tables give sets of integer chamfer mask weights for bcc grids with the scale factor allowing comparison of the final weighted distance map with an euclidean one composed with real numbers . they also give the maximum relative error that can occur between the weighted distance map and the euclidean one . for the two first cases , we also give the optimal real weights computed using the equations of section [ sec : coeffsopt ] . balls obtained by using different number of weights ( shown in bold below ) are shown in figure [ balls ] . 1 . one weight + for the mask corresponding to the 8-neighborhood , using only one weight , the equations of section [ sec : coeffsopt ] give the following result : + + [ cols= " < , < " , ] + + we have presented a general theory for weighted distances . this allows application of the weighted distance transform to any point - lattice . the optimal weight calculation and the construction of a chamfer algorithm to produce distance maps are straight - forward by the procedure in section [ sec : coeffsopt ] and theorem [ chamf - alg - theorem ] . for the chamfer algorithm to construct correct distance - maps , the only limitation is that the image must be wedge - preserving , definition [ def : wedge_pres ] , or have all border points in the background , definition [ def : borderpoint ] . with these conditions which can easily be satisfied , images in any dimension can be considered . it is also worth mentioning that despite its popularity , to our knowledge , no proof of the correctness of the two - scan algorithm has been published until now , except for the 2d 3 @xmath579 3 mask algorithm proof from 1966 @xcite . also , the important general formula for the weighted distance between two grid points in theorem [ theorem : value ] is new . the application of these properties and of the chamfer algorithm are numerous for the cubic and parallelepiped grids in the literature . to illustrate how this theory can be applied , section [ sec : bccfcc ] gives examples for the fcc and bcc grids . but it can be applied to any point - lattice , for example the 3d grid where the voxels are hexagonal cylinders as suggested in @xcite or 4d a grid as in @xcite , or even other grids as long as they satisfy the module conditions . j. cai , j. chu , d. recine , m. sharma , c. nguyen , r. rodebaugh , v. saxena , a. ali , ct and pet lung image registration and fusion in radiotherapy treatment planning using the chamfer - matching method . , international journal of radiation oncology biology physics 43 ( 4 ) ( 1999 ) 883891 . l. ibanez , c. hamitouche , c. roux , determination of discrete sampling grids with optimal topological and spectral properties , in : proceedings of @xmath580 conference on discrete geometry for computer imagery , lyon , france , vol . 1176 of lecture notes in computer science , springer , 1996 , pp . 181192 . c. o. s. sorzano , r. marabini , j. velzquez - muriel , j. r. bilbao - castro , s. h. w. scheres , j. m. carazo , a. pascual - montano , xmipp : a new generation of an open - source image processing package for electron microscopy , journal of structural biology 148 ( 2004 ) 194204 . c. fouard , g. malandain , s. prohaska , m. westerhoff , f. cassot , c. mazel , d. asselot , j .- marc - vergnes , skeletonization by blocks for large datasets : application to brain microcirculation , in : international symposium on biomedical imaging : from nano to macro ( isbi04 ) , ieee , arlington , va , usa , 2004 . c. r. maurer , r. qi , v. raghavan , a linear time algorithm for computing exact euclidean distance transforms of binary images in arbitrary dimensions , ieee transactions on pattern analysis and machine intelligence 25 ( 2 ) ( 2003 ) 265270 . e. remy , e. thiel , look - up tables for medial axis on squared euclidean distance transform , in : proceedings of @xmath581 conference on discrete geometry for computer imagery , naples , italy , vol . 2886 of lecture notes in computer science , springer , 2003 , pp . 224235 . l. dorst , p. w. verbeek , the constrained distance transformation : a pseudo - euclidian , recursive implementation of the lee - algorithm , in : proc . european signal processing conference 1986 , the hague , the netherlands , 1986 , pp .

in this article , we study the null controllability of linearized compressible navier - stokes system in one and two dimension . we first study the one - dimensional compressible navier - stokes system for non - barotropic fluid linearized around a constant steady state . we prove that the linearized system around , with is not null controllable by localized interior control or by boundary control . but the system is null controllable by interior controls acting everywhere in the velocity and temperature equation for regular initial condition . we also prove that the the one - dimensional compressible navier - stokes system for non - barotropic fluid linearized around a constant steady state , with is not null controllable by localized interior control or by boundary control for small time next we consider two - dimensional compressible navier - stokes system for barotropic fluid linearized around a constant steady state we prove that this system is also not null controllable by localized interior control . * key words . * linearized compressible navier - stokes system , null controllability , localized interior control , boundary control , gaussian beam . * ams subject classifications . * 35q30 , 93c20 , 93b05

1410.4351

control of fluid flow has been an important area of research and has many practical applications . the question about controllability of fluid flows has attracted the attention of many researchers , more for incompressible flow but much less for compressible flow . in this paper we are interested in controllability properties of linearized compressible navier - stokes system . for a compressible , isothermal barotropic fluid ( density is a function of only pressure ) , the navier - stokes system in @xmath8 consists of equation of continuity @xmath9 = 0 , \end{array}\ ] ] and the momentum equation @xmath10 \\ [ 3.mm ] \displaystyle = -\nabla p(x , t ) + \mu \bigtriangleup { \bf u}(x , t ) + ( \lambda + \mu ) \nabla [ \mbox{div } { \bf u}(x , t ) ] , \end{array}\ ] ] where @xmath11 is the density of the fluid , @xmath12 denotes the velocity vector in @xmath13 and @xmath14 is an external force field in @xmath15 the pressure satisfies the following constitutive law @xmath16 for some constants @xmath17 , @xmath18 the viscosity coefficients @xmath19 and @xmath20 are assumed to be constant satisfying the following thermodynamic restrictions , @xmath21 , @xmath22 for non - barotropic fluid ( when density is a function of pressure and temperature of the fluid ) , the navier - stokes system consists of the equation of continuity , the momentum equation and an additional thermal energy equation @xmath23(x , t ) + \theta(x , t ) \frac{\partial p}{\partial \theta}(x , t ) \mbox{div } { \bf u}(x , t ) \\ [ 3.mm ] \displaystyle = \kappa \bigtriangleup \theta ( x , t ) + \lambda ( \mbox{div } { \bf u}(x , t))^2 + 2 \mu \sum_{i , j=1}^n \frac{1}{4 } \left(\frac{\partial { \bf u}_i}{\partial x_j } + \frac{\partial { \bf u}_j}{\partial x_i}\right)^2 , \end{array}\ ] ] where @xmath24 denotes the temperature of the fluid , @xmath25 is the specific heat constant and @xmath26 is the heat conductivity constant . for ideal gas , the pressure is given by boyle s law : @xmath27 where @xmath28 is the universal gas constant ( see @xcite ) . in this article , we first consider the compressible navier - stokes system for non - barotropic fluid in a bounded interval @xmath29 linearized around a constant steady state @xmath30 with @xmath31 and @xmath32 more precisely we consider the system @xmath33 where @xmath34 is the characteristic function of an open subset @xmath35 we choose the following initial and boundary conditions for the system : @xmath36 u(0,t ) = 0 , \quad u(l , t ) = 0 \quad \forall \ \ t > 0 , \\ [ 2.mm ] \theta(0,t ) = 0 , \quad \theta(l , t ) = 0 \quad \forall \ \ t > 0 . \end{array}\ ] ] in - , @xmath37 , @xmath38 and @xmath39 are distributed controls . we are interested in the following question : given @xmath40 and @xmath41 can we find interior control functions such that the solution @xmath42 of - satisfies @xmath43 our first main result regarding interior null controllability is the following , [ thm1 ] let @xmath44 i.e. , @xmath45 is a proper subset of @xmath46 let us assume @xmath47 . the system - is not null controllable in any @xmath48 by the interior controls @xmath49 , @xmath50 and @xmath51 acting on density , velocity and temperature equation respectively . the above negative result can be extended to the case of less regular interior controls or boundary control . see remark [ rem : lessregcontrol ] and remark [ rem : bdycontrol ] for more details . our next positive result shows that , if initial density @xmath52 lies in a more regular space then the linearized system is null controllable by velocity and temperature controls acting everywhere in the domain . [ thm2 ] let @xmath53 in . let us denote @xmath54 let us assume @xmath55 . then for any @xmath56 there exist controls @xmath57 and @xmath58 acting everywhere in the velocity and temperature equation respectively , such that the solution of - satisfies @xmath59 the next result shows that the above result is sharp as null controllability can not be achieved by localized interior velocity and temperature controls . [ thm3 ] let @xmath53 in . let @xmath60 i.e. , @xmath61 is a proper subset of @xmath46 let us assume @xmath55 . the system - is not null controllable in any @xmath48 by the interior controls @xmath62 and @xmath63 acting on velocity and temperature equation respectively . the proof of these results relies on the observability inequality . we know that the null controllability of a linear system is equivalent to a certain observability inequality for the solutions of adjoint system ( see @xcite , chapter 2 ) . to prove the negative results , we will construct particular solutions for the adjoint system such that the observability inequality can not hold . in order to do that first we will consider the adjoint system in @xmath64 as a terminal value problem . we will construct highly localized solutions known as `` gaussian beam '' . similar kind of construction has been used for hyperbolic equations by ralston ( @xcite ) and for wave equations by maci and zuazua ( @xcite ) . we will prove that solutions are localized in a small neighbourhood of any @xmath65 thus given an observation set , we can always find an interval away from the observation set such that the solutions are localized in that interval . using this we are able to prove the negative results . to the author s best knowledge these are new results regarding controllability issues of navier - stokes system for non - barotropic fluid . in theorem [ thm3 ] , we proved a positive result when controls acting everywhere in the equation . the question then arises : whether positive results could be obtained by using control supported in a small , but moving region , as in @xcite . rosier and rouchon in @xcite proved that the structurally damped wave equation in one dimension is not null controllable by a boundary control . later on martin , rosier and rouchon in @xcite proved that the same equation in one dimension with periodic boundary conditions , is null controllable with a moving distributed control for sufficiently large time . chaves - silva , rosier and zuazua in @xcite extend the above result to higher dimension . the structure of the system considered by the authors in @xcite , in some sense , is similar to the linearized compressible , barotropic navier - stokes system in one dimension as well as in higher dimension . these issues will be discussed in a future work which is in progress . we have studied in this paper the null controllability of the linearized compressible navier - stokes system only . the `` gaussian beam '' construction is used to show negative results . however one may use other techniques based on the use of nonlinearity ( see @xcite for example ) to achieve controllability results for the full nonlinear system . there have been some results regarding the control of compressible barotropic fluid models in recent years . amosova in @xcite considers compressible navier - stokes system for viscous barotropic fluid in one dimension in lagrangian coordinates in a bounded domain @xmath66 with dirichlet boundary condition . she proves local exact controllability to trajectories for the velocity in any time @xmath56 using a localized interior control on the velocity equation , provided that the initial density is already on the targeted trajectory and initial condition lies in @xmath67 ervedoza , glass , guerrero and puel in @xcite consider the compressible navier - stokes system in one space dimension in a bounded domain @xmath68 . they prove local exact controllability to constant states @xmath69 with @xmath70 using two boundary controls , both for density and velocity , in time @xmath71 when initial condition lies in @xmath72 chowdhury , ramaswamy and raymond in @xcite consider the compressible barotropic navier - stokes system linearized around a constant steady state @xmath73 with @xmath74 in a bounded domain @xmath75 . they proved that the linearized system is not null controllable by a localized control or by boundary control . they also proved that the linearized system is null controllable by an interior control acting everywhere in the velocity equation when initial condition lies in @xmath76 chowdhury in @xcite considers the compressible barotropic navier - stokes system linearized about a constant steady state @xmath77 with @xmath78 in @xmath79 with dirichlet boundary condition and an interior control on the velocity equation acting on open subset @xmath80 he proves that the system is approximately controllable in @xmath81 when @xmath82 he also proves a similar result in two dimension . chowdhury , mitra , ramaswamy and renardy in @xcite consider the compressible barotropic navier - stokes system linearized about a constant steady state @xmath77 with @xmath78 in @xmath83 with periodic boundary condition . they proved that the linearized system is null controllable by a localized velocity control when @xmath84 and initial condition lies in @xmath85 our linearized system - is similar to the linearized system considered by the authors in @xcite . so we expect similar controllability results . their method is based on explicit expression for eigenfunctions and the behaviour of the spectrum of the linearized operator . they proved that there is an accumulation point in the spectrum of the linearized operator . this system behaves very badly with respect to controllability properties and a similar type of controllability behaviour is also observed in @xcite for different types of systems where an accumulation point is present in the spectrum of linearized operator . but the method used in @xcite does not seem to fit very well in our case . in fact one can prove that there is an accumulation point in the spectrum of the linearized operator considered here , for certain boundary condition . but the expressions of eigenvalues and eigenfunctions are complicated . so here we use gaussian beam approach to achieve the negative results . this technique does not require the knowledge of the spectrum and it seems to extend to higher dimension also . the controllability properties are completely different , if we consider compressible navier - stokes system linearized around non null velocity . for barotropic fluid , the system linearized around @xmath86 is not controllable in any time @xmath87 by localized interior control but the system linearized around @xmath88 is null controllable by localized interior control for large time @xmath6 it is interesting to note that there is no accumulation point in the spectrum of the linearized operator in the latter case and better controllability behaviour at least for @xmath87 large enough(see @xcite ) . but the question remains what happens when time @xmath87 is small enough . our results answer this question in the negative . we consider the compressible non - barotropic navier - stokes equation linearized around constant steady state @xmath89 @xmath90 @xmath91 \displaystyle u_t - \frac{\lambda + 2\mu}{{\bar \rho } } u_{xx } + \frac{r { \bar \theta}}{{\bar \rho}}\;\rho_x + \bar v u_x + r \theta_x \ ; = g \chi_{\mathcal{o}_2 } , \mbox { in } ( 0,l ) \times ( 0,t ) , \\[3.mm ] \displaystyle \theta_t ( x , t ) -\frac{\kappa}{{\bar \rho}c_v } \theta_{xx } + \frac{r { \bar \theta}}{c_v } u_x + \bar v \theta_x \ ; = h \chi_{\mathcal{o}_3 } , \mbox { in } ( 0,l ) \times ( 0,t ) , \\[3.mm ] \displaystyle \rho(0 ) = \rho_0,\quad \quad u(0)= u_0 , \mbox { and } \quad \theta(0 ) = \theta_0 , \quad \mbox { in } ( 0,l ) , \\ [ 2.mm ] \displaystyle \rho(0,t ) = 0 , \;\;\ ; u(0,t ) = 0 = u(l , t ) , \;\ ; \forall \ t \in ( 0,t ) , \\ [ 2.mm ] \displaystyle \theta(0,t ) = 0 = \theta(l , t ) , \;\ ; \forall \ t \in ( 0,t ) . \end{array}\ ] ] we prove the following theorem . [ thm1.4 ] let @xmath92 i.e. , @xmath45 is a proper subset of @xmath46 let us assume @xmath47 . if @xmath93 then the system is not null controllable by localized interior controls @xmath94 @xmath50 and @xmath51 acting on density , velocity and temperature equation respectively . as a corollary of the above theorem , one can rule out null controllability of compressible barotropic navier - stokes system linearized around constant steady state @xmath95 in small time @xmath96 using a boundary control or localized interior control . we consider compressible barotropic navier - stokes system linearized around a constant steady state @xmath95 , @xmath97 in @xmath98 or in @xmath99 as in @xcite . * for initial condition belonging to @xmath100 the system with dirichlet boundary condition is not null controllable at any time @xmath40 by a interior control acting only in the velocity equation . the control may act in a non empty open subset of @xmath79 or in the whole domain @xmath101 * the system with periodic boundary condition is not null controllable by interior control localized in @xmath102 acting only in the velocity equation when initial condition lies in @xmath103 and time @xmath104 * for initial condition belonging to @xmath100 the same system is not null controllable by boundary control if time @xmath105 from the above corollary we see that the condition @xmath106 in ervedoza , glass , guerrero and puel ( @xcite ) is natural . next we will show that , our method can be extended to higher dimension also . for simplicity we consider the compressible barotropic navier - stokes system in two - dimensional bounded domain @xmath107 , linearized around a constant steady state solution @xmath108 , @xmath1 @xmath109 \displaystyle { \bf u } - \frac{\mu}{{\bar \rho } } \delta { \bf u } - \frac{\lambda + \mu}{{\bar \rho}}\nabla[\mathrm{div } \ { \bf u } ] + \;a \gamma { \bar \rho}^{\gamma-2 } \nabla \rho \;=\;{\bf g } \chi_{{\mathcal o}_2 } , \mbox { in } \omega \times ( 0,t ) , \\[2.mm ] \displaystyle \rho(0 ) = \rho_0 \quad \mbox{and}\quad { \bf u}(0)= { \bf u}_0 , \quad \mbox { in } \omega , \\ [ 2.mm ] \displaystyle { \bf u } = { \bf 0 } \ \ \mbox { on } \partial \omega \times ( 0,t ) . \end{array}\ ] ] where @xmath110 and @xmath111 are open subsets of @xmath112 we obtain the following negative null controllability result for the system . [ thm1.5 ] let @xmath113 i.e. , @xmath45 is a proper open subset of @xmath112 let us assume that @xmath114 then the system is not null controllable in time any @xmath40 by interior controls @xmath115 and @xmath116 the plan of the paper is as follows . in section 2 , we study the control system linearized around a constant steady state @xmath0 in one dimension . we prove theorem [ thm1 ] , theorem [ thm2 ] and theorem [ thm3 ] here . in section 3 , we study the control system linearized around a constant steady state @xmath117 theorem [ thm1.4 ] is proved here . in section 4 we consider the linearized system in two dimension around constant steady state @xmath7 we prove theorem [ thm1.5 ] here . * acknowledgement * : the author would like to thank prof . sylvain ervedoza for providing important references on gaussian beams . the author also would like to thank him and prof . mythily ramaswamy for very useful discussions which improved the initial version . the author acknowledges the financial support under the project `` pde control '' from the indo french centre for applied mathematics ( ifcam ) . in this section , we will discuss interior null controllability of the system - . we introduce the positive constants @xmath118 let us define @xmath119 endowed with the inner product @xmath120 the following proposition about existence and uniqueness of the system - follows easily from semigroup theory . let @xmath121 . let us assume that @xmath115 , + @xmath122 and @xmath123 . then - has a unique solution @xmath42 with @xmath124 , @xmath125 and @xmath126 . moreover @xmath127;z).$ ] the idea is to use the adjoint system to derive certain identity which can be used to obtain an observability inequality , equivalent to null controllability . ( see @xcite , chapter 2 ) . for this we consider the following adjoint problem , @xmath128 \displaystyle - v_t - \nu_0 v_{xx } - \frac{r { \bar \theta}}{{\bar \rho}}\;\sigma_x - r \phi_x \ ; = 0 , \mbox { in } { ( 0,l)}\times ( 0,t ) , \\[3.mm ] \displaystyle - \phi_t - k_0 \phi_{xx } - \frac{r { \bar \theta}}{c_v } v_x \ ; = 0,\ \mbox { in } { ( 0,l)}\times ( 0,t ) , \\ [ 2.mm ] \displaystyle \sigma(t ) = \sigma_t,\quad v(t ) = v_t , \quad \phi(t ) = \phi_t , \mbox { in } { ( 0,l ) } , \\ [ 2.mm ] v(0,t ) = 0 = v(l , t ) \quad \forall \ \ t > 0,\quad \phi(0,t ) = 0 = \phi(l , t ) \quad \forall \ \ t > 0 . \end{array}\ ] ] with @xmath129 the adjoint system is also well posed in @xmath130 . in fact we have let @xmath129 the system has a unique solution with @xmath131 , @xmath132 and @xmath133 . moreover @xmath134 belongs to @xmath135;z).$ ] let us first assume that @xmath136 @xmath137 + @xmath138 @xmath139 @xmath140 and let @xmath42 and @xmath141 be the solutions of and respectively . taking inner product in @xmath130 of with @xmath141 and integrating we obtain @xmath142 an integration by parts and use of gives @xmath143 dx + { \bar \rho}^2 \int_0^l [ u(x , t ) v_t(x ) - u_0(x ) v(x,0 ) ] dx \notag \\ + \frac{{\bar \rho}^2 c_v}{{\bar \theta } } \int_0^l [ \theta(x , t ) \phi_t(x ) - \theta_0(x ) \phi(x,0 ) ] dx = r { \bar \theta}\int_0^t \int_{\mathcal{o}_1 } f \sigma \ ; dx dt \ \notag \\ + { \bar \rho}^2 \int_0^t \int_{\mathcal{o}_2 } g v \ dx dt + \frac{{\bar \rho}^2 c_v}{{\bar \theta } } \int_0^t \int_{\mathcal{o}_2 } h \phi \ dx dt . \end{aligned}\ ] ] the above relation leads us to the identity equivalent to null controllability . for each initial state @xmath144 , the solution of the system - can be driven to rest by interior controls @xmath145 , @xmath146 and @xmath147 in time @xmath87 if and only if @xmath148 for all @xmath149 where @xmath134 is the solution of the adjoint system . by a density argument we deduce that for any @xmath121 and @xmath150 the identity holds . thus from , it follows that holds if and only if - is null controllable and @xmath151 are the corresponding controls . one can use the identity to get an observability inequality which is also equivalent to null controllability . more precisely we have the following proposition ( see @xcite section 2 and @xcite chapter 2 ) . the system - is null controllable in @xmath130 in time @xmath40 if and only if there exists a constant @xmath152 such that for any terminal condition @xmath153 , @xmath141 , the solution of the adjoint problem satisfies the following observability inequality @xmath154 we now want to prove that the system - is not null controllable in @xmath130 when controls are localized . our idea is to show that the observability inequality does not hold in this case . for this we first consider the adjoint problem in whole real line : @xmath155 \displaystyle - v_t - \nu_0 v_{xx } - \frac{r { \bar \theta}}{{\bar \rho}}\;\sigma_x - r \phi_x \ ; = 0 , \mbox { in } \mathbb{r } \times ( 0,t ) , \\[3.mm ] \displaystyle - \phi_t ( x , t ) - k_0 \phi_{xx } - \frac{r { \bar \theta}}{c_v } v_x \ ; = 0 , \mbox { in } \mathbb{r } \times ( 0,t ) , \\ [ 2.mm ] \displaystyle \sigma(t ) = \sigma_t,\quad v(t ) = v_t , \quad \phi(t ) = \phi_t \quad \mbox { in } \mathbb{r}. \end{array}\ ] ] first we will construct a particular solution of the above adjoint problem which is localized in a small neighbourhood of any @xmath65 for this we would like to have a fourier representation formula for the solution of . let us assume that @xmath156 and @xmath157 we define @xmath158 the fourier transform of @xmath159 in the space variable @xmath160 for each @xmath161 as follows , @xmath162 we also define @xmath163 and @xmath164 in a similar manner . applying the fourier transform in , we obtain the following system of ode satisfied by @xmath165 , @xmath166 and @xmath167 @xmath168 \displaystyle - \partial_t { \hat v}(\xi , t)- \nu_0 ( - \xi^2 ) { \hat v}(\xi , t ) - \frac{r { \bar \theta}}{{\bar \rho}}\;(i\xi){\hat \sigma}(\xi , t ) - r ( i\xi){\hat \phi}(\xi , t ) \ ; = 0 , \qquad \xi \in { \mathbb r } , t \in ( 0,t ) , \\[3.mm ] \displaystyle - \partial_t { \hat \phi}(\xi , t ) - k_0 ( - \xi^2){\hat \phi}(\xi , t ) - \frac{r { \bar \theta}}{c_v}(i\xi ) { \hat v}(\xi , t ) \ ; = 0 , \qquad \xi \in { \mathbb r } , t \in ( 0,t ) , \\ [ 2.mm ] \displaystyle { \hat \sigma}(\xi , t ) = { \hat \sigma}_t(\xi),\quad { \hat v}(\xi , t ) = { \hat v}_t(\xi ) , \quad { \hat \phi}(\xi , t ) = { \hat \phi}_t(\xi ) \quad \xi \in { \mathbb r}. \end{array}\ ] ] let us define @xmath169 then can be written in the following form @xmath170 the unique solution of the above system of ode can be written as @xmath171 we will now discuss some properties of the eigenvalues of @xmath172 let @xmath173 , @xmath174 and @xmath175 be three smooth functions . let @xmath176 be a cubic polynomial . let us define the discriminant of the above cubic polynomial @xmath177 now the roots of the above cubic polynomial are given by the following formula @xmath178 where @xmath179 are the three cubic roots of unity , and @xmath180{\frac{d_1(\xi ) + \sqrt{d_1(\xi)^2 - 4 d_0(\xi)^3}}{2 } } \qquad\ ] ] with @xmath181 and when @xmath182 in this formula , @xmath183 and @xmath184{~~}$ ] denote any choice for the square or cube roots , but one has to be consistent with the choice for all @xmath185 . if @xmath186 and @xmath187 for some @xmath188 the sign of @xmath189 has to be chosen to have @xmath190 i.e. one should define @xmath191 in this case the roots are given by @xmath192{d_1(\xi)}\right ) , \mbox { for } k=1,2,3.\ ] ] if @xmath193 and @xmath187 for some @xmath188 the three roots are equal @xmath194 if @xmath193 and @xmath195 for some @xmath188 there is a double root @xmath196 and a simple root @xmath197 as @xmath198 @xmath199 and @xmath200 are differentiable functions of @xmath188 it is easy to deduce that real part and the complex part of the roots are also differentiable for all @xmath201 we have the following lemma about the properties of the eigenvalues of @xmath172 [ lem:2.5 ] the eigenvalues of @xmath202 always have non positive real part for all @xmath203 let @xmath204 , be the eigenvalues of @xmath202 , where @xmath205 and @xmath206 . there exists a constant @xmath207 such that for @xmath208 one of the eigenvalues , say @xmath209 satisfies @xmath210 where @xmath211 the eigenvalues of @xmath202 are given by the roots of the characteristic polynomial @xmath212 since it is a polynomial of degree three , it will always have a real root . now the polynomial @xmath213 is stable i.e. all the roots have negative real part if and only if @xmath214 ( theorem 2.4 , part i of @xcite ) . for our case , @xmath215 and it is easy to verify that @xmath216 for all @xmath217 . so all the eigenvalues have negative real part for all @xmath217 . for @xmath218 the characteristic polynomial has only one root @xmath219 let @xmath204 , be the eigenvalues of @xmath220 given by the formula . thus @xmath221 comparing the coefficients of the characteristic polynomial we obtain , @xmath222 it is well known that , if the discriminant @xmath223 then the polynomial has three distinct real roots and if @xmath224 then the equation has one real root and two complex conjugate roots . in our case the discriminant is @xmath225 \xi^{10 } + o(\xi^8).\end{aligned}\ ] ] we also have @xmath226 this leads us to consider the following two cases . * case i. * let @xmath227 in this case , there exists a positive constant @xmath228 such that @xmath229 and @xmath230 for all @xmath231 hence for @xmath208 @xmath232 and @xmath233 are all real , positive and distinct . let us define , @xmath234 in , letting @xmath235 we obtain @xmath236 as @xmath237 are all positive , from the first equation of , we deduce that @xmath238 are all bounded . they are also continuous for @xmath231 as @xmath239 are all distinct and continuous for @xmath208 without loss of generality we assume that @xmath240 from the last equation of , we obtain that @xmath241 converges to @xmath242 as @xmath243 from we obtain , @xmath244 therefore , @xmath245 and @xmath246 both converge as @xmath247 and one of them converges to @xmath248 and the other one to @xmath249 without loss of generality we assume that @xmath250 therefore @xmath251 from the last two equations of , we have @xmath252 letting @xmath235 we deduce that @xmath253 * case ii . * let @xmath254 in this case , @xmath255 thus there exists a constant @xmath228 such that @xmath256 and @xmath195 for all @xmath257 hence for all @xmath258 we have one real root and two complex conjugate roots . let @xmath259 hence , for all @xmath208 @xmath260 and @xmath209 are all real and positive . let us define , @xmath261 from we obtain @xmath262 as @xmath263 are all positive , from the first two equations of , we deduce that @xmath264 @xmath265 are all bounded for @xmath266 . they are also continuous . from the last equation of , we obtain @xmath267 multiplying the second equation of by @xmath268 and letting @xmath247 we obtain @xmath269 = 0.\ ] ] again multiplying the second equation of by @xmath270 and letting @xmath247 we obtain @xmath271 = 0.\ ] ] now we can proceed as in case i , to obtain @xmath272 in fact more can be said about @xmath273 we have [ lem:2.6 ] let @xmath202 be defined as in and @xmath274 be the eigenvalue satisfying @xmath275 then there exists a constant @xmath276 such that for all @xmath277 @xmath278 further for @xmath208 where @xmath279 as in lemma [ lem:2.5 ] , @xmath209 is differentiable and we have @xmath280 for some positive constant @xmath152 . from lemma [ lem:2.5 ] , @xmath209 lies on the right half side of the complex plane and it is bounded for all @xmath203 thus there exists a constant @xmath281 such that @xmath278 the coefficients of the characteristic polynomial are differentiable and from the formula it is easy to deduce that , the real and complex parts of the roots are differentiable for @xmath231 for @xmath282 @xmath209 is real and hence differentiable . as @xmath274 is a root of the characteristic equation , we have @xmath283 differentiating the equation with respect to @xmath185 and using the fact that @xmath209 is bounded for all @xmath185 we obtain the estimate . [ lem : eigenfunction ] let @xmath202 be defined as in and @xmath274 is the eigenvalue satisfying , @xmath275 the eigenfunction of @xmath202 corresponding to @xmath274 is @xmath284 where + @xmath285 for @xmath286 sufficiently large , we have @xmath287 for some positive constant @xmath152 . from , we first obtain @xmath288 hence for @xmath286 sufficiently large , we obtain @xmath289 and the estimate of @xmath290 follows . we now want to give a representation formula for solution of . we have the following proposition . let us consider @xmath291 for a suitable @xmath292 such that @xmath293 then @xmath294 is the solution of , corresponding to the terminal condition @xmath295 denote @xmath296 to be the solution of corresponding to the terminal condition defined in . then from , the solution of can be written as @xmath297 it is easy to verify that @xmath298 by taking the inverse of fourier transform , we obtain , @xmath141 defined as in , is the solution of system corresponding to the terminal condition . we are now interested in the construction of some particular solutions of the equation , which are localized in a neighbourhood of some @xmath299 . for that , we choose @xmath300 in using a suitable cut off function @xmath301 and analyze the solution @xmath302 of given by . [ thm4 ] let @xmath301 be a smooth function compactly supported in @xmath303 and of unit @xmath304 norm . for any @xmath305 , sufficiently small and for any @xmath306 , let us take @xmath307 let @xmath308 and @xmath309 be as in and respectively . then they satisfy the following * @xmath310 , * for any @xmath311 , there exists a constant @xmath152 independent of @xmath312 such that @xmath313 * for some positive constant @xmath152 independent of @xmath312 , @xmath314 let us denote @xmath315 first we verify that @xmath316 we have @xmath317 thus @xmath318 . now @xmath319 so we have @xmath320 . similarly we can show that @xmath321 thus the representation formula is well defined . now we will prove ( i ) . note that @xmath322 by parseval s relation we have @xmath323 using lemma [ lem:2.6 ] we have @xmath324 hence we have proved ( i ) . to prove ( ii ) we have using a change of variable formula @xmath325 note that @xmath326 thus for @xmath327 we have , after integration by parts @xmath328 now @xmath329 thus for @xmath312 small , we have for some @xmath152 independent of @xmath312 @xmath330 therefore , for @xmath312 small enough and @xmath331 we have @xmath332 thus there exists a positive constant @xmath333 , such that @xmath334 this proves ( ii ) . we have @xmath335 similarly we can show that @xmath336 let us give some more properties of @xmath337 which we will need later on to prove the main results . [ lem : improved - regularity ] let @xmath338 and @xmath309 be as in theorem [ thm4 ] . then @xmath339 and @xmath340 lie in @xmath341 moreover @xmath342 let us first show that @xmath343 @xmath344 thus @xmath343 in a similar way we can show that @xmath345 and @xmath346 belong to @xmath347 next we will show @xmath348 first we have @xmath349 next we have @xmath350 thus we have shown @xmath351 similarly we can show that @xmath352 and @xmath353 also belong to @xmath354 let @xmath355 be as in theorem [ thm4 ] . then for any @xmath356 we have the following estimate @xmath357 first by lemma [ lem : improved - regularity ] , we conclude that @xmath358 proceeding in a similar way as in theorem [ thm4 ] , we first obtain @xmath359 for some positive constant @xmath360 independent of @xmath361 therefore @xmath362 for @xmath312 small . [ rem : trace ] let @xmath352 and @xmath353 be as in theorem [ thm4 ] . then @xmath363 for any @xmath364 by lemma , we have @xmath365 and @xmath366 lie in @xmath367 for any @xmath364 we have @xmath368 and @xmath369 thus @xmath370 similarly , we can show @xmath371 now we will use the above construction to prove theorem [ thm1 ] . first we prove the following theorem . theorem [ thm1 ] will be a direct consequence of this theorem . [ thm5 ] let @xmath45 be a proper subset of @xmath79 and @xmath372 then there exists a terminal condition @xmath373 such that the solution of , corresponding to this terminal condition , satisfy the following estimates * @xmath374 + and * @xmath375 as @xmath45 is a proper subset of @xmath376 we choose @xmath377 and @xmath378 such that @xmath379 let us choose @xmath380 is as in theorem [ thm4 ] with the above choice @xmath381 let @xmath382 and @xmath383 are as in and respectively . let us set @xmath384 let @xmath385 be the restriction of @xmath309 to @xmath98 and @xmath386 be the restriction of @xmath387 to @xmath101 then @xmath388 , @xmath389 and @xmath390 satisfy the following system , @xmath391 \displaystyle - \breve v_t^\epsilon - \nu_0 \breve v^\epsilon_{xx } - \frac{r { \bar \theta}}{{\bar \rho}}\;\breve \sigma^\epsilon_x - r \breve \phi^\epsilon_x \ ; = 0 , \mbox { in } { ( 0,l)}\times ( 0,t),\\[3.mm ] \displaystyle - \breve \phi_t^\epsilon - k_0 \breve \phi^\epsilon_{xx } - \frac{r { \bar \theta}}{c_v } \breve v^\epsilon_x \ ; = 0 , \mbox { in } { ( 0,l)}\times ( 0,t ) , \\ [ 2.mm ] \displaystyle \breve \sigma^\epsilon(t ) = \breve\sigma^\epsilon_t,\quad \breve v^\epsilon(t ) = \breve v^\epsilon_t , \quad \breve \phi^\epsilon(t ) = \breve \phi^\epsilon_t , \\ [ 2.mm ] \breve v^\epsilon(0,t ) = q^\epsilon_0(t ) , \quad \breve v^\epsilon(l , t ) = q^\epsilon_l(t ) \quad \forall \ \ t > 0 , \\ [ 2.mm ] \breve\phi^\epsilon(0,t ) = r^\epsilon_0(t ) , \quad \breve\phi^\epsilon(l , t ) = r^\epsilon_l(t ) \quad \forall \ \ t > 0 . \end{array}\ ] ] note that by lemma [ lem : improved - regularity ] and lemma [ rem : trace ] , we have @xmath392 and @xmath393 the above system has a unique solution @xmath394 with @xmath395;z).$ ] they satisfy the following estimates : * from , we have @xmath396 * from theorem [ thm4 ] , @xmath397 * from lemma [ rem : trace ] , @xmath398 let @xmath399 , @xmath400 and @xmath401 satisfy the following system @xmath402 \displaystyle - \widetilde v_t^\epsilon ( x , t)- \nu_0 \widetilde v^\epsilon_{xx } - \frac{r { \bar \theta}}{{\bar \rho}}\;\widetilde\sigma^\epsilon_x - r \widetilde\phi^\epsilon_x \ ; = 0 , \mbox { in } { ( 0,l)}\times ( 0,t ) , \\[3.mm ] \displaystyle - \widetilde\phi_t^\epsilon - k_0 \widetilde\phi^\epsilon_{xx } - \frac{r { \bar \theta}}{c_v } \widetilde v^\epsilon_x \ ; = 0 , \mbox { in } { ( 0,l)}\times ( 0,t ) , \\ [ 2.mm ] \displaystyle \widetilde \sigma^\epsilon(t ) = 0,\quad \widetilde v^\epsilon(t ) = 0 , \quad \widetilde \phi^\epsilon(t ) = 0 , \mbox { in } { ( 0,l ) } , \\ [ 2.mm ] \widetilde v^\epsilon(0,t ) = - q^\epsilon_0(t ) , \quad \widetilde v^\epsilon(l , t ) = - q^\epsilon_l(t ) \quad \forall \ \ t > 0 , \\ [ 2.mm ] \widetilde \phi^\epsilon(0,t ) = - r^\epsilon_0(t ) , \quad \widetilde \phi^\epsilon(l , t ) = - r^\epsilon_l(t ) \quad \forall \ \ t > 0 . \end{array}\ ] ] as @xmath403 the above system has a unique solution @xmath404;z).$ ] using , we obtain @xmath405 and @xmath406;l^2(0,l ) ) } \leq c \epsilon^{3/4}.\end{aligned}\ ] ] let us set @xmath407 then , @xmath408 satisfy the system and the estimate holds . using the above theorem now we can prove theorem [ thm1 ] . * proof of theorem [ thm1 ] . * theorem [ thm5 ] shows , in the observability inequality , l.h.s @xmath409 and r.h.s @xmath410 hence can not hold . hence system - is not null controllable . [ rem : lessregcontrol ] the above result is established when @xmath411 , @xmath412 and @xmath413 . we can extend this negative result to less regular control . more precisely we can take @xmath414 and @xmath415 in order to give a sense to the r.h.s . of , we need to replace @xmath416 and @xmath417 in by @xmath418 and @xmath419 respectively . the observability inequality becomes @xmath420 where @xmath134 is the solution of the adjoint system . using the same construction as above one can prove that this observability inequality does not hold and hence the system is not null controllable . [ rem : bdycontrol ] one can use the above gaussian beam construction to rule out null controllability using a boundary control . let us consider the system - with @xmath421 and @xmath422 in this case null controllability is equivalent to the following observability inequality @xmath423 where @xmath134 is the solution of the adjoint system . we can use the above construction to show that the above observability inequality does not hold and hence the system is not null controllable by a boundary control in any time @xmath40 . we proved that the system - is not null controllable when initial condition lies in @xmath130 . so the natural question is if the system is null controllable or not when the initial conditions are regular . in case of barotropic fluid , if initial density lies in @xmath424 then the system is null controllable ( see @xcite and @xcite ) . in this section we choose @xmath55 . we will first show that - is null controllable by velocity and temperature control only ( i.e. when @xmath53 ) acting everywhere in the domain with this initial regular condition . then we will show that we can not achieve null controllability by localizing velocity and temperature control . we consider the following interior control system @xmath425 \displaystyle u_t - \nu_0 u_{xx } + \frac{r { \bar \theta}}{{\bar \rho}}\;\rho_x + r \theta_x \ ; = g \chi_{\mathcal{o}_2 } , \mbox { in } ( 0,l ) \times ( 0,t ) , \\[3.mm ] \displaystyle \theta_t - k_0 \theta_{xx } + \frac{r { \bar \theta}}{c_v } u_x \ ; = h \chi_{\mathcal{o}_3},\ \mbox { in } ( 0,l ) \times ( 0,t ) , \\ [ 2.mm ] \displaystyle \rho(0 ) = \rho_0,\quad u(0 ) = u_0 , \quad \theta(0 ) = \theta_0 , \mbox { in } ( 0,l ) \\ [ 2.mm ] u(0,t ) = 0 = u(l , t ) \quad \forall \ \ t > 0 , \quad \theta(0,t ) = 0 = \theta(l , t ) \quad \forall \ \ t > 0 . \end{array}\ ] ] here @xmath38 and @xmath39 are velocity and temperature control respectively . let us first explain why we need average zero condition for initial density . integrating the density equation of in @xmath79 and using the boundary conditions we deduce that @xmath426 therefore @xmath427 thus if the system is null controllable in time @xmath40 then necessarily @xmath428 let us define @xmath429 we have the following lemma about existence and uniqueness of solution to the system follows easily from semigroup theory . given @xmath430 @xmath57 and @xmath431 the system has a unique solution @xmath432 with @xmath433 and @xmath434 moreover @xmath42 belongs to @xmath435;v).$ ] * proof of theorem [ thm2 ] . * let us first take the following system with interior control @xmath436 @xmath437 \displaystyle u_t - \nu_0 u_{xx } + \frac{r { \bar \theta}}{{\bar \rho}}\;\rho_x \ ; = \tilde g , \mbox { in } ( 0,l ) \times ( 0,t ) , \\[3.mm ] \displaystyle \rho(0 ) = \rho_0,\quad u(0 ) = u_0 , \mbox { in } ( 0,l ) , \\ [ 2.mm ] u(0,t ) = 0 = u(l , t ) \quad \forall \ \ t > 0 . \end{array}\ ] ] by theorem 5.1 of @xcite , we know that for every @xmath438 there exists a control @xmath439 such that the solution of satisfies @xmath440 now we consider the following heat equation @xmath441 \displaystyle \theta(x,0 ) = \theta_0(x ) , x \in { ( 0,l ) } , \ \ \theta(0,t ) = 0 = \theta(l , t ) \quad \forall \ \ t > 0 . \end{array}\ ] ] by theorem 2.66 of @xcite , for every @xmath442 there exists a control @xmath443 such that the solution of satisfies @xmath444 now define @xmath445 where @xmath446 , @xmath447 are the solutions of and respectively . thus @xmath448 is the solution of the system with @xmath38 and @xmath39 defined as above and it satisfies @xmath449 therefore the theorem follows . now we want to show that even if @xmath450 null controllability can not be achieved by localized velocity and temperature controls , i.e. , we want to prove theorem [ thm3 ] . first we will derive an observability inequality . in order to do this we introduce , @xmath451 differentiating the first equation of @xmath452 we obtain the following system @xmath453 \displaystyle u_t - \nu_0 u_{xx } + \frac{r { \bar \theta}}{{\bar \rho}}\;\alpha + r \theta_x \ ; = g \chi_{\mathcal{o}_2 } , \mbox { in } ( 0,l ) \times ( 0,t ) , \\[3.mm ] \displaystyle \theta_t ( x , t ) - k_0 \theta_{xx } + \frac{r { \bar \theta}}{c_v } u_x \ ; = h \chi_{\mathcal{o}_3 } , \mbox { in } ( 0,l ) \times ( 0,t ) , \\ [ 2.mm ] \displaystyle \alpha(0 ) = \alpha_0 : = ( \rho_0)_x,\quad u(0 ) = u_0 , \quad \theta(0 ) = \theta_0 , \mbox { in } ( 0,l ) , \\ [ 2.mm ] u(0,t ) = 0 = u(l , t ) \quad \forall \ \ t > 0 , \quad \theta(0,t ) = 0 = \theta(l , t ) \quad \forall \ \ t > 0 . \end{array}\ ] ] here @xmath454 . the system is well posed in @xmath130 . note that , to prove theorem [ thm3 ] , it is enough to show system is not null controllable in @xmath130 by localized interior controls @xmath38 and @xmath39 . as before we have the following proposition , for every @xmath455 the system is null controllable in time @xmath87 by localized interior controls @xmath456 and @xmath457 if and only if for every @xmath373 the solution of the following adjoint system @xmath458 \displaystyle - v_t - \nu_0 v_{xx } + \frac{r { \bar \theta}}{{\bar \rho}}\;\sigma_{xx } - r \phi_x \ ; = 0 , \mbox { in } ( 0,l ) \times ( 0,t ) , \\[3.mm ] \displaystyle - \phi_t - k_0 \phi_{xx } - \frac{r { \bar \theta}}{c_v } \phi_x \ ; = 0 , \mbox { in } ( 0,l ) \times ( 0,t ) , \\ [ 2.mm ] \displaystyle \sigma(t ) = \sigma_t , \quad v(t ) = v_t , \quad \phi(t ) = \phi_t , \mbox { in } ( 0,l ) , \\ [ 2.mm ] \sigma(0,t ) = 0 = \sigma(l , t ) \quad \forall \ \ t > 0 , \quad v(0,t ) = 0 = v(l , t ) \quad \forall \ \ t > 0 , \\ [ 2.mm ] \phi(0,t ) = 0 = \phi(l , t ) \quad \forall \ \ t > 0 . \end{array}\ ] ] satisfies @xmath459 the system is well posed in @xmath130 . let us consider the adjoint problem in @xmath64 @xmath460 \displaystyle - v_t - \nu_0 v_{xx } + \frac{r { \bar \theta}}{{\bar \rho}}\;\sigma_{xx } - r \phi_x \ ; = 0 , \mbox { in } \mathbb{r } \times ( 0,t ) , \\[3.mm ] \displaystyle - \phi_t - k_0 \phi_{xx } - \frac{r { \bar \theta}}{c_v } \phi_x \ ; = 0 , \mbox { in } \mathbb{r } \times ( 0,t ) , \\ [ 2.mm ] \displaystyle \sigma(t ) = \sigma_t , \quad v(t ) = v_t , \quad \phi(t ) = \phi_t , \mbox { in } \mathbb{r}. \end{array}\ ] ] applying the fourier transform in , we obtain the following system of ode @xmath461 where @xmath462 eigenvalues of @xmath463 are the same as eigenvalues of @xmath202 . let @xmath274 be the eigenvalue satisfying @xmath272 the eigenfunction of @xmath463 corresponding to @xmath274 is @xmath464 where @xmath290 is as in lemma [ lem : eigenfunction ] . thus if we choose @xmath465 the solution of can be written in the following way @xmath466 we have the following theorem . [ thm8 ] let @xmath467 be as in theorem [ thm4 ] and @xmath309 be as in . they satisfy the following * @xmath310 * for any @xmath311 , there exists a constant @xmath152 independent of @xmath312 such that @xmath468 * @xmath469 + for some positive constant @xmath152 independent of @xmath312 . the proof is similar to theorem [ thm4 ] . now we can proceed in a similar way as in theorem [ thm5 ] to prove theorem [ thm3 ] . in this section we will discuss null controllability of system . we want to prove theorem [ thm1.4 ] . first we have the following proposition about existence and uniqueness of the system . let @xmath121 . let us assume that @xmath115 , + @xmath122 and @xmath470 then has a unique solution @xmath42 with @xmath124 , @xmath125 and @xmath471 moreover @xmath42 belongs to @xmath435;z).$ ] proceeding as before we have the following proposition about the equivalence of null controllability and observability inequality . for every initial state @xmath41 the system is null controllable in any time @xmath87 by localized interior controls @xmath94 @xmath456 and @xmath457 if and only if for every @xmath472 @xmath473 the solution of the following adjoint system @xmath474 \displaystyle - v_t ( x , t)- \nu_0 v_{xx } - \frac{r { \bar \theta}}{{\bar \rho}}\;\sigma_x - \bar v v_x - r \phi_x \ ; = 0 , \mbox { in } { ( 0,l)}\times ( 0,t ) , \\[3.mm ] \displaystyle - \phi_t -\frac{k}{{\bar \rho}c_v } \phi_{xx } - \frac{r { \bar \theta}}{c_v } v_x - \bar v \phi_x(x , t ) \ ; = 0 , \mbox { in } { ( 0,l)}\times ( 0,t ) , \\[3.mm ] \displaystyle \sigma(t ) = \sigma_t,\quad \quad v(t)= v_t , \mbox { and } \phi(t ) = \phi_t , \qquad \mbox { in } { ( 0,l)}\times ( 0,t ) , , \\ [ 2.mm ] \displaystyle \sigma(l , t ) = 0 , \;\;\ ; v(0,t ) = 0 = v(l , t ) , \;\ ; t \in ( 0,t ) , \\ [ 2.mm ] \displaystyle \phi(0,t ) = 0 = \phi(l , t ) , \;\ ; t \in ( 0,t ) . \end{array}\ ] ] satisfies the following observability inequality @xmath475 the above adjoint system is well posed in @xmath476 . we want to show the the observability inequality does not hold for small time @xmath87 . let us first consider the adjoint problem , in @xmath477 as a terminal value problem only @xmath478 \displaystyle - v_t - \nu_0 v_{xx } - \frac{r { \bar \theta}}{{\bar \rho}}\;\sigma_x - \bar v v_x - r \phi_x \ ; = 0 , \mbox { in } \mathbb{r } \times ( 0,t ) , \\[3.mm ] \displaystyle - \phi_t -\frac{k}{{\bar \rho}c_v } \phi_{xx } - \frac{r { \bar \theta}}{c_v } v_x - \bar v \phi_x \ ; = 0 , \mbox { in } \mathbb{r } \times ( 0,t ) , \\[3.mm ] \displaystyle \sigma(t ) = \sigma_t,\quad \quad v(t)= v_t , \mbox { and } \phi(t ) = \phi_t , \qquad \mbox { in } \mathbb{r}. \end{array}\ ] ] we have the following theorem . let @xmath338 be as in theorem [ thm4 ] . let @xmath383 be the solution of with this terminal condition . then they satisfy the following estimates * @xmath479 * for any @xmath311 , there exists a constant @xmath152 independent of @xmath312 such that @xmath480 where @xmath481 * there exists a constant @xmath360 independent of @xmath482 such that @xmath483 let @xmath134 be a solution of . let us define the transformed functions for @xmath484 @xmath485 then @xmath486 is a solution of the system . thus from theorem [ thm4 ] , we easily prove the theorem . we now give the proof of theorem [ thm1.4 ] . * proof of theorem [ thm1.4 ] . * we will proceed in a similar way as in theorem [ thm5 ] . in order to obtain a cotradiction to the observability inequality , we need @xmath487 and does not intersect the set @xmath488 for all @xmath489 in particular @xmath490 must lie inside @xmath79 and does not intersect @xmath491 thus we have to choose @xmath492 and @xmath493 properly so that @xmath490 is either a proper subset of @xmath494 or @xmath495 . let us first choose @xmath496 and @xmath497 such that @xmath498 . thus if we want @xmath499 to be a proper subset of @xmath494 , we need @xmath500 this implies @xmath501 similarly if we choose @xmath502 then we need @xmath503 thus when @xmath504 the observability inequality does not hold and hence the system is not null controllable . in this section , we will prove theorem [ thm1.5 ] . we will construct highly localized solutions , as we did in one dimension . let us introduce the following constants @xmath505 the system is well posed in @xmath506 as before we have the equivalence between null controllability and observability inequality for every initial state @xmath507 the system is null controllable by localized interior controls @xmath115 and @xmath508 if and only if for every @xmath509 the solution of the following adjoint system @xmath510 \displaystyle - { \bf v}_t - \mu_0 \delta { \bf v } - \gamma_0 \nabla \mathrm{div } \ { \bf v } - b_1 \nabla \sigma \;=\;0 , \mbox { in } \omega \times ( 0,t ) , \\[2.mm ] \displaystyle \sigma(t ) = \sigma_t\quad \mbox{and}\quad { \bf v}(t)= { \bf v}_t , \quad \mbox { in } \omega , \\ [ 2.mm ] \displaystyle { \bf v } = { \bf 0 } \ \ \mbox { on } \partial \omega \times ( 0,t ) . \end{array}\ ] ] satisfies the observability inequality @xmath511 here @xmath512 the above adjoint system is well posed in @xmath506 we will now use `` gaussian beam '' construction to show that the observability inequality does not hold . let us first consider the adjoint problem in @xmath513 as a terminal value problem only . applying the fourier transformation , we obtain the following system of ode * the eigenvalues of @xmath516 are @xmath517 @xmath518 * for all @xmath519 there exists a constant @xmath520 , such that @xmath521 * there exists @xmath522 such that for all @xmath523 , @xmath524 and @xmath270 are all real and distinct . the eigenvalues satisfy @xmath525 * for @xmath258 the eigenvalues are differentiable and we have @xmath526 for some positive constant @xmath152 . * the eigenfunction of @xmath516 corresponding to @xmath270 is @xmath527 let @xmath531 with @xmath532 and @xmath533 let @xmath534 be a smooth function compactly supported in the unit ball and of unit @xmath535 norm . for any @xmath305 , define @xmath536 and @xmath537 as in . then they satisfy the following estimates * @xmath538 * for any @xmath311 , there exists a constant @xmath152 independent of @xmath312 such that @xmath539 * there exists a positive constant @xmath152 independent of @xmath312 such that @xmath540 thus for @xmath312 small , we have @xmath544 notice that the constant @xmath152 is independent of @xmath361 therefore for @xmath312 small enough and @xmath545 we have @xmath546 thus there exists a positive constant @xmath333 , such that @xmath334 this proves ( ii ) .

let denote the thompson group with standard generators , . it is a long standing open problem whether is an amenable group . by a result of kesten from 1959 , amenability of is equivalent to and to where in both cases the norm of an element in the group ring is computed in via the regular representation of . by extensive numerical computations , we obtain precise lower bounds for the norms in and , as well as good estimates of the spectral distributions of and of with respect to the tracial state on the group von neumann algebra . our computational results suggest , that it is however hard to obtain precise upper bounds for the norms , and our methods can not be used to prove non - amenability of . supported by the erc advanced grant no . oafpg 247321 , and partially supported by the danish national research foundation ( dnrf ) through the centre for symmetry and deformation at the university of copenhagen , and the danish council for independent research , natural sciences . ] supported by the erc advanced grant no . oafpg 247321 , and by the center for experimental mathematics at the university of copenhagen . ]

1409.1486

the thompson group @xmath0 is the group of homeomorphisms @xmath14\to[0,1]$ ] for which : * the endpoints satisfy @xmath15 , @xmath16 , * it is piecewise linear with finitely many break points on the diadic numbers @xmath17\cap ( 0,1)$ ] , * all slopes of @xmath18 are in the set @xmath19 . @xmath0 is a countable group , and it is generated by the elements @xmath20 whose graphs are shown in fig . [ f : generatorsaboff ] . of @xmath0.[f : generatorsaboff],title="fig : " ] of @xmath0.[f : generatorsaboff],title="fig : " ] moreover , it has a finite presentation in terms of @xmath20 , namely @xmath21=[ab^{-1},a^{-2}ba^2]=e\rangle,\ ] ] where the group commutator @xmath22=ghg^{-1}h^{-1}$ ] is defined as usual . recall that elementary amenability implies amenability , and a copy of the free group @xmath23 ( on two generators ) inside a group implies non - amenability of the group . it is known that @xmath0 is not elementary amenable , i.e. @xmath0 can not be obtained from finite or abelian groups by taking subgroups , quotients , extensions , and direct limits . on the other hand , by a result of brin and squier @xcite , @xmath0 does not contain a copy of @xmath23 . for more information on the thompson group @xmath0 , see the survey paper by cannon , floyd and perry @xcite . it is a main open problem to decide whether the thompson group @xmath0 is amenable . recently , monod @xcite has constructed examples of groups of homeomorphisms of @xmath24 $ ] which are non - amenable , but which also do not contain a copy of @xmath23 . these groups resemble @xmath0 . moreover , olesen and the second named author has shown in @xcite that if the reduced @xmath25-algebra @xmath26 of the ( non - amenable ) thompson group @xmath27 is simple , then @xmath0 is non - amenable . both of the above mentioned results suggest that @xmath0 might not be amenable , and extrapolations of our computational results point in the same ( non - amenability ) direction . the present paper grew out of an attempt to test the amenability problem for @xmath0 by using computers to estimate norms of certain elements in the group ring @xmath5 of @xmath0 . by the norm @xmath28 ( see also section [ s : two ] ) of an element @xmath29 in the group ring of a discrete group @xmath30 we mean @xmath31 where @xmath32 is the left regular representation of @xmath30 . as explained in section [ s : two ] , it is standard to write @xmath33 instead of @xmath34 , for any @xmath35 , and we will continue with this tradition . our starting point is the following two theorems due to kesten and lehner : ( see section [ s : six ] for a more detailed discussion ) . [ t:1.2 ] let @xmath30 be a discrete group with a generating set @xmath36 such that @xmath37 and @xmath38 . then , @xmath39 moreover , the upper bound is attained if and only if @xmath30 is amenable , and the lower bound is attained if and only if @xmath40 generates @xmath30 freely . [ t:1.3 ] let @xmath30 be a discrete group with a generating set @xmath36 such that @xmath37 and @xmath41 . then , @xmath42 moreover , the upper bound is attained if and only if @xmath30 is amenable , and the lower bound is attained if and only if @xmath40 generates @xmath30 freely . hence for the thompson group @xmath0 we get [ c:1.4 ] let @xmath43 and @xmath44 be the standard generators of @xmath0 , and let @xmath45 denote the unit element of @xmath0 . then @xmath46 @xmath47 moreover , in both cases the upper bound is attained if and only if @xmath0 is amenable . let @xmath48 denote the von neumann algebra of a discrete group @xmath30 , i.e. @xmath48 is the von neumann algebra in @xmath49 generated by @xmath50 . then @xmath51 defines a normal faithful tracial state on @xmath48 ( see e.g. section 6.7 in @xcite ) . hence @xmath52 ( cf . section [ s : four ] ) . hence if we knew all the numbers @xmath53 we could also compute the norm @xmath54 . in practice we can only compute a finite number of the moments @xmath55 . in this paper , we develop efficient methods , both mathematically and computationally to compute the numbers @xmath55 in the case @xmath56 for any finite set @xmath57 in a discrete group @xmath30 . we then apply the methods to the elements @xmath58 and @xmath59 in the group ring @xmath5 of the thompson group @xmath0 given by @xmath60 as a result , we have been able to compute the moments @xmath61 for @xmath62 and the moments @xmath63 for @xmath64 . using the spectral theorem to the self - adjoint operators @xmath65 ( cf . section [ s : two ] ) , one gets that there are unique probability measures @xmath66 to @xmath67 with @xmath68 $ ] such that @xmath69 is invariant under the reflection @xmath70 , and such that @xmath71 moreover , @xmath72 . using methods from the theory of orthonormal polynomials applied to these two measures ( cf . section [ s : four ] ) we obtain from our moment calculations good lower bounds for @xmath73 , @xmath74 , namely @xmath75 and @xmath76 in fact for each @xmath77 ( resp . @xmath78 ) , we find a lower estimate of @xmath79 ( resp . @xmath80 ) , and a suitable extrapolation of those two finite series of numbers suggests that the actual norms are much closer to 3 ( resp 4 ) , namely @xmath81 furthermore , based on our moment calculations , we have also been able to estimate the lebesgue densities of the measures @xmath82 and @xmath83 with fairly high precision . this shows that the measures @xmath82 and @xmath83 are very close to zero on the interval @xmath84 $ ] and @xmath85 $ ] respectively , but we can not rule out , that the measures have very thin tails " stretching all the way up to @xmath86 and @xmath87 , respectively , which would imply that @xmath0 is amenable ( cf . corollary [ c:1.4 ] ) . in comparison , one gets for the free group @xmath23 on two generators @xmath88 that @xmath89 and @xmath90 the measures @xmath82 , @xmath83 based on @xmath88 instead of @xmath20 , will be denoted by @xmath91 ( @xmath74 ) , and they can be computed explicitly ( see section [ s : six ] ) : @xmath92}(x ) \ , dx\\ & & \mu_2^{\mathrm{free}}= \frac 2{\pi } \frac{\sqrt{12-x^2}}{16-x^2 } 1_{[-2\sqrt 3,2\sqrt3]}(x ) \ , dx.\end{aligned}\ ] ] since 2007 a number of papers has been published about computational approaches to problems related to the thompson group @xmath0 , including the amenability problem ( cf . @xcite,@xcite,@xcite,@xcite,@xcite ) . our paper is the first that considers the moments of @xmath93 for @xmath94 . in @xcite burillo , cleary and wiest use probabilistic methods to estimate the moments of @xmath63 for @xmath95 for @xmath96 . moreover , elder , rechnitzer and wong compute in @xcite the first 22 cogrowth coefficients of @xmath0 with respect to the symmetric set of generators @xmath97 , from which the exact values of @xmath98 @xmath99 can easily be computed . we will comment in more detail on the results of @xcite and @xcite at the end of section [ s : four ] . the thompson group @xmath0 has been the key motivation for the research presented in this paper . however , we should stress , that the methods developed in this paper can be applied to do similar computations in any other finitely generated group , provided there is a reasonable fast algorithm to decide whether two words in the generators correspond to the same element of the group . the rest of the paper is organized as follows : section [ s : two ] contains background material on the reduced @xmath25-algebra @xmath100 and the group von neumann algebra @xmath48 associated to a discrete group @xmath30 . in section [ s : three ] , we consider for any given finite set @xmath57 in a discrete group @xmath30 , the element @xmath101 and explain how one can compute the moments @xmath102 from computing much smaller number @xmath103 , @xmath104 , @xmath105 , @xmath106 ( @xmath107 ) associated to the pair @xmath108 . we next apply this to the special case of @xmath109 , the thompson group @xmath0 , and @xmath110 or @xmath111 in order to compute numerically the moments @xmath112 for @xmath113 in the first case and for @xmath114 for the second case . in sections [ s : four ] and [ s : fourb ] we apply methods from the theory of orthogonal polynomials to estimate the norms @xmath115 and @xmath116 , and to estimate the probability measures @xmath82 , @xmath83 , associated to @xmath117 and @xmath118 via formula ( [ e:4plus ] ) above . the sections [ s : five ] and [ s : six ] are the main theoretical sections of the paper . in the general setting of a finite subset @xmath57 of a discrete group @xmath30 , we derive in section [ s : five ] the formulas ( used in section [ s : three ] ) , that allow us to pass back and forth between the 5 sequences of numbers @xmath103 , @xmath104 , @xmath105 , @xmath106 and @xmath112 ( @xmath107 ) . in section [ s : six ] we formulate and explain kesten s and lehner s results ( theorem [ t:1.2 ] and theorem [ t:1.3 ] ) in the setting of leinert sets ( cf . theorem [ t : lowerboundleinersetsupperboundamenability ] ) . moreover , we make comparison between our numbers @xmath103 , @xmath104 , @xmath105 , @xmath106 ( @xmath107 ) and the cogrowth coefficients due to cohen @xcite and grigorchuk @xcite , and show that amenability of the group @xmath119 generated by @xmath120 can be decided from the asymptotics of each of the 4 sequences mentioned above ( cf . corollary [ c:7.9 ] ) . let @xmath30 be a countable discrete group . we consider the left regular representation @xmath32 of @xmath30 on @xmath121 i.e. @xmath122 we will also use the letter @xmath32 for the extension of the regular representation to the group ring @xmath123 of @xmath30 i.e. @xmath124 for any finite subset @xmath125 and any set of complex numbers @xmath126 indexed by @xmath30 . the reduced @xmath25-algebra of @xmath30 is the @xmath25-algebra generated by @xmath50 , i.e. @xmath127 and the group von neumann algebra @xmath48 is the von neumann algebra generated by @xmath50 , i.e. @xmath128 where the closure is taken in the strong operator topology on @xmath129 . note that @xmath48 can also be expressed as @xmath130 ( double commutant ) . moreover , @xmath131 with isometric inclusions . we define the norm of an element @xmath132 by @xmath133 let @xmath134 for @xmath135 . then @xmath136 form an orthonormal basis for @xmath121 . the functional @xmath137 is a normal faithful tracial state on @xmath48 . in particular , @xmath138 @xmath139 @xmath140 @xmath141 since @xmath142 is one - to - one , we may consider @xmath123 as a subalgebra of @xmath143 . then the trace @xmath11 restricted to @xmath123 satisfies @xmath144 in particular , @xmath145 and @xmath146 for @xmath147 . moreover , for @xmath148 , @xmath149 for all the above see e.g. section 6.7 of @xcite or section 2.5 of @xcite . let @xmath150 be a self - adjoint operator in @xmath48 . then @xmath151 given by @xmath152 is a positive functional on @xmath153 such that @xmath154 . hence , there is a unique probability measure @xmath155 on @xmath156 for which @xmath157 since @xmath11 is a faithful trace , @xmath158 we will also consider @xmath155 as a probability measure on all of @xmath67 concentrated on @xmath156 . since @xmath159 , the spectral radius of @xmath160 , we have @xmath161 where @xmath162 and @xmath163 . by the moments of @xmath160 we mean the numbers @xmath164 note that @xmath165 are also the moments of the measure @xmath155 , i.e. @xmath166 and since @xmath155 has a compact support , @xmath155 is uniquely determined by its moments . for a not necessarily self - adjoint element @xmath167 , it is convenient to consider the self - adjoint operator @xmath168 and its moments @xmath169 and spectral distribution @xmath170 with respect to the normal faithful trace @xmath171 on @xmath172 given by @xmath173 , where @xmath174 on @xmath175 , i.e. @xmath176\in m_2(l(\gamma)).\ ] ] using the trace properties of @xmath11 , one has @xmath177 since all the odd moments vanish , @xmath178 is symmetric . i.e. @xmath179 , where @xmath180 in particular , @xmath181 and hence by ( [ e : normsabsalphaabsbeta ] ) @xmath182 and moreover @xmath183 note also that @xmath184 is equal to the image measure @xmath185 of @xmath170 by the map @xmath186 because @xmath184 and @xmath185 are both compactly supported and by ( [ e : mntautilde ] ) they have the same moments . let @xmath30 be a discrete group and let @xmath125 be a finite set with @xmath187 elements ( @xmath188 ) . our main example will be @xmath109 , the thompson group , and @xmath57 either @xmath189 or @xmath97 , but we will for some time stick to the general case in order to treat the two particular cases in a similar way . our goal is to compute as many as possible of the moments @xmath190 where @xmath191 and @xmath11 is the trace on @xmath123 coming from the group von neumann algebra @xmath48 as in section [ s : two ] . recall that @xmath192 hence @xmath193 , and for @xmath107 , the numbers @xmath194 are @xmath195 where @xmath196 denotes the number of elements in a set @xmath40 . for composing elements of the thompson group @xmath0 , we used the belk and brown forest algorithm from @xcite . consider the subsets @xmath197 @xmath198 @xmath199 we say that @xmath200 is cyclic " as all the cyclic rotations of each element in @xmath200 are contained in the set itself . since these subsets are smaller , it takes less time to compute the reduced " numbers @xmath201 and even less time to compute the cyclic " numbers @xmath202 the relationship between these numbers and the moments are derived in section [ s : five ] . in particular , we have @xmath203 which shows that the first @xmath204 moments ( @xmath205 , , @xmath112 ) can be computed from either @xmath206 or @xmath207 . in the case of a symmetric set @xmath57 , i.e. @xmath208 , the reduced numbers @xmath105 are the even co - growth coefficients @xmath209 introduced by cohen @xcite and grigorchuk @xcite . in the notation of elder - rechnitzer - wong @xcite , @xmath210 and @xmath211 . ( still for a symmetric set @xmath212 in cohen words , @xmath105 is the number of reduced " words in @xmath57 of length @xmath213 which represent the unit element in @xmath30 . it is much faster to compute the @xmath105 s or @xmath106 s , than to compute the moments @xmath112 directly from ( [ e : mncomp ] ) . however , we found a different approach to compute the moments , which speeded up the computations much further : let @xmath214 and recall that @xmath215 is the 2-norm on @xmath48 associated with the trace @xmath11 . then @xmath216 and the reduced number @xmath105 can be computed from the square of the 2-norm of @xmath217 by the following two equations @xmath218 ( cf . section [ s : five ] ) , and hence the moment series @xmath219 can also be computed from the numbers @xmath220 . in practice we computed @xmath103 as follows . note first that @xmath221 where @xmath222 is the set of all distinct terms in the sum ( [ e:3.5 ] ) , and @xmath223 is the multiplicity of the occurrence of @xmath224 in the sum ( [ e:3.5 ] ) . note that @xmath225 , @xmath226 , @xmath227 , @xmath228 , etc . since @xmath229 we have @xmath230 the advantage of computing the squared 2-norms @xmath103 instead of the reduced numbers @xmath105 or the cyclic numbers @xmath106 is that we only have to consider @xmath231 words of length @xmath204 instead of @xmath232 words of length @xmath213 . this made it possible for us to almost double the number of moments we could compute in the two cases @xmath110 and @xmath233 for the thompson group @xmath0 . the only drawback was that we need first to store all the terms of the sum ( [ e:3.5 ] ) , and next to sort the list in order to compute the multiplicities @xmath234 , @xmath224 . we wrote two programs , both using parallel computing , to calculate the squared 2-norms @xmath103 , which can be downloaded at : + + https://github.com/shaagerup/thompsongroupf/ + https://github.com/mariars/thompsongroupf/ + http://www.math.ku.dk/~haagerup/thompsongroupf/ + http://www.math.ku.dk/~mrs/thompsongroupf/ + .the series of numbers for @xmath235 ( case 1).[t : case1seriesofnumbers ] [ cols="<,<,<,<,<",options="header " , ] .[f : iabgraphs ] ] [ f : abainvbinvgraphs ] ] recall that from proposition [ p:4.1 ] and proposition [ p:4.7 ] that @xmath236 and all three sequences on the left hand side converges monotonically to @xmath54 as @xmath237 . note that in both cases @xmath238 and @xmath239 are poor lower estimates ( see tables [ t : case1numbersforestimatingthenorm ] , [ t : case2numbersforestimatingthenorm ] and figs . [ f : iabgraphs ] , [ f : abainvbinvgraphs ] ) . for the listed range of integers @xmath204 , they both stay well below the known lower bound @xmath240 in case 1 , ( resp . @xmath241 in case 2 ) , while the lower estimates @xmath242 stay above this value for @xmath243 in case 1 ( resp . for @xmath244 in case 2 ) . the best exact lower bound for @xmath54 in case 1 ( @xmath245 ) we can obtain from our results is @xmath246 note however that by proposition [ p:4.4 ] , @xmath247 and since @xmath248 appear to be monotonically increasing for @xmath249 our computation results make it very likely that actually @xmath250 to get some prediction of the actual value of @xmath115 , we made a least squares fitting of the 26 numbers @xmath251 to a function of the following form @xmath252 and found that the optimal values of the parameters @xmath253 were @xmath254 in particular , this extrapolation argument predicts that @xmath255 however , we can in no way rule out that @xmath256 , i.e. that @xmath0 is amenable . in the same way , we get in case 2 , ( @xmath257 ) the precise lower bound @xmath258 and that most likely we have @xmath259 moreover , by making a least squares fitting of the @xmath260 numbers @xmath261 to a function of the form @xmath262 we found the values of @xmath253 to be @xmath263 in particular , this extrapolation method predicts that @xmath264 but again we can not rule out that @xmath0 is amenable . in @xcite , burillo , cleary and wiest used probabilistic methods to estimate the moments @xmath112 in case 2 ( @xmath265 ) for @xmath266 . in their notation , @xmath267 and @xmath268 . they found that ( see table 1 in @xcite ) @xmath269 and @xmath270 since @xmath271 is increasing slowly for @xmath272 we also have @xmath273 their estimates are based on random samples of words in the generators , and are therefore not precise lower bound for @xmath116 . in comparison , we found an exact lower bound @xmath274 of the norm based only on @xmath275 and a very likely lower bound @xmath276 based on the same list of moments . in @xcite , elder , rechnitzer and wong also worked on estimating the norm @xmath116 . they found the lower bound @xmath277 based on computing numerically the largest eigenvalue of the adjacency matrix for the cayley graph of @xmath0 restricted to balls in @xmath0 of size @xmath278 , which corresponds to consider elements of @xmath0 of distance up to @xmath279 from the identity element . moreover , they computed cohen s cogrowth coefficients @xmath280 for @xmath281 ( cf . table 3 in @xcite ) . note that @xmath280 is equal to our reduced " number @xmath105 ( cf . section [ s : three ] ) . in the notation of @xcite , @xmath282 , and @xmath210 , and they use a different method ( based on power series ) to pass from the @xmath283-series to the @xmath284-series . in this section , we will estimate the spectral distribution measures @xmath285 for @xmath235 ( case 1 ) and @xmath286 ( case 2 ) , based on the moment sequences listed in section [ s : three ] . this is done by computing the ( possibly signed ) measures @xmath287 given by @xmath288 where @xmath289 is the unique polynomial on @xmath67 of degree @xmath290 for which @xmath291 where @xmath292 $ ] in case 1 , and @xmath293 $ ] in case 2 . we do not know , whether @xmath285 has a density with respect to the lebesgue measure , but if it has a density @xmath294 and if @xmath294 happens to be in @xmath295 , then @xmath289 is simply the orthogonal projection of @xmath294 onto the subspace of @xmath295 spanned by @xmath296 , @xmath297 , @xmath298 , @xmath299 , @xmath300 . let @xmath301 be the sequence of legendre polynomials . then it is is well known that the sequence @xmath302 form an orthonormal basis for @xmath303,dt)$ ] . hence @xmath304 form an orthonormal basis for @xmath295 , where @xmath305 in case 1 and @xmath306 in case 2 . hence in the case @xmath285 has density @xmath307 , @xmath289 is the orthogonal projection of @xmath294 onto @xmath308 , i.e. @xmath309 where @xmath310 note that ( [ e:4.rhon ] ) can also be written as @xmath311 the latter formula also makes sense if @xmath285 does not have an @xmath312-density , and it is not hard to check that ( [ e:4.rhonseconddefn ] ) provides the unique solution to ( [ e:4.munyapprox ] ) also when @xmath285 does not have an @xmath312-density . based on our moment calculations we can compute the polynomials @xmath313 for @xmath314 in case 1 and for @xmath315 in case 2 . since @xmath285 is a symmetric measure , all the odd terms in ( [ e:4.rhonseconddefn ] ) hence @xmath289 is an even polynomial of degree at most @xmath290 . case 1 : in fig . [ f : iablebesguedensity ] we have for @xmath235 plotted @xmath316 for @xmath317 together with the corresponding free " density ( cf . section [ s : one ] ) and in figs . [ f : iablebesguedensitytail]-[f : iablebesguedensitytailsumoddeven ] we have plotted first @xmath318 and @xmath316 in the tail interval @xmath319 $ ] and next @xmath320 in the same interval . the reason is that the plot in fig . [ f : iablebesguedensitytail ] shows that for @xmath297 close to @xmath86 , @xmath318 and @xmath316 oscillates around zero with opposite signs , and therefore we expect that @xmath321 gives a better approximation to the measure @xmath285 . figure [ f : iablebesguedensitytailsumoddeven ] indicates , that @xmath285 has very little mass on the interval @xmath322 $ ] , and hence @xmath323 could be any number in the interval @xmath322 $ ] , including the number 2.95 found in section [ s : four ] by an extrapolation argument . recall from the end of section [ s : two ] that @xmath324 is the image measure of @xmath285 by the map @xmath325 . hence figs . [ f : iablebesguedensity]-[f : iablebesguedensitytailsumoddeven ] can also be used to compute an approximation to the spectral density of @xmath9 . , where @xmath235 [ f : iablebesguedensity ] ] , where @xmath235 [ f : iablebesguedensitytail ] ] , where @xmath235 [ f : iablebesguedensitytailsumoddeven ] ] , where @xmath326 [ f : abainvbinvlebesguedensity ] ] , where @xmath326 [ f : abainvbinvlebesguedensitytail ] ] , where @xmath326 [ f : abainvbinvlebesguedensitytailsumoddeven ] ] case 2 : in fig . [ f : abainvbinvlebesguedensity ] we have for @xmath286 plotted @xmath327 for @xmath328 together with the corresponding free " density , and in figs . [ f : abainvbinvlebesguedensitytail]-[f : abainvbinvlebesguedensitytailsumoddeven ] we have plotted @xmath329 , @xmath327 and @xmath330 in the tail interval @xmath331 $ ] . again @xmath332 appear to give the best approximation to the measure @xmath285 in the tail region , and it shows that @xmath285 must have very little mass in the interval @xmath333 $ ] . hence @xmath323 could be any number in the interval @xmath333 $ ] including the number @xmath334 found in section [ s : four ] by an extrapolation argument . note finally , that in the case of @xmath286 , @xmath335 . to see this , we just have to check that @xmath336 and @xmath285 have the same moments . the even moments coincide because @xmath337 and therefore @xmath338 moreover , @xmath339 where the first equality follows because @xmath285 is symmetric and the second equality follows because no word of odd length in @xmath340 can represent the identity in @xmath0 . [ [ sss : qnconstantq ] ] let @xmath341 be a fixed natural number . define the polynomials @xmath342 in @xmath343 $ ] recursively by @xmath344 since @xmath345 is an even polynomial for @xmath204 even and an odd polynomial for @xmath204 odd , there exists polynomials @xmath346 and @xmath347 in @xmath343 $ ] such that @xmath348 next we write the @xmath345 polynomials in terms of chebyshev polynomials . [ p : qnrelatedtochebyshevpolys ] let @xmath349 and @xmath350 be the chebyshev polynomials of first and second kind . then * @xmath351 * @xmath352 . let @xmath353 . then the recursion formula @xmath354 holds for all @xmath107 . letting @xmath355 we get @xmath356 the chebyshev polynomials also satisfy an identical recursion formula : @xmath357 hence , if we can choose constants @xmath358 such that @xmath359 then @xmath360 would hold for all @xmath361 . indeed , @xmath362 and @xmath363 is the only solution since @xmath364 , @xmath365 , @xmath366 , @xmath367 . this proves ( @xmath368 because @xmath369 the proof of @xmath8 is an adaptation from the proof of formula ( 2.6 ) in @xcite . from eq . ( 2 ) p. 184 in @xcite , the chebyshev polynomials satisfy @xmath370 it follows that @xmath371 for @xmath372 and hence for all @xmath373 because we are dealing with polynomials . since @xmath374 , @xmath373 , @xmath375 , we get @xmath376 from this we get @xmath377 simplifying , @xmath378 therefore @xmath379 the last equation together with eq . ( [ e : qnrelatedtoqn ] ) yields @xmath8 . [ [ section ] ] let @xmath30 be a discrete group , let @xmath125 be a finite set with @xmath187 elements @xmath380 , and let @xmath381 as in section [ s : three ] . define @xmath345 , @xmath382 , @xmath383 as in eqs . ( [ e : qnconstantq ] ) and ( [ e : q1nq2n ] ) for this value of @xmath384 . define the sequences @xmath385 , @xmath386 in the group ring @xmath123 of @xmath30 by @xmath387 @xmath388 in all cases , @xmath389 means @xmath390 @xmath391 , and @xmath392 . notice that @xmath393 and @xmath394 . in fact @xmath395 for @xmath204 odd , and for @xmath204 even @xmath396 and @xmath397 . [ t : hnconnectedtoqn ] let @xmath385 , @xmath386 be the sequences defined in eqs . ( [ e : hn ] ) and ( [ e : kn ] ) . then in the algebra @xmath398 of @xmath399 matrices over @xmath123 we have * @xmath400 for @xmath204 odd . * @xmath401 for @xmath204 even . these formulas can also be expressed in the group ring @xmath402 as * @xmath403 , + @xmath404 @xmath405 . * @xmath406 , + @xmath407 @xmath408 . we start by proving @xmath7 and @xmath8 by induction in @xmath409 . note first that @xmath7 holds for @xmath410 and that @xmath8 holds for @xmath411 because @xmath412 and similarly @xmath413 . we next prove the following 4 recursion formulas . @xmath414 @xmath415 to prove the first formula in eq . ( [ e : hnplus1knplus1even ] ) note that @xmath416 is odd . hence @xmath417 the second formula in eq . ( [ e : hnplus1knplus1even ] ) follows from this by replacing @xmath57 with @xmath418 . the two formulas in eq . ( [ e : hnplus1knplus1odd ] ) can be proven in exactly the same way noticing that in this case @xmath416 is even . the formulas in eqs . ( [ e : hnplus1knplus1even ] ) and ( [ e : hnplus1knplus1odd ] ) can be rewritten as @xmath419 for @xmath420 , @xmath204 even ; and @xmath421 for @xmath422 , @xmath204 odd . by induction in @xmath204 we get @xmath423 once we rewrite the above two matrix equations in terms of @xmath7 and @xmath8 , which is precisely the definition of @xmath345 given in formula @xmath424 . hence @xmath7 and @xmath8 hold . since @xmath425 we have by eq . ( [ e : q1nq2n ] ) @xmath426 for @xmath427 , proving @xmath428 . similarly for @xmath408 , we have @xmath429 proving @xmath430 . if @xmath208 then @xmath337 and @xmath431 for all @xmath409 . [ p : hnstarhneqh2nplus ] for all @xmath409 we have @xmath432 for @xmath410 we have from the proof of theorem [ t : hnconnectedtoqn ] that @xmath433 . consider now @xmath420 . if @xmath204 is odd , then @xmath434 where @xmath435 is obtained by summing over only those @xmath436 for which @xmath437 and @xmath438 when @xmath439 , and @xmath440 when @xmath441 . if @xmath442 then @xmath443 and thus @xmath444 . if @xmath441 then @xmath445 and thus @xmath446 . for @xmath447 , @xmath448 odd , we have @xmath449 because for fixed @xmath450 , @xmath451 with @xmath452 , @xmath453 can be chosen in @xmath454 ways , namely @xmath455 can first be chosen in @xmath456 ways , and next @xmath457 , @xmath458 etc . can each be chosen in @xmath459 ways . the same holds for @xmath448 even and/or @xmath204 even by obvious modifications of the above proof . we have @xmath460 for @xmath204 odd , @xmath395 by theorem [ t : hnconnectedtoqn ] . thus @xmath461 . for @xmath204 even , @xmath462 , theorem [ t : hnconnectedtoqn ] yields @xmath463 as @xmath464 for any @xmath465 . [ [ section-1 ] ] define the group ring sequence @xmath466 by @xmath467 where @xmath468 . we could say that @xmath469 is the cyclic version of @xmath470 defined in eq . ( [ e : hn ] ) . recall that @xmath471 , where @xmath125 is a finite set with @xmath187 elements . we assume @xmath472 . define the number sequences @xmath473 , @xmath474 , @xmath475 by @xmath476 [ p : xietazetabounded ] for @xmath409 , @xmath104 , @xmath105 , @xmath106 are integers and @xmath477 the numbers @xmath104 , @xmath105 , @xmath106 are integers because @xmath478 and @xmath479 for all @xmath480 . since @xmath481 for all @xmath482 , and @xmath217 is a sum on @xmath483 while @xmath469 is the same sum but on a subset of @xmath483 , we have @xmath484 . by proposition [ p : hnstarhneqh2nplus ] @xmath485 finally , @xmath486 [ p:6.7page29a ] @xmath487 we have @xmath488 where @xmath489 and where for @xmath447 , @xmath490 denotes the subset of @xmath491 for which @xmath492 and for @xmath493 : @xmath494 @xmath495 clearly @xmath496 . moreover @xmath497 because @xmath498 as @xmath499 . for @xmath447 ( @xmath448 odd ) we can write @xmath500 where the summation is over all @xmath501 and @xmath502 for which @xmath503 for fixed @xmath504 there are exactly @xmath454 choices of @xmath505 , namely first @xmath506 can be chosen in @xmath456 ways because @xmath507 and next each of @xmath508 , @xmath509 , @xmath299 , @xmath510 can be chosen in @xmath511 ways . since @xmath512 it follows that @xmath513 the same formula holds for @xmath448 even ( @xmath447 ) by an obvious modification of the proof . this proves the proposition . [ p : xietazetarelatedtoeachother ] for @xmath514 we have * @xmath515 * @xmath516 . similarly , * @xmath517 * @xmath518 consider the power series , @xmath519 by proposition [ p : xietazetabounded ] , they are all convergent for all @xmath520 with @xmath521 . we have already seen that @xmath7 follows from proposition [ p : hnstarhneqh2nplus ] . by @xmath7 we have for @xmath522 : @xmath523 hence for @xmath524 , we have @xmath525 by comparing the coefficients of @xmath526 in the power series expansion of @xmath527 and of @xmath528 we get ( @xmath529 ) . note next that @xmath428 follows from proposition [ p:6.7page29a ] . hence , as in the proof of ( @xmath530)@xmath531(@xmath529 ) we get @xmath532 which implies @xmath430 . [ [ section-2 ] ] let @xmath125 be a finite subset with @xmath187 elements and let as before @xmath533 and @xmath534 recall that the trace @xmath535 on @xmath398 was defined in eq . ( [ e : tautilde ] ) . let @xmath285 be the spectral measure of @xmath536 on the interval @xmath537 $ ] with respect to the trace @xmath171 , i.e. the unique probability measure on @xmath537 $ ] satisfying @xmath538 ( see section [ s : two ] ) . since @xmath539 and @xmath540 we have @xmath541 and @xmath542 the latter condition implies via riesz representation theorem that @xmath285 is symmetric . i.e. @xmath543 , where @xmath544 is the image measure of @xmath285 with respect to the map @xmath70 . ( this could also be shown algebraically ) . let @xmath545 notice that @xmath193 and @xmath546 . [ p : etazetaintermsofchebyshevpols ] for @xmath409 , let @xmath105 , @xmath106 be as in eq . ( [ e : xirhozetan ] ) , and @xmath547 , @xmath548 be the chebyshev polynomials of first and second kind . then * @xmath549 * @xmath550 . since @xmath551 , @xmath405 , we have for every polynomial @xmath552 $ ] : @xmath553 hence , by theorem [ t : hnconnectedtoqn](@xmath554 ) and eq . ( [ e : q1nq2n ] ) @xmath555 by proposition [ p : qnrelatedtochebyshevpolys](@xmath530 ) we get ( @xmath530 ) , and by proposition [ p : qnrelatedtochebyshevpolys](@xmath529 ) and proposition [ p : xietazetarelatedtoeachother]@xmath430 we get ( @xmath529 ) . define @xmath556 later on , we will show that these are the even moments for the measure in ( [ e : mutildeh1 ] ) . [ p : momentscontainingfreepart ] for @xmath107 , we have @xmath557 by euler s formula and the binomial theorem , we have @xmath558 substituting @xmath559 , we get @xmath560 for @xmath561 , and hence also for all @xmath373 as both sides of the equality are polynomials . substituting @xmath297 with @xmath562 we get @xmath563 integrating both sides with respect to @xmath285 we get by eq . ( [ e : mn ] ) and proposition [ p : etazetaintermsofchebyshevpols](@xmath529 ) that @xmath564 which proves the proposition . a simple reformulation of ( [ e : mnfromzetas ] ) yields a formula for computing the cyclic numbers @xmath106 from the moments @xmath112 : we have @xmath565 , and @xmath566 [ [ section-3 ] ] a subset @xmath57 of a group @xmath30 is called a leinert set if for all @xmath409 , all tuples @xmath567 satisfy @xmath568 by theorem iiif and theorem iid(b ) in @xcite we have * if @xmath125 and @xmath569 then @xmath570 is a leinert set if and only if @xmath57 generates freely a copy of the free group @xmath571 with @xmath572 generators inside @xmath30 . * if @xmath125 and @xmath573 then @xmath574 is a leinert set if and only if @xmath57 generates freely a copy of @xmath571 inside @xmath30 . by the following theorem , due to kesten and lehner , the norm of @xmath575 from section [ s : five ] is bounded by two values , the lower bound is related to leinert sets and the upper bound to amenability . [ t : lowerboundleinersetsupperboundamenability ] let @xmath57 be a finite set in a discrete group @xmath30 with @xmath187 elements @xmath576 . let @xmath577 . then * @xmath578 * @xmath579 if and only if the subgroup @xmath580 generated by @xmath120 is amenable . * @xmath581 if and only if @xmath57 is a leinert set . ( @xmath530 ) : the upper bound is trivial since each @xmath582 is a unitary operator in @xmath48 . the proof of the lower bound is the following : write @xmath583 . apply now proposition 2 and proposition 5 of @xcite to @xmath584 , @xmath585 , and @xmath586 the unique group homomorphism for which @xmath587 where @xmath588 are the generators of @xmath589 . then @xmath590 + ( @xmath529 ) : we have @xmath591 where @xmath592 for all @xmath593 because @xmath594 is self - adjoint . moreover , @xmath120 is a symmetric set ( i.e. @xmath595 ) , and @xmath596 hence , by section 3 of @xcite , @xmath597 is amenable if and only if @xmath598 . this proves @xmath8 because @xmath599 . + ( @xmath600 ) : the proof of @xmath428 follows from theorem 9 of @xcite . note that in view of remark [ r:7.3 ] below , theorem [ t:1.2 ] and theorem [ t:1.3 ] in the introduction are both special cases of theorem [ t : lowerboundleinersetsupperboundamenability ] . [ r:7.3 ] if @xmath57 is a symmetric set ( i.e. @xmath208 ) it follows immediately from kesten s theorem , that @xmath579 if and only if the subgroup @xmath601 generated by @xmath57 is amenable . this follows also from @xmath8 because @xmath602 is a subgroup of @xmath603 of index at most @xmath604 . ( recall that , amenability of a group is always preserved by any subgroup . on the other hand , amenability of a subgroup is preserved by the group if the index is finite . ) [ t : leinertxirhozetamnconnection ] let @xmath125 be a finite subset in a discrete group @xmath30 with @xmath187 elements , where @xmath472 . then , with the notation from section [ s : five ] , the following are equivalent : * @xmath57 is a leinert set . * @xmath605 for all @xmath409 . * @xmath606 for all @xmath409 . * @xmath607 for all @xmath409 . * @xmath608 for all @xmath409 . * @xmath609 , where @xmath610 is given by eq . ( [ e : mqn ] ) . * @xmath611 where @xmath612}(t)\,dt.\ ] ] @xmath613 : recall that @xmath614 since @xmath615 for @xmath616 and a leinert set omits the identity the equivalence follows . + @xmath617 this follows by definition of @xmath104 . @xmath618 : this follows from proposition [ p : xietazetarelatedtoeachother ] . @xmath619 : this follows from proposition [ p : momentscontainingfreepart ] . @xmath620 : by ( [ e : rhoeqintq2ndmuhtilde ] ) we have for @xmath621 that @xmath622 where @xmath623 $ ] . moreover , since @xmath285 is a symmetric measure ( cf . section [ s : two ] ) , and @xmath624 is an odd polynomial when @xmath625 is odd , @xmath626 hence @xmath430 is equivalent to @xmath627 since @xmath285 is also a probability measure @xmath628 since @xmath629 is the set of all polynomials in @xmath630 $ ] it follows from weierstrass approximation theorem and riesz representation theorem that @xmath631 is uniquely determined by ( [ e:36 ] ) and ( [ e:37 ] ) . in @xcite pp . 283 - 284 ( see also @xcite ) a sequence of polynomials @xmath632 is defined by @xmath633 for a fixed number @xmath634 , and it is proven in @xcite p. 284 that @xmath635 where @xmath636 for @xmath637 and @xmath638 . letting now @xmath639 , then for @xmath107 , @xmath640 coincide with our polynomials @xmath641 . hence using ( [ e:38 ] ) first with @xmath637 and @xmath642 , and next with @xmath643 we get @xmath644 and @xmath645 hence , the measure @xmath646 defined in @xmath647 satisfies ( [ e:36 ] ) and ( [ e:37 ] ) . therefore @xmath430 is equivalent to that @xmath611 . the odd moments of @xmath646 are zero . the even moments of @xmath646 are given by @xmath648 , i.e. @xmath649 where @xmath650 let @xmath57 be a leinert set with @xmath187 elements . by theorem [ t : leinertxirhozetamnconnection ] , @xmath611 and @xmath609 . hence @xmath651 [ [ section-4 ] ] @xmath652 + let @xmath40 be a finite set of generators of a group @xmath30 such that @xmath41 and @xmath653 . let @xmath654 and @xmath655 . in @xcite and @xcite , cohen and grigorchuk independently introduced the notion of cogrowth coefficients @xmath656 for @xmath657 , by putting @xmath658 equal to the number of elements in the set @xmath659 as cohen puts it , @xmath658 is the number of reduced words in @xmath57 of length @xmath204 , which represent the unit element of @xmath30 . since @xmath615 , @xmath616 we have @xmath660 note that since @xmath208 , we have @xmath661 according to the definition of the reduced numbers @xmath105 in ( @xmath662 ) . cohen proved in pp . 302 - 303 in @xcite that if @xmath653 and @xmath40 does not generate @xmath30 freely , then @xmath663 exists and @xmath664 $ ] . moreover if we let @xmath381 , ( assuming still that @xmath40 does not generate @xmath30 freely ) by theorem 3 in @xcite : @xmath665 since @xmath666 , it follows that @xmath667 we next prove the following extension of the above : [ t:2.6 ] let @xmath668 be a finite set with @xmath187 elements . let @xmath669 as in theorem [ t : lowerboundleinersetsupperboundamenability ] , and let @xmath670 , @xmath283 @xmath671 be as in ( @xmath662 ) and let @xmath672 then @xmath673 moreover , if @xmath57 is not a leinert set , and @xmath188 then @xmath674 and @xmath675 by theorem [ t : lowerboundleinersetsupperboundamenability ] @xmath676 hence @xmath677 . let @xmath678 then @xmath679 hence @xmath680 the function @xmath681 is strictly increasing on the interval @xmath682 , and @xmath683 . hence @xmath684 , which proves ( [ e:2.5 ] ) . assume next that @xmath188 and @xmath57 is not a leinert set . then by theorem [ t : lowerboundleinersetsupperboundamenability ] , @xmath685 . hence @xmath686 , which shows that @xmath687 . we next prove @xmath688 . since @xmath689 ( proposition [ p : xietazetabounded ] ) , it is sufficient to show that @xmath690 note that ( [ e:2.7 ] ) follows immediately from the following lemma : [ l:2.7 ] for all @xmath691 , @xmath692 by proposition [ p : etazetaintermsofchebyshevpols ] we have @xmath693 where @xmath285 is the spectral distribution of @xmath694 . in particular , @xmath695.\ ] ] by formula ( 2 ) in page 184 of @xcite @xmath696\\ & & t_k(\cosh u)=\cosh(k u),\qquad u\ge 0.\end{aligned}\ ] ] hence , @xmath697 for @xmath698 $ ] , @xmath699 , and @xmath700 is strictly increasing on @xmath701 . since also @xmath702 , we have @xmath703 hence by ( [ e:2.9 ] ) and ( [ e:2.10 ] ) @xmath704 by ( [ e : gammagesqrtq ] ) @xmath705 is non negative . then @xmath706 which implies that @xmath707 this proves that @xmath708 where the last inequality follows from the inequality @xmath709 . from proposition [ p : xietazetarelatedtoeachother](@xmath600 ) we get , using again @xmath709 that @xmath710 in particular , @xmath711 , @xmath712 . applying proposition [ p : xietazetarelatedtoeachother](@xmath530 ) to this inequality we get in the same way that @xmath713 . hence @xmath714 we prove now ( [ e:2.8 ] ) . recall that @xmath715 , and let @xmath716 be arbitrary . let @xmath717 since @xmath718 is strictly increasing on @xmath682 , we have @xmath719 since , by ( [ e : pmnormtinsupport ] ) , @xmath720 we also have @xmath721)>0.$ ] we have previously seen that @xmath722 for @xmath723 $ ] and that @xmath724 is positive and increasing on @xmath701 . hence @xmath725).\end{aligned}\ ] ] let now @xmath726 . then @xmath727 hence @xmath728 which shows that @xmath729).\ ] ] since @xmath730 it follows that @xmath731 and since @xmath732 was arbitrary we have @xmath733 proving ( [ e:2.8 ] ) . [ r:6.8 ] if @xmath57 is a leinert set , then @xmath734 by theorem [ t : lowerboundleinersetsupperboundamenability](@xmath600 ) . hence by theorem [ t : leinertxirhozetamnconnection ] @xmath735 while @xmath736 [ c:7.9 ] let @xmath30 be a discrete group and let @xmath125 be a finite set with @xmath187 elements @xmath576 , such that @xmath120 generates @xmath30 . then the following are equivalent : * @xmath30 is amenable * @xmath737 * @xmath738 * @xmath739 * @xmath740 . note as well that @xmath30 is amenable if and only if @xmath741 . if @xmath30 is amenable , then by theorem [ t : lowerboundleinersetsupperboundamenability](@xmath529 ) , @xmath579 and thus @xmath742 . since @xmath188 , @xmath743 . hence by theorem [ t : lowerboundleinersetsupperboundamenability](@xmath600 ) @xmath57 is not a leinert set , and @xmath744 in the corollary follows now from formula ( [ e:2.6 ] ) in theorem [ t:2.6 ] . conversely , if one of the statements @xmath8 , @xmath428 , @xmath430 , or @xmath745 in the corollary holds , then by remark [ r:6.8 ] , @xmath57 is not a leinert set , so using again formula ( [ e:2.6 ] ) in theorem [ t:2.6 ] we have @xmath746 , and hence by ( [ e:2.5 ] ) , @xmath747 , which by theorem [ t : lowerboundleinersetsupperboundamenability](@xmath529 ) implies that @xmath30 is amenable . by proposition [ p:4.1 ] @xmath748 which proves the last statement in the corollary . in @xcite farley proved that the thompson group @xmath0 admits a proper affine isometric action on a hilbert space or equivalently , @xmath0 has the haagerup property ( cf . hence by the result of higson and kasparov @xcite , @xmath0 satisfies the baum - connes conjecture with coefficients , which in turn implies that @xmath0 satisfies the kadison - kaplansky conjecture , i.e. ( since @xmath0 is torsion free ) the reduced @xmath25-algebra @xmath749 has no projections other than @xmath750 or @xmath296 ( see e.g. @xcite ) . here projections mean self - adjoint idempotents ( @xmath751 ) . this implies that the spectrum of every self - adjoint operator @xmath752 is a connected subset of @xmath67 . the following two propositions are simple applications of this : [ p : a1case1 ] let @xmath235 and @xmath753 . then there exists a @xmath754 such that @xmath755\cup [ \delta , ||h||].\ ] ] since @xmath594 is a self - adjoint element of @xmath749 , its spectrum is a connected subset of @xmath67 . hence @xmath756 $ ] , where @xmath757 and @xmath758 moreover , since @xmath759 , @xmath760 since @xmath324 is the image measure of the symmetric measure @xmath285 by the map @xmath761 ( see the end of section [ s : two ] ) , we have @xmath762\cup [ \sqrt{\alpha } , ||h||].\ ] ] put @xmath763 . then @xmath764 $ ] . to see that @xmath765 , note first that @xmath766 because @xmath189 is an orthonormal set with respect to the inner product @xmath767 . hence @xmath768 hence @xmath769 by corollary [ c:1.4 ] . [ p : acase2 ] let @xmath286 and @xmath770 . then @xmath771.\ ] ] we know that @xmath335 ( see end of section [ s : fourb ] ) . hence @xmath772 and from section [ s : two ] , we have @xmath773\ ] ] and @xmath774 but since @xmath775 , @xmath776 is a connected subset of @xmath67 . hence @xmath777,\ ] ] proving the proposition . [ r : a3 ] by some extra work , it can be proved also for case 1 that @xmath778.\ ] ] however , in the rest of this section , we shall only need the following immediate corollary of proposition [ p : a1case1 ] and proposition [ p : acase2 ] : [ c : a4 ] in both case 1 and case 2 , @xmath779 is an infinite subset of @xmath67 without isolated points . let @xmath780 be a hilbert space , and let @xmath781 be the set of compact operators on @xmath780 and let @xmath782 denote the quotient map @xmath783 . then the essential spectrum @xmath784 of an element @xmath785 is the spectrum @xmath786 of @xmath787 , see e.g. p. 30 in @xcite . note that @xmath788 and for every @xmath789 , @xmath790 . for a self - adjoint operator @xmath791 , it is easy to see that @xmath792 if and only if for every @xmath793 , the spectral projection @xmath794 of @xmath160 is infinite dimensional . the following result is well known ( cf . theorem vii.10 in @xcite ) . [ p : a5 ] let @xmath791 . then @xmath795 if and only if the following two conditions hold * @xmath32 is an isolated point in @xmath156 . * @xmath32 is an eigenvalue of @xmath160 and @xmath796 . [ c : a6 ] if @xmath791 and @xmath156 has no isolated points then @xmath797 . [ p : a7 ] let @xmath798 be an operator of the form @xmath799 where @xmath800 for all @xmath621 and assume that @xmath801 has no isolated points . then @xmath802 the inequality on the left hand side follows from the proof of proposition [ p:4.4 ] ( without any assumption on @xmath801 ) . put @xmath803 then @xmath804 , where @xmath345 denotes the projection on @xmath805 , and @xmath806 is the standard basis for @xmath807 . but since @xmath808 is a compact perturbation of @xmath809 , @xmath810 so by corollary [ c : a6 ] @xmath811 and since the norm of a self - adjoint operator is equal to its spectral radius , also @xmath812 so altogether , @xmath813 by the proof of proposition [ p:4.4 ] , @xmath814 hence @xmath815 [ c : a8 ] let @xmath235 ( case 1 ) or @xmath286 ( case 2 ) and let @xmath799 be the operator in @xmath816 built from @xmath285 as in section [ s : four ] , then @xmath817 by the proof of proposition [ p:4.3 ] , @xmath809 is unitarily equivalent to the multiplication operator @xmath818 on @xmath819 given by @xmath820 , @xmath821 . hence @xmath822 . formula ( [ e:90 ] ) now follows from proposition [ p : a7 ] and corollary [ c : a4 ] . this section contains a proof of the test for computational errors mentioned at the end of section [ s : three ] . let @xmath30 be a group and @xmath125 a finite subset of @xmath30 with @xmath187 elements @xmath576 . moreover , let @xmath112 , @xmath823 , @xmath104 , @xmath105 , @xmath106 be the sequences of numbers introduced in sections [ s : three ] and [ s : five ] . then the following holds : * @xmath824 and @xmath825 . * for @xmath420 the numbers @xmath104 , @xmath105 , @xmath106 , @xmath103 are all even integers , while the @xmath112 numbers have the same parity as the number @xmath826 . * let @xmath827 denote the mbius function , i.e. @xmath828 and for @xmath829 , @xmath830 if @xmath30 is torsion free , then the numbers @xmath831 are non - negative integers divisible by @xmath213 . in particular , if @xmath832 is a prime number , then @xmath833 and hence @xmath834 is divisible by @xmath835 . @xmath7 : by definition @xmath836 hence , by propositions [ p : xietazetarelatedtoeachother ] and ( [ e : xirhozetan ] ) , @xmath837 and @xmath825 . + @xmath8 : recall that for @xmath409 : @xmath838 and @xmath839 where @xmath840 define the reversing map @xmath841 by @xmath842 then @xmath843 and @xmath844 . moreover , for each @xmath845 , @xmath846 , because @xmath847 . hence , all the orbits in @xmath848 under the action of the group @xmath849 have size 2 . since @xmath844 , @xmath839 is an even number . thus , by proposition [ p : xietazetarelatedtoeachother ] , @xmath105 and @xmath104 are also even . hence for @xmath420 , @xmath850 is also even . the last statement in @xmath8 follows from ( [ e : mqn ] ) and ( [ e : mnfromzetas ] ) . + ( @xmath600 ) : let the reversing map @xmath851 and @xmath852 be as in the proof of ( @xmath529 ) , and define @xmath853 by @xmath854 then clearly @xmath855 , @xmath856 , @xmath857 , and the group @xmath858 of transformations of @xmath848 generated by @xmath859 and @xmath782 is equal to @xmath860 where @xmath861 . moreover , @xmath862 , the dihedral group with @xmath213 elements . we prove next the following claim : for each @xmath845 , the stabilizer group @xmath863 is contained in @xmath780 . + to prove the claim , we just have to show that @xmath864 . from the proof of @xmath8 , we already know that @xmath865 . let now @xmath866 and assume that @xmath867 . then @xmath868 by comparing the @xmath448th terms of the two tuples , we get @xmath869 , which is a contradiction . hence @xmath864 proving the claim . since @xmath870 is a subgroup of @xmath871 , @xmath870 is of the form @xmath872 where @xmath873 is a divisor of @xmath204 . note that @xmath874 . in particular @xmath875 . assume now that @xmath30 is torsion free and write @xmath876 where @xmath877 if @xmath878 , for a divisor @xmath879 of @xmath204 , for which @xmath880 , then @xmath881 where @xmath882 is repeated @xmath883 times , and @xmath884 for @xmath885 . since @xmath886 , @xmath887 and since @xmath30 is torsion free , also @xmath888 together with @xmath889 for @xmath885 , this shows that @xmath890 . conversely , if @xmath890 , then @xmath582 given by ( [ e : apendixbestar ] ) will be in @xmath891 . hence @xmath891 and @xmath892 have the same number of elements proving that @xmath893 hence , by the mbius inversion formula , ( cf . theorem 2.9 in @xcite ) we have @xmath894 recall that @xmath895 hence the orbit @xmath896 has @xmath213 elements for all @xmath897 , which implies that @xmath898 is divisible by @xmath213 proving ( @xmath600 ) .

the solar system started to form about 4.56 gyr ago and despite the long intervening time span , there still exist several clues about its formation . the three major sources for this information are meteorites , the present solar system structure and the planet - forming systems around young stars . in this introduction we give an overview of the current understanding of the solar system formation from all these different research fields . this includes the question of the lifetime of the solar protoplanetary disc , the different stages of planet formation , their duration , and their relative importance . we consider whether meteorite evidence and observations of protoplanetary discs point in the same direction . this will tell us whether our solar system had a typical formation history or an exceptional one . there are also many indications that the solar system formed as part of a star cluster . here we examine the types of cluster the sun could have formed in , especially whether its stellar density was at any stage high enough to influence the properties of today s solar system . the likelihood of identifying siblings of the sun is discussed . finally , the possible dynamical evolution of the solar system since its formation and its future are considered . november 2014

1501.03101

for decades the solar system was assumed to be the prototype for planetary system formation . with the detection of over a thousand confirmed exoplanets and many more candidates , it has become apparent that many planetary systems exist that differ substantially in their structural properties from our solar system . nevertheless the formation of the solar system is still of special interest for several reasons . first , it is only for the solar system that we can directly examine material that is left over from the formation process in the form of meteorites . second , only for the solar system do we have detailed structural information about the entire system including its smaller bodies . last but not least , it is only for the solar system that we know for sure that life exists . the three major sources about the formation of the solar system are meteorites , the present solar system structure and contemporary young planet - forming systems . we start by reviewing the current status of meteorite research concerning the chronology of early solar system formation including the formation of the terrestrial planets in section 2 . in this context the question of the origin of short - lived radioactive nuclei in these meteorites is of special interest . some of these can only be produced in supernovae events of high - mass stars - different possibilities are discussed in section 3 . other sources of information are young stars surrounded by accretion discs from which planetary systems might form . in section 4 the properties of these discs - masses , gas content and chemical composition - are discussed . estimates of the life times of these discs are given and the consequences for planet formation scenarios are discussed . section 5 provides a closer look at the different stages of planet formation . starting from dust grains , then considering pebble - sized objects to planetismals the current state of research is presented . this is followed by the final step in which planets form . many of these young systems are part of a cluster of stars . there are several indications that our own solar system also formed as part of a star cluster . section 6 gives the arguments for such an early cluster environment and discusses the possibilities of finding today stars that formed in the same cluster as our sun did . not only the location and masses of the planets but also those of the asteroid and kuiper belt are characteristics of our solar system that might potentially give clues to its formation . in section 7 the early dynamical evolution of the kuiper belt is illustrated . possible scenarios for the late heavy bombardment between 4.0 and 3.7 gyr ago are discussed . it is still an open question to what degree the solar system characteristics changed since its formation and how stable the solar system is in the long - run . the likely long - term evolution of the solar and other planetary systems is discussed in section 8 . this is followed by a summary in section 9 . first , we look at the information that meteorites give about the formation of the solar system . in order to do so a relative age dating of these meteorites is necessary . studying meteorites from our solar system is the only way to directly constrain timescales of its protoplanetary disc evolution . most meteorites are older than 4.5 billion years and originate from the asteroid belt . the asteroid belt between mars and jupiter provides the only vestige of the planetesimals which were the first larger objects in the protoplanetary disc that provided the building materials for the larger planets . simulations indicate that it is most likely that the larger planets formed via collisions of such first generation planetesimals ( e.g. wetherill 1990 , chambers 2003 ) . the different groups of meteorites sample these first generation planetesimals and cover the different evolutionary steps of early solar system evolution in great detail . in general , three major groups of meteorites can be distinguished . chondrites represent unprocessed , brecciated early solar system matter , whereas differentiated meteorites such as achondrites and iron meteorites originate from asteroids that have undergone melting and internal differentiation . these asteroidal melting events were triggered by either decay of short - lived @xmath0al or by impact events . due to the short half life of @xmath0al ( 0.7 myr ) , the first heating mechanism is confined to the first 5 million years of solar system evolution . the oldest dated solar system matter are ca , al - rich inclusions ( cais ) in chondritic meteorites that have been dated by the u - pb method to 4.567 - 4.568 billion years ( amelin et al . 2002 , 2011 ; bouvier et al . cais are an important anchor point to constrain the abundance of significant short - lived nuclides such as @xmath0al or @xmath1hf at the beginning of the solar system . in addition to the long lived u - pb chronometer , short - lived nuclides with their half - lifes of less than 100 million years enable dating of meteorites and their components at an age resolution as low as several tens of thousands of years . based on combined u - pb and al - mg chronometry , the ages of chondrules , a major component of chondrites , has been constrained to as late as up to 4 million years after solar system formation ( e.g. bizzarro et al . 2004 ; villeneuve et al . it is currently contentious , as to whether there is a circa 1.5 million years age gap between the formation of the first cais and the formation of the first chondrules ( see villneuve et al . 2009 ; amelin et al . 2011 ; larsen et al 2011 ; connelly et al . there is , however , now consensus that the undifferentiated asteroidal parent bodies of chondrites themselves accreted ca . 2 - 4 million years after the beginning of the solar system ( e.g. bizzarro et al . 2005 ; kleine et al . because of their younger accretion ages , chondrites escape internal heating triggered by @xmath0al ( trieloff et al . 2003 , henke et al . 2013 ) , and thus they preserved their pristine structure . recent high precision dating using hf - w and al - mg chronometry has shown that some differentiated meteorite groups do in fact predate formation of chondrites . beside cais , some groups of magmatic iron meteorites that represent the cores of differentiated asteroids have been shown to constitute the oldest solar system objects ( kleine et al . if corrected for cosmic ray exposure , magmatic iron meteorites can be shown to have differentiated within the first 2 million years of solar system history , and their asteroidal parent bodies must have accreted earlier than 0.5 million years after solar system formation ( kruijer et al . 2013 , 2014 ; wittig et al . some older groups of achondrites that sample the outer silicate portion of differentiated asteroids differentiated only slightly later than iron meteorites . the particularly pristine group of angrites ( achondrites consisting mostly of the mineral augite ) is inferred to have differentiated by around 3 - 5 million years after the cais ( larsen et al . 2011 , brennecka & wadhwa 2012 ; kleine et al . 2012 ) . likewise , achondrite groups such as acapulcuites , eucrites , and winonaites as well as non - magmatic iron meteorites differentiated in the time interval between 3 - 5 million years after solar system formations , with inferred accretion ages between 1 and 2 million years after solar system formation ( schulz et al . 2009 , 2010 ) . in summary , there is now clear evidence from short - lived nuclide and long - lived u - pb chronometry that cais are the oldest solar system objects , followed by magmatic iron meteorites ( differentiation of 1 - 2 million years after solar system formation ) and most achondrite groups and non - magmatic iron meteorites ( differentiation 3 - 5 million years after solar system formation ) . differentiation of these asteroidal bodies was likely driven by internal heat sources , with decay of @xmath0al being the most important one . the parent asteroids of chondrites accreted much later ( 2 - 4 million years after solar system formation ) than those of the differentiated asteroids ( less than 2 million years after solar system formation ) . it is therefore likely that the chondritic parent bodies did not undergo internal differentiation , because the heat supply from short - lived nuclides such as @xmath0al was insufficient , or , alternatively , their parent body size was too small . alternatively , undifferentiated chondrites may sample the outer layers of planetesimals that differentiated in their interiors . the larger rocky planets in the inner solar system formed much later than the small asteroids , at timescales of tens of millions of years . hf - w systematics in terrestrial samples and chondrites constrain the age for the earth s growth to at least 38 million years ( kleine et al . 2002 , yin et al . 2002 ; knig et al . 2011 ) and probably as late as 120 million years after solar system formation ( allegre et al . 2008 ; rudge et al . there are fewer indicators to constrain the age of mars , but estimates reach from 2 to 10 million years after solar system formation ( nimmo & kleine 2007 ; dauphas & pourmand 2011 ) . the uncertainties of these estimates originate from the parameters that are used to model planetary growth and internal differentiation into metal cores and silicate mantles . the small size of mars , for instance , has been explained by radial migration of jupiter , suggesting that mars represents a `` starved '' planet that could not accrete further because of interaction with jupiter ( walsh et al . 2011 ) . for constraining the age of the earth , the moon - forming giant impact event is of great importance , and physical models explaining the formation of the moon have dramatically changed over the past years ( e.g. canup & asphaug 2001 ; cuk & stewart 2012 ; canup 2012 ) . likewise , the hf - w and u - pb chronometry of earth s formation has undergone significant reinterpretation . most earlier studies agreed that both the hf - w and u - pb chronometers date segregation of the earth s core during the final stages of earth s accretion ( e.g. allegre et al . 2008 ; rudge et al . 2010 , wood & halliday 2005 ) . a combination of experimental and geochemical studies now rather argue that volatile elements such as pb were delivered late to the growing earth , implying that the u - pb systematics of the earth s silicate mantle bear no chronological significance ( albarede 2009 ; ballhaus et al . 2013 ; marty et al . 2012 ; schnbchler et al . 2010 ) . the geochemical and isotopic inventory of meteorites provides a unique chronological record of early solar system evolution . it took only less than one million years after condensation of the first solid matter to form the first generation of planetesimals . these planetesimals underwent internal heating between 1 and 4 million years after solar system formation , largely caused by @xmath0al decay , triggering their differentiation into metal cores and silicate mantles . the parent asteroids of undifferentiated chondritic meteorites formed 1 - 2 million years later , and therefore these bodies did not differentiate . formation of the larger terrestrial planets in the inner solar system occurred via collision of smaller asteroidal bodies . the timescales involved in planetary growth are in the order of tens of million years , with mars being a possible exception due to gravitational interaction with nearby jupiter . there is growing evidence , that the volatile element inventory of the inner solar system was added late , i.e. during the last stages of planetary growth . chondrites have not endured any geological activity since they formed during the first million years of our solar system . as such , they record physical processes in the solar protoplanetary disc ( spd ) but also of the molecular cloud stage , which preceded the formation of the proto - sun @xcite . chondrites are made of cais , chondrules and iron - nickel metal . all these components are cemented by a fine - grained matrix ( fig . 1 ) . a key property of chondrites is the past presence within their components of short - lived radionculides ( slrs ) . slrs are radioactive elements , which have half - lives , @xmath2 , smaller than 200 myr and were present in the early solar system @xcite . though they have now decayed , their past presence in chondrites and other meteorites is demonstrated by excess of their daughter isotopes . since 1962 and the discovery of @xmath3i ( @xmath2 = 15.7 myr ) , a long list of slrs has been identified in meteorites ( see the table of @xcite ) . slrs are important because they can potentially help to build an early solar system chronology and are the most likely heat source of the building blocks of planets , planetesimals @xcite . in addition , they can help decipher the astrophysical context of our solar system formation . in other words , assuming the sun was born as most stars in a ( giant ) molecular cloud , elucidating slrs origin is the only tool to trace back what was the relationship of the sun with other stars born in the same molecular cloud . the presence in the early solar system of slrs having the longest half - lives is not surprising . similarly to stable isotopes they have been produced in the interior of stars , which have preceded our sun in the galactic history and were delivered to the interstellar medium ( ism ) by supernova ( sn ) explosions and massive star winds . their abundance is roughly in line with galactic evolution models @xcite . the presence of slrs with shorter half - lives ( @xmath2 @xmath4 10 myr ) is more difficult to understand because their half - life is comparable or smaller than the typical timescales of star formation processes @xcite . obviously , constraining the origin of slrs depends also on their initial abundance , which is not always clearly known . some slrs have been synthesised via the irradiation of the spd gas / dust by proto - solar cosmic - rays @xcite . as we saw in section 2 this is especially the case of @xmath5be ( @xmath2 = 1.4 myr ) whose abundance in cais is variable @xcite . independent evidence for early solar system irradiation is given by the variable and intense x - ray emission of protostars @xcite . other slrs such as @xmath6cl ( @xmath2 = 0.3 myr ) or @xmath7ca ( @xmath2 = 0.1 myr ) could also have been produced by irradiation according to some models @xcite . we note that the situation for these slrs is complicated by the fact that their initial abundance in the solar system is not well constrained @xcite . aluminium-26 ( @xmath2 = 0.72 myr ) and iron-60 ( @xmath2 = 2.6 myr ) are probably the slrs , which have attracted the most attention during the last decade . iron-60 is too rich in neutrons to be produced by irradiation processes @xcite . as @xmath8fe is absent from massive star winds , its only possible source in the early solar system is - at least one - supernova , which is known to produce large abundances of that slr @xcite . because the initial abundance of @xmath8fe has long been considered to be elevated ( @xmath8fe/@xmath9fe @xmath10 10@xmath11 @xcite ) , the presence of @xmath8fe has been interpreted as evidence for a nearby supernova @xcite . according to supernova nucleosynthetic models , the sn had to be located within a few pc from the nascent solar system @xcite . in n - body simulations of the birth environment of the sun it is often assumed that the progenitor of this supernova was a star with a mass of @xmath12 25 ( see section 6 ) . however , from the cosmochemical side there are some difficulties with this scenario . it is very unlikely to find a protoplanetary disc or a dense core that close to a sn . before they explode as sne , massive stars carve large ionized regions in the ism ( called hii regions ) where the gas density is too low and temperature too high for star formation to take place . observations show that even around a massive star that still needs to evolve for 2 myr before it becomes a sn , discs and cores are found only several parsecs away @xcite , too far to incorporate @xmath8fe at the solar abundance . in addition , because sne ejecta are vastly enriched in @xmath8fe relative to @xmath0al and their respective solar abundances ( woosley & weaver 2007 ) , all models relying on sn injection lead to a @xmath0al/@xmath8fe ratio far lower than the initial solar ratio , unless specific and ad hoc conditions are adopted @xcite . finally , the initial abundance of @xmath8fe has been recently revised downwards by almost two orders of magnitude @xcite . the revised @xmath8fe abundance is compatible with a * somewhat * enhanced galactic background due to stars born and died in the same molecular cloud as the sun but in a prior generation ( gounelle et al . , 2009 ) . al in a dense shell created by a massive star wind . phases 1 and 2 correspond to the collection of interstellar gas and injection of @xmath0al by the wind ( arrows ) . phase 3 corresponds to the gravitational collapse of the shell and the formation of a new , @xmath0al - rich star and other stars including the sun ( yellow ) . the whole process lasts a few myr ( see text).,scaledwidth=38.0% ] because supernovae vastly overproduce @xmath8fe relative to @xmath0al , @xmath0al can not originate from a single supernova as first proposed by @xcite . it also implies @xmath0al can not have a background origin because at the molecular cloud scale , slrs are mainly due to supernovae . in addition , the @xmath0al distribution in the early solar system seems to have been heterogeneous @xcite , which is incompatible with a global ( molecular cloud ) scale origin @xcite . the most likely origin for @xmath0al is therefore local , i.e. the wind of a single massive star @xcite . @xcite proposed that a single runaway wolf - rayet ( wr ) star injected @xmath0al in a shock - induced dense shell . given the required velocity of the star ( 20 pc / myr ) and gas ambient density ( n @xmath10 100 @xmath13 ) it is however likely the star will have escaped the molecular cloud well before the star collects enough gas and enters the wr phase during which @xmath0al is injected . this model suffers from the brevity and therefore the rarity of the wr phase . have proposed that @xmath0al originated in a wind - collected shell around a massive star ( figure 2 ) . in this model , the dense shell collected by the massive star wind is continuously enriched by @xmath0al brought to the surface of the star by convection promoted by rotation @xcite . after several myr , the collected shell reaches densities large enough to make it gravitationally unstable and a new star generation forms in the shell . in this model , the sun is a second generation star and the massive star having provided @xmath0al can be seen as its parent star . this mechanism is generic in the sense that this mode of star formation , though not the rule , is relatively common @xcite . this model constraints the astrophysical context of our sun s formation as it requires that the parent cluster to which belongs the parent star ( baptised _ coatlicue _ ) contained roughly 1200 stars . in conclusion , though the origin of slrs is still a complex and debated issue ( see davis et al . 2014 for a more detailed review ) it seems that three star formation spatial scales are relevant . slrs with the longest half - lives originate at the scale of star complexes , while @xmath8fe ( @xmath2 = 2.6 myr ) originates at the scale of the molecular cloud and @xmath0al ( @xmath2= 0.72 myr ) at the local massive star wind scale . the distribution of slrs in the solar system is the record of hierarchical ( in termes of spatial and temporal scales ) star formation in the ism . another source of potential information about the formation of the solar system are young stars that are currently forming planets from the disc material surrounding them @xcite . these discs form as a consequence of angular momentum conservation during the star formation process . in the early stages of the star formation process the disc mediates accretion from the parental core to the forming protostar , while at later stages it provides the natural location and material for the planet formation process . the observational study of the properties and evolution of protoplanetary discs is thus a direct probe of the initial conditions for planet formation . the presence of protoplanetary discs around young stars was originally inferred from the emission in excess to the stellar photosphere at infrared- and submillimetre - wavelengths ( eg . * ? ? ? * ; * ? ? ? it was very rapidly understood that the excess emission was likely due to a disc - like distribution of dust located around and heated by the young star at the centre , this interpretation was also consistent with the dynamics of the molecular gas , which could be interpreted as orbiting the star in keplerian fashion . the inferred sizes ( up to few times 100 au in radius ) and masses ( up to 10 - 20% of the central star mass ) were also consistent with the expected values for a pre - solar nebula . these ideas were then spectacularly confirmed by the hst silhouette optical absorption images a few years later @xcite . following these initial observations , protoplanetary discs around young solar analogues have been extensively studied in nearby star forming regions . the recent review by summarizes most of our current knowledge of the general properties of protoplanetary disc populations in nearby star forming regions . the current best estimate of the lifetime of the disc , or the timescale for planetesimal and gas giant formation , is derived from the variation of the fraction of stars , which show infrared excess as a function of the age of the star forming regions ( e.g. * ? ? ? * ) , which is also found to be consistent with the variation of the fraction of young stars that show signs of interactions with the inner disc . the fraction of young stars with inner disc is found to drop exponentially with an e - folding time of @xmath103 myr . this estimate is uncertain because of a number of possible systematic and environmental effects , e.g. the uncertainties in the age determination of young stars @xcite , or environmental disc dispersal mechanisms in clusters , which may not be applicable to the progenitors of the bulk of the field stellar population @xcite . nevertheless , it is obvious that if the majority of stars host planetary systems , the formation process has to occur within a few myr , at least for planet types , which contain significant gaseous envelopes ( gas giants , ice giants and low - density super - earths ) . in large star forming complexes in the magellanic clouds and our own galaxy , there is controversial evidence for stars with circumstellar discs at relatively old ages ( up to and even beyond 30 myr * ? ? ? * ; * ? ? ? while in nearby regions the few existing candidates have not been confirmed so far , it is possible that a small but non - negligible fraction of the stellar population born in large complexes reaccrete a disc and possibly have multiple chances of forming planetary systems . young protoplanetary discs are gas rich and most of the raw material available for planet formation is in the form of molecular gas , just like in clouds and cores . most of this mass is in the observationally elusive form of h@xmath14 and can only be traced , indirectly , through the emission of less abundant species . the molecular gas content , chemistry and emission from protoplanetary discs has been recently reviewed by @xcite , and we refer to that work for a detailed account . the key feature for constraining the conditions for planet formation is that around young solar analogues most of the disc midplane is at low temperature and the most abundant molecular species ( with the obvious exception of h@xmath14 ) are frozen inside the ice mantles of grains . in these conditions , the chemistry is likely to be dominated by reactions on ices and may lead to significant production of complex organic and pre - biotic molecules , as observed in laboratory experiments . while complex organic compounds have been revealed in solar system objects , the direct detection and abundance measurements in protoplanetary discs is challenging . nevertheless , in the cold disc midplane cosmic - ray induced desorption of ices may release a small fraction of the complex molecules in the gas phase , a process similar to the one that is thought to occur in cold protostellar cores @xcite . alternative promising locations to search for these complex compounds are the snowlines of the major molecular species , which could carry in the gas phase complex organic molecules as impurities as part of the sublimation process . the major snowlines that may produce such an effect are the co snowline at @xmath1020k and the h@xmath14o one at @xmath10150k . the co snowline is now within grasp of the alma observatory in nearby star forming regions , and much work dedicated to the characterisation of this important transition in protoplanetary discs is expected in the coming years . future instruments may help quantifying gas to dust ratio by combining high spatial and spectral resolution . as a consequence of the complexity of the molecular gas chemistry in discs and the difficulty in observing h@xmath14 , the most reliable mass estimates for protoplanetary discs still come from the measurements of the thermal dust emission . most of the disc is optically thin at submm and longer wavelengths . therefore , the thermal dust emission is directly proportional to the mass , which can be readily estimated from the observed fluxes assuming a dust opacity coefficient and a temperature structure for the disc . the latter is normally computed using a self consistent thermal balance model , for typical accretion rates measured in young pre - main sequence objects the heating from the central star dominates the disc heating and the thermal structure of the dust is usually well constrained by models ( see e.g. * ? ? ? * ; * ? ? ? * ) the disc masses derived with this method are typically between 0.2% and 20% of the mass of the central star , implying that a large fraction of young solar analogues are hosting a disc with a mass well above the minimum mass solar nebula and in principle capable of forming a planetary system similar to our own . high angular resolution observations allow to resolve the disc emission and estimate the distribution of matter in the disc . the surface density of solids , traced by the thermal dust emission , are typically found to fall off as @xmath15 with radial distance @xmath16 until a critical radius and then drop sharply , consistently with an exponential fall off . the observed surface densities of the most massive discs are broadly consistent with the average surface density of our own solar system ( andrews et al . 2009 ) . however , most systems that have a total mass above the minimum mass solar nebula threshold are larger than our own solar system , resulting in a lower surface density . these estimates are not without uncertainties : firstly , the solids are assumed to trace approximately 1% of the total mass ( following the estimates of the dust to gas ratio in the interstellar medium ) , however this number is highly uncertain in discs and also expected to evolve with time and location , as gas and dust evolve following different pathways ; secondly , the dust opacity coefficient depend on the composition , size , shape , and porosity of the dust grains @xcite . measurements of proto planetary disc masses are relatively rare as of today : up - coming alma observations may be able to provide a statistical view of the distribution of dust surface densities and possibly also gas surface densities . as the dust emission is optically thin , multi - wavelength observations at submillimetre wavelength allow us to probe the dust opacity coefficient , which in turn can be a powerful probe of the grain population properties . in a protoplanetary disc , grains are expected to grow and settle onto the midplane ( for a recent review , see * ? ? ? as grains grow to sizes of the same order or larger than observing wavelengths , the dust opacity coefficient dependence with wavelength changes from the typical value of the sub - micron size interstellar grains ( @xmath17 ) to a much shallower dependence , approaching @xmath18 in the limit of a population of grains all much larger than the observing wavelength @xcite . this argument has been used to constrain the level of grain growth in protoplanetary discs , finding that in most discs around young stars the dust has grown to pebble - size aggregates . the observational results can be understood in the framework of global dust evolution models in discs , which describe the evolution of dust in the disc environment based on constraints from laboratory measurements of the grain - grain collision outcomes . more specifically , the emerging observational constraints on the dust grain size distribution as a function of disc radius matches the expectations from the evolutionary models . this apparent success hides a number of difficulties that still need to be solved , in particular , smooth disc models still predict a radial drift and fragmentation of the large grains that is inconsistent with the observations and require to artificially slow down the radial drift . the planet formation process then requires that additional growth barriers are overcome , a variety of possibility exist , from pressure traps generated by disc instabilities , localised growth across snowlines , or the efficient accretion of pebbles onto existing planetary cores @xcite . some of these models predict observational signatures that are within reach of the current generation of observing facilities . indeed the effect of pressure traps induced by planet - disc interaction has been observed in some systems with alma ( e.g. * ? ? ? an intriguing development in recent years has been the observation by several groups of the possibility of significant dust evolution before the disc stage in prestellar cores and protostars . if confirmed , these findings would imply a significant revision of our assumptions on the initial conditions for planet formation in protoplanetary discs . planets from the gas and dust present in protoplanetary discs . in the first stage of planet formation dust and ice particles collide and stick together to form larger aggregates @xcite . the aggregates maintain a high porosity during the growth , since collision speeds are initially too low to cause restructuring . particles obtain increasingly higher collision speeds as they grow in size and their frictional coupling with the gas diminishes ( voelk et al . , 1980 ; ormel & cuzzi , 2007 ) , so that particle pairs decouple from the smallest eddies and thus will have different velocity vectors even when at the same location . an additional contribution to the collision speed is the relative radial and azimuthal drift between different - sized particles @xcite . silicate dust aggregates in the inner protoplanetary disc are compactified by collisions when they reach approximately mm sizes @xcite . the combination of high collision speeds and low porosity halts growth by direct sticking , since compact particles can not dissipate enough energy during a collision to allow sticking @xcite . this is referred to as the _ bouncing barrier_. ice monomers are generally stickier than silicate monomers and this increases their resistance against compactification @xcite . the size of the constituent monomers also plays an important role in determining the collision outcome . ice aggregates , which consist of very small monomers of @xmath19 @xmath20 m sizes , can stick at up to 50 m / s collision speeds @xcite . it is nevertheless not known whether such small ice monomers are prevalent in protoplanetary discs or whether sublimation and condensation cycles drive ice particles to much larger sizes . the continued growth from mm - sizes faces the formidable _ radial drift barrier_. the gas in the protoplanetary disc is slightly pressure supported in the radial direction , since viscous heating and stellar irradiation heat the inner regions more strongly @xcite . this outwards - directed push on the gas causes the gas to orbit slower than the keplerian speed , by approximately 50 m / s @xcite . solid particles do not sense the pressure difference from front to back ( since their material density is much higher than the gas density ) , so particles would orbit at the keplerian speed in absence of gas drag . the drag from the slower moving gas drains particles of their angular momentum and causes them to spiral inwards towards the star . in the asteroid belt the radial drift peaks for m - sized particles , which fall towards the star in a few hundred years , to be destroyed at the silicate sublimation line close to the star . the peak of the radial drift occurs for cm - sized particles in the outer protoplanetary disc where the giant planets form . there are three main ways by which particles can cross the radial drift barrier : _ mass transfer_. at collision speeds from 1 to 25 m / s particles can grow by mass transfer . the projectile is destroyed in the collision but leaves up to 50% of its mass attached to the target @xcite . particles stuck at the bouncing barrier do not collide at such high speeds . however , artificially injected cm - sized seeds can grow from the sea of bouncing barrier particles through mass transfer @xcite . the growth rate is nevertheless too low to compete with the radial drift , so formation of planetesimals by direct sticking requires a reduction in the radial drift speed of the particles , e.g. through the presence of a long - lived pressure bump . such a pressure bump may arise at the inner edge of the dead zone @xcite or around the water ice line where the jump in dust density has been proposed to cause a jump in the turbulent viscosity @xcite , although this requires a very steep viscosity transition at the ice line @xcite . _ fluffy particles_. ice aggregates consisting of small monomers acquire very low densities during their growth , down to @xmath21 @xcite . particle growth is only able to outcompete the radial drift if the particle grows larger than the mean free path of the gas molecules and enters the stokes drag force regime . here the friction time is proportional to the particle radius squared ; hence decoupling is much more rapid than in the epstein regime valid for small particles @xcite . very fluffy particles are large enough to be in the stokes regime and can cross the radial drift barrier within 10 au by growing in size while drifting radially @xcite . hence this is a way to cross the radial drift barrier , if particles can remain fluffy and avoid compactification and erosion by the gaseous headwind . _ particle concentration_. particles can become strongly concentrated in the turbulent gas , triggering the formation of planetesimals by a gravitational collapse of the overdense regions . particles concentrate either _ passively _ or _ actively_. for passive concentration the general mechanism is that particles pile up where the gas pressure is high , since any rotating structure in the turbulent gas flow must be in semi - equilibrium between the dominant forces , namely the pressure gradient force and the centrifugal / coriolis force ( barge & sommeria 1995 ) . particles do not sense gas pressure and must therefore move towards the direction of increasing gas pressure . on the smallest scales of the turbulent flow , approximately km under prevalent conditions in the asteroid belt , mm - sized particles pile up between rapidly over - turning eddies @xcite . the largest scales of the protoplanetary disc are dominated by the coriolis force , and the turbulence organises into axisymmetric pressure bumps , surrounded by zonal flows @xcite , or elongated vortices @xcite . in the streaming instability scenario the particles actively drive concentration , by piling up in filaments , which locally accelerate the gas towards the keplerian speed and hence do not drift radially @xcite . strong particle concentration by the streaming instability is triggered at a metallicity slightly higher than solar @xcite . the gravitational collapse of overdense regions leads to the formation of planetesimals of sizes from 100 to 1000 km @xcite . the solar system contains remnant planetesimals from the epoch of planet formation , in the asteroid belt and in the kuiper belt . these populations of planetesimals can be studied to infer constraints on the actual processes that led to planetesimal formation in the solar system . the relative lack of asteroids smaller than 100 km in diameter , compared to a direct extrapolation from the largest sizes , does not agree with the formation of asteroids by hierarchical coagulation in a population of km - sized planetesimals ( * ? ? ? * but see @xcite for an alternative view ) . asteroids may instead have had birth sizes larger than 100 km , in agreement with gravitational collapse models , while the smaller asteroids that are found in the asteroid belt today are mainly collisional fragments @xcite . large birth sizes have also proposed to explain a similar lack of small kuiper belt objects @xcite . planetesimals are the building blocks of both terrestrial planets and the cores of the giant planets . in the _ core accretion _ scenario giant planets form as gas from the protoplanetary disc collapses onto a core that has grown to approximately 10 earth masses by accumulation of planetesimals @xcite . in this picture , the ice giants uranus and neptune only manage to attract a few earth masses of gas before the gaseous protoplanetary disc dissipates after a few million years . core accretion by planetesimals is nevertheless very slow because the number density of planetesimals is low in the region where the giant planets form . an enhancement of 4 - 6 over the minimum mass solar nebula is needed to form jupiter and saturn within 10 million years in the classical core accretion scenario @xcite . for the considered protoplanet mass , and feed a particle accretion disc orbiting the protoplanet . figure adapted from @xcite.,width=302 ] core accretion timescales can be decreased when accreting planetesimal _ fragments_. small fragments have their scale heights damped by gas drag and hence the growing core can accrete a much larger fraction of the solid material @xcite . however , global disc simulations show that a large fraction of the slowly drifting fragments are trapped in resonances with the cores @xcite . the formation of a system of giant planets is further complicated by the excitation of planetesimal eccentricities and inclinations when scattered by the growing cores . this leads to a slow oligarchic growth phase with growth rates that are too low to form cores within the life - time of the protoplanetary disc @xcite . pebbles left over from the planetesimal formation process can be accreted very efficiently by the growing cores @xcite . the hill sphere of the core denotes the radial distance over which an incoming particle on a faster ( interior ) or slower ( exterior ) orbit is scattered gravitationally by the core . the core only extends a small fraction of its hill radius ( 0.1% at 5 au , 0.01% at 50 au ) . gravitational focussing of the incoming planetesimals makes the accretion radius much larger than the physical radius , but planetesimals are still only accreted from about 3% of the hill radius at 5 au . pebbles of mm - cm sizes experience strong drag during the gravitational scattering and lose enough energy to be gravitationally bound to the core , for any impact parameter up to approximately the hill radius ( see figure [ f : figure6 ] ) . this leads to very high accretion rates , about 1,000 times faster than classical core accretion at 5 au and 10,000 times faster at 50 au . such high growth rates are needed to explain the formation of gas giants observed in wide orbits around some young stars ( e.g.hr8799 , * ? ? ? * ) assuming they are not formed from other means ( like gravitational instability ) . as detailed in section 4.4 , protoplanetary discs observed at mm- and cm - wavelengths show signs of large populations of mm - cm - sized pebbles , which can boost planet formation . many of the young stars surrounded by discs are part of a cluster of stars , indeed it seems that most stars are born in a clustered environment ( lada & lada 2003 ) . in the next section we will discuss the possible birth environment of the sun . in particular we consider whether we can constrain the properties of the birth cluster , indeed its possible identity , and whether we can identify solar siblings , i.e. the stars that were born from the same molecular cloud as the sun . analysis of the isotopic abundances of meteorites reveal that they contain the decay products of the radioactive isotopes @xmath0al and @xmath8fe , which have half lives of 0.7 myr and 2.6 myr @xcite . as mentioned in section 3 the most likely source of these isotopes are supernovae explosions ( e.g. * ? ? ? how many is still a matter of debate . given the observed distribution of stellar masses of newly - formed stars follows a power law favouring less - mass stars ( @xmath22 - * ? ? ? * ) one would need a fairly large group of stars reasonably close to the forming sun to explain the observed abundancies . there are other indications that the sun once was part of a fairly dense and therefore large stellar cluster . first , the fairly abrupt cut - off in the mass distribution at 30 - 50au , which is usually not observed in protoplanetary discs , but could be caused by a fly - by during the protoplanetary disc phase . second , the trans - neptunian objects sedna and 2012 vp@xmath23 ( morbidelli & levison 2004 , kenyon & bromley 2004 , brasser et al . 2006 , trujillo & sheppard 2014 ) with their high excentricities are another property that could be explained by an encounter during the planet forming phase . calculations of encounter probabilities in clusters suggest that the sun was formed in a stellar cluster containing 2000 to 10000 stars ( portegies zwart 2009 ) . further evidence for the sun having been in a cluster , comes from the measured tilt of the sun s rotation , one suggestion being that the protoplanetary disc was titled by a star - disc interaction ( heller 1993 ; thies et al . 2005 ) . stellar clusters are potentially dangerous environments for planetary systems . close encounters between stars are relatively frequent in such crowded places . the timescale for a particular star to have another star pass within some distance @xmath24 can be approximated by @xcite @xmath25 one can distinguish between different phases during which such an encounter could happen : ( i ) the early phase when the star is surrounded by an accretion disc ( first few myr ) , ( ii ) the planet growth phase , and ( up to 100 myr ) , and ( iii ) after the solar system was fully formed ( up to now ) . close encounters occurring whilst stars still possess protoplanetary discs could lead to their truncation . in addition , illumination by massive stars may lead to the early evaporation of protoplanetary discs around young stars . thus , clusters may inhibit the planet formation process . planetary systems , once formed , are also affected by other stars within clusters . fly - by encounters with other stars may perturb otherwise stable planetary systems ( e.g. * ? ? ? * ; * ? ? ? * ) changes in planetary orbits within a system can lead to the growth in eccentricities until orbits cross . scattering can then lead to the ejection of planets leaving those remaining on more bound and eccentric orbits . for lower - mass planets on tighter orbits , orbit crossing tends to lead to collisions between planets ( e.g. * ? ? ? alternatively , stars hosting planetary systems may exchange into ( wide ) binaries . the stellar companion may then perturb planets orbits via the kozai - lidov mechanism , where planets orbits periodically pass through phases of high eccentricity . planetary orbits may then cross leading to scattering or collisions ( e.g. * ? ? ? * ; * ? ? ? * ) . in the early phases the probability of encounters are highest , as afterwards clusters expand and therefore the encounter likelihood decreases . we have seen above how the measured isotopic abundances in meteorites are well explained by the enrichment of the protosolar - system material from a nearby supernova formed from a relatively massive star or from a combination of supernovae and a massive @xmath26-producing star . the rarity of such objects in turn led us to argue that the solar system likely formed within a stellar cluster containing at least 2000 stars . as just discussed , stellar clusters are hazardous environments . the rate of destructive close stellar encounters tends to increase with increasing cluster mass ( see eq.[mbd_spz_equation1 ] ) . therefore , there is an optimum stellar cluster containing around a few thousand stars where pollution from supernovae is at least possible whilst at the same time a reasonable fraction of planetary systems may survive unperturbed @xcite . one can perform n - body simulations of stellar clusters containing one 25 m@xmath27 star to determine the number of solar - like stars , which are 0.1 - 0.3 pc from the 25 m@xmath27 star when it explodes as a supernova . one can then follow the subsequent trajectories of these polluted stars within the cluster to determine the fraction of them , which avoid both perturbing fly - by encounters and exchange encounters into binaries . such numerical studies of clusters containing 2100 stars ( including one 25 m@xmath27 star ) reveal that some 25 percent of clusters contain enriched , unperturbed solar - like stars ( i.e. g - dwarfs ) , and usually only one or two per cluster out of a total of 96 g - dwarfs @xcite . therefore , roughly one percent of g - dwarfs from such clusters are enriched whilst being unperturbed . if the sun was indeed born in an open cluster , it is rather natural to ask whether this cluster is still around today , and if so , whether we can identify it . there have been a number of observations of star clusters of different ages dedicated to find planets in them ( meibom et al . 2013 , and references therein ) . however , planet detection in clusters provides some added difficulties , for example , even nearby clusters are often more distant than the field stars one finds usually planets around ) , thus the frequency of planets in cluster environments compared to that in the field is still an open question . however , probably most clusters in the solar neighbourhood loose a considerable fraction ( @xmath28 70% ) of their stars ( pfalzner & kaczmarek 2013 ) . measuring detailed abundance patterns in stellar atmospheres through high - resolution spectroscopy offers us the hope of identifying other stars , which originate from the same gas cloud as the sun . even if the cluster has long since dissolved , stars from it will still share similar orbits whose properties could be measured astrometrically ( e.g. * ? ? ? the open cluster m 67 is one candidate which we will consider here , though there must have been a number of cluster produced in the milky way having similar ages and galactocentric radii . m 67 has an age . of about 3.54.8 gyr and the stars within it also have a composition similar to the sun . have identified one star in the cluster , which is a better match to our sun than most solar - like stars in the solar neighbourhood . the analysis of 13 additional stars in m 67 confirm that the abundances of the sun and m 67 are similar . is the cylindrical radius from the galactic centre and @xmath29 is the distance from the galactic plane ( figure 7 of * ? ? ? * reproduced with permission of the aas).,width=453 ] however , as pointed out by @xcite the current orbits of m 67 and the sun are very different , as shown in fig . [ mbd_spz_figure1 ] , where we plot their current orbits in cylindrical coordinates ( figure 7 from * ? ? ? the radial distance from the galactic centre , @xmath30 , for the sun and m67 are similar . earlier mention of abundance enhancements in the sun compared to its distance from the galactic centre would indicate that the sun experienced a radial migration of 0.84.1kpc . in that case m 67 should also have migrated over the same distance . possibly , the migration happened while the sun was still a member of its birth cluster . extensive numerical analysis of the orbit of the sun in the galactic potential however indicates that this radial migration in the current galactic potential would be negligible @xcite . one can see from fig . [ mbd_spz_figure1 ] that the orbit of m 67 takes the cluster significantly further out of the galactic plane to much larger values of @xmath29 than is experienced by the sun . if one assumes that m 67 has always had its current orbit and that the sun was ejected from m 67 , this would imply that the sun received a significant kick in order to leave it on its current orbit , residing in the plane of the galaxy @xcite . the point here is that it is not possible to impart such a large kick whilst keeping the solar system intact . a possible solution to this problem has been suggested in @xcite . the key idea is that the orbit of m 67 has changed over time . the orbit could originally have been more in the galactic plane ( i.e. with lower maximum values of @xmath29 ) allowing the sun to escape from it with a relatively low velocity . scattering off giant molecular clouds ( gmcs ) could then leave the cluster on the orbit we see today . simulations of scattering due to gmcs show that it is possible to put m 67 on its current orbit beginning with an orbit similar to the sun s @xcite . in that case the sun must have left m 67 before the scattering event with a gmc took place . beyond neptune lies a reservoir of remnant icy planetesimals from the formation of the solar system analogous to the main asteroid belt between mars and jupiter . known as the kuiper belt , this distant population displays an intricate orbital structure that has inspired and constrained a number of ideas on the early evolution of our planetary system . the high abundance of kuiper belt objects ( kbos ) in mean - motion resonances with planet neptune led to the idea that the ice giant has migrated through the original disc of planetesimals , sweeping a significant fraction into stable resonant orbits @xcite . the superposition of dynamically hot ( high inclination and eccentricity ) and cold ( low inclination and nearly circular ) orbits suggests that while some kbos have been scattered onto their current orbits @xcite , likely by the outer planets , others have formed in - situ and remained unperturbed @xcite . an alternative explanantion is provided by an encountering star that scattered the kuiper belt ( ida et al . the current morphology of the scattered kuiper belt could only be reproduced if the mass of the encountering star @xmath31 had an impact parameter of @xmath32au ( punzo et al . however , such an encounter would like scatter the entire kuiperbelt . in addition , cooling off part of the hot population to re - populate the cold population would require some 1000 pluto - mass objects , which are not observed . the need to explain the dynamical structure of the kuiper belt and to reconcile the orbits of the giant planets with the discovery of hot - jupiters were main drivers in the evolution of planet migration theories . in the solar system , the current paradigm considers that the giant planets emerged from the protoplanetary disc in a more compact , near - resonant configuration , all within 15 au from the sun @xcite . angular momentum exchanges with the surrounding disc of planetesimals led to the outward migration of neptune , uranus , and saturn , as they scattered small bodies inwards , while jupiter , sufficiently massive to eject planetesimals from the solar system , migrated predominantly towards the sun @xcite . the diverging migrations of jupiter and saturn eventually resulted in a mutual resonance crossing and triggered a dynamical instability throughout the solar system , which impulsively expanded the orbits of uranus and neptune to near their current locations and violently scattered the planetesimal disc @xcite . the late heavy bombardment was also likely caused by this instability event @xcite . geochronological studies of the lunar rocks returned by the apollo missions yielded the surprising result that all large multi - ring basins on the earth s moon they are also called mare : orientale , crisium , imbrium , nectaris , humorum , serenitatis originated by giant impacts in an extremely confined time interval between 4.0 and 3.7 gyr ago . this process is called the lunar ( or late ) heavy bombardment ( lhb ) , as it was completed only 800 myr after solar system formation . it requires that long after final formation of the terrestrial planets , there was an increased flux of small bodies ( up to about 100 km in size ) in the inner solar system 3.8 gyr ago . possible evidence of the lhb is also inferred from he d meteorites , which are inferred to come from vesta , and martian meteorites like alha84001 @xcite , although this is not undisputed . two general scenarios appear plausible for the early solar system impact history : \i ) the lhb was the final phase of terrestrial planet accretion this requires that the small body flux must have been considerably higher before ( implying somewhat unrealistic total mass estimations of initial small body populations ) , and that the record of earlier impacts must have been erased by the late impact phase 3.8 gyr ago . \ii ) the lhb was an episodic spike of the influx of asteroidal or cometary bodies , caused by dynamical excitation of small body populations in the asteroid or kuiper belts . as general explanation for the dynamical excitation , planetary migration processes are promising candidates , in particular the aforementioned resonance crossing of jupiter and saturn @xcite , i.e. in more general terms , the giant planet instability ( morbidelli et al . , 2007 , levison et al . , 2011 ) . advanced models consider a more prolonged bombardment history induced by planet migration processes ( morbidelli et al . 2012 ) . however , there is increasing evidence that before the time of the lhb 3.8 gyr ago , there were other episodic times of increased impact rates on the earth - moon system , registered by , e.g. ar - ar and zircon chronology , and also by other meteorite parent bodies , e.g. 4.2 gyr impact events affecting the ll chondrite parent body @xcite . it was suggested @xcite to consider close stellar encounters as a possible reason for repeated episodic dynamical excitation . however , there is increasing evidence that before the time of the lhb 3.8 gyr ago , there were other episodic times of increased impact rates on the earth - moon system , with concomitant impact episodes on other inner solar system bodies . for example , both ar - ar and zircon chronology ( fernandes et al . 2013 , nemchin et al . 2008 ) evidence an increased impact cratering rate 4.2 ga ago , which was also experienced by the ll chondrite parent body ( trieloff et al . 1989 , 1994 , dixon et al . although the chronological data base has still to be improved , age clustering can only hardly be reconciled with stochastical bias effects , as multiple celestial bodies were affected . as giant planet instabilities can only once trigger bombardments , it was suggested ( trieloff 2014 ) to consider close stellar encounters as a possible reason for repeated episodic dynamical excitation . however , the corresponding dynamical scenarios are only poorly explored . kbos possess the most diverse surfaces of any small body population in the solar system @xcite . this is puzzling , given the low and slowly varying temperatures in the kuiper belt region . the migration - driven instability event described above offers a plausible solution to this problem by allowing planetesimals formed in the region to be scattered to the current kuiper belt . however , the stochastic character of this process complicates the mapping of current kbo surface properties to their formation location . a recent survey of kbos and centaurs using _ herschel _ has built on an earlier _ spitzer _ survey @xcite to produce the largest sample of albedos and sizes for this type of object ( table [ table.one ] ) . these data reveal that kbos are also extremely diverse in reflectivity , with albedos spanning more than an order of magnitude from about 0.03 to 0.40 . an interesting trend emerges when combining the kbo colour and albedo data @xcite : an albedo - colour plot ( figure [ fig.as ] ) shows that kbos fall into two main groups : a broad bright - red ( br ) cluster , centred around albedo @xmath100.15 and colour spectral slope @xmath1030% and a more compact dark - neutral ( dn ) clump near albedo @xmath100.05 and colour slope @xmath1010% . importantly , kbos in dynamical classes believed to originate beyond neptune occupy only the bright - red group , while kbos that may have been scattered out to the kuiper belt from nearer the sun are found in both the bright - red and the dark - neutral groups . this result hints at a compositional gradient that was present in the planetesimal disc prior to the planetary instability event . @lllc dynamical class & albedo & colour & surface types + scattered disc & @xmath33 & @xmath34 & dn , br + centaurs & @xmath35 & @xmath36 & dn , br + hot classicals & @xmath37 & @xmath38 & dn , br + plutinos & @xmath39 & @xmath40 & dn , br + outer resonants & @xmath41 & @xmath42 & br + cold classicals & @xmath43 & @xmath44 & br + detached kbos & @xmath45 & @xmath46 & br + planetary system development is often considered in the context of _ dynamical stability _ , or the ability of a planet , moon , or asteroid to adhere to its regular motion i.e. approximately retain its original orbit over time . past or future instability can help constrain and explain observations and place our solar system today into context @xcite . in this section , we ask , are planetary systems - and in particular the solar system - stable as their host stars evolve ? the future orbital evolution of the solar system , like all other planetary systems , can not be determined to infinite precision . instead , we obtain likelihoods for particular futures through suites of numerical simulations . qualitatively , the evolutionary behaviour can be split into two parts depending on the state of the sun : ( i ) the main sequence and ( ii ) post - main - sequence phases of solar evolution . the sun will exist in its present state for about a further 6 gyr before turning off the main sequence . during these 6 gyr , the eight planets will technically evolve chaotically @xcite but realistically are very likely to avoid dynamical instability . in fact , through a suite of simulations , @xcite determined that the inner four planets have a 98%-99% chance of surviving and not having planetary orbits cross ; 100% of all his simulations yielded survival of the outer four planets . figure 6 shows how the eccentricity of the earth and mercury evolve over this timescale . earth s eccentricity remains nearly constant . mercury s eccentricity gradually increases , but rarely becomes large enough for its orbit to cross that of venus . after the sun leaves the main sequence , it will simultaneously shed about half of its mass while becoming a red giant star that is nearly 1 au in radius . the increase in radius can more easily allow the sun to tidally capture objects encountered by the solar envelope @xcite , whereas the mass loss will push out and potentially warp the orbits of surviving bodies @xcite . the result on the inner three planets will be dramatic , and on the outer five less so . mercury and venus will be swallowed and earth will suffer an uncertain fate . our home planet will lie on the cusp of being tidally drawn into the sun @xcite . mars s orbit will simply expand . all 4 outer planets will also survive and expand their orbits such that the ratios of their mutual separations will remain unchanged @xcite . beyond neptune and the kuiper belt , the sun s mass loss will trigger instability in the oort cloud , allowing for objects beyond @xmath47 au to escape the solar system @xcite . observational data for extrasolar planets are significantly less accurate and precise than those for the solar system planets @xcite . consequently , the future of exoplanetary systems is not as well constrained . however , every exoplanetary system so far discovered contains fewer planets than the solar system , simplifying the analysis . the majority of known planetary systems contain either two or three planets ( www.exoplanets.org ) , motivating study of these two cases in detail . during the main sequence evolution of the host star two planets will never cross orbits if they are sufficiently far away from each other to be _ hill stable _ @xcite . a hill stable system could potentially become _ lagrange unstable _ if the inner planet collides with the star or the outer planet escapes the system @xcite . empirical and analytical estimates for these stability boundaries , even when scaled down in mass to the test particle limit , show good agreement with numerical simulations of the long - term evolution @xcite . three - planet systems do not appear to admit analytical stability boundaries , and therefore investigations of these systems require a numerical approach . these studies have yielded portraits of instability times as a function of initial separation , as well as empirical estimates for these times @xcite . these instabilities can occur any time during the main sequence evolution , and fundamentally change the resulting dynamical architecture , perhaps leading to the establishment of a population of hot jupiters and eccentric planets @xcite . the study of systems with a higher number of exoplanets produce a similarly - wide variety of outcomes to the three - planet case @xcite . on the y - axis represents the single - planet hill radius . the planet mass and the innermost semi - major axis are fixed at 1 @xmath48 and 10 au , respectively . systems below the black solid line will be unstable on the ms . systems in the tiny strip between the black solid and the black dotted lines will become unstable at some point between the end of the ms and the end of the asymptotic giant branch ( agb ) . systems below the red solid line will be unstable during the wd phase , assuming that the star formed at the big bang ; systems above this line must be stable for the age of the universe . the red dotted and dashed lines show the stability boundaries for wd cooling ages of 1 and 5 gyr . this plot demonstrates that after the star has become a white dwarf , many more systems can become unstable.,width=302 ] we know that planetary systems do survive the giant branch phases of stellar evolution based on abundant evidence of atmospheric metal pollution in white dwarfs ( wds ) @xcite ; wds , which are the size of the earth but with the mass of the sun , represent the end - product of stellar evolution . in over 35 systems , wd atmospheric metal pollution is accompanied by compact accretion discs of gas and/or dust @xcite . although the precise origin of the pollution and discs have yet to be identified , they must arise from currently dynamically active planetary system remnants . during giant branch mass loss , the dynamical stability limits of a multi - body system are changed @xcite . consequently , two - planet systems , which remained stable throughout the entire main sequence may become unstable either during the giant branch phase , or the wd phase @xcite . for three - planet systems , the situation becomes more complex , allowing for multiple instabilities to occur during different phases of stellar evolution @xcite . see figure 7 for a phase portrait of stability limits through all phases of stellar evolution for three - planet systems . planets residing close to the giant star will tidally interact in a complex way depending on the pulsations of the giant @xcite . a likely progenitor of the wd discs are exo - asteroids . although unobservable during main sequence and giant branch phases , their evolution through these epochs will crucially determine their placement in wd systems . the effect of mass loss on a single planet and single exo - asteroid belt has been considered in the context of an exo - kuiper belt at about 30 au @xcite and an exo - asteroid belt at about 5 au @xcite . combining the evidence from these different sources for the early history of the solar system the following picture emerges : the geochemical and isotopic inventory of meteorites provides a unique chronological record of early solar system evolution . it took less than one million years after condensation of the first solid matter to form the first generation of planetesimals . these planetesimals underwent internal heating between 1 and 4 million years after the solar system formation , largely caused by @xmath0al decay , triggering their differentiation into metal core and silicate mantle . the parent asteroids of undifferentiated chondritic meteorites formed 1 - 2 million years later , and therefore these bodies did not differentiate . formation of the larger terrestrial planets in the inner solar system likely occurred via collisions of smaller asteroidal bodies . the timescales involved in planetary growth are of the order tens of millions of years , with mars being a possible exception due to gravitational interaction with nearby jupiter . there is growing evidence that the volatile element inventory of the inner solar system was added later , i.e. during the last stages of planetary growth . the presence of the remnants of short - lived radioactive nuclei in meteorites was for a long time interpreted as evidence for the formation of the sun in a large star cluster . it was usually assumed that the different radioactive nuclei were formed in one single event in a supernova explosion . however , there now exist arguments to support that a sequence of events took place . here the slrs with the longest half - lives originate from star complexes , while @xmath8fe ( @xmath2 = 2.6 myr ) originates from a molecular cloud and @xmath0al ( @xmath2 = 0.72 myr ) from the wind of a local massive star . in this picture the distribution of slrs in the solar system is the result of sequential star formation in the ism . the observation of young disc - surrounded stars also imposes constraints on the timescale of planet formation . the disc fractions in star clusters strongly indicate that most stars dissipate their discs within 2 - 3 myr which would favour rapid planet formation - at least for the gas giants . however , recent works argue in favour of longer disc dissipation times , of the order of 5 - 10 myr . here a comparison with the meteorite record could possibly bring clarification in the future . comparing the typical masses of the discs around young stars it is found a large fraction of young solar analogues are hosting a disc with a mass well above the minimum mass solar nebula and in principle are capable of forming a planetary system similar to our own . for the most massive of these discs the observed surface densities distribution of the disc mass is roughly consistent with the average surface density of our own solar system . however , most systems are larger than our own solar system , resulting in a lower surface density . it is currently unclear whether this difference is real or due to the large uncertainties in the observations . the formation of planets from dust and ice particles likely proceeds in the initial stages by sticking collision , the so - formed porose aggregates eventually compactify . the continued growth from mm - sized particles faces the the formidable bouncing and radial drift barrier . planetesimals form despite these difficult circumstances , likely by a combination of particle sticking and gravitational collapse of regions overdense in pebble - sized particles . the various mechanisms which allow particles to avoid this process have been suggested . future studies of the properties of the remnant planetesimals in the asteroid belt and in the kuiper belt will be crucial to shed light on this open issue . in addition , studying protoplanetary discs will help to understand this phase of planet formation better , as populations of mm - cm sized pebbles have been detected in discs around young stars . it is presently unclear to what degree nearby others stars influence the formation of planetary systems . from the slr abundances and the size of the solar system it can be concluded that the birth cluster of the sun contained at least 2000 stars . measuring detailed abundance patterns several groups have recently started the search for other stars that originate from the same gas cloud as the sun . there are currently some candidate siblings , however further studies are required to confirm those results . the orbital structure of the kuiper belt and the cratering record on the moon are evidence for a dynamically active early solar system . planetary migration and resonance - driven instabilities have acted on the planetesimal disc to enhance impact rates onto the inner solar system , to boost the mixing of populations formed at different distances from the sun , and to sculpt the dynamics of the main planetesimal reservoirs . on the other hand , external forces from close stellar encounters within the suns parental cluster could have been the cause of episodic dynamical excitations of minor body populations and impact rates . the current challenge for understanding the origin of the solar system ( and other planetary systems ) is to seek clues that allow us to see past these dynamical events , to the time when the first planetesimals began to form . what will happen to the solar system in the future ? after the sun leaves the main sequence mercury and venus will be swallowed and earth will suffer an uncertain fate as it will lie on the cusp of being tidally drawn into the sun . mars and the 4 outer planets will also survive and expand their orbits . in summary , in recent years we have seen many new insights into the formation of the solar system . however , there are still a considerable number of open questions where different competing theories exist . so far the different fields contributing in the investigations to the origin of the solar system have generally worked fairly orthogonally . however , it would be timely to start combining these seperate efforts , using them to build a more complete pictures . progress can be expected in determining the actual chronology of events . we want to thank both referees for their very constructive comments . sp is supported by the minvera program of the max - planck society . mbd is supported by the swedish research council ( grant 2011 - 3991 ) . aj was supported by the knut and alice wallenberg foundation , the swedish research council ( grant 2010 - 3710 ) and the european research council ( erc starting grant 278675 ) . lt is partly supported by the italian ministero dellistruzione , universita e ricerca through the grant ialma - progetti premiali 2012 ( cup c52i13000140001 ) . mt acknowledges support by deutsche forschungsgemeinschaft ( dfg ) , and klaus tschira stiftung ggmbh . dv is supported from the european research council under the european union s seventh framework programme ( fp/2007 - 2013 ) / erc grant agreement n. 320964 ( wd tracer ) . , a. & testi , l. 2004 , in astronomical society of the pacific conference series , vol . 323 , star formation in the interstellar medium : in honor of david hollenbach , ed . d. johnstone , f. c. adams , d. n. c. lin , d. a. neufeeld , & e. c. ostriker , 279 , j. , grundy , w. , brown , m. , cruikshank , d. , spencer , j. , trilling , d. , & margot , j .- 2008 , in the solar system beyond neptune , ed . m. a. barucci , h. boehnhardt , d. p. cruikshank , & a. morbidelli ( the univ . of arizona press , 2008 ) , 161179

high - accuracy astrometry permits the determination of not only stellar tangential motion , but also the component along the line - of - sight . such non - spectroscopic ( i.e. astrometric ) radial velocities are independent of stellar atmospheric dynamics , spectral complexity and variability , as well as of gravitational redshift . three methods are analysed : ( 1 ) changing annual parallax , ( 2 ) changing proper motion and ( 3 ) changing angular extent of a moving group of stars . all three have significant potential in planned astrometric projects . current accuracies are still inadequate for the first method , while the second is marginally feasible and is here applied to 16 stars . the third method reaches high accuracy ( km s ) already with present data , although for some clusters an accuracy limit is set by uncertainties in the cluster expansion rate .

astro-ph9907145

for well over a century , radial velocities for objects outside the solar system have been determined through spectroscopy , using the ( doppler ) shifts of stellar spectral lines . the advent of high - accuracy ( sub - milliarcsec ) astrometric measurements , both on ground and in space , now permits radial velocities to be obtained by alternative methods , based on geometric principles and therefore independent of spectroscopy . the importance of such _ astrometric radial velocities _ stems from the fact that they are independent of phenomena which affect the spectroscopic method , such as line asymmetries and shifts caused by atmospheric pulsation , surface convection , stellar rotation , stellar winds , isotopic composition , pressure , and gravitational potential . conversely , the differences between spectroscopic and astrometric radial velocities may provide information on these phenomena that can not be obtained by other methods . although the theoretical possibility of deducing astrometric radial velocities from geometric projection effects was noted already at the beginning of the 20th century ( if not earlier ) , it is only recently that such methods have reached an accuracy level permitting non - trivial comparison with spectroscopic measurements . we have analysed three methods by which astrometric radial velocities can be determined ( fig . [ fig : methods ] ) . two of them are applicable to individual , nearby stars and are based on the well understood secular changes in the stellar trigonometric parallax and proper motion . the third method uses the apparent changes in the geometry of a star cluster or association to derive its kinematic parameters , assuming that the member stars share , in the mean , a common space velocity . in sects . [ sec : pidot ] to [ sec : mcm ] we describe the principle and underlying assumptions of each of the three methods and derive approximate formulae for the expected accuracy of resulting astrometric radial velocities . for the first and second methods , an inventory of nearby potential target stars is made , and the second method is applied to several of these . however , given currently available astrometric data , only the third ( moving - cluster ) method is capable of yielding astrophysically interesting , sub - km s@xmath1 accuracy . in subsequent papers we develop in detail the theory of this method , based on the maximum - likelihood principle , as well as its practical implementation , and apply it to a number of nearby open clusters and associations , using data from the hipparcos astrometry satellite . in the following sections , @xmath2 , @xmath3 and @xmath4 denote the trigonometric parallax of a star , its ( total ) proper motion , and its radial velocity . the components of @xmath3 in right ascension and declination are denoted @xmath5 and @xmath6 , with @xmath7 . the dot signifies a time derivative , as in @xmath8 . the statistical uncertainty ( standard error ) of a quantity @xmath9 is denoted @xmath10 . ( we prefer this non - standard notation to @xmath11 , since @xmath9 is itself often a subscripted variable . ) @xmath12 is used for the physical velocity dispersion in a cluster . @xmath13 km is the astronomical unit ; the equivalent values @xmath14 km yr s@xmath1 and @xmath15 mas km yr s@xmath1 are conveniently used in equations below ( cf . table 1.2.2 in vol . 1 of esa @xcite ) . other notations are explained as they are introduced . in estimating the potential accuracy of the different methods , we consider three hypothetical situations : * case a : a quasi - continuous series of observations over a few years , resulting in an accuracy of @xmath16 mas ( milliarcsec ) for the trigonometric parallaxes and @xmath17 mas yr@xmath1 for the proper motions . * case b : similar to case a , only a thousand times better , i.e. @xmath16 @xmath3as ( microarcsec ) and @xmath17 @xmath3as yr@xmath1 . * case c : _ two _ sets of measurements , separated by an interval of 50 yr , where each set has the same accuracy as in case b. the much longer - time baseline obviously allows a much improved determination of the accumulated changes in parallax and proper motion . the accuracies assumed in case a are close to what the hipparcos space astrometry mission ( esa @xcite ) achieved for its main observation programme of more than 100@xmath18000 stars . current ground - based proper motions may be slightly better than this , but not by a large factor . this case therefore represents , more or less , the state - of - the - art accuracy in optical astrometry . accuracies in the 1 to 10 @xmath3as range are envisaged for some planned or projected space astrometry missions , such as gaia ( lindegren & perryman @xcite ) and sim ( unwin et al . @xcite ) . the duration of such a mission is here assumed to be about 5 years . using the longer - time baselines available with ground - based techniques , similar performance may in the future be reached with the most accurate ground - based techniques ( pravdo & shaklan @xcite ; shao @xcite ) . case b therefore corresponds to what we could realistically hope for within one or two decades . case c , finally , probably represents an upper limit to what is practically feasible in terms of long - term proper - motion accuracy , not to mention the patience of astronomers . the most direct and model - independent way to determine radial velocity by astrometry is to measure the secular change in the trigonometric parallax ( fig . [ fig : methods]a ) . the distance @xmath19 ( from the solar system barycentre ) is related to parallax @xmath2 through @xmath20 . since @xmath21 , the radial velocity is @xmath22 where @xmath23 is the astronomical unit ( sect . [ sec : not ] ) . the equivalent of eq . ( [ eq : pidot ] ) was derived by schlesinger ( @xcite ) , who concluded that the parallax change is very small for every known star . however , although extremely accurate parallax measurements are obviously required , the method is not as unrealistic as it may seem at first . to take a specific , if extreme , example : for barnard s star ( gl 699 = hip 87937 ) , with @xmath24 mas and @xmath25 km s@xmath1 , the expected parallax rate is @xmath26as yr@xmath1 . according to our discussion in sect . [ sec : acc ] this will almost certainly be measurable in the near future . it can be noted that the changing - parallax method , in contrast to the methods described in sect . [ sec : mudot ] and [ sec : mcm ] , does not depend on the object having a large and uniform space motion , and would therefore be applicable to all stars within a few parsecs of the sun . lrcc @xmath2 & @xmath27 & + [ mas ] & & case b & case c + 740 & 3 & 1.2 & 0.05 + 300 & 14 & 7.5 & 0.3 + 200 & 60 & 17 & 0.7 + 100 & 326 & 68 & 2.8 + [ tab : epspidot ] the accuracy in @xmath4 is readily estimated from eq . ( [ eq : pidot ] ) for a given accuracy in @xmath28 , since the contribution of the parallax uncertainty to the factor @xmath29 is negligible . the achievable accuracy in @xmath28 depends both on the individual astrometric measurements and on their number and distribution in time . concerning the temporal distribution of the measurements we consider two limiting situations : _ quasi - continuous observation . _ the measurements are more or less uniformly spread out over a time period of length @xmath30 centred on the epoch @xmath31 . this is a good approximation to the way a single space mission would typically be operated ; for example , hipparcos had @xmath32 yr and @xmath33 . in such a case there exist simple ( mean ) relations between how accurately the different astrometric parameters of the same star can be derived , depending on @xmath30 . for instance , @xmath34 , if @xmath35 is the accuracy of the proper motion in declination and @xmath36 that of the declination at @xmath31 . this approximation is applicable to case a and b as defined in sect . [ sec : acc ] . _ two - epoch observation . _ two isolated parallax or proper - motion measurements are taken , separated by a long time interval ( say , @xmath37 years ) during which no observation takes place . each measurement must actually be the result of a series covering at least a year or so , but the duration of each such series is assumed to be negligible compared with @xmath37 . this could be two similar space missions separated by several decades and is applicable to case c in sect . [ sec : acc ] . for quasi - continuous observation we may assume that the parallax variation is linear over the observation period @xmath30 . thus , @xmath38 , where @xmath39 and @xmath28 are two parameters to be determined from the observations . there exists an approximate relation between the accuracies of these two parameters that is similar to that between the proper motion and the position at the mean epoch , viz . moreover , the estimates of the two parameters are uncorrelated , so @xmath41 equals the accuracy @xmath42 of a parallax determination in the absence of the parallax - change term ; thus @xmath43 in the case of a two - epoch observation , let us assume that independent parallax measurements @xmath44 and @xmath45 are made at epochs @xmath46 and @xmath47 . the estimated rate of change is @xmath48 . with @xmath49 and @xmath50 denoting the accuracies of the two measurements , we have @xmath51^{1/2}/t$ ] and consequently @xmath52^{1/2 } \ , . \label{eq : epspidotb}\ ] ] for given observational errors we find , from both eq . ( [ eq : epspidota ] ) and ( [ eq : epspidotb ] ) , that the radial - velocity error is simply a function of distance . the number of potential target stars for a certain maximum radial - velocity uncertainty is therefore given by the total number of stars within the corresponding maximum distance . table [ tab : epspidot ] gives the actual numbers of such stars , and the observational accuracies that may be reached . to a good approximation , single stars move with uniform linear velocity through space . for a given linear tangential velocity , the angular velocity ( or proper motion @xmath3 ) , as seen from the sun , varies inversely with the distance to the object . however , the tangential velocity changes due to the varying angle between the line of sight and the space - velocity vector ( fig . [ fig : methods]b ) . as is well known ( e.g. van de kamp @xcite , murray @xcite ) the two effects combine to produce an apparent ( perspective ) acceleration of the motion on the sky , or a rate of change in proper motion amounting to @xmath53 . with @xmath54 we find @xmath55 schlesinger ( @xcite ) derived the equivalent of this equation , calculated the perspective acceleration for kapteyn s and barnard s stars ( cf . table [ tab : epsmudot ] ) and noted that , if accurate positions are acquired over long periods of time , `` we shall be in position to determine the radial velocities of these stars independently of the spectroscope and with an excellent degree of precision '' . the equation for the perspective acceleration was earlier derived by seeliger ( @xcite ) and used by ristenpart ( @xcite ) in an ( unsuccessful ) attempt to determine @xmath56 observationally for groombridge 1830 . a major consideration for ristenpart seems to have been the possibility to derive the parallax from the apparent acceleration in combination with a spectroscopic radial velocity . such a determination of ` acceleration parallaxes ' was also considered by eichhorn ( @xcite ) . lrrlrccccl cns3 & hd & hip & sp & @xmath57 & @xmath19 & @xmath58 & & remark + & & & & & [ pc ] & [ arcsec@xmath59 yr@xmath1 ] & case b & case c + gl 699 & & 87937 & sdm4 & 9.5 & 1.8 & 5.69 & 0.13&0.01 & barnard s star + gl 551 & & 70890 & m5ve & 11.0 & 1.3 & 2.98 & 0.25&0.01 & @xmath61 cen c ( proxima ) + gl 559b & 128621 & 71681 & k0v & 1.4 & 1.3 & 2.76 & 0.27&0.01 & @xmath61 cen b + gl 559a & 128620 & 71683 & g2v & 0.0 & 1.3 & 2.75 & 0.28&0.01 & @xmath61 cen a ( ab : @xmath62 yr ) + gl 191 & 33793 & 24186 & m0v & 8.9 & 3.9 & 2.21 & 0.34&0.01 & kapteyn s star + gl 887 & 217987 & 114046 & m2ve & 7.4 & 3.3 & 2.10 & 0.36&0.01 & + gl 406 & & & m6 & 13.5 & 2.4 & 1.96 & 0.38&0.01 & wolf 359 + gl 411 & 95735 & 54035 & m2ve & 7.5 & 2.5 & 1.88 & 0.40&0.01 & + gl 820a & 201091 & 104214 & k5ve & 5.2 & 3.5 & 1.52 & 0.50&0.02 & 61 cyg a + gl 820b & 201092 & 104217 & k7ve & 6.1 & 3.5 & 1.48 & 0.51&0.02 & 61 cyg b ( ab : @xmath63 yr ) + gl 1 & 225213 & 439 & m4v & 8.6 & 4.4 & 1.40 & 0.54&0.02 & + gl 845 & 209100 & 108870 & k5ve & 4.7 & 3.6 & 1.30 & 0.58&0.02 & @xmath64 ind + gl 65a & & & dm5.5e & 12.6 & 2.6 & 1.28 & 0.59&0.02 & + gl 65b & & & dm5.5e & 12.7 & 2.6 & 1.28 & 0.59&0.02 & uv cet ( ab : @xmath65 yr ) + gl 273 & & 36208 & m3.5 & 9.8 & 3.8 & 0.98 & 0.77&0.03 & luyten s star + gl 866ab & & & m5e & 12.3 & 3.4 & 0.96 & 0.79&0.03 & @xmath66 yr + gl 412a & & 54211 & m2ve & 8.8 & 4.8 & 0.93 & 0.81&0.03 & + gl 412b & & & m6e & 14.4 & 5.3 & 0.86 & 0.88&0.03 & wx uma + gl 825 & 202560 & 105090 & m0ve & 6.7 & 3.9 & 0.88 & 0.86&0.03 & + gl 15a & 1326 & 1475 & m2v & 8.1 & 3.6 & 0.82 & 0.92&0.03 & gx and + gl 15b & & & m6ve & 11.1 & 3.6 & 0.84 & 0.90&0.03 & gq and ( ab : @xmath67 yr ) + gl 166a & 26965 & 19849 & k1ve & 4.4 & 5.0 & 0.81 & 0.94&0.03 & 40 eri a ( @xmath68 eri ) + gl 166b & 26976 & & da4 & 9.5 & 4.8 & 0.84 & 0.90&0.03 & 40 eri b ( bc : @xmath69 yr ) + gl 166c & & & dm4.5e & 9.5 & 4.8 & 0.84 & 0.90&0.03 & dy eri + gl 299 & & & dm5 & 12.8 & 6.8 & 0.77 & 0.98&0.04 & ross 619 + gl 451a & 103095 & 57939 & g8vi & 6.4 & 9.2 & 0.77 & 0.98&0.04 & groombridge 1830 + gl 451b & & & & 12 & 9.2 & 0.82 & 0.92&0.03 & cf uma ( non - existent star ? ) + gl 35 & & 3829 & dz7 & 12.4 & 4.4 & 0.68 & 1.1&0.04 & van maanen 2 + gl 725b & 173740 & 91772 & dm5 & 9.7 & 3.6 & 0.66 & 1.2&0.04 & ab : @xmath70 yr + gl 725a & 173739 & 91768 & dm4 & 8.9 & 3.6 & 0.63 & 1.2&0.04 & + gl 440 & & 57367 & dq6 & 11.5 & 4.6 & 0.58 & 1.3&0.05 & + gl 71 & 10700 & 8102 & g8vp & 3.5 & 3.6 & 0.53 & 1.4&0.05 & @xmath71 cet + gl 754 & & & m4.5 & 12.2 & 5.7 & 0.52 & 1.5&0.05 & + gl 139 & 20794 & 15510 & g5v & 4.3 & 6.1 & 0.52 & 1.5&0.05 & 82 eri + gl 905 & & & dm6 & 12.3 & 3.2 & 0.51 & 1.5&0.05 & + gl 244a & 48915 & 32349 & a1v & @xmath72 & 2.6 & 0.51 & 1.5&0.05 & @xmath61 cma ( sirius ) + gl 244b & & & da2 & & 2.6 & 0.51 & 1.5&0.05 & ab : @xmath73 yr + gl 53a & 6582 & 5336 & g5vi & 5.2 & 7.6 & 0.50 & 1.5&0.06 & @xmath3 cas + gl 53b & & & & 11 & 7.6 & 0.51 & 1.5&0.06 & ab : @xmath74 yr + [ tab : epsmudot ] subsequent attempts to determine the perspective acceleration of barnard s star by lundmark & luyten ( @xcite ) , alden ( @xcite ) and van de kamp ( @xcite ) yielded results that were only barely significant or ( in retrospect ) spurious . meanwhile , russell & atkinson ( @xcite ) suggested that the white dwarf van maanen 2 might exhibit a gravitational redshift of several hundred km s@xmath1 and that this could be distinguished from a real radial velocity through measurement of the perspective acceleration . the astrophysical relevance of astrometric radial - velocity determinations was thus already established ( oort @xcite ) . in relatively recent times , the perspective acceleration was successfully determined for barnard s star by van de kamp ( @xcite , @xcite , @xcite , @xcite , @xcite ) ; for van maanen 2 by van de kamp ( @xcite ) , gatewood & russell ( @xcite ) and hershey ( @xcite ) ; and for groombridge 1830 by beardsley et al . ( @xcite ) . among these determinations the highest precisions , in terms of the astrometric radial velocity , were obtained for barnard s star ( corresponding to @xmath75 km s@xmath1 ; van de kamp @xcite ) and van maanen 2 ( @xmath76 km s@xmath1 ; gatewood & russell @xcite ) . our application of the method , combining hipparcos measurements with data in the astrographic catalogue , yielded radial velocities for 16 objects , as listed in table [ tab : ac ] . the accuracy of the radial velocity calculated from eq . ( [ eq : mudot ] ) can be estimated as in sect . [ sec : epspidot ] . it depends on the parallax - proper - motion product @xmath58 . the most promising targets for this method are listed in table [ tab : epsmudot ] , which contains the known nearby stars ranked after decreasing @xmath58 . for quasi - continuous observation during a period of length @xmath30 we may use a quadratic model for the angular position @xmath77 of the star along the great - circle arc : here @xmath79 is the proper motion at the central epoch @xmath31 . the estimates of @xmath79 and @xmath56 are found to be uncorrelated and their errors related by @xmath80 . consequently , @xmath81 where @xmath82 is the accuracy of proper - motion measurements in the absence of temporal changes . we neglect the ( small ) contribution to @xmath83 from the uncertainty in the denominator @xmath58 . for a two - epoch observation , consider proper - motion measurements @xmath84 and @xmath85 made around @xmath46 and @xmath47 . the estimated acceleration is @xmath86 . provided the two observation intervals centred on @xmath46 and @xmath87 do not overlap , the measurements are independent , yielding the standard error @xmath88^{1/2}/t$ ] . for the radial velocity this gives @xmath89^{1/2 } \ , . \label{eq : epsmudotb}\ ] ] based on these formulae , table [ tab : epsmudot ] gives the potential radial - velocity accuracy for the two cases b and c defined in sect . [ sec : acc ] . in a two - epoch observation we normally have , in addition , a very good estimate of the _ mean _ proper motion between @xmath46 and @xmath87 , provided the positions @xmath90 and @xmath91 at these epochs are accurately known . in the previous quadratic model we may take the reference epoch to be @xmath92 and find @xmath93 with standard error @xmath94^{1/2}/t$ ] . the three proper - motion estimates @xmath79 , @xmath84 and @xmath85 ( referred to @xmath31 , @xmath46 and @xmath87 ) are mutually independent and may be combined in a least - squares estimate of @xmath56 . if @xmath95 ( equal weight at @xmath46 and @xmath87 ) , then it is found that @xmath79 does not contribute at all to the determination of @xmath56 , and the standard error is still given by eq . ( [ eq : epsmudotb ] ) . if , on the other hand , the two observation epochs are not equivalent , then some improvement can be expected by introducing the position measurements . an important special case is when there is just a position ( no proper motion ) determined at one of the epochs , say @xmath46 . this is however equivalent to the two independent proper - motion determinations @xmath79 at @xmath92 , and @xmath85 at @xmath87 , separated by @xmath96 . applying eq . ( [ eq : epsmudotb ] ) on this case yields @xmath97^{1/2 } \ , . \label{eq : epsmudotc}\ ] ] this formula is applicable on the combination of a recent position and proper - motion measurement ( e.g. by hipparcos ) with a position derived from old photographic plates ( e.g. the astrographic catalogue ) . taking @xmath98 , @xmath99 mas as representative for the astrographic catalogue , and @xmath100 , @xmath101 mas , @xmath102 mas yr@xmath1 for hipparcos , we find @xmath103 . with such data , moderate accuracies of a few tens of km s@xmath1 can be reached for several stars ( sect . [ sec : pm - results ] ) . the perspective - acceleration method depends critically on the assumption that the star moves with uniform space motion relative the observer . the presence of a real acceleration of their relative motions , caused by gravitational action of other bodies , would bias the calculated astrometric radial velocity by @xmath104 , where @xmath105 is the tangential component of the relative acceleration . the acceleration towards the galactic centre caused by the smoothed galactic potential in the vicinity of the sun is @xmath106 km s@xmath107 . for a hypothetical observer near the sun but unaffected by this acceleration , the maximum bias would be 0.06 km s@xmath1 for barnard s star , and 0.17 km s@xmath107 for proxima . however , since real observations are made relative the solar - system barycentre , which itself is accelerated in the galactic gravitational field , the observed ( differential ) effect will be very much smaller . in the case of proxima the acceleration towards @xmath61 cen ab is of a similar magnitude as the galactic acceleration . for the several orbital binaries in table [ tab : epsmudot ] the curvature of the orbit is much greater than the perspective acceleration . application of this method will therefore require careful correction for all known perturbations : the possible presence of long - period companions may introduce a considerable uncertainty . among other effects which may have to be considered are light - time effects which , to first order in @xmath108 , may require a correction of @xmath109 on the right - hand side of eq . ( [ eq : mudot ] ) , where @xmath110 is the tangential velocity . for typical high - velocity ( population ii ) stars the correction is 0.10.2 km s@xmath1 . at this accuracy level , the precise definition of the radial - velocity concept itself requires careful consideration ( lindegren et al . @xcite ) . llcccl hip & & & remark + & no . & epoch & astrom . & spectr . & + 439 & 3152964 & 1912.956 & @xmath111 & @xmath112 & + 1475 & 1406215 & 1898.435 & @xmath113 & @xmath114 & gx and + 5336 & 1721511 & 1913.868 & @xmath115 & @xmath116 & @xmath3 cas + 15510 & 3488626 & 1901.018 & @xmath117 & @xmath118 & 82 eri + 19849 & 2125614 & 1892.970 & @xmath119 & @xmath120 & 40 eri + 24186 & 3505363 & 1899.058 & @xmath121 & @xmath122 & kapteyn s star + 36208 & 282902 & 1908.859 & @xmath123 & @xmath124 & luyten s star + 54035 & 1340883 & 1930.895 & @xmath125 & @xmath126 & + 54211 & 1463341 & 1895.620 & @xmath127 & @xmath128 & + 57367 & 4195112 & 1924.492 & @xmath129 & & + 57939 & 1342199 & 1930.260 & @xmath130 & @xmath116 & groombridge 1830 + 87937 & 146626 & 1905.979 & @xmath131 & @xmath132 & barnard s star + 104214/217 & 1382645/649 & 1921.699 & @xmath133 & @xmath134 & 61 cyg + 104214/217 & 1382645/649 & 1921.699 & @xmath135 & & + 105090 & 3462277 & 1905.316 & @xmath136 & @xmath137 + 108870 & 4384302 & 1901.189 & @xmath138 & @xmath139 & @xmath64 ind + 114046 & 3355101 & 1913.368 & @xmath140 & @xmath141 & + past determinations of the perspective acceleration , e.g. by van de kamp ( @xcite ) and gatewood & russell ( @xcite ) , were based on photographic observations collected over several decades , in which the motion of the target star was measured relative to several background ( reference ) stars . one difficulty with the method has been that the positions and motions of the reference stars are themselves not accurately known , and that small errors in the reference data could cause a spurious acceleration of the target star ( van de kamp @xcite ) . the hipparcos catalogue ( esa @xcite ) established a very accurate and homogeneous positional reference frame over the whole sky . using the proper motions , this reference frame can be extrapolated backwards in time . it is then possible to re - reduce measurements of old photographic plates , and express even century - old stellar positions in the same reference frame as modern observations . this should greatly facilitate the determination of effects such as the perspective acceleration , which are sensitive to systematic errors in the reference frame . as part of the _ carte du ciel _ project begun more than a century ago , an astrographic programme to measure the positions of all stars down to the 11th magnitude was carried out and published as the astrographic catalogue , ac ( see eichhorn @xcite for a description ) . after transfer to electronic media , the position measurements have been reduced to the hipparcos reference frame ( nesterov et al . @xcite , urban et al . the result is a positional catalogue of more than 4 million stars with a mean epoch around 1907 and a typical accuracy of about 200 mas . we have used the version known as ac 2000 ( urban et al . @xcite ) , available on cd - rom from the us naval observatory , to examine the old positions of all the stars with hip identifiers in table [ tab : epsmudot ] . for the stars in table [ tab : ac ] we successfully matched the ac positions with the positions extrapolated backwards from the hipparcos catalogue and hence could calculate the astrometric radial velocities . other potential targets in table [ tab : epsmudot ] were either outside the magnitude range of ac 2000 ( e.g. @xmath61 cen and proxima ) or lacked an accurate proper motion from hipparcos ( e.g. van maanen 2 and hip 91768 + 91772 ) . the basic procedure was as follows . the rigorous epoch transformation algorithm described in sect . 1.5.5 of the hipparcos catalogue , vol . 1 , was used to propagate the hipparcos position and its covariance matrix to the ac 2000 epoch relevant for each star . this extrapolated position was compared with the actual measured position in ac 2000 , assuming a standard error of 200 mas in each coordinate for the latter . a @xmath142 goodness - of - fit was then calculated from the position difference and the combined covariance of the extrapolated and measured positions . the epoch transformation algorithm requires that the radial velocity is known . the radial velocity was therefore varied until the @xmath142 attained its minimum value . the @xmath1431@xmath144 confidence interval given in the table was obtained by modifying the radial velocity until the @xmath142 had increased by one unit above the minimum . for some of the stars , data had to be corrected for duplicity or known orbital motion . the solutions for the resolved binary 61 cyg ( hip 104214 + 104217 ) refer to the mass centre , assuming a mass ratio of @xmath145 , as estimated by means of standard isochrones from the absolute magnitudes and colour indices of the components ( sderhjelm , private communication ) . for the astrometric binary @xmath3 cas ( hip 5336 ) the hipparcos data explicitly refer to the mass centre using the orbit by heintz & cantor ( @xcite ) ; the same orbit was used to correct the ac position of the primary to the mass centre . no correction for orbital motion was used for gx and and 40 eri . table [ tab : ac ] gives two solutions for 61 cyg . the first solution was obtained as described above , using only the hipparcos data plus the ac positions for the two components . the second solution , marked with an asterisk in the table , was derived by including also the observations by bessel ( @xcite ) from his pioneering determination of the star s parallax . bessel measured the angular distances from the geometrical centre ( half - way between the components ) of 61 cyg to two reference stars , called @xmath146 and @xmath19 in his paper . after elimination of aberration , proper motion and parallax , he found the distances @xmath147 arcsec and @xmath148 arcsec for the beginning of year 1838 ( b1838.0 = j1838.0022 ) . the uncertainties are our estimates ( standard errors ) based on the scatter of the residuals in bessel s solution ` ii ' . we identified the reference stars in ac 2000 and in the tycho catalogue ( esa @xcite ) as @xmath146 = ac 1382543 = tyc 3168 708 1 and @xmath19 = ac 1382712 = tyc 3168 1106 1 . extrapolating the positions from these catalogues back to b1838 allowed us to compute the position of the geometrical centre of 61 cyg in the hipparcos / tycho reference frame . this could then be transformed to the position of the mass centre , using bessel s own measurement of the separation and position angle in 61 cyg and the previously assumed mass ratio . actually , all the available data were combined into a @xmath142 goodness - of - fit measure and the radial velocity was varied in order to find the minimum and the @xmath1431@xmath144 confidence interval . this gave @xmath149 km s@xmath1 . table [ tab : ac ] also gives the spectroscopic radial velocities when available in the literature . a comparison between the astrometric and spectroscopic radial velocities is made in fig . [ fig : res ] . given the stated confidence intervals , the agreement is in all cases rather satisfactory . the exercise demonstrates the basic feasibility of this method , but also hints at some of the difficulties in applying it to non - single stars . the moving - cluster method is based on the assumption that the stars in a cluster move through space essentially with a common velocity vector . the radial - velocity component makes the cluster appear to contract or expand due to its changing distance ( fig . [ fig : methods]c ) . the relative rate of apparent contraction equals the relative rate of change in distance to the cluster . this can be converted to a linear velocity ( in km s@xmath1 ) if the distance to the cluster is known , e.g. from trigonometric parallaxes . in practice , the method amounts to determining the space velocity of the cluster , i.e. the convergent point and the speed of motion , through a combination of proper motion and parallax data . once the space velocity is known , the radial velocity for any member star may be calculated by projecting the velocity vector onto the line of sight . the method can be regarded as an inversion of the classical procedure ( e.g. binney & merrifield @xcite ) by which the distances to the stars in a moving cluster are derived from the proper motions and ( spectroscopic ) radial velocities : if instead the distances are known , the radial velocities follow . the first application of the classical moving - cluster method for distance determination was by klinkerfues ( @xcite ) , in a study of the ursa major system . the possibility to check spectroscopic radial velocities against astrometric data was recognised by klinkerfues , but could not then be applied to the ursa major cluster due to the lack of reliable trigonometric parallaxes . this changed with hertzsprung s ( @xcite ) discovery that sirius probably belongs to the ursa major moving group . the relatively large and well - determined parallax of sirius , combined with its considerable angular distance from the cluster apex , could lead to a meaningful estimate for the cluster velocity and hence for the radial velocities . rasmuson ( @xcite ) and smart ( @xcite ) appear to have been among the first who actually made this computation , although mainly as a means of verifying the cluster method for distance determination . later studies by petrie ( @xcite ) and petrie & moyls ( @xcite ) reached formal errors in the astrometric radial velocities below 1 km s@xmath1 . the last paper concluded `` there does not appear to be much likelihood of improving the present results until a substantial improvement in the accuracy of the trigonometric parallaxes becomes possible . '' one of the purposes of the petrie & moyls study was to derive the astrometric radial velocities of spectral type a in order to check the victoria system of spectroscopic velocities . the method was also applied to the hyades ( petrie @xcite ) but only with an uncertainty of a few km s@xmath1 . given the expected future availability of more accurate proper motions and trigonometric parallaxes , petrie ( @xcite ) envisaged that one or two moving clusters could eventually be used as primary radial - velocity standards for early - type spectra . such astrometric data are now in fact available . in sect . [ sec:3prec ] we derive a rough estimate of the accuracy of the method and survey nearby clusters and associations in order to find promising targets for its application . an important consideration is to what extent systematic velocity patterns in the cluster , in particular cluster expansion , will limit the achievable accuracy . this is discussed in sect . [ sec:3syst ] and appendix a. in sect . [ sec : indpar ] we briefly consider the improvement in the distance estimates for individual stars resulting from the moving - cluster method . the present discussion of the moving - cluster method is only intended to highlight its theoretical potential and limitations . its actual application requires a more rigorous formulation , which is developed in a second paper . lcccccccccccc name & iau & @xmath27 & age & @xmath150 & @xmath151 & @xmath4 & @xmath110 & & & & @xmath152 + & designation & & [ myr ] & [ arcmin ] & [ pc ] & & & ( a ) & ( b ) & & [ km s@xmath1 ] + cassiopeia taurus & & 83 & 25 & 1800 & 190 & @xmath1536&21 & & 0.24 & 0.06 & & @xmath1547.3 + upper centaurus lupus & & 221 & 13 & 670 & 140 & @xmath1535&21 & & 0.25 & 0.09 & & @xmath15410 + ursa major & & 40 & 300 & 4300 & 25 & @xmath15411&5 & & 0.11 & 0.10 & & @xmath1540.08 + lower centaurus crux & & 180 & 10 & 560 & 118 & @xmath15312&19 & & 0.30 & 0.12 & & @xmath15411 + hyades & c 0424@xmath153157 & 380 & 625 & 560 & 46 & @xmath15343&25 & & 0.19 & 0.14 & & @xmath1540.07 + perseus ob3 ( @xmath61 per ) & c 0318@xmath153484 & 186 & 50 & 350 & 180 & @xmath1541&29 & & 0.64 & 0.18 & & @xmath1543.4 + ` hip 98321 ' & & 59 & 60 & 740 & 300 & @xmath15417&4 & & 1.1 & 0.19 & & @xmath1544.8 + upper scorpius & & 120 & 5 & 325 & 145 & @xmath1545&18 & & 0.71 & 0.24 & & @xmath15428 + lacerta ob1 & & 96 & 16 & 350 & 370 & @xmath15413&8 & & 1.8 & 0.26 & & @xmath15422 + collinder 121 & & 103 & 5 & 290 & 540 & @xmath15326&15 & & 3.3 & 0.32 & & @xmath154103 + collinder 70 & c 0533@xmath154011 & 345 & & 140 & 430 & & & & 2.7 & 0.33 & & + cepheus ob2 & & 71 & 5 & 320 & 615 & @xmath15421&12 & & 4.1 & 0.34 & & @xmath154120 + vela ob2 & & 93 & 20 & 260 & 415 & @xmath15318&20 & & 2.8 & 0.36 & & @xmath15420 + perseus ob2 & & 41 & 7 & 340 & 300 & @xmath15320&14 & & 2.5 & 0.43 & & @xmath15446 + pleiades & c 0344@xmath153239 & 277 & 130 & 120 & 125 & @xmath1537&29 & & 1.1 & 0.43 & & @xmath1540.92 + coma berenices & c 1222@xmath153263 & 273 & 460 & 120 & 87 & 0&9 & & 0.84 & 0.43 & & @xmath1540.18 + ngc 3532 & c 1104@xmath154584 & 677 & 290 & 50 & 480 & @xmath1537&27 & & 6.0 & 0.66 & & @xmath1541.6 + praesepe & c 0837@xmath153201 & 161 & 830 & 70 & 160 & @xmath15333&26 & & 3.1 & 0.98 & & @xmath1540.18 + ngc 2477 & c 0750@xmath154384 & 1911 & 1260 & 20 & 1150 & @xmath1537&17 & & 21 & 0.98 & & @xmath1540.87 + ic 4756 & c 1836@xmath153054 & 466 & 830 & 39 & 390 & @xmath15418&5 & & 7.6 & 1.0 & & @xmath1540.45 + ic 4725 & c 1828@xmath154192 & 601 & 41 & 29 & 710 & @xmath1533&20 & & 16 & 1.2 & & @xmath15417 + trumpler 10 & & 23 & 15 & 150 & 370 & @xmath15321&26 & & 8.6 & 1.2 & & @xmath15424 + cepheus ob6 & & 20 & 50 & 150 & 270 & @xmath15420&22 & & 6.8 & 1.3 & & @xmath1545.2 + ngc 752 & c 0154@xmath153374 & 77 & 3300 & 75 & 360 & @xmath1543&20 & & 9.0 & 1.3 & & @xmath1540.10 + ngc 6618 & c 1817@xmath154162 & 660 & & 25 & 1500 & &26 & & 38 & 1.3 & & + ngc 2451 & c 0743@xmath154378 & 153 & 41 & 50 & 315 & @xmath15327&27 & & 8.4 & 1.4 & & @xmath1547.3 + ngc 7789 & c 2354@xmath153564 & 583 & 1780 & 25 & 1800 & @xmath15454&27 & & 49 & 1.4 & & @xmath1541.0 + ngc 2099 & c 0549@xmath153325 & 1842 & 200 & 14 & 1300 & @xmath1538&45 & & 35 & 1.4 & & @xmath1546.4 + ngc 6475 & c 1750@xmath154348 & 54 & 130 & 80 & 240 & @xmath15412&7 & & 6.8 & 1.5 & & @xmath1541.8 + ngc 2264 & c 0638@xmath153099 & 222 & 10 & 39 & 800 & @xmath15322&17 & & 23 & 1.5 & & @xmath15476 + stock 2 & c 0211@xmath153590 & 166 & 170 & 45 & 300 & @xmath1532&30 & & 8.6 & 1.5 & & @xmath1541.7 + ic 2602 & c 1041@xmath154641 & 33 & 29 & 100 & 150 & @xmath15322&14 & & 4.6 & 1.5 & & @xmath1545.0 + [ tab : clus ] the accuracy of the astrometric radial velocity potentially achievable by the moving - cluster method can be estimated as follows . let @xmath19 be the ( mean ) distance to the cluster and consider a star at angular distance @xmath155 from the centre of the cluster , as seen from the sun . the projected linear distance of the star from the centre of the cluster is @xmath156 , provided the angular extent of the cluster is not very large . as the cluster moves through space , its linear dimensions remain constant , so that @xmath157 . putting @xmath158 ( the proper motion relative to the cluster centre ) , @xmath159 , and @xmath54 , gives @xmath160 . now suppose that the parallaxes and proper motions of @xmath27 cluster stars are measured , each with uncertainties of @xmath42 and @xmath82 . standard error propagation formulae give the expected accuracy in @xmath4 as @xmath161^{1/2 } \ , \label{eq : epsmc0}\ ] ] where @xmath150 is in radians ; @xmath23 is the astronomical unit ( sect . [ sec : not ] ) . the expression within the square brackets derives from the uncertainty in the mean cluster distance , by which the derived radial velocity scales . for the type of ( space ) astrometry data considered here ( case a and b ) , @xmath162 is on the order of a few years ( for hipparcos the mean ratio is @xmath163 yr ) . the factor in brackets can then be neglected except for the most extended ( and nearby ) clusters . under certain circumstances it is not the accuracy of proper - motion measurements that defines the ultimate limit on @xmath83 , but rather internal velocity dispersion among the cluster stars . assuming isotropic dispersion with standard deviation @xmath12 in each coordinate , one must add @xmath164 quadratically to the measurement error @xmath82 in eq . ( [ eq : epsmc0 ] ) . thus @xmath165^{1/2 } \nonumber \\ & & \times\left [ 1 + \left(\frac{v_r\rho_{\rm rms}\epsilon(\pi ) } { a\epsilon(\mu)}\right)^2 \right]^{1/2 } \ , , \label{eq : epsmc1}\end{aligned}\ ] ] is the accuracy achievable for the radial velocity of the cluster centroid . for the radial velocity of an individual star this uncertainty must be increased by the internal dispersion . the internal velocity dispersion will dominate the error budget for nearby clusters , viz . if @xmath166 . assuming a velocity dispersion of 0.25 km s@xmath1 and a proper - motion accuracy of 1 mas yr@xmath1 ( as for hipparcos ) , this will be the case for clusters within 50 pc of the sun . for an observational accuracy in the 110 @xmath3as yr@xmath1 range the internal dispersion will dominate in practically all galactic clusters and eq . ( [ eq : epsmc1 ] ) can be simplified to @xmath167 . in this case the achievable accuracy becomes independent of the astrometric one . table [ tab : clus ] lists some nearby clusters and associations , with estimates of the achievable accuracy in the radial velocity of the cluster centroid , assuming current ( hipparcos - type ) astrometric performance ( case a in sect . [ sec : acc ] ) as well as future ( microarcsec ) expectations ( case b ) . as explained above , increasing the astrometric accuracy still further gives practically no improvement ; this is why case c is not considered in the table . the entry ` hip 98321 ' refers to the possible association identified by platais et al . ( @xcite ) and named after one of its members . of dubious status , it was included as an example of the extended , low - density groups that may exist in the general stellar field , but are difficult to identify with existing data . blaauw ( @xcite ) showed that the proper motion pattern for a linearly expanding cluster is identical to the apparent convergence produced by parallel space motions . astrometric data alone therefore can not distinguish such expansion from a radial motion . if such an expansion exists , and is not taken into account in estimating the astrometric radial velocity , a bias will result , as examined in appendix a ( eq . [ eq : bias ] ) . the gravitationally unbound associations are known to expand on timescales comparable with their nuclear ages ( de zeeuw & brand @xcite ) . but also for a gravitationally bound open cluster some expansion can be expected as a result of the dynamical evolution of the cluster ( see mathieu @xcite and wielen @xcite for an introduction to this complex issue ) . in either case the inverse age of the cluster or association may be taken as a rough upper limit on the cluster s relative expansion rate @xmath168 [ yr@xmath1 ] . ( [ eq : bias ] ) then gives @xmath169}{\rm age~[myr]}~{\rm km}~{\rm s}^{-1}\ , \label{eq : deltaexp}\ ] ] for the bias of a star near the centre of the cluster . ( for an expanding cluster , @xmath170 is always negative . ) resulting values , in the last column of table [ tab : clus ] , are adequately small for a few nearby , relatively old clusters . in other cases the potential bias is very large and will certainly limit the applicability of the method . the ob associations are particularly troublesome , not only because they are young objects ( implying large values of @xmath168 ) , but also because they sometimes appear to expand significantly faster than their photometric ages would suggest ( de zeeuw & brand @xcite ) . however , it should be remembered that the ultimate limitation set by cluster expansion depends on how accurately the expansion rate @xmath168 can be estimated by some independent means . for instance , if @xmath168 can somehow be estimated to within 10 per cent of its value , then the residual biases would still be on the sub - km s@xmath1 level for most of the objects in table [ tab : clus ] . numerical simulation of the dynamical evolution of clusters might in principle provide such estimates of @xmath168 , as could the spectroscopic radial velocities as function of distance . the use of spectroscopic data would not necessarily defeat the purpose of the method , i.e. to determine _ absolute _ radial velocities , since the expansion is revealed already by _ relative _ measurements . in a rigorous estimation of the space motion of a moving cluster , such as will be presented in a second paper , the distances to the individual member stars of the cluster appear as parameters to be estimated . a by - product of the method is therefore that the individual distances are improved , sometimes considerably , compared with the original trigonometric distances ( dravins et al . @xcite ; madsen @xcite ) . the improvement results from a combination of the trigonometric parallax @xmath171 with the kinematic ( secular ) parallax @xmath172 derived from the star s proper motion @xmath3 and tangential velocity @xmath110 , the latter obtained from the estimated space velocity vector of the cluster . the accuracy of the combined parallax estimate @xmath173 can be estimated from @xmath174 . in calculating @xmath175 we need to take into account the observational uncertainty in @xmath3 and the uncertainty in @xmath110 from the internal velocity dispersion . the result is @xmath176^{-1/2 } \ , . \label{eq : epspihat}\ ] ] from the data in table [ tab : clus ] we find that , in case a , the moving - cluster method will be useful to resolve the depth structures of the hyades cluster and of the associations cassiopeia taurus , upper centaurus lupus , lower centaurus crux , perseus ob3 and upper scorpius . in case b , all the clusters and associations are resolved by the trigonometric parallaxes , so the kinematic parallaxes will bring virtually no improvement . calculation of kinematic distances to stars in moving clusters is of course a classical procedure ( e.g. , klinkerfues @xcite and van bueren @xcite ) ; what is new in our treatment is that such distances are derived without recourse to spectroscopic data . with improved astrometric data , further methods for radial - velocity determinations may become feasible . the moving - cluster method could in principle be applied to any geometrical configuration of a fixed linear size . to reach an accuracy of 1 km s@xmath1 in the astrometric radial velocity of an object at 10 pc distance requires a dimensional ` stability ' of the order of @xmath177 yr@xmath1 ; at a distance of 1 kpc the requirement is @xmath178 yr@xmath1 . since these numbers are greater than or comparable with the inverse dynamical timescales of many types of galactic objects , there is at least a theoretical chance that the method could work , given a sufficient measurement accuracy . we consider briefly two such possibilities . according to the previous argument it would be possible to ignore the relative motion of the components in a binary if the period is longer than some @xmath179 yr . this implies an linear separation of at least some 50000 astronomical units , or a few degrees on the sky at 10 pc distance . in principle , then , this case is equivalent to a moving cluster with @xmath180 stars . in the opposite case of a ( relatively ) short - period binary , the radial velocity might be obtained from apparent changes of the orbit . the projected orbit will not be closed , but form a spiral on the sky : slightly diverging if the stars are approaching , slightly converging if they recede . for a system at a distance of 10 pc , say , with a component separation of 10 astronomical units , a radial velocity of 100 km s@xmath1 will change the apparent orbital radius of 1 arcsec by 10 @xmath3as per year . the relative positions would need to be measured during at least a significant fraction of an orbital period , or some 20 years in our example , resulting in an accumulated change by about 0.2 mas . since only relative position measurements between the same stars are required , the observational challenges are not as severe as in some other cases . for a binary with a few arcsec separation , the isoplanatic properties of the atmosphere imply that the cross - correlation distance between the speckle images of both stars should be stable to better than one mas . averaging very many exposures should reduce the errors into the @xmath3as range , with practical limits possibly set by differential refraction ( mcalister @xcite ) . the moving - cluster method ( sect . [ sec : mcm ] ) could in principle be applied also to globular clusters . globular clusters differ markedly from open clusters in that ( potentially ) many more stars could be measured , and through a much larger velocity dispersion ( @xmath181 km s@xmath1 ; peterson & king @xcite ) . the higher number of stars partly compensates the larger dispersion . however , all globular clusters are rather distant , making their angular radii small . as discussed in sect . [ sec:3prec ] the approximate formula @xmath182 applies in the case when the internal motions are well resolved . taking @xmath18320 arcmin as representative for the angular radii of comparatively nearby globular clusters , we find that averaging over some @xmath184 to @xmath185 member stars is needed to reach a radial - velocity accuracy comparable with @xmath12 , i.e. several km s@xmath1 . furthermore , in view of the discussion in sect . [ sec:3syst ] , it is not unlikely that the complex kinematic structures of these objects ( e.g. norris et al . @xcite ) would bias the results . thus , globular clusters remain difficult targets for astrometric radial - velocity determination . the theoretical possibility to use astrometric data ( parallax and proper motion ) to deduce the radial motions of stars has long been recognised . with the highly accurate ( sub - milliarcsec ) astrometry already available or foreseen in planned space missions , such radial - velocity determinations are now also a practical possibility . this will have implications for the future definition of radial - velocity standards , as the range of geometrically determined accurate radial velocities , hitherto limited to the solar system and to solar - type spectra , is extended to many other stellar types represented in the solar neighbourhood . we have identified and analysed three main methods to determine astrometric radial velocities . the first method , using the changing annual parallax , is the intuitively most obvious one , but requires data of an accuracy beyond current techniques . it is nevertheless potentially interesting in view of future space missions or long - term observations from the ground . the second method , using the changing proper motion or perspective acceleration of stars , has a long history , and was previously applied to a few objects , albeit with modest precision in the resulting radial velocity . new results for a greater number of stars , obtained by combining old and modern data , were given in table [ tab : ac ] and fig . [ fig : res ] , thus proving the concept . however , to realise the full potential of the method again requires the accuracies of future astrometry projects . in both these methods the uncertainty in the astrometric radial velocity increases , statistically , with distance squared . they are therefore in practice limited to relatively few stars very close to the sun and , in the second method , to stars with a large tangential velocity . in the general case , the two methods could actually be combined to yield a somewhat higher accuracy , but at least for the stars considered in tables [ tab : epspidot ] and [ tab : epsmudot ] this would only bring a marginal improvement . the third method , using the changing angular extent of a moving cluster or association , is an inversion of the classical moving - cluster method , by which the distance to the cluster was derived from its radial velocity and convergence point . if the distance is known from trigonometric parallaxes , one can instead calculate the radial velocities . it appears to be the only method by which astrophysically interesting accuracies can be obtained with existing astrometric data . in future papers we will develop and exploit this possibility in full , using data from the hipparcos mission . a by - product of the method is that the distance estimates to individual cluster stars may be significantly improved compared with the parallax measurements . one would perhaps expect the moving - cluster method to become extremely powerful with the much more accurate data expected from future astrometry projects . unfortunately , this is not really the case , as internal velocities ( both random and systematic ) become a limiting factor as soon as they are resolved by the proper - motion measurements . nevertheless , even the limited number of clusters within reach of such determinations contain a great many stars spanning a wide range in spectral type and luminosity . this project is supported by the swedish national space board and the swedish natural science research council . we thank dr . s. sderhjelm ( lund observatory ) for providing information on double and multiple stars , prof . de zeeuw and co - workers ( leiden observatory ) for advance data on nearby ob associations , and the referee , prof . a. blaauw , for stimulating criticism of the manuscript . in this appendix we examine how sensitive the moving - cluster method is to _ systematic _ velocity patterns in the cluster , and to what extent such patterns can be determined independently of the astrometric radial velocity . for this purpose we may ignore the random motions as well as the observational errors and we consider only a linear ( first - order ) velocity field . let @xmath186 be the position of the cluster centroid relative to the sun and @xmath187 the position of a member star relative to the centroid . the space velocity of the star is @xmath188 , where @xmath189 is the peculiar velocity . the velocity field is described by the tensor @xmath190 such that @xmath191 . in cartesian coordinates the components of this tensor are simply the partial derivatives @xmath192 for @xmath193 . these nine numbers together describe the three components of a rigid - body rotation , three components of an anisotropic expansion or contraction , and three components of linear shear . it is intuitively clear that certain components of the linear velocity field , such as rotation about the line of sight , can be determined purely from the astrometric data . if the corresponding components of @xmath190 are included as parameters in the cluster model , they can be estimated and will not produce a systematic error ( bias ) in the astrometric radial velocity derived from the model fitting . such components of @xmath190 are ` observable ' and in principle not harmful to the method . let us now examine more generally the extent to which @xmath190 is observable by astrometry . suppose there exists a non - zero tensor @xmath190 such that the velocity fields @xmath194 and @xmath195 produce identical observations for some vector @xmath196 . since the cluster velocity @xmath197 is already a parameter of the model , the observational effects of the velocity field @xmath190 could then entirely be absorbed by adjusting @xmath197 . in this case @xmath190 would be a non - observable component of the general velocity field . moreover , if there exists such a component in the real velocities , then the estimated cluster velocity will have a bias equal to @xmath196 . we now need to calculate the effect of the arbitrary field @xmath190 on the observables . since the parallaxes are not affected , only the proper motion vector @xmath198 needs to be considered . in this equation @xmath199 is the unit vector from the sun towards the star , @xmath200 is the rate of change of that direction , and @xmath201 is the tensor representing projection perpendicular to @xmath199 [ @xmath202 is the unit tensor ; thus @xmath203 is the tangential component of the general vector @xmath204 . with @xmath205 we can write @xmath206 . @xmath190 is non - observable if the space velocities @xmath194 and @xmath207 produce identical tangential velocities for every star , i.e. if @xmath208 = ( \vec{i } - \vec{r}\vec{r}^\prime)(\vec{v}_0+\delta\vec{v})\ ] ] for all directions @xmath199 and distances @xmath19 . this is equivalently written @xmath209 in order that this should be satisfied for all @xmath19 , it is necessary that @xmath210 and @xmath211 are separately satisfied for all unit vectors @xmath199 . the latter equation implies that @xmath212 the former equation can be written @xmath213 , which shows that @xmath199 is an eigenvector of @xmath190 ( with eigenvalue @xmath214 ) . but the only tensor for which every unit vector is an eigenvector is the isotropic tensor , @xmath215 for the arbitrary scalar @xmath168 . it follows that the only non - observable component of @xmath190 is of the form @xmath216 , parametrised by the single scalar @xmath168 , and that consequently eight linearly independent components of @xmath190 can , in principle , be determined from the astrometric observations . the non - observable field @xmath215 describes a uniform isotropic expansion ( @xmath217 ) or contraction ( @xmath218 ) of the cluster with respect to its centroid . these effects are observationally equivalent to an approach or recession of the cluster , i.e. to a different value of its radial velocity . @xmath196 is the bias for the centroid velocity . for any given star , the bias vector @xmath219 is the difference between the derived ( apparent ) space velocity vector @xmath220 and the true vector @xmath221 . using @xmath187 and eq . ( [ eq : deltav ] ) we find @xmath222 the resulting bias in the astrometric radial velocity is @xmath223 isotropic expansion ( @xmath224 ) , in particular , gives the bias @xmath225 for a uniformly expanding cluster @xmath226 equals the expansion age , i.e. the time elapsed since all the stars were confined to a very small region of space . gliese w. , jahrei h. , 1991 , preliminary version of the third catalogue of nearby stars , astron . rechen - institut , heidelberg ( available from centre de donnes astronomiques de strasbourg , http://vizier.u-strasbg.fr/ ) nesterov v. , gulyaev a. , kuimov k. , et al . , 1998 , in : mclean b.j . , golombek d.a . , hayes j.j.e . , payne h.e . iau symp . 179 , new horizons from multi - wavelength sky surveys . kluwer , dordrecht , p. 409

our current view of galaxies considers them as systems of stars and gas embedded in extended halos of dark matter , much of it formed by the infall of smaller systems at earlier times . the true extent of a galaxy remains poorly determined , with the virial radius " ( ) providing a characteristic separation between collapsed structures in dynamical equilibrium and external infalling matter . other physical estimates of the extent of gravitational influence include the gravitational radius , gas accretion radius , and galactopause " arising from outflows that stall at 100 - 200 kpc over a range of outflow parameters and confining gas pressures . physical criteria are proposed to define bound structures , including a more realistic definition of for stellar mass and halo mass , half of which formed at assembly redshifts " ranging from . we estimate the extent of bound gas and dark matter around galaxies to be kpc . the new virial radii , with mean kpc , are 40 - 50% smaller than values estimated in recent hst / cos detections of and absorbers around galaxies . in the new formalism , the milky way stellar mass , , would correspond to kpc for half - mass halo assembly at . the frequency per unit redshift of low - redshift absorption lines in qso spectra suggests absorber sizes kpc when related to intervening galaxies . this formalism is intended to clarify semantic differences arising from observations of extended gas in galactic halos , circumgalactic medium ( cgm ) , and filaments of the intergalactic medium ( igm ) . astronomers should refer to _ bound gas _ in the galactic halo or cgm , and _ unbound _ gas at the cgm - igm interface , on its way into the igm .

1401.5799

over the last several decades , with new evidence , the objects we call galaxies " have become much larger . extended dark - matter halos " were proposed to produce flat rotation curves at large radii in disk galaxies ( rubin et al . 1980 ) , and a corona " of hot interstellar gas at the galaxy s virial temperature was predicted by spitzer ( 1956 ) to provide pressure confinement of high - latitude clouds . more recently , astronomers have observed galactic kinematic tracers ( blue horizontal branch stars , globular clusters , satellite galaxies ) to distances of 50 - 250 kpc , and x - ray absorption - line ( ) spectroscopy and stacked soft x - ray emission have provided evidence of large reservoirs of hot ionized gas in milky way halo ( miller & bregman 2013 ) and the outskirts of external galaxies ( soltan 2006 ; anderson et al . 2013 ) . in ultraviolet spectroscopy , the cosmic origins spectrograph ( cos ) on the _ hubble space telescope _ ( hst ) has recently detected extended ( 100 - 150 kpc ) reservoirs of highly ionized oxygen ( ) around star - forming galaxies ( tumlinson et al . 2011 , 2013 ; stocke et al . 2013 ) likely created by outflows of metal - enriched gas from star formation . thus , our picture of galaxies has evolved to a system of stars and gas embedded in an extended dark - matter halo , often associated with an even larger gaseous circumgalactic medium ( cgm ) . galaxies must also be viewed in a cosmological context , in which most of the baryonic matter in the universe is unseen ( persic & salucci 1992 ) and likely distributed through the intergalactic medium ( igm ) in a cosmic web " shaped by dark - matter structure and inefficient galaxy formation ( cen & ostriker 1999 ; dav et al . 1999 ; smith et al . 2011 ) . observations and modeling ( shull et al . 2012 ) suggest that 60 - 80% of the cosmological baryons reside in the low - redshift , multi - phase igm , with perhaps 20% in collapsed form ( galaxies , groups , clusters ) . although these baryon fractions have some uncertainty , current research is devoted to understanding the physical conditions , spatial extent , and evolution of the gas at distances of 100 kpc to a few mpc from galaxies . at low redshift ( @xmath13 ) the influence of galactic winds and metal injection out to @xmath14 mpc has been inferred from the association of qso absorption systems of and with nearby galaxies ( penton et al . 2002 ; prochaska et al . 2011 ; stocke et al . 2006 , 2013 ) . galactic outflows have been detected in absorption toward intermediate - redshift galaxies ( steidel et al . 2004 ; martin et al . 2012 ; tripp et al . 2011 ) , while at higher redshifts , @xmath15 , rudie et al . ( 2012 ) found an enhancement of circumgalactic ( @xmath16 kpc ) absorbers in a sample of 886 star - forming galaxies probed by 15 background qsos . thus , the connection between galaxies and extended absorption systems seems secure . perhaps because of the recent nature of these discoveries , a semantic problem has arisen regarding the proposed structures : halo , cgm , igm . when does gas cease to be circumgalactic and become intergalactic ? are the edges of galaxies defined by gravity or gas outflows ? are quasar absorption lines the extended halos of intervening galaxies ( bahcall & spitzer 1969 ) or filaments of intergalactic gas ( sargent et al . similarly , the phrase circumgalactic medium " appears to have replaced the concept of a gaseous halo " or galactic corona " of hot interstellar gas at the galaxy s virial temperature . ionized gas with high covering factor has been detected above the galactic disk in uv absorption - line surveys in metal ions such as ( sembach et al . 2003 ) and ( shull et al . 2009 ; lehner & howk 2011 ) and in soft x - ray absorption lines of or at @xmath17 ( nicastro et al . 2002 ; mckernan et al . 2005 ; wang et al . 2005 ) . the uv absorbers appear to come primarily from gas within 2 - 10 kpc of the disk plane , elevated by supernovae and star formation in the disk . the galactic x - ray absorption suggests hot gas at @xmath18 k , but its radial extent is controversial . it may come from a 50-kpc halo ( anderson & bregman 2010 ; gupta et al . 2012 ) although absorption toward background agn ( fang et al . 2006 ; hagihara et al . 2010 ) and x - ray binaries ( yao & wang 2005 ; hagihara et al . 2011 ) suggests that much of the absorption comes within several kpc of the disk . the disk model is consistent with both x - ray observations and total mass considerations ( collins et al . 2005 ; fang et al . 2006 ) . the intent of this paper is to improve the definition of galaxy halos as regions of strong gravitational influence , using dynamical principles and observational constraints . section 2 discusses physical measures of the spatial extent of large ( @xmath19 ) galaxies including the milky way and andromeda . we discuss the somewhat arbitrary and occasionally misused definition of virial radius " . somewhat better defined are the gravitational radius " , @xmath20 , derived from galactic mass and potential energy , the gravitational sphere of influence , accretion radius , and tidal radius . section 2.3 develops a physically realistic definition of @xmath21 for halos of mass @xmath3 , assembled primarily at redshifts @xmath4 with further mass accretion down to the present epoch . these new virial radii are typically 50 - 60% the sizes used to analyze hot halo gas with hst / cos . section 3 discusses four estimates of galaxy extent : ( 1 ) recent kinematical studies of the milky way and andromeda ( m31 ) ; ( 2 ) the galactopause " where outflow ram pressure balances thermal pressure of the cgm ; ( 3 ) qso absorber cross sections derived from metal absorption - line frequency in redshift ; and ( 4 ) virial radii and halo masses obtained from galaxy abundance - matching . we also discuss recent estimates of the mass and size of the milky way and andromeda halos from kinematics of stars , galactic satellites , and the local group . most of our estimates suggest a smaller spatial extent ( @xmath6 kpc ) for galaxies of mass @xmath22 , comparable to the milky way and m31 . section 4 concludes with applications of the new definition of virial radius to observations of extended gas around galaxies made with hst / cos , and to recent mass and size measurements for the milky way and m31 . this section begins with an overview of the cosmological context , in which galaxies form from density perturbations that turn around from the hubble expansion background and collapse into dynamical ( virial ) equilibrium , but with continued infall of dark matter and gas . the environment of these structures will be influenced by feedback " from star formation within galaxies , through outflows that escape to the igm or become stalled and return to the galaxy . both infall and outflow processes depend on the gravitational attraction of the galaxy and its halo , as measured by parameters such as escape velocity , accretion radius , and tidal radius . below , these measures of gravitational influence are explored , leading to a practical definition of virial radius , consistent with collapse in the past ( @xmath23 ) with continued mass infall to the present . the classic overdensity @xmath24 ( often rounded to 200 ) was originally derived analytically for a spherical top - hat density perturbation in an einstein - desitter universe . the mean background density , @xmath25 , following the behavior of the expansion parameter @xmath26 . the perturbation begins its collapse at a turn - around " density , @xmath27 , at a redshift @xmath28 . for example , if all the matter in the local group were associated with a spherically distributed mass @xmath3 , the proper size of the mass perturbation at turnaround would be @xmath29 ^{1/3 } \approx ( 1.2~{\rm mpc } ) \left ( \frac { m_{\rm lg } } { 5\times10^{12}~m_{\odot } } \right ) ^{1/3 } \left [ \frac { 1.5 } { 1+z_{\rm turn } } \right ] \ ; .\ ] ] here , we have scaled to the local group mass , @xmath30 at a turnaround redshift @xmath31 , corresponding to the recently measured dynamical history of the milky way and andromeda system ( van der marel et al . these galaxies turned around from the hubble flow even earlier , perhaps at @xmath32 , and had correspondingly smaller radii , probably @xmath33 mpc . the fact that a given galaxy halo is assembled over a range of redshifts suggests an improved definition of virial radius , @xmath34 , for a galaxy observed at redshift @xmath35 but assembled primarily at redshift @xmath36 , when half its halo mass had collapsed . this change in definition increases the background matter density , @xmath37 , and makes the virial radii smaller . we return to this new formalism in sections 2.3 and 3.4 . for extended self - gravitating systems without a sharp boundary , it is useful to define a characteristic gravitational radius " ( binney & tremaine 2008 ) as @xmath38 , where @xmath39 is the gravitational potential energy , @xmath40 the gravitational radius also bears a close relation to the half - light radius " , @xmath41 , defined by spitzer ( 1969 ) for spherical stellar systems . for many mass distributions , @xmath42 , so that @xmath43 and @xmath44 . for example , a homogeneous sphere of mass @xmath45 and radius @xmath46 has @xmath47 and @xmath48 . for the density distribution of plummer s model , @xmath49^{-5/2}$ ] , @xmath50 and @xmath51 , where @xmath52 is the plummer scale length . the structure of collapsed halos has been well studied through cosmological n - body simulations ( navarro , frenk , & white 1997 ; springel et al . 2005 ; klypin et al . these numerical experiments show that dark - matter halos collapse into structures with cuspy cores and extended halos . the collapsed structures in these simulations have been fitted to various radial profiles , such as nfw ( navarro , frenk , & white 1997 ) , for which the density , potential , and enclosed mass are , @xmath53 ^2 } \\ \phi(r ) & = & - \left ( 4 \pi g \rho_0 r_s^2 \right ) \frac { \ln ( 1 + r / r_s ) } { r / r_s } \\ m(r ) & = & \left ( 4 \pi \rho_0 r_s^3 \right ) \left [ \ln \left ( 1 + \frac{r}{r_s } \right ) - \frac { r / r_s}{1 + r / r_s } \right ] \ ; .\end{aligned}\ ] ] here @xmath54 is a characteristic radius defined by the break in slope and related to the virial radius through @xmath55 , with the concentration parameter " , @xmath56 , typically between 520 . the mass enclosed with the virial radius is written as @xmath57 \ ; , \ ] ] where @xmath58 and @xmath56 are found to be correlated in cosmological simulations ( navarro et al . 1997 ; prada et al . 2012 ) . increasing the concentration , @xmath56 , adds mass and extent to the halo . for example , changing @xmath56 from 5 to 20 doubles the enclosed mass , @xmath59 . the circular velocity , @xmath60 , reaches a maximum @xmath61 at radius @xmath62 , and equal to @xmath63 } \ ; .\ ] ] for the nfw model , the mass diverges logarithmically with radius , and the gravitational radius is formally undefined without a cut - off . by integrating equation ( 2 ) out to the virial radius , @xmath64 , we can define a finite nfw potential energy and gravitational radius , @xmath65 } { \left [ \ln ( 1+c ) - \frac { c}{1+c } \right]^2 } \\ r_g ^{\rm ( nfw ) } & = & gm_{\rm vir}^2 / \vert w_{\rm vir } \vert = ( 2 r_s ) \ ; \frac { \left [ \ln ( 1+c ) - \frac { c}{1+c } \right]^2 } { \left [ 1 - \frac { \ln ( 1+c ) } { ( 1+c ) } - \frac { 1}{(1+c ) } \right ] } \ ; .\end{aligned}\ ] ] this nfw gravitational radius ranges from @xmath66 ( for @xmath67 ) to @xmath68 ( for @xmath69 ) . table 1 summarizes the ratio of gravitational radius to characteristic scale lengths for several commonly used density distributions . another observationally useful characteristic radius is the gravitational radius of influence " ( ferrarese & ford 2005 ; binney & tremaine 2008 ) defined as the radius at which the kepler velocity around a central mass @xmath45 equals the transverse velocity dispersion , @xmath70 , of surrounding stars , @xmath71 in the galactic context of this paper , we assume that @xmath45 is the total mass of the galaxy within radius @xmath72 , scaled as @xmath73 . the gravitational influence of this mass can traced by luminous halo stars or galaxy satellites with dispersion @xmath74 . the dynamical dependence of @xmath75 ( observed in the stellar component ) on @xmath45 will soften the linear dependence of @xmath76 . in the milky way , radial velocity dispersions , @xmath77 km s@xmath78 , of different halo populations have been measured out to 80 kpc ( xue et al . 2008 ; gnedin et al . 2010 ) . from radial velocities of blue horizontal branch stars between 100 - 150 kpc , deason et al . ( 2012 ) find even lower dispersions , @xmath79 km s@xmath78 , although values for satellites are somewhat larger . thus , depending on assumptions about the ratio , @xmath80 , the value of @xmath72 for the milky way halo may be somewhat larger than 200 kpc . this formula is analogous to that used in bondi accretion of gas with isothermal sound speed @xmath81 , for which one defines an accretion radius " @xmath82 here again , we have scaled to a @xmath83 galaxy surrounded by a gaseous halo at its virial temperature @xmath84 with @xmath85 for mean atomic weight @xmath86 appropriate for a fully ionized plasma with @xmath87 by number . the fact that these two raidii are similar is a consequence of the chosen gas temperature and stellar velocity dispersion , which are expected to be in near equilibrium . in standard cosmological terminology , the virial radius , @xmath0 , is defined such that the mean density of a halo of mass @xmath3 equals the virial overdensity " ( variously denoted as @xmath88 or @xmath89 ) times the mean density of the universe . unfortunately , some confusion has arisen in the proper value of @xmath88 , much of it over convention : whether the overdensity is expressed relative to the critical ( closure ) density , @xmath90 , or to the matter density , @xmath91 . both conventions are used in the literature , but if one is not careful in using a consistent definition of @xmath88 , the estimated virial radius will be too small by a factor @xmath92 and the mean matter density of the halo too large by @xmath93 . there are other misapplications of the virial radius because the initial collapse to dynamical equilibrium occurred in the past , when @xmath94 was higher , followed by additional infall during galaxy assembly . for these reasons , we develop a physically motivated definition of @xmath95 for a halo of mass @xmath3 , observed at redshift @xmath35 but assembled at an earlier epoch , @xmath36 . the redshift of halo assembly @xmath96 is taken as the epoch of major virialization ( @xmath97 ) when half the halo mass had collapsed , a useful convention introduced by tacchella , trenti , & carillo ( 2013 ) hereafter denoted ttc13 . the basic concept of virialization follows from the assumption of conservation of kinetic energy and potential energy during collapse ( @xmath98 ) together with the virial theorem ( @xmath99 ) . the gravitating system is assumed to collapse to half its turnaround radius at twice the turnaround time for cycloidal collapse . the collapsed system has then increased in mean density by a factor of 8 since turnaround . in an einstein - desitter universe ( @xmath100 ) , the expanding background density , @xmath25 , has dropped by a factor of 4 for an expansion parameter @xmath26 . thus , one obtains the classic virial overdensity , @xmath101 , where @xmath102 and @xmath103 are evaluated at the times of turnaround and virialization , respectively . cole & lacey ( 1996 ) explored the structure of halos in n - body simulations of clustering in an @xmath100 universe . they found that the radius , @xmath104 enclosing a mean overdensity of 178 , accurately separated the virialized halo interior , in approximate dynamic equilibrium , from the exterior where matter was still falling in . this property was noted by binney & tremaine ( 2008 ) , who recommended using @xmath105 . the situation changes in a flat @xmath106cdm cosmology ( @xmath107 ) where both the expanding background and collapse dynamics differ from the @xmath100 case . eke et al . ( 1996 ) studied virialization dynamics for flat models with @xmath108 , but they defined overdensity relative to _ closure _ density . for values of @xmath109 at @xmath110 , they found @xmath111 . in figure 1 of their paper , they plotted the dependence of @xmath89 on @xmath112 , which serves as a proxy for higher redshift , since @xmath113 approaches 1 at the higher redshifts when most systems collapsed . bryan & norman ( 1998 ) provided an approximate fit , @xmath114 $ ] where @xmath115 . this expression must be divided by @xmath113 to be consistent with the convention that overdensity is relative to _ matter _ density . we define @xmath0 through the relation @xmath116 , where the mean matter density at the virialization ( or assembly ) redshift is @xmath117 , and the parameters @xmath118 and @xmath119 are defined at the current epoch ( @xmath120 ) . further discussion is given by klypin et al . ( 2011 ) and van der marel et al . ( 2012 ) , both of whom follow the convention that @xmath88 is relative to the mean _ matter _ density . their expressions are scaled to a product , @xmath121 , evaluated at @xmath120 for which @xmath122 and @xmath123 . however , many papers in the literature do not account for the fact that virial collapse occurred at earlier times , at redshift @xmath97 when @xmath124 . for a flat @xmath106cdm universe with @xmath125 at @xmath110 , one finds @xmath126 ( relative to @xmath94 ) and a product @xmath127 . for the more typical collapse at higher redshifts , the product @xmath128 , appropriate for the decrease in @xmath129 as @xmath130 . this overdensity of 200 is slightly above the classical value @xmath131 for dynamical reasons discussed by binney & tremaine ( 2008 ) . consider a galaxy observed at redshift @xmath35 , associated with halo mass @xmath132 and assembled ( at @xmath96 ) with half its current mass . the virialization redshift @xmath97 is taken to be @xmath96 , determined from standard cosmological collapse criteria ( lacey & cole 1993 ; sheth & tormen 1999 ) and computed as in trenti , perna , & tacchella ( 2013 ) . table 2 lists the assembly redshifts , which range from @xmath133 ( at @xmath134 ) to @xmath135 ( at @xmath136 ) for various galaxies observed at @xmath137 with stellar masses @xmath2 . based on these new criteria for galaxy collapse and assembly , the virial radius is : @xmath138 ^{-1/3 } \left ( 1+z_a \right)^{-1 } \ ; .\ ] ] although the relation between @xmath139 exhibits a small anti - correlation ( table 2 ) , it does contain some numerical scatter which should be kept in mind . this expression differs from often - used formulae by the over - density scaling and by the factor @xmath140 , reflecting the fact that most galaxies underwent virialization in the past , not at @xmath120 . after this initial ( half - mass ) assembly , their proper size changes gradually because of continued mass infall into the halo . to lowest order , this slight growth in @xmath0 ( @xmath141 ) is captured in eq . ( 12 ) by using the _ current _ halo mass ( @xmath3 ) . second - order effects of late infall or adiabatic compression may occur in the radial mass distribution , which in the nfw formalism changes the scale parameter @xmath54 and concentration @xmath56 . some of the infalling matter may not attain virial equilibrium , as the orbital timescale is quite long , @xmath142 . this formalism assumes spherical infall and a constant density profile ; both approximations change in more realistic collapse scenarios . for example , after accounting for central condensation of the collapsed structure , rubin & loeb ( 2013 ) find a reduced virial overdensity . in addition to infall , galaxy halos can be probed by fast stars injected radially outward . we therefore discuss the radius at which the fastest galactic stars reach maximum apogalactic distance in a dark matter halo potential . as a simple model that illustrates the basic effect , we assume an isothermal sphere distribution in density , @xmath143 and enclosed mass , @xmath144 leading to a constant circular velocity @xmath145 . for this density and mass distribution , the equation of motion is @xmath146 , for a star launched with radial velocity @xmath147 from radius @xmath148 , leading to the first integral , @xmath149 = - ( 2 \sigma^2 ) \ln ( r / r_0 ) \ ; .\ ] ] if the star coasts to rest ( @xmath150 at @xmath151 ) one has @xmath152 and thus @xmath153 \ ; .\ ] ] equation ( 14 ) illustrates the sensitivity of @xmath154 to launching conditions ( @xmath155 and @xmath156 ) and halo mass ( velocity dispersion @xmath75 ) . for example , a star with @xmath157 km s@xmath78 injected radially from @xmath158 kpc into a halo with @xmath159 km s@xmath78 would have a maximum extent of just 40 kpc . the stellar ejection distance doubles ( @xmath160 kpc ) if the launch velocity @xmath161 km s@xmath78 . these distances are not surprising for velocities comparable to the escape speed from the solar circle within the milky way ( @xmath162 kpc ) , recently estimated from rave stellar survey data to be @xmath163 km s@xmath78 ( piffl et al . 2014 ) . repeating the above calculation for the maximum radius in an nfw mass distribution and potential ( equations 3 , 4 , 5 ) we find , @xmath164 } \left [ \frac { \ln \left ( 1 + r_0 / r_s \right ) } { r_0 / r_s } - \frac { \ln \left ( 1 + r_{\rm max } / r_s \right ) } { r_{\rm max } / r_s } \right ] \ ; .\ ] ] this equation can be solved for the velocity @xmath155 required for a star to reach radius @xmath165 from initial radius @xmath156 . combined with equation ( 7 ) for the maximum circular velocity , @xmath61 , in an nfw halo , we can write the velocity @xmath155 needed to escape to @xmath166 as @xmath167 ^{1/2 } \ ; , \ ] ] independent of @xmath56 . if we take @xmath168 as the radius at @xmath61 , we find @xmath169 . this agrees reasonably well with the milky way values , if we assume @xmath170 km s@xmath78 and obtain an escape velocity @xmath171 km s@xmath78 . a related size estimate comes from the hypothetical radius , @xmath172 , to which one must extend an isothermal density distribution to maintain a constant ( flat ) circular velocity , @xmath173 \ ; .\ ] ] of course , rotation curves do not remain flat at large radii , but @xmath172 provides a lower limit on the extent of the isothermal model , as noted by smith et al . ( 2007 ) . for @xmath163 km s@xmath78 and @xmath174 km s@xmath78 , one finds a truncation radius @xmath175 kpc for @xmath158 kpc , with a large range ( 40 to 110 kpc ) arising from uncertainties in determining @xmath176 and @xmath177 . at larger radii , the circular velocity , @xmath178 likely declines , particularly if the halo density distribution falls off as @xmath179 as in the nfw model . the total ( virial ) masses of galaxies include substantial mass beyond 50 - 80 kpc , particularly for halos with @xmath180 . extra mass accretion may account for the discrepancy between low and high values for the milky way mass estimates . it is also consistent with the physically motivated revision to the method in which the virial radius is calculated at the galaxy assembly time " at redshift @xmath36 ( section 2.3 ) . we return to these observational issues in section 3.4 . with the current state of observations , these kinematic constraints suggest halo sizes of 150 - 200 kpc for @xmath5 galaxies . future observations of rotation curves , @xmath178 , and more distant kinematic tracers ( blue horizontal branch stars , hypervelocity stars , globular clusters , and satellites ) will be of great help . the rave survey data will be greatly improved by the _ gaia _ mission , leading to better determinations of escape velocity from the solar circle , and later with lsst , through studies of galaxy satellites , bhb stars , and rr - lyrae stars . despite their proximity , the milky way ( mw ) and andromeda ( m31 ) galaxies remain surprisingly uncertain in their inferred mass and extent . recent mw mass estimates range from just under @xmath83 ( smith et al . 2007 ; xue et al . 2008 ; gnedin et al . 2010 ; deason et al . 2012 ) to values between @xmath181 ( boylan - kolchin et al . 2013 ; nesti & salucci 2013 ) . some of the lower masses are inferred for stellar tracer distances well within @xmath0 . for example , gnedin et al . ( 2010 ) use hypervelocity stars at distances beyond 25 kpc to find @xmath182 . xue et al . ( 2008 ) use 2400 blue horizontal branch stars to estimate @xmath183 . however , they extrapolate to find a virial mass @xmath184 for two models of the circular velocity curve . busha et al . ( 2011 ) use lmc and smc kinematics to infer @xmath185 . boylan - kolchin et al . ( 2013 ) provide a thorough discussion of the previous mw mass measurements , both large and small , but they base their mass on the high probability that the leo i satellite is gravitationally bound . for a leo i distance ( @xmath186 kpc ) and galactocentric space velocity ( 200 km s@xmath78 ) , they find a median milky way virial mass @xmath187 with an asymmetric 90% confidence interval of @xmath188\times10^{12}~m_{\odot}$ ] and @xmath189 kpc for the median mass . nesti & salucci ( 2013 ) use dark - matter halo models to estimate @xmath190 but with substantial uncertainties for both burkert and nfw density profiles . piffl et al . ( 2014 ) use the inferred escape speed at the solar circle , @xmath191 km s@xmath78 from the rave survey , to estimate @xmath192 and a virial radius @xmath193 kpc . watkins et al . ( 2010 ) find @xmath194 within 300 kpc from 26 satellite galaxies , and mcmillan ( 2011 ) proposes a mass model with @xmath195 . based on the recent dynamical estimates out to larger distances , we conclude that the milky way virial mass is likely close to @xmath196 . mass estimates for the andromeda ( m31 ) galaxy include @xmath197 within 300 kpc from 23 satellite galaxies ( watkins et al . 2010 ) and @xmath198 within 200 kpc from kinematics of 53 globular clusters ( veljanoski et al . tamm et al . ( 2012 ) used photometry and the m31 rotation curve , together with several mass distributions , to find a dark - matter halo mass @xmath199 , stellar mass @xmath200 , and virial radius 189 - 213 kpc . fardal et al . ( 2013 ) used bayesian simulation models of the giant southern stream around andromeda to derive @xmath201 . they note that this mass , and those recently inferred for the milky way , alleviate the tension between virial mass estimates and the local group timing mass , recently estimated at @xmath202 ( van der marel et al . 2012 ) . because both the milky way and andromeda halos have likely accreted substantial mass beyond their observable radii ( 60 - 80 kpc ) , their virial masses are taken here to be @xmath203 and @xmath204 , with a sum , @xmath205 , close to the inferred timing mass of the local group . simulations suggest that the remaining mass may consist of diffuse dark matter " produced by the destruction of sub - halos ( gao et al . cosmic scatter could reduce the local group mass from the value above , as has been suggested by van der marel et al . ( 2012 ) and gonzalez , kravtsov , & gnedin ( 2013 ) . another measure of gravitational influence is from tidal effects between nearby galaxies . we define the characteristic tidal distance in terms of masses and densities of the two galaxies , treated as deformable fluid spheres ( chandrasekhar 1963 ) of masses ( @xmath206 , @xmath207 ) and radii ( @xmath208 , @xmath209 ) , @xmath210 in the case of the milky way ( mw ) and andromeda ( m31 ) system , internal tidal influences are seen in the form of large - scale streams of stars ( fardal et al . 2013 ; ibata et al . 2014 ) , but no obvious tails or bridges produced by external tides are detected . ibata et al . ( 2014 ) find a smooth stellar - light halo in m31 out to 150 kpc , with internal streams from tidally stripped satellites . usually the roche formula is applied to situations with @xmath211 . however , in the case of the mw - m31 system , we assume @xmath212 to estimate that @xmath213 for each galaxy . here , @xmath208 can be regarded as the effective radius of either galaxy , each of mass @xmath214 . if the distance between the two galaxy centers is @xmath215 kpc ( van der marel et al . 2012 ) , the absence of strong mutual tidal effects limits their tidal radius to @xmath216 . equation ( 18 ) then implies that most of the stellar and gaseous matter in the two galaxies is confined within radius @xmath217 kpc . the fluid approximations in making this estimate are large , and more accurate , time - dependent models and searches for external tidal effects would be helpful to set a better limit . many galaxies undergoing active star formation have been spectroscopically observed at @xmath218 ( steidel et al . 2004 ) to have large - scale outflows with bulk velocities of 200 - 300 km s@xmath78 . for a sample of over 200 star - forming galaxies at redshifts @xmath219 , blueshifted absorption at greater than 100 km s@xmath78 is seen toward 20% of the population ( martin et al . because the outflows are collimated , the outflow fraction " must be signiifcantly higher than the blueshifted fraction " . some 2.5% have velocities greater than 200 km s@xmath78 , with the largest doppler components suggesting velocities up to 500 km s@xmath78 . the outflow fraction depends on the observed star - formation rate ( sfr ) and is three times higher for galaxies with sfr @xmath220 compared to those at @xmath221 . if these outflows persist at the observed rates for @xmath14 gyr , as martin et al . ( 2012 ) suggest , they could remove a large mass of baryons , up to @xmath222 . such metal - enriched outflows could be responsible for the large baryon reservoirs of gas and metals around star - forming galaxies observed in absorption as part of the cos - halos project . tumlinson et al . ( 2011 , 2013 ) found substantial column densities , @xmath223 @xmath224 out to distances of 100 - 150 kpc , toward a sub - sample of 28 galaxies with active star formation . these -bearing halos surround galaxies with sfrs of @xmath225 yr@xmath78 and specific sfrs greater than @xmath226 yr@xmath78 . how far will these galactic winds expand before stalling ? several simulations ( furlanetto & loeb 2003 ; oppenheimer & dav 2008 ) estimate that these outflows expand to @xmath227 kpc , distances inferred observationally in metal - line auto - correlation of absorbers ( martin et al . a simple estimate illustrates the main effects , as winds encounter surrounding gas in the cgm and produce a galactopause " where the wind ram pressure , @xmath228 , equals the confining gas pressure of the cgm , @xmath229 . here , @xmath230 is the mass - loss rate of a conical outflow into total solid angle @xmath231 , @xmath232 is the mass density , and @xmath233 is the wind velocity . observers and modelers typically relate @xmath234 to the sfr through a mass - loading factor " , defined here as @xmath235 . this mass loading has been inferred from optical and x - ray data to lie in the range @xmath236 ( strickland & heckman 2009 ) . in a careful study of the dwarf starburst galaxy ngc 1569 , martin , kobulnicky , & heckman ( 2002 ) use observations of the sfr , h@xmath237 emission , metallicities , and x - ray emission from the outflow to estimate @xmath238 for the mass ratio of x - ray emitting wind to stellar ejecta . with standard assumptions about the ratio of ejecta mass to star formation , this range corresponds to mass loading factors @xmath239 . for the cgm gas pressure in low - redshift , actively star - forming galaxies , we use the observed column densities ( @xmath240 @xmath224 ) and spatial impact parameters ( 100 - 150 kpc ) from cos - halos data , which suggest total hydrogen column densities @xmath241 @xmath224 for solar oxygen abundances , ( o / h ) @xmath242 , and ionization fractions @xmath243 . stocke et al . ( 2013 ) estimated a mean pressure @xmath244 @xmath245 k ( with a range from 3 to 60 @xmath245 k ) for warm clouds in the cgm of low - redshift galaxies . using x - ray absorption through the galactic halo , miller & bregman ( 2013 ) modeled thermal pressures @xmath246 $ ] @xmath245 k for halos with characteristic scales of [ 50 , 100 ] kpc and gas temperature @xmath247 . thus , in the inner portions of the cgm that would initially confine the outflows , the gas pressure is scaled to @xmath248 , appropriate for hydrogen number density @xmath249 , halo virial temperature @xmath250 , and fully ionized gas with he / h @xmath251 by number . for wind speeds @xmath252 , the approximate radius of the galactopause is @xmath253 the combination of parameters ( @xmath254 ) is expected to be of order unity , while bipolar wind solid angles @xmath231 typically cover 10 - 40% of @xmath255 sterradians for starburst galaxies ( veilleux , cecil , & bland - hawthorn 2005 ) . if the galaxy wind breaks through the higher - density cgm , the igm pressure will be considerably lower , and the metal - enriched winds will expand well beyond 200 kpc , depending on the duration of the starburst outflow . simulations of the low - redshift igm ( e.g. , smith et al . 2011 ) find a wide range of intergalactic gas pressures in the ly@xmath237 forest ( @xmath256 ) and hotter -bearing gas ( @xmath257 ) . in the above estimate , the star - formation rates were scaled to @xmath258 . in the sample of active star - forming galaxies at @xmath259 ( martin et al . 2012 ) , the observed sfrs range from 20 - 100 @xmath260 , while in the sample of 28 active cos - halos galaxies at @xmath261 ( tumlinson et al . 2013 ) the sfrs are 110 @xmath260 . these rates likely reflect the general decline in sfr history among all galaxies , from @xmath262 to the present epoch . indirect estimates of the sizes of extended gaseous halos and metal - enriched outflows around galaxies can be made from observations of qso absorption lines . in a pioneering paper , bahcall & spitzer ( 1969 ) proposed that these absorbers are produced in extended halos of normal galaxies " and used the line frequency , @xmath263 , the number of absorbers per unit redshift , to constrain the halo cross section . their estimate , @xmath264 kpc , for the radius of a spherical halo is comparable to recent estimates from uv spectra of low - redshift qso absorbers obtained from hst ( shull et al . 1996 ; stocke et al . 2006 ; prochaska et al . 2011 ) . those uv studies related line frequency of absorbers in ( ly@xmath237 ) or the doublet ( 1032 , 1038 ) to the absorber cross section and space density , @xmath265 , of galaxies , using a luminosity function to extrapolate to dwarf galaxy scales . in this formalism , the igm absorbers are associated with the nearest - neighbor galaxy , a procedure expected to be correct statistically , even though some absorbers have alternate possible galaxies within 200600 kpc ( stocke et al.2013 ) . at redshifts @xmath13 and ignoring cosmological effects , the number of of absorbers per unit redshift is @xmath266 , from which one infers an absorber size @xmath267 ^{1/2 } \approx \left ( 157~{\rm kpc } \right ) \left [ \frac { d{\cal n } / dz } { 20 } \right]^{1/2 } \left [ \frac { \phi_{\rm gal } } { 0.06~{\rm mpc}^{-3 } } \right ] ^{-1/2 } \\ \ ; .\ ] ] this formula uses the observed @xmath268 down to 10 m equivalent width for the 1032 absorption line in low - redshift absorber surveys ( danforth & shull 2008 ; tilton et al . we adopt a hubble constant @xmath269 , a hubble length @xmath270 , and a galaxy luminosity function , @xmath271 with normalization @xmath272 at b - band absolute magnitude @xmath273 and faint - end slope @xmath274 given by the millennium galaxy catalog ( driver et al . 2005 ) and converted to @xmath275 scale . the galaxy space density is scaled to 0.06 mpc@xmath276 , a value appropriate for extrapolation of @xmath277 down to dwarf galaxies of luminosity @xmath278 , which have @xmath279 times the space density of galaxies with @xmath280 for @xmath281 . these sizes are consistent with surveys that attempt to match the nearest - neighbor galaxy , in 3d space , to low - redshift absorbers ( stocke et al . 2006 , 2013 ) . the galaxy spatial distributions in the surveys suggest that the nearest galaxies to absorbers lie at distances @xmath14 mpc from an @xmath5 galaxy and at @xmath6 kpc from a galaxy at @xmath12 . observers often lack direct measurements of galaxy masses . instead , they estimate halo mass , @xmath282 , from the inferred stellar mass @xmath2 or galaxy luminosity @xmath283 ( shankar et al . 2006 ) using the technique of abundance matching " ( behroozi et al . 2010 ; moster et al . 2010 , 2013 ) . by matching the abundance of dark - matter halos in simulations to observed galaxy distributions in stellar mass and luminosity , they attempt to characterize the relationship between the stellar masses of galaxies and the masses of the dark matter halos in which they live . moster et al . ( 2010 ) obtain a parameterized stellar - to - halo mass relation by populating halos and subhalos in an n - body simulation with galaxies and requiring that the observed stellar mass function be reproduced . in a different approach , tacchella , trenti , & carollo ( ttc13 ) assume that the rest - frame uv luminosity and stellar mass of a galaxy are related to its dark - matter halo assembly and gas infall rate . galaxies are assumed to experience a burst of star formation at the halo assembly time ( redshift @xmath284 ) followed by a constant star formation rate sustained by steady gas accretion that provides the dominant contribution to the uv luminosity at all redshifts . the model is calibrated by constructing a galaxy luminosity versus halo mass relation at @xmath285 via abundance matching . after this luminosity calibration , the model naturally fits the @xmath285 stellar mass function and correctly predicts the evolution of both luminosity and stellar mass functions from @xmath286 to @xmath110 . generically , halo - matching methods have many similarities , including the flattening at high @xmath2 of the relation that maps stellar mass @xmath2 into halo mass @xmath3 . the similar relation between luminosity @xmath283 of the central galaxy and the dark - matter halo has been explained ( cooray & milosavljevic 2005 ) as the result of a decline in merging efficiency for the accretion of satellites . flattening of the @xmath2 curve at high @xmath3 predicts that ( @xmath110 ) galaxies with @xmath287 reside in very massive halos , @xmath288 , although errors in @xmath2 ( or @xmath283 ) can create large uncertainties in the inferred @xmath3 . a more subtle effect is that scatter in @xmath2 is amplified asymmetrically in @xmath3 ; simply adopting the mean relation for @xmath289 leads to errors ( behroozi et al . various halo - matching methods differ in specific details , including the redshift evolution of the ratio @xmath290 . here , we adopt the ttc13 formalism , which has a clear physical basis for mass assembly and rest - frame uv luminosity evolution . the cos - halos study ( tumlinson et al . 2013 ; werk et al . 2013 ) used the moster et al . ( 2010 ) technique , whereas this paper adopts the ttc13 formalism . both methods find that galaxies with @xmath291 correspond to large halo masses . however , as seen in table 3 , there are differences between the two methods in the estimated halo masses and virial radii of the 44 galaxies in the cos - halos sample . the inferred values of @xmath3 and @xmath0 are particularly uncertain for galaxies with @xmath291 . comparing values in columns ( 5)(8 ) , one sees that the new method gives virial radii smaller by factors 0.4 - 0.8 . statistically , the ratio of mean virial radii computed with the two methods is @xmath292 , with the largest change occurring for lower - mass halos , @xmath293 ( all masses are expressed in @xmath294 units ) . table 4 summarizes the statistical differences for the 44 galaxies in this sample , including subsets of galaxies at low , moderate , and high stellar masses . this decrease in @xmath0 occurs primarily because we evaluate the galaxy collapse and virialization at overdensity , @xmath88 , at the redshifts of galaxy assembly , @xmath295 , rather than at the observed redshifts @xmath137 . in addition , the conversion from @xmath2 to @xmath3 using modeling or abundance matching differs between methods , particularly in its dependence on redshift . another difference is that we define @xmath88 , relative to the _ matter _ density rather than the _ critical _ density . in a few instances with @xmath291 and large stellar masses , @xmath296 , we find large virial radii , @xmath297 kpc , and anomalously large massive halos , @xmath298 . this result occurs because halo - matching algorithms and simulated luminosity functions find large stellar masses to be rare and associate them with very massive halos . applied to the large local group galaxies , halo - matching gives high values for the milky way ( @xmath299 and @xmath9 kpc for @xmath300 ) and m31 ( @xmath301 and @xmath302 kpc for @xmath303 ) . unfortunately for cgm studies , halo - matching for @xmath291 ( high @xmath2 ) galaxies remains an uncertain technique , as demonstrated in figure 1 of stocke et al . this paper reviews current observations and inferences about the spatial extent of galaxies , and the extent of their gravitational influence . these included several dynamical measures : the gravitational radius " @xmath38 , the radius of influence " @xmath304 , and the accretion radius " @xmath305 . clearly , these definitions are related for systems near dynamical equilibrium in the gravitational potential of the galactic halo . for example , in section 2.1 and table 1 , we found a ratio @xmath306 for nfw halos with concentrations ranging from @xmath69 to @xmath67 . various estimates of radial extent from rotation curves and injection of fast stars into extended halos suggest 150 - 200 kpc extents for halo masses of @xmath83 . kinematic observations of the milky way and m31 give masses @xmath307 within 60 - 80 kpc . the inferred galactic escape velocity , @xmath163 km s@xmath78 , together with the local circular velocity @xmath308 km s@xmath78 , is consistent with halos extending beyond 80 - 100 kpc . somewhat larger halo masses , @xmath309 and @xmath310 ( all masses in @xmath294 units ) have been inferred for the milky way and m31 out to larger radii , using globular clusters , satellite galaxies , and stellar streams . their mass sum , @xmath311 , is comparable to recent estimates of the timing mass of the local group , @xmath312 ( van der marel 2012 ) , although a recent study ( gonzalez et al . 2013 ) used cosmological simulations of halo pairs and a likelihood correction to find a lower value , @xmath313 . for either of these masses , the milky way and andromeda are likely to have radial extents @xmath6 kpc . the answer to the question posed in the title to this paper depends on how one wishes to define or observe the edge of a galaxy " . does a galaxy end because of gravity or outflows ? this paper suggests that the best definition of @xmath0 is in terms of gravitational binding . dark matter halos have very large extents , but galactic outflows can drive gas out of the stellar disk , into the ( bound ) halo , and often into the ( unbound ) cgm and igm . we have formulated a new definition of virial radius in this spirit , based on galaxy assembly in the past , when the background density was higher . as a result , we find virial radii smaller by factors of 0.5 - 0.6 , with important consequences for assessing whether the extended absorption seen with cos is bound halo gas , or unbound gas on its way to the igm . * a revised formulation of virial radius , @xmath314 , expresses the critical overdensity @xmath88 relative to the background matter density , @xmath315 , evaluated at @xmath96 , the redshift of galaxy assembly when half the halo mass had collapsed . here , @xmath96 depends on @xmath3 and is inferred from stellar mass @xmath2 . for dynamical consistency , the circumgalactic medium of large galaxies is defined by their gravitationally bound regions , estimated from the virial radius or gravitational radius , @xmath316 . * for galaxies with halo mass @xmath317 observed by hst / cos at redshifts @xmath318 , the assembly redshifts are @xmath319 . because virialization was largely determined by early collapse , when the background density was higher , we adopt @xmath320 and @xmath0 is reduced by a factor @xmath321 compared to @xmath110 . * halo masses can be estimated from stellar masses by halo - matching based on a physically motivated model ( ttc13 ) for galaxy assembly consistent with rest - frame uv luminosity functions from @xmath322 . applied to 44 cos - halos galaxies , this formalism yields virial radii smaller by factors of 0.5 - 0.6 . galaxies with @xmath323 ( @xmath324 ) have @xmath325 kpc , consistent with their observed radial extent . * if confined by cgm gas pressure , the wind galactopause is predicted to occur at distances @xmath326 kpc for sfr = 10 - 30 @xmath294 yr@xmath78 , mass - loading factors @xmath327 , and the observed range of cgm gas pressures and wind velocities . stronger galaxy winds could break out into the lower density igm , with radial extents beyond 200 kpc determined by the strength , collimation , and duration of the outflow . * for dynamical masses of the large local group galaxies , we adopt : milky way ( @xmath328 ) and m31 ( @xmath329 ) with ( gravitational ) radial extents of 150 - 200 kpc . in our new formalism , the milky way stellar mass , @xmath8 , would correspond to somewhat higher halo mass @xmath299 and virial radiius @xmath9 kpc for assembly at @xmath10 . the inferred @xmath3 for m31 is anomalously high ( @xmath301 ) , which suggests caution in using halo - matching to obtain @xmath3 and @xmath0 for massive , luminous galaxies ( @xmath291 ) . as noted earlier , the new definition of @xmath0 is important in deciding whether gas in the vicinity of galaxies is gravitationally bound or unbound . whether galaxies are open or closed boxes " helps to determine the metal evolution of galaxies and the extent of metal - pollution of the igm . it also suggests that astronomers should be careful in making interchangeable use of the terms cgm and halo . matter in the galactic halo is _ gravitationally bound _ , while the circumgalactic ( gaseous ) medium may in fact be _ unbound outflows _ that will merge into the igm . what many observers are calling the cgm is probably gas at the cgm igm interface . in observations of large metal - enriched reservoirs around star - forming galaxies ( tumlinson et al . 2013 ; stocke et al . 2013 ) , the radial offset distances @xmath46 of the qso sight lines were normalized to @xmath0 , with much of the metal - enriched gas within @xmath330 . with the revised ( smaller ) virial radii ( table 3 and figure 1 ) some of this metal - enriched gas extends beyond the region of strong gravitational influence , probably on its way out to the igm . this trend is seen in the 28 actively star - forming galaxies in the cos - halos sample , whereas the undetected around 16 passive galaxies all had sight lines with @xmath330 . in future work , both data sets will be examined carefully with the new formalism . i thank crystal martin , michele trenti , massimo ricotti , john stocke , brian keeney , blair savage , and vasily belokurov for helpful discussions , and evan tilton and joshua moloney for comments on the manuscript . this research was supported by the astrophysical theory program at the university of colorado boulder ( grant nnx07-ag77 g from nasa ) . i am grateful to the institute of astronomy at cambridge university for their stimulating scientific environment and support through the sackler visitor program . lccc uniform sphere & @xmath331 & @xmath46 & @xmath332 + plummer model & @xmath333^{-5/2}$ ] & @xmath52 & @xmath334 + jaffe model & @xmath335 & @xmath336 & @xmath337 + hernquist model & @xmath338 & @xmath336 & @xmath339 + nfw model ( @xmath340 ) & @xmath341 & @xmath54 & @xmath342 + nfw model ( @xmath343 ) & @xmath341 & @xmath54 & @xmath344 + nfw model ( @xmath345 ) & @xmath341 & @xmath54 & @xmath346 + nfw model ( @xmath347 ) & @xmath341 & @xmath54 & @xmath348 + 9.00 & 1.709 & & & 12.25 & 1.112 + 9.25 & 1.666 & & & 12.50 & 1.070 + 9.50 & 1.620 & & & 12.75 & 1.026 + 9.75 & 1.572 & & & 13.00 & 0.979 + 10.00 & 1.531 & & & 13.25 & 0.936 + 10.25 & 1.486 & & & 13.50 & 0.891 + 10.50 & 1.439 & & & 13.75 & 0.844 + 10.75 & 1.394 & & & 14.00 & 0.806 + 11.00 & 1.346 & & & 14.25 & 0.765 + 11.25 & 1.302 & & & 14.50 & 0.722 + 11.50 & 1.255 & & & 14.75 & 0.684 + 11.75 & 1.216 & & & 15.00 & 0.645 + 12.00 & 1.155 & & lccccccrrr j0226 + 0015(268,22 ) & 0.22744 & 0.51 & 10.8 & 12.62 & 12.73 & 303 & 178 & 1.03 & 80 + j0401 - 0540 ( 67,24 ) & 0.21969 & 0.37 & 10.2 & 12.08 & 11.86 & 200 & 85 & 1.19 & 86 + j0803 + 4332(306,20 ) & 0.25347 & 1.63 & 11.3 & 13.48 & 13.84 & 581 & 462 & 0.83 & 79 + j0910 + 1014 ( 34,46 ) & 0.14274 & 0.60 & 10.6 & 12.48 & 12.39 & 279 & 133 & 1.09 & 116 + j0910 + 1014(242,34 ) & 0.26412 & 2.16 & 11.5 & 13.76 & 14.40 & 716 & 747 & 0.74 & 139 + j0914 + 2823 ( 41,27 ) & 0.24431 & 0.37 & 9.8 & 11.87 & 11.48 & 169 & 61 & 1.26 & 104 + j0925 + 4004(196,22 ) & 0.24745 & 1.48 & 11.3 & 13.45 & 13.84 & 569 & 462 & 0.83 & 86 + j0928 + 6025(110,35 ) & 0.15400 & 0.52 & 10.8 & 12.65 & 12.73 & 317 & 178 & 1.03 & 94 + j0935 + 0204 ( 15,28 ) & 0.26228 & 0.79 & 11.0 & 12.88 & 13.13 & 365 & 251 & 0.95 & 114 + j0943 + 0531(106,34 ) & 0.22839 & 0.74 & 10.8 & 12.61 & 12.73 & 300 & 178 & 1.03 & 125 + j0943 + 0531(216,61 ) & 0.14311 & 0.72 & 11.0 & 12.80 & 13.13 & 382 & 251 & 0.95 & 154 + j0943 + 0531(227,19 ) & 0.35295 & 0.24 & 9.6 & 11.66 & 11.31 & 141 & 69 & 1.29 & 95 + j0950 + 4831(177,27 ) & 0.21194 & 1.44 & 11.2 & 13.30 & 13.59 & 511 & 371 & 0.88 & 94 + j1009 + 0713(204,17 ) & 0.22784 & 0.28 & 9.9 & 11.90 & 11.56 & 174 & 66 & 1.24 & 62 + j1009 + 0713(170,9 ) & 0.35569 & 0.29 & 10.3 & 12.06 & 11.98 & 189 & 94 & 1.16 & 45 + j1016 + 4706(274,6 ) & 0.25195 & 0.23 & 10.2 & 12.10 & 11.86 & 202 & 85 & 1.19 & 24 + j1016 + 4706(359,16 ) & 0.16614 & 0.44 & 10.5 & 12.35 & 12.24 & 251 & 117 & 1.12 & 46 + j1112 + 3539(236,14 ) & 0.24670 & 0.52 & 10.3 & 12.17 & 11.98 & 214 & 94 & 1.16 & 54 + j1133 + 0327(110,5 ) & 0.23670 & 1.94 & 11.2 & 13.19 & 13.59 & 515 & 371 & 0.88 & 14 + j1133 + 0327(164,21 ) & 0.15449 & 0.19 & 10.1 & 12.09 & 11.76 & 206 & 78 & 1.21 & 56 + j1157 - 0022(230,7 ) & 0.16378 & 0.55 & 10.9 & 12.72 & 12.91 & 334 & 207 & 1.00 & 20 + j1220 + 3853(225,38 ) & 0.27371 & 0.66 & 10.8 & 12.53 & 12.73 & 279 & 178 & 1.03 & 159 + j1233 + 4758(94,38 ) & 0.22210 & 0.70 & 10.8 & 12.58 & 12.73 & 295 & 178 & 1.03 & 137 + j1233 - 0031(168,7 ) & 0.31850 & 0.37 & 10.6 & 12.30 & 12.39 & 230 & 133 & 1.09 & 33 + j1241 + 5721(199,6 ) & 0.20526 & 0.20 & 10.2 & 12.10 & 11.86 & 205 & 85 & 1.19 & 21 + j1241 + 5721(208,27 ) & 0.21780 & 0.20 & 10.1 & 12.02 & 11.76 & 192 & 78 & 1.21 & 92 + j1245 + 3356(236,36 ) & 0.19248 & 0.20 & 9.9 & 11.81 & 11.56 & 178 & 66 & 1.24 & 117 + j1322 + 4645(349,11 ) & 0.21418 & 0.58 & 10.8 & 12.50 & 12.73 & 303 & 178 & 1.03 & 39 + j1330 + 2813(289,28 ) & 0.19236 & 0.24 & 10.3 & 12.22 & 11.98 & 225 & 94 & 1.16 & 91 + j1342 - 0053(157,10 ) & 0.22702 & 1.08 & 11.0 & 12.67 & 13.13 & 345 & 251 & 0.95 & 37 + j1342 - 0053(77,10 ) & 0.20127 & 0.25 & 10.5 & 12.34 & 12.24 & 247 & 117 & 1.12 & 34 + j1419 + 4207(132,30 ) & 0.17925 & 0.55 & 10.6 & 12.46 & 12.39 & 272 & 133 & 1.09 & 92 + j1435 + 3604(126,21 ) & 0.26226 & 0.33 & 10.4 & 12.20 & 12.01 & 218 & 97 & 1.15 & 86 + j1435 + 3604(68,12 ) & 0.20237 & 1.37 & 11.1 & 13.08 & 13.35 & 433 & 302 & 0.92 & 40 + j1437 + 5045(317,38 ) & 0.24600 & 0.46 & 10.2 & 12.06 & 11.86 & 196 & 85 & 1.19 & 149 + j1445 + 3428(232,33 ) & 0.21764 & 0.29 & 10.4 & 12.26 & 12.01 & 230 & 97 & 1.15 & 117 + j1514 + 3619(287,14 ) & 0.21223 & 0.18 & 9.7 & 11.81 & 11.39 & 164 & 57 & 1.28 & 49 + j1550 + 4001(197,23 ) & 0.31247 & 1.78 & 11.4 & 13.50 & 14.10 & 578 & 577 & 0.79 & 106 + j1550 + 4001(97,33 ) & 0.32179 & 0.82 & 10.9 & 12.54 & 12.91 & 311 & 207 & 1.00 & 155 + j1555 + 3628(88,11 ) & 0.18930 & 0.54 & 10.5 & 12.38 & 12.24 & 254 & 117 & 1.12 & 35 + j1617 + 0638(253,39 ) & 0.15258 & 2.65 & 11.5 & 14.03 & 14.40 & 912 & 747 & 0.74 & 104 + j1619 + 3342(113,40 ) & 0.14137 & 0.19 & 10.1 & 12.12 & 11.76 & 211 & 78 & 1.21 & 100 + j2257 + 1340(270,40 ) & 0.17675 & 0.67 & 10.9 & 12.78 & 12.91 & 348 & 207 & 1.00 & 120 + j2345 - 0059(356,12 ) & 0.25389 & 0.80 & 10.9 & 12.63 & 12.91 & 304 & 207 & 1.00 & 48 + all galaxies & 44 & 10.61 & 12.53 & 12.55 & 315 & 200 & 0.563 + @xmath349 & 12 & 10.23 & 12.12 & 11.89 & 207 & 81 & 0.422 + @xmath350 & 10 & 10.84 & 12.62 & 12.80 & 309 & 190 & 0.613 + @xmath351 & 7 & 11.34 & 13.53 & 13.97 & 626 & 534 & 0.845 +

one key goal of the _ hubble space telescope _ cosmic assembly near - infrared deep extragalactic legacy survey is to track galaxy evolution back to . its two - tiered `` wide and deep '' strategy bridges significant gaps in existing near - infrared surveys . here we report on galaxy candidates selected as f105w - band dropouts in one of its deep fields , which covers 50.1 arcmin to 4 ks depth in each of three near - infrared bands in the great observatories origins deep survey southern field . two of our candidates have mag , and are mag brighter than any previously known f105w - dropouts . we derive constraints on the bright - end of the rest - frame ultraviolet luminosity function of galaxies at , and show that the number density of such very bright objects is higher than expected from the previous schechter luminosity function estimates at this redshift . another two candidates are securely detected in _ spitzer _ infrared array camera images , which are the first such individual detections at . their derived stellar masses are on the order of a few , from which we obtain the first measurement of the high - mass end of the galaxy stellar mass function at . the high number density of very luminous and very massive galaxies at , if real , could imply a large stellar - to - halo mass ratio and an efficient conversion of baryons to stars at such an early time .

1112.6406

in the recent years , deep near - infrared imaging surveys have begun to yield a significant number of candidate galaxies at very high redshifts . the wide - field surveys from the ground have produced a handful of bright candidates at @xmath6 ( e.g. , ouchi et al . 2009 ; hickey et al . 2010 ; castellano et al . 2010 ; capak et al . 2011 ; hsieh et al . 2012 ; hathi et al . 2012 ) , while the pencil - beam survey by the _ hubble space telescope _ ( hst ) wide field camera 3 ( wfc3 ) within the historical advanced camera for surveys ( acs ) hubble ultra deep field ( hereafter acs - hudf ; beckwith et al . 2006 ) and its two parallel fields ( hudf09 ; pi . illingworth ) have probed the fainter populations at @xmath68 ( oesch et al . 2010 ; bouwens et al . 2010 , b10 ; bunker et al . 2010 ; mclure et al . 2010 , m10 ; yan et al . 2010 , y10 ; finkelstein et al . 2010 ) and possibly out to @xmath7 ( yan et al . 2010 ; wyithe et al . 2011 ; bouwens et al . the wfc3 early release science program ( ers , pi . oconnell ; windhorst et al . 2011 ) has played an important role in connecting the `` wide '' and the `` deep '' ends of the exploration , allowing construction of @xmath68 samples at intermediate brightness levels ( wilkins et al . 2011 ; b11 ; lorenzoni et al . 2011 , l11 ) . meanwhile , the hubble infrared pure parallel imaging extragalactic survey ( hippies ; yan et al . 2011 ) and the brightest of reionizing galaxies survey ( trenti et al . 2011a ) have been exploring the bright end of the population through a large number of random , discrete wfc3 pointings obtained during the _ hst _ parallel orbits . surveys in foreground cluster fields utilizing gravitational lensing magnification have also resulted in a handful of promising candidates at @xmath89 ( e.g. , richard et al . 2006 ; bradley et al . 2008 , 2012 ; laporte et al . 2011 ; hall et al . 2012 ) , which complement the surveys in blank - sky fields . the cosmic assembly near - infrared deep extragalactic legacy survey ( candels ; pis : faber & ferguson ; see grogin et al . 2011 and koekemoer et al . 2011 ) employs a two - tiered , `` wide and deep '' strategy , which makes it uniquely positioned in bridging the significant gaps among existing surveys . in particular , its `` deep '' component will cover @xmath9 arcmin@xmath1 in two fields upon completion , and its data will be ideally suited for studying galaxies at very high redshifts . in this paper , we report our preliminary results from the deep observations in the great observatories origins deep survey ( goods ; giavalisco et al . 2004 ) southern field , where we use the nearly complete data set to study the galaxy population at @xmath0 . we use the following cosmological parameters throughout : @xmath10 , @xmath11 and @xmath12 km s@xmath13 mpc@xmath13 . the quoted magnitudes are all in the ab system . as detailed in grogin et al . ( 2011 ) , the candels / deep region in the goods - s field is located in its middle stripe where we have the deepest goods _ spitzer _ coverage , which also includes the acs - hudf . the data used in this study include 100% of the observations in f105w ( hereafter @xmath14 ) and 83% of the observations in f125w ( @xmath15 ) and f160w ( @xmath16 ) scheduled for this field . the typical exposure times per sky position are 8090 , 7450 , and 7770 seconds in @xmath14 , @xmath15 and @xmath16 , respectively . our present analysis is confined to the region where the effective integration times are at least 4000 seconds in all three wfc3 bands , which is 62.9 arcmin@xmath1 in size . for the reason described below , we further exclude the region covered by the acs - hudf data , and the final effective area in our current study is 50.1 arcmin@xmath1 . the products of the candels data reduction ( koekemoer et al . 2011 ) include the science mosaics and the associated `` root mean square '' ( rms ) maps . these rms maps account for correlations between pixels that are introduced during the data reduction procedures ( see e.g. , dickinson et al . such correlations tend to reduce the measured pixel - to - pixel background noise in the _ hst _ images , and can lead photometry programs to underestimate the actual photometric uncertainties . we avoid this problem by computing uncertainties using the normalized rms maps , summing their values in quadrature over the measurement apertures , and including an additional term for poisson noise from the source flux ( generally negligible for very faint dropout galaxies ) . the values of the candels rms maps have been validated both by comparison with autocorrelation measurements of the image noise , and by analysis of the dispersion of measurements in apertures randomly distributed within empty regions of the images . our final mosaics have a pixel scale of @xmath17 , and are tied to the same astrometric frame as the acs data in goods . the goods acs data are used in our study to confirm the optical non - detections of the candidates ( see 3.1 ) . these data have been reprocessed to make mosaics at the same @xmath17 pixel scale for candels . on average , the 5-@xmath18 sensitivities ( as measured from the rms maps ) within a @xmath19-radius aperture in acs f435w , f606w , f775w , f850lp ( hereafter @xmath20 , @xmath21 , @xmath22 , @xmath23 ) and wfc3 @xmath14 , @xmath15 , and @xmath16 are 28.04 , 28.25 , 27.65 , 27.48 , 28.15 , 28.05 , and 27.82 mag , respectively . the hudf region presents a complexity in our analysis . both the hudf09 wfc3 ir data and the acs - hudf data are much deeper than their counterparts that cover the rest ( and much wider ) area of interest . if we were to use these deeper data in our candidate selection , it would be very difficult to uniformly carry out the simulations that are necessary for the follow - up analysis ( see 3.4 ) . for simplicity , we therefore exclude the acs - hudf region from this current work . in addition to the goods acs data , we also have new acs data in f814w ( hereafter @xmath24 ) that candels took in coordinated parallel mode to the wfc3 ir observations ( see grogin et al . these new acs data suffer severely from the degraded charge transfer efficiency ( cte ) of the aged ccds , however , and the empirical correction of anderson & bedin ( 2010 ) were applied first before they were processed by the candels reduction pipeline . the photometry in wfc3 and acs was done together by running the sextractor software ( bertin & arnouts 1996 ) in dual - image mode . we used the sum of the @xmath15 and the @xmath16 mosaics , each weighted by the inverse square of their rms maps , as the detection image . a 3-pixel fwhm , 7@xmath257 gaussian filter was used to convolve the detection image , the threshold was set to 0.8-@xmath18 , and at least 4 connected pixels above this threshold are required . sextractor provides a number of methods to measure flux , among which mag_aper , mag_iso and mag_auto are commonly used in dropout selections . mag_aper magnitudes are calculated within a circular aperture of fixed size . as most high - redshift galaxies are compact , using these magnitudes can still be appropriate when the circular aperture size is carefully chosen so that it maximizes the s / n and minimizes the loss of light for the majority of the sources of interest . while it has the advantage of being robust in extracting very faint sources , mag_aper has the disadvantage that real galaxies , which have different light profiles , have different fractions of light lost for a fixed aperture . mag_iso magnitudes are calculated within isophotal areas , which are set by the detection threshold and hence trace the different shapes of real galaxies . for this reason , mag_iso usually can be tuned to maximize s / n for galaxies of different morphologies . however , it does not capture the total light of a galaxy . mag_auto magnitudes are calculated using adaptive elliptical apertures based on the first moment of the light distributions of galaxies ( kron 1980 ) , and are the most suitable for deriving total magnitudes if an appropriate `` kron factor '' is chosen . however , using a large kron factor to capture total light usually means that the apertures are set too large for very faint sources such that their s / n are not optimized . bearing both the advantages and the caveats in mind , we adopted the mag_auto magnitudes ( using the default [ kron factor , minimum radius]@xmath26[2.5 , 3.5 ] ) in this work . dropout selection uses hard boundaries in color space ( see 3.1 ) . as a result , photometric errors cause objects be scattered in and out of the selection area : an object whose intrinsic colors meet the selection threshold could be missed because of the errors in its measured colors , and vice versa . while this effect can be taken into account statistically by applying the correction for the `` effective volume '' ( see 3.4 ) , the candidate samples constructed using different photometry could differ significantly on an object - by - object basis . to demonstrate this effect , we also constructed a separate sample using mag_iso magnitudes , which is presented in the appendix and is compared to the mag_auto sample . in both cases , we only kept the sources that have s / n @xmath27 5.0 in at least one band among @xmath15 and @xmath16 , measured within the mag_auto or the mag_iso apertures . we also make extensive use of the goods _ spitzer _ irac data ( dickinson et al . 2004 ) , which also have the same sky projection as other data here , and have a pixel scale of @xmath28 . we mostly concentrate on the @xmath29 and @xmath30 @xmath31 m channels ( hereafter [ 3.6 ] and [ 4.5 ] ) in this study . we used a version of the goods photometry where the detection was done with a `` mexhat '' filter to optimize de - blending . we adopted mag_aper within a @xmath32-radius ( i.e. , 2.5 pixels ) aperture , and obtained the `` total magnitudes '' by applying the corrections as determined in the goods program where the aperture of this size has been used ( e.g. , yan et al . 2005 ) , which are @xmath33 and @xmath34 mag for [ 3.6 ] and [ 4.5 ] , respectively . in this section , we describe our samples of candidate galaxies at @xmath0 selected as @xmath14-dropouts and the possible biases in our selections . our first color criterion for @xmath14-dropout selection is @xmath35 mag . this is the same as that employed by b10 and y10 , and is significantly larger than that of @xmath36 mag used in b11 and oesch et al . ( 2012 ) where those authors have opted to included more galaxies at lower redshifts ( starting at @xmath37 ) in their @xmath14-dropout selection . our larger color decrement threshold is sensitive to lyman - break at @xmath38 . when the objects have s / n @xmath39 in @xmath14 , we use their 2 @xmath18 upper limits and only keep those that have @xmath40 mag as calculated using these conservative upper limits . our second criterion is that a @xmath14-dropout should have s / n @xmath41 in all four bands of goods acs images ( @xmath42 ) . this is to implement the invisibility requirement in these `` veto '' images , which is a necessary condition ( though not sufficient ) that the detected @xmath43 color decrements in our candidates are not features in galaxies at low redshifts ( for example , the 4000 break at @xmath44 2 ) that could mimic the lyman - break , because such low - redshift galaxies could have detectable fluxes in these bands . setting the threshold to a quantitative value of s / n @xmath41 is somewhat arbitrary , however . based on experience , adopting s / n @xmath41 for invisibility usually is a good compromise between minimizing the incompleteness and keeping the size of the initial candidate sample manageable for the later visual inspection ( see below ) . to exclude as much as possible the contamination from the red galaxy population at low redshifts , we impose a third criterion of @xmath45 mag . this also largely limits our selection window to @xmath46 , and could be biased against genuine galaxies at @xmath0 that have unusually large dust reddening or extraordinarily old stellar populations should they exist . figure 1 shows our selection in the @xmath47 space . each of these criteria introduces incompleteness , for which we will correct statistically ( see 3.4 ) . the initial candidates have been visually inspected to ensure that they are legitimate sources in the ir images and that there is no reason to suspect that their photometry might be unreliable . examples of problems that are not accounted for in the statistical noise model ( and thus require visual inspection ) include diffraction spikes , poorly rejected cosmic rays , inappropriate deblending , and bad background subtraction due to nearby neighbors . while such issues are rare in the entire photometric catalog , they can be selected when looking for objects with unusual colors , such as dropouts . visual inspection step is necessary to reduce the rate of contamination , and is commonly adopted in dropout selections by various groups . it is somewhat subjective ; different inspectors might not agree on a particular candidate , and even the same inspector might not be able to completely reproduce at different time his / her results on a large set of samples . nevertheless , visual inspection is irreplaceable to get a reliable sample . ideally , the simulation discussed in 3.4 should also include this step when deriving the incompleteness correction . however , it is impractical to examine the 90,000 artificial galaxies inserted during the simulation . nevertheless , we do not think that many of these would have been inappropriately rejected by visual inspectors , because the inserted galaxies do not resemble the artifacts or those with skewed photometry that we typically reject . relative to the 90,000 inserted objects , we suspect only a handful would have these problems . thus we do not think that the failure to visually inspect the artificial galaxies has a significant effect on our computation of the incompleteness and the effective volume . in our visual inspection process , the initial candidates were first examined by one of us ( hy ) to get rid of the most obvious contaminators ( @xmath48 objects ) , such as image defects close to the field edges , diffraction spikes of bright stars , bad pixels , residuals of ir persistence , etc . the remaining candidates ( @xmath49 ) were passed to six persons for independent inspections using a three - grade system of firm acceptance " , firm rejection " and border - line case " . to be included in the final samples , a candidate should be firmly accepted by at least three inspectors , and should not be firmly rejected by more than one inspector . as described below , this approach has allowed us to study the possible systematic biases caused by visual inspections . we note that the contemporaneous candels @xmath24 images have also been used in our current selection process . however , the flux upper limit estimate is subtle in these cte - restored images when a source is undetected in single exposures , and a rigorous treatment to be adopted by our team is still yet to be finalized . to avoid possible changes to our sample in the future because of this reason , at this stage we refrain from directly using the @xmath24 upper limits or s / n values in the quantitative veto process , and just use the @xmath24 images in the visual inspection step ( nevertheless , we do confirm that our current final candidates indeed formally satisfy the invisibility requirement in @xmath24 ; see 3.3 ) . in addition to the usual examination of the mosaics , we have taken extra steps for assurance . in the ir , we have followed yan et al . ( 2011 ) and examined the final candidates in their single - epoch wfc3 science images and the associated `` data - quality '' flags to confirm that they are not due to instrument defects . in the optical , we have constructed the `` @xmath50 '' sum image of the four goods acs @xmath42 bands for our candidates . based on szalay et al . ( 1999 ) , the @xmath50 sum of these images optimally combines the signals from different bands and extend to a fainter limit that achieved by individual bands . the @xmath50 images of our final candidates all have @xmath51 , which is consistent with being drawn from a random sky distribution ( szalay et al . 1999 ) and thus further confirms their non - detections in the acs . the final candidates in our sample are listed in table 1 . their images are displayed in figure 2 . to investigate the possible contamination of our sample by noise spikes due to background fluctuations , we carried out the so - called negative image test ( e.g. , dickinson et al . 2004 ) . the image mosaics described in 2 , which all have zero mean background by construction ( see koekemoer et al . 2011 ) , were multiplied by @xmath52 to make any positive pixels to negative and vise versa . sextractor was run in dual - image mode again on these negative images in the same way as in 2 , using the weighted sum of the the negative @xmath15 and @xmath16 images as the detection image . after applying the same selection criteria for our @xmath14-dropouts as in 3.1 , only three `` objects '' survived , all of which have @xmath53 mag , and all only show up in the negative @xmath15 image . their morphologies have odd shapes and are typical of noise spikes , and would be easily rejected during the visual inspection step . therefore , we conclude that the contamination due to noise spikes is negligible in this work . our stringent requirements of non - detection in the acs ( individual bands and @xmath50 sum ) as well as @xmath45 mag are effective in removing contaminations from red galaxies at intermediate redshifts ( see figure 1 ) . in addition , the irac data for the isolated candidates support that they are unlikely contaminators of this kind , because such objects , if very weak in optical , should have very red colors at 14 @xmath31 m such that @xmath54\gtrsim 2 $ ] mag ( e.g. , yan et al . our candidates that have irac flux measurements are all at @xmath54\lesssim 1.3 $ ] mag . another possible contaminant could be cool galactic dwarf stars . our criterion of @xmath55 mag is large enough to exclude most such contaminations . in addition , the number density of m / l / t - dwarfs in a high galactic latitude region such as the goods - s is likely negligible ( ryan et al . 2011 ) . furthermore , our detailed morphological analysis in the accompanying paper ( r. ryan et al . 2012 , apj submitted ) shows that our brightest candidates are more consistent with extended light profiles than point - like . using the 4 ms _ chandra _ catalog of xue et al . ( 2011 ) , we find no x - ray counterparts to our candidates within @xmath56 . there is thus no evidence that these objects are agn rather than normal galaxies . another paper based on the similar data set as ours was submitted by an independent group right after ours ( oesch et al . 2012 , hereafter o12 ) , and here we compare our sample to theirs based on the latest update available to us ( p. oesch , private communication , 2012 ) . in the overlapping region , the wfc3 ir data are essentially the same in both studies . however , o12 s reduction of the candels wfc3 data is different from ours , and hence the end products of the science mosaics and the associated rms maps are different . the photometry is also different . o12 s detection image is a @xmath50 image derived from the @xmath15 and @xmath16 images , whereas ours is the weighted sum of the @xmath15 and @xmath16 . while they used mag_auto as we also did , they adopted a smaller aperture ( 1.2 kron ) to compute colors . in addition , they stacked all the available acs data in this region . we refrained from combining the late acs data that suffer from the cte problem with the earlier goods v2.0 @xmath42 acs data , and for the same reason we only used the @xmath24 data in the visual inspection step ( see 3.2 ) . finally , o12 adopted rather generous color criteria of @xmath36 mag and @xmath57 mag , while we used @xmath55 mag and @xmath45 mag . all these factors contribute to the differences between the o12 sample and ours . it is interesting and instructive to compare the inclusion of the o12 candidates in our samples , and vice versa . based on the photometry reported in o12 , nine of their 14 @xmath14-dropouts in the same field would be expected to enter our samples . our sample , as shown in table 1 , include five of them . among the four candidates not in our sample , three of them , candy-2209246371 , 2432246169 and 2277945141 have @xmath58 in @xmath15 and @xmath16 and are not included in our mag_auto catalog , and the other one , candy-2272447364 , has @xmath59 based on our photometry . we note that one of our brightest candidates , auto_100 , is not in the o12 sample . o12 believes that this object has a 2.4 @xmath18 detection in @xmath24 , derived in the conventional way using sextractor . while our selection does not use the formal s / n estimates in @xmath24 ( see 3.2 ) , we have also derived such estimates and confirmed after - the - fact that all our final candidates formally have s / n@xmath41 . this particular object has s / n@xmath41 in both the mag_auto and the mag_iso apertures ( see also appendix ) . we attribute this discrepancy to the differences in photometry as outlined in 3.3.2 . to conclude , the level of overlap and discrepancy between these two samples is all well expected given the level of differences in their constructions . we further demonstrate this in the appendix . a given choice of lyman - break color criteria , applied to a particular data set , leads to a probability distribution as a function of redshift and magnitude , @xmath60 for selecting galaxies as dropout candidates . this probability distribution , in turn , corresponds to an effective volume ( @xmath61 ) for lyman - break selection as a function of magnitude ( steidel et al . we have carried out extensive simulations to derive @xmath60 and @xmath61 , in order to correct for the sample incompleteness that is inherent in the dropout technique . the results are the redshift selection functions at different brightness levels , which are shown in figure 3 . to derive these functions , we put 90,000 artificial galaxies into the wfc3 and the acs mosaics and calculated their rate of recovery @xmath60 after applying the same color criteria as we did when constructing our dropout samples . these artificial galaxies have different fluxes and morphologies , with 70% being exponential disks and 30% having the de vaucouleurs profile . their sizes follow a log - normal distribution and @xmath62 , where @xmath63 is the peak size at flux @xmath64 and is normalized to @xmath19 at @xmath65 mag . the input spectrum is from the models of bruzual & charlot ( 2003 ; bc03 ) , and has a constant star formation history ( sfh ) and an age of 50 myr . we adopted the extinction law of calzetti ( 2001 ) , and assumed that the distribution of e(b - v ) is a gaussian function with mean @xmath66e(b - v)@xmath67 mag and scatter @xmath68 mag , restricted to e(b - v ) @xmath69 . the seds of the simulated galaxies were attenuated by the line - of - slight h i absorption according to the recipe of meiksin ( 2006 ) , which is very close to that of madau ( 1995 ) . by calculating the ratio of the number of the recovered galaxies and that of the input ones , we obtained the selection function for the incompleteness correction . figure 3 presents it as a function of @xmath15 mag_auto magnitudes , as these magnitudes are taken as the total magnitudes , and are the closest to the input total magnitudes of the simulated galaxies . we have searched for irac counterparts of our @xmath14-dropouts within @xmath70 , and the results are also summarized in table 1 . only two objects are securely detected in irac , namely , auto_035 and auto_293 . as we will discuss later , their wfc3 versus irac colors are very consistent with being at high redshift . objects auto_204 , 212 , 368 are in isolated regions , and they are invisible in the goods irac data . for all these objects , their 2-@xmath18 flux upper limits in @xmath71 $ ] and @xmath72 $ ] channels are calculated within a @xmath73 aperture . all other objects in our samples are either blended with ( denoted by `` b '' in table 1 ) or severely contaminated ( `` c '' ) by foreground neighbors in irac and thus no useful information can be obtained . -dropouts in the @xmath47 color - color diagram . the small dots are field objects that have s / n@xmath41 in acs . the grey area is our selection region , where the brighter candidates ( @xmath74 mag ) are plotted as red filled squares , and the fainter candidates as red open squares . the blue - magenta track shows a young galaxy ( using a bc03 model of age 100 myr and no dust extinction ) at @xmath7510 , with the blue section for @xmath76 and the magenta section for @xmath77 . the orange stars show the colors of galactic brown dwarfs ( leggett et al . 2002 ) . the green symbols show the colors of a typical red galaxy at @xmath783 simulated using a bc03 model ( @xmath79 myr and age of 2.0 gyr ) . ] -dropouts in the auto sample , each @xmath80 on a side . the locations of the candidates are at the center marked by red circles , which are @xmath81 in radius . the common objects in the comparison iso sample ( see appendix ) are also labeled . ] . ] our sample suffer from the same kinds of incompleteness and contamination that are intrinsic to any dropout samples because of their selection technique . while a number of these issues have been mentioned in the above sections , we review them here before proceeding to discuss the statistical implications of our samples to astrophysical questions . there are two main factors that contribute to the incompleteness , which cause significant differences among samples as demonstrated in 3.3.1 and 3.3.2 . first of all , an object whose intrinsic brightness is above our sensitivity limit could be undetected because its observed brightness could fall below the detection threshold due to noise . secondly , a genuine @xmath0 galaxy that is detected in source extraction could still escape our color selection because it could be `` scattered '' out of the color selection region ( see figure 1 ) due to photometric error , or it could be incorrectly `` vetoed '' because of some unrelated events within its vicinity ( for example , noise spikes ) that accidentally increase its measured s / n in the veto images above the adopted quantitative veto threshold . both effects have been taken into account by our simulations , and therefore can be corrected statistically . the major factor that causes the contamination is the inclusion of low - redshift objects with spectral features that could mimic the lyman - break at @xmath0 . such contaminators could be selected because their @xmath43 colors meet our color threshold , either intrinsically , or after being boosted by photometric error ( i.e. , being scattered into the selection region ) . using the @xmath16 data , our secondary @xmath82 criterion reduces the number of contaminators by rejecting those that have red seds . for other contaminators , the best way to reject them is to check their visibility in the veto images . as discussed earlier , the quantitative criterion of s / n @xmath41 gets rid of most of such objects , and the final step is the visual inspection . the caveat here is that we implicitly assume that the veto images are deep enough in detecting the contaminators in the blue bands , but this is not guaranteed . currently we do not have a reliable way to assess this effect . we point out that the rejection of contaminators based on their visibility in the veto images also introduces incompleteness , and that this happens both in the quantitative pre - screening ( requiring s / n @xmath41 ) step and in the last visual inspection step . while the former can be statistically corrected by the simulations , the latter can not . our samples are unique in two aspects . ( 1 ) we have two very bright candidates at @xmath83 mag , which are at least @xmath84 mag brighter than the brightest @xmath14-dropouts previously found . they are comparable in brightness to the brightest @xmath85-dropouts reported by yan et al . ( 2011 ) and trenti et al . ( 2011a , b ) in their wfc3 parallel surveys , but could be at higher redshifts because the redshift window of this current @xmath14-dropout selection is higher than that of the existing @xmath85-dropouts . ( 2 ) two other @xmath14-dropouts , which are not among the brightest in the wfc3 bands , are securely detected in irac , which is in sharp contrast to previous results where only non - detection in irac have been reported for @xmath14-dropouts . here we discuss the implications of both . we investigate the stellar populations of the irac - detected @xmath14-dropouts , auto_035 and auto_293 , by analyzing their spectral energy distributions ( sed ) through template fitting . the approach here is similar to that of finkelstein et al . we first use the eazy code ( brammer et al . 2008 ) and the photometry of the goods acs & candels wfc3 data to estimate photometric redshifts ( @xmath86 ) . the full sed ( including the irac photometry ) is then fitted to a suite of templates based on the updated models of bc03 ( the so - called `` cb07 '' models ) . a salpeter initial mass function ( salpeter 1955 ) and the metallicities of 0.021@xmath87 are adopted . the templates have a range of exponentially decreasing and increasing sfhs ( see also papovich et al . 2011 ) , and can include nebular lines based on the number of ionizing photons and metallicity of a given model ( b. salmon et al . in preparation ) . we assume the extinction law of calzetti ( 2001 ) with @[email protected] mag and the h i absorption as formulated in madau ( 1995 ) . the results are summarized in figure 4 . we note that fitting redshift and other properties simultaneously does not signficantly change these results , and the differences are captured in the errors that we quote here . the best - fit photometric redshifts are @xmath89 and @xmath90 for auto_035 and 293 , respectively , which are consistent with the redshift window of our color selection . adding the formal @xmath24 limit to the fitting process does not change the results . we fit the seds with and without the contribution from nebular lines , and both results suggest high stellar mass ( @xmath91 ) in the range of @xmath92 to @xmath93 for both objects . including nebular emission lines , we obtain @xmath94 and @xmath95 @xmath5 for auto_035 and 293 respectively , while without nebular emission lines these are @xmath96 and @xmath97 @xmath5 , respectively . the error bars reflect the 68% confidence level of the fit . labb et al . ( 2010a ) derived the stellar masses of galaxies at lower redshifts of @xmath6 using models without nebular emission , and the average value is @xmath98 @xmath5 . the two objects in our sample have comparable or even larger values using similar models , which are surprisingly high at such an early epoch . using the @xmath91 estimates , we obtain the first measurement of the mass function ( mf ) of galaxies at @xmath0 at the high - mass end . as we only have two objects whose @xmath91 values are reasonably similar , we opt to count them within one mass bin of a @xmath99 dex bin size . the two different sets of models result in significantly different mass estimates , and therefore the bin center is different for each case . assuming a top - hat selection function of our survey within @xmath100 , we get @xmath101 mpc@xmath102 dex@xmath13 if we adopt the values derived using the models with nebular emission . if we apply the effective volume ( @xmath61 ) correction of the survey instead ( see 3.4 ) , we obtain @xmath103 mpc@xmath102 dex@xmath13 . here we use @xmath104 , where @xmath105 is the unit co - moving volume at redshift @xmath106 , and @xmath107 is the redshift selection function at different magnitudes @xmath108 as derived through simulations ( see figure 3 ) . the error bars here reflect the 68% interval of the uncertainties caused by the poisson noise in the sample . for the case of using the @xmath91 estimates based on the models without nebular lines , all the above values are applicable at @xmath109 . the contribution of these objects to the global stellar mass density , @xmath110 , at @xmath0 within our survey volume is @xmath111 and @xmath112 @xmath5 mpc@xmath102 based on the models with and without nebular emission lines , respectively , if assuming a top - hat selection function at @xmath113 . the errors reflect the 68% confidence level of the fitting results . if we apply the same correction for @xmath61 as above , we obtain @xmath114 and @xmath115 @xmath5 mpc@xmath102 . we derive the mass - to - light ratio of these two objects using their 4.5 @xmath31 m flux as a proxy to @xmath116 . using the fit results with contribution from nebular lines , we obtain @xmath117 and @xmath118 for auto_035 and 293 , respectively . without the contribution from nebular lines , we obtain @xmath119 and @xmath120 . the latter values are in general agreement with those obtained by labb et al . ( 2010a , b ) through stacking analysis of the irac data of the @xmath85-dropouts and the @xmath14-dropouts in the wfc3 ers field ( see also gonzlez et al . 2011 ) and the hudf using similar models without nebular lines . on the other hand , labb et al . ( 2010b ) estimated that nebular lines have a small impact on @xmath121 , reducing it by @xmath122 dex . our results suggest a stronger effect , reducing @xmath121 by @xmath1230.6 dex . m image ( @xmath124 ) . the bottom panels show the likelihood functions of @xmath86 and @xmath125 , and the best - fit values are indicated by the vertical lines . ] here we use our sample to constrain the lf at @xmath0 . we adopt the @xmath15 mag_auto values of the candidates as their total magnitudes . after applying the correction for @xmath61 , we obtain the stepwise lf in the five 0.5-mag bins ( 25.45 , 25.95 , 26.45 , 26.95 , 27.45 ) mag as @xmath126 mpc@xmath102 mag@xmath13 . these number densities are shown in figure 5 , and compared to the predictions from a number of schechter lf estimates at @xmath0 ( m10 ; y10 ; b11 ; l11 ) . the black squares are our observed densities , while the red squares are the densities after correcting for @xmath61 . the error bars represent a bayesian 68% credible interval , indicating the central 68 percentile range for the posterior distribution of the true number density assuming poisson statistics . these uncertainties account for the incompleteness , but do not account for any systematic uncertainties from the contamination due to low - z interlopers . for ease of direct comparison to observations , surface density and apparent magnitude scales are also provided on the same figures to present these results in terms of differential number densities versus apparent @xmath15 magnitudes . for comparision , we also plot the _ observed _ densities ( i.e. , before corrections for their corresponding @xmath61 values ) extracted from the sample of y10 in the hudf09 proper and those of b11 in the hudf09 , hudf09p1 and hudf09p2 _ after _ applying the additional criteria of @xmath127 mag and @xmath45 mag . based on the sample of @xmath14-dropouts in candels / deep from this current work . the black squares are the differential surface densities inferred from the number counts in our sample , while the red squares are those after correction for @xmath61 with respect to the volume within @xmath100 . the count predictions over @xmath100 are from the schechter lf estimates at @xmath0 from b11 , m10 , y10 , and l11 are shown as various black curves , together with a non - evolution one at @xmath6 from b11 as the blue curve . for comparison , different symbols at the faint - end show the raw ( observed ) surface densities of @xmath14-dropouts based on the following samples : y10 ( hudf stars ) , b11 ( hudf crosses ; hudfp1 pentagons ; hudfp2 triangles ) . ] an intriguing feature that our samples reveal is that there could be an excess of bright objects at @xmath128 mag with respect to any of the existing schechter lf estimates at @xmath0 in the literature would predict . the inferred number density at this bright end is even higher than the prediction from a non - evolution @xmath6 lf ( e.g. , the one from b11 ) . this excess is still evident with the observed counts _ before _ applying the incompleteness correction . the observed counts at @xmath128 mag in figure 5 comes from objects auto_048 and auto_100 . based on their photometry in @xmath15 and assuming @xmath129 , they have @xmath130 of @xmath131 and @xmath132 , respectively . while such high luminosities have been observed in the spectroscopically confirmed @xmath6 galaxies of ono et al . ( 2012 ) , they are 59@xmath25 more luminous than any previously known @xmath14-dropouts . such an excess is unlikely due to cosmic variance , whose effect is expected to be small compared to the poisson errors for our observed number density if we use the formalism of trenti & stiavelli ( 2008 ) . we note that , however , that o12 do not see such an excess , even though their sample also includes our brightest candidate auto_048 . the main reason for this discrepancy could be that o12 use a larger volume in the calculation . they have included part of the candels wide field in the goods - s region , which is a major contributor to their larger volume . however , this area does not have sufficiently deep @xmath14 data to apply our selection criterion of @xmath133 mag down to @xmath134 mag . o12 have adopted a less stringent criterion of @xmath135 mag , which allows them to construct a @xmath14-dropout sample in this area at the cost of including candidates at lower redshifts ( @xmath136 ) . the advantage of their approach is that they can incorporate the wfc3 ers field in the goods - s region as well , which further increases their survey volume . while the wfc3 ers field has @xmath85 but not @xmath14 data , the redshift range probed by their @xmath85-dropouts in this area overlaps that of their @xmath14-dropouts elsewhere , and therefore o12 has combined the @xmath85- and the @xmath14-dropouts to address the lf averaged over a wide redshift range ( @xmath68.7 ) . in contrast , our color criteria select galaxies in a higher range of redshifts , @xmath137 , and the bright - end excess ( with respect to a smooth schechter function ) that we find suggests that our understanding of the lf at @xmath0 is still far from being complete . the significance of this excess is still subject to small number statistics , and we can not rule out the possibility that one or both of our brightest objects could be interlopers at low redshifts . however , we should point out that similar bright - end excess at @xmath138 has also been suggested in other studies , for example in the bright @xmath85-dropout study in hippies ( yan et al . 2011 ) at similar depths , in the ground - based survey of laporte et al . ( 2011 ) at a brighter level , and in the much wider survey in capak et al . ( 2011 ) at an even brighter limit . ( @xmath139 ) for the candidates in capak et al . ( 2011 ) , the true nature of these objects still remains uncertain . ] our @xmath0 candidate galaxy samples have two very luminous objects and two high stellar mass objects . we have presented tentative evidence that the bright end of the galaxy lf at @xmath0 might not follow the exponential cut - off of the schechter function . while the stellar mass function at this redshift has not yet been determined , we have provided the first , albeit still tentative measurement at the high - mass end and shown that some galaxies with stellar masses as high as a few @xmath4 @xmath5 might already in place at @xmath0 . the physical origin of the exponential cut - off at the bright end of the lf or the high mass end of the mf seen at lower redshifts remains a matter of debate . at @xmath140 , star formation is `` quenched '' in halos with masses greater than @xmath141 @xmath5 ( `` quenching mass '' ) , perhaps by agn feedback ( e.g. , croton et al . 2006 ; somerville et al . 2008 ; gabor et al . however , at higher redshift ( @xmath142 ) , the quenching mass may be higher ( dekel et al . 2009 ; behroozi et al . 2010 ) , and the existence of massive , rapidly star forming galaxies at @xmath143 is well established . at @xmath144 , the observed exponential cut - off in the rest - frame uv lf appears to be due to dust ( reddy et al . 2010 ) rather than quenching . at very high redshifts , it has been noted before that the very blue rest - frame uv colors of @xmath6 candidate galaxies ( some of them confirmed ) suggest that they may contain little dust ( e.g. , oesch et al . 2010 ; m10 ; y10 ; bunker et al . thus , at @xmath0 , when ( 1 ) the quenching mass is much higher than the typical halo mass , and ( 2 ) dust has little effect , perhaps we should not expect to see an exponential cutoff in the luminosity or mass functions of uv - bright galaxies . it is interesting that the estimated stellar masses and number densities could imply a rather high efficiency of conversion of baryons into stars . for the lower stellar mass estimates , which are obtained using the models with nebular emission ( the average is @xmath145 @xmath5 ) , and assuming that @xmath146% of the available baryons have been converted to stars , the implied halo masses are on the order of @xmath147 @xmath5 , for which the expected number density of dark matter halos in the currently favored lcdm cosmology is comparable to the observed number density of objects . this is consistent with the predicted star formation efficiencies and host halo masses at @xmath0 from cosmological hydrodynamic simulations . however , if the true stellar masses are higher , or the star formation efficiency is lower , a rapidly growing tension arises between the number density of dark matter halos and the observed number density of galaxies ( above halo mass @xmath147 @xmath5 , the halo number density declines by about two orders of magnitude for a factor of @xmath148 increase in mass ) . while converting 20% of the available baryons into stars may not sound excessive , this is in fact the _ maximum _ value that has been inferred at any epoch . due to the presumably very low metallicity of the gas in these early objects , we might have expected much lower star formation efficiencies than are seen locally ( e.g. krumholz & dekel 2010 ) . in this work , we search for candidate galaxies at @xmath0 in the candels deep goods - s field and study their properties . our sample of @xmath14-dropouts hints that the number density of @xmath0 galaxies at the bright - end might be higher than expected from the previous schechter lf estimates , which lends support to the suggestion made by a number of earlier studies that there could be a bright - end excess in the galaxy number density at very high redshifts . furthermore , two of our candidates are securely detected at 3.6 and 4.5 @xmath31 m in _ spitzer _ data . these are the first @xmath14-dropouts individually detected at these wavelengths . their derived stellar masses are on the order of @xmath98 @xmath5 , from which we obtain the first measurement of the high - mass end of the galaxy stellar mass function at @xmath0 . if the high number densities of very luminous and very massive galaxies at @xmath0 are real , they could imply a large stellar - to - halo mass ratio and an efficient conversion of baryons to stars at very early time in the cosmic stellar mass assembly history . we thank the referee for the useful comments , which improved the quality of this work . we also thank b. mobasher and j. dunlop for their comments on an earlier version of this paper . acknowledges the support of nasa grant hst - go-11192.1 . support for program number hst - go-12060 was provided by nasa through a grant from the space telescope science institute , which is operated by the association of universities for research in astronomy , incorporated , under nasa contract nas5 - 26555 . anderson , j. & bedin , l. r. 2010 , , 122 , 1035 beckwith , s. v. w. , stiavelli , m. , koekemoer , a. m. , et al . 2006 , , 132 , 1729 behroozi , p. s. , conroy , c. & wechsler , r. h. 2010 , , 717 , 379 bertin , e. & arnouts , s 1996 , , 117 , 393 bouwens , r. j. , illingworth , g. d. , labbe , i. , et al . 2011a , nature , 469 , 504 bouwens , r. j. , illingworth , g. d. , oesch , p. a. , et al . 2010 , apj , 709 , l133 ( b10 ) bouwens , r. j. , illingworth , g. d. , oesch , p. a. , et al . 2011b , apj , 737 , 90 ( b11 ) bowler r. a. a. ; dunlop , j. s. ; mclure , r. j. ; et al . 2012 , mnras , 426 , 2772 bradley , l. d. , bouwens , r. j. , ford , h. c. , et al . 2008 , apj , 678 , 647 bradley , l. d. , bouwens , r. j. , zitrin , a. , et al . 2012 , apj , 747 , 3 brammer , g. b. , van dokkum , p. g. , & coppi , p. 2008 , , 686 , 1503 bruzual , g. & charlot , s. 2003 , , 344 , 1000 bunker , a. j. , wilkins , s. , ellis , r. s. , et al . 2010 , mnras , 409 , 855 calzetti , d. 2001 , , 113 , 1449 capak , p. , mobasher , b. , scoville , n. z. , et al . 2011 , apj , 730 , 68 castellano , m. , fontana , a. , paris , d. , et al . 2010 , a&a , 524 , a28 chabrier , g. 2003 , , 115 , 763 croton , d. j. , springel , v. , white , s. d. m. , et al . 2006 , mnras , 365 , 11 dekel , a. , birnboim , y. , engel , g. , et al . 2009 , nature , 457 , 451 dickinson , m. , stern , d. , giavalisco , m. , et al . 2004 , apj , 600 , l99 finkelstein , s. l. , papovich , c. , giavalisco , m. , et al . 2010 , apj , 719 , 1250 finkelstein , s. l. , papovich , c. , salmon , b. , et al . 2012 , apj , 756 , 164 gabor , j. m. , dav , r. , oppenheimer , b. d. & finlator , k. 2011 , , 417 , 2676 giavalisco , m. , ferguson , h. c. , koekemoer , a. m. , et al . 2004 , apj , 600 , l93 gonzlez , v. , labb , i. , bouwens , r. j. , et al . 2011 , , 735 , l34 grogin , n. a. , kocevski , d. d. , faber , s. m. , et al . 2011 , apjs , 197 , 35 hall , n. , brada , m. , gonzalez , a. h. , et al . 2012 , , 745 , 155 hathi , n. , mobasher , b. , capak , p. , wang , w .- h . , & ferguson , h. c. 2012 , apj , 757 , 43 hickey , s. , bunker , a. , jarvis , m. j. , chiu , k. & bonfield , d. 2010 , , 404 , 212 hsieh , b .- q . , wang , w .- h . , yan , h. , et al . 2012 , apj , 749 , 88 koekemoer , a. m. , faber , s. m. , ferguson , h. c. , et al . 2011 , , 197 , 36 kron , r. g. 1980 , , 43 , 305 krumholz , m. r. & dekel , a. 2010 , , 406 , 112 labb , i. , gonzlez , v. , bouwens , r. j. , et al . 2010a , , 708 , l26 labb , i. , gonzlez , v. , bouwens , r. j. , et al . 2010b , , 716 , l103 laporte , pell , r. , schaerer , d. , n. , et al . 2011 , a&a , 531 , 74 leggett , s. k. , golimowski , d. a. , fan , x. , et al . 2002 , apj , 564 , 452 lorenzoni , s. , bunker , a. j. , wilkins , s. m. , et al . 2011 , , 414 , 1455 ( l11 ) madau , p. 1995 , , 441 , 18 mclure , r. j. , dunlop , j. s. , cirasuolo , m. , et al . 2010 , , 403 , 960 ( m10 ) meiksin , a. 2006 , , 365 , 807 oesch , p. a. , bouwens , r. j. , carollo , c. m. , et al . 2010 , , 709 , l21 oesch , p. a. , bouwens , r. j. , illingworth , g. d. , et al . 2012 , , 759 , 135 ono , y. , ouchi , m. , mobasher , b. , et al . 2012 , , 744 , 83 ouchi , m. , mobasher , b. , shimasaku , k. , et al . 2009 , , 706 , 1136 papovich , c. , finkelstein , s. l. , ferguson , h. c. , lotz , j. m. & giavalisco , m. 2011 , , 412 , 1123 reddy , n. a. , erb , d. k. , pettini , m. , steidel , c. c. , & shapley , a. e. 2010 , , 712 , 1070 richard , j. , pell , r. , schaerer , d. , le borgne , j .- f . & kneib , j .- 2006 , a&a , 456 , 861 ryan , r. e. , thorman , p. a. , yan , h. , et al . 2011 , , 739 , 83 salpeter , e. e. 1955 , , 121 , 161 somerville , r. s. , hopkins , p. f. , cox , t. j. , robertson , b. e. & hernquist , l. 2008 , , 391 , 481 steidel , c. c. , adelberger , k. l. , giavalisco , m. , dickinson , m. & pettini , m. 1999 , , 519 , 1 steidel , c. c. , adelberger , k. l. , shapley , a. e. , et al . 2003 , apj , 592 , 728 szalay , a. s. , connolly , a. j. & szokoly , g. p. , 1999 , , 117 , 68 trenti , m. , bradley , l. d. , stiavelli , m. , et al . 2011a , , 727 , l39 trenti , m. & stiavelli , m. 2008 , , 676 , 767 trenti , m. , bradley , l. d. , stiavelli , m. , et al . 2012 , apj , 746 , 55 wilkins , s. m. , bunker , a. j. , lorenzoni , s. & caruana , j. 2011 , , 411 , 23 windhorst , r. a. , cohen , s. h. , hathi , n. p. , et al . 2011 , , 193 , 27 wyithe , j. s. b. , yan , h. , windhorst , r. a. & mao , s. 2011 , , 469 , 181 xue , y. q. , luo , b. , brandt , w. n. , et al . 2011 , , 195 , 10 yan , h. , dickinson , m. , eisenhardt , p. r. m. , et al . 2004 , , 616 , 63 yan , h. , dickinson , m. , stern , d. , et al . , 2005 , , 634 , 109 yan , h. , windhorst , r. a. , hathi , n. p. , et al . 2010 , res . astrophys . , 10 , 867 ( y10 ) yan , h. , yan , l. , zamojski , m. a. , et al . 2011 , , 728 , l22 cccccccl auto_048 & 3:32:49.936 @xmath14927:48:18.101 & [email protected] & [email protected] & @[email protected] & c & c & iso_085 ; o12 - 2499448181 + auto_100 & 3:32:41.417 @xmath14927:44:37.831 & [email protected] & [email protected] & [email protected] & b & b & iso_164 + auto_212 & 3:32:20.965 @xmath14927:51:37.073 & [email protected] & [email protected] & [email protected] & @xmath151 & @xmath152 & o12 - 2209751370 + auto_293 & 3:32:20.981 @xmath14927:48:53.467 & [email protected] & @xmath153 & @[email protected] & [email protected] & [email protected] & iso_071 ; o12 - 2209848535 + auto_204 & 3:32:18.185 @xmath14927:52:45.566 & [email protected] & [email protected] & [email protected] & @xmath154 & @xmath152 & o12 - 2181852456 + auto_094 & 3:32:40.675 @xmath14927:45:11.624 & [email protected] & [email protected] & [email protected] & c & c + auto_035 & 3:32:34.998 @xmath14927:49:21.623 & [email protected] & @xmath155 & [email protected] & [email protected] & [email protected] & iso_063 ; o12 - 2350049216 + due to the nature of the dropout selection technique , it is not unusual that samples constructed by independent groups can differ significantly on an object - by - object basis ( see 3 ) . to further demonstrate this point , we constructed a separate @xmath14-dropout sample to mimic an independent study . we used exactly the same data as in 2 , and therefore any difference in this sample should be attributed to the selection technique . we followed the same procedures as in 3 , but used mag_iso instead when calculating colors . the motivation of adopting mag_iso in this exercise is that isophotal apertures are commonly used in measuring lyman - break galaxy colors ( e.g. , steidel et al . 2003 ) because they often lead to higher @xmath156 , even if they do not necessarily include all of the light from each galaxy . hereafter we refer to this sample as the `` iso sample '' to separate from the one described in the main text , which we refer to as the `` auto sample '' . the initial candidates in the iso sample were also visually examined by the same group of six inspectors . the visual inspection of this sample was done about one month later than that for the auto sample , and therefore even when we were examining the same object that is also in the auto sample we did not keep the memory of the result from the last time , and hence we mimicked an `` independent '' study as much as we could . the final iso sample consists of eleven candidates . 6 shows their selection in the @xmath47 color space , while fig . 7 shows the redshift selection function . images of these candidates are displayed in fig . table 2 lists their photometric information . -dropout sample . the labeled magnitude ranges are still based on mag_auto in @xmath15 , as the simulation only has control over the input total magnitudes , which are best represented by mag_auto . to be consistent , the correction of @xmath157 for the iso sample is done at the corresponding mag_auto ranges . , width=226 ] for the sake of completeness , we also compared the iso sample to the o12 s sample . based on the photometry reported in o12 , nine of their candidates are expected to meet our color criteria ( see also however , only three of them are included in our iso sample ( see table 2 ) . among the five that are not in our sample , o12 s candy-2432246169 has @xmath58 in @xmath15 and @xmath16 and is not included in our mag_iso catalog . candy-2272447364 has @xmath158 in @xmath21 and does not satisfy our optical non - detection criterion . candy-2277945141 and 2320345371 have @xmath159 and 0.57 mag based on our photometry and hence do not meet out color criteria . the other two , o12 s candy-2209751370 and 2181852456 , meet all our color criteria , but were rejected during the visual inspection when constructing our iso sample . note that these latter two objects actually are in our auto sample ( auto_212 and 204 , respectively ) and survived the visual inspection for our auto sample ( see below ) . a more constructive comparision is that between this iso sample and the auto sample , which shows that they have four objects in common ( see table 2 ) . there are seven iso candidates not in the auto sample . five of them were rejected in the auto selection because they have s / n @xmath160 within their mag_auto apertures in at least one veto image ( iso_011 , 017 , 157 , 158 , 160 ) . one was rejected because its mag_auto @xmath161 color limit does not satisfy @xmath162 mag ( iso_008 ) . the other one ( iso_082 ) satisfies all the quantitative criteria in the auto selection , however it was rejected in the visual inspection step during the auto selection run . on the other hand , four auto candidates are not in the iso sample . two of them were rejected because they have mag_auto @xmath161 color or limit below the @xmath162 mag threshold ( auto_094 and 368 ) . the other two satisfy all the quantitative criteria in the iso selection , however they were rejected in the visual inspection step during the iso selection run ( auto_204 and 212 ) . this internal comparison of our two samples thus further demonstrates the points addressed earlier : ( 1 ) adoption of different photometry result in samples that can be significantly different on an object - by - object basis , and ( 2 ) visual inspections at different time , especially the inspections of the veto images , can also result in differences in this sense because such inspections are not guaranteed to be fully repeatable for a large sample at the s / n@xmath41 level . nevertheless , the main statistical trends revealed by the iso sample are very similar to those inferred from the auto sample . both samples include the same two bright @xmath14-dropouts ( auto_048/iso_085 and auto_100/iso_184 ) , and the same two irac - detected ones ( auto_293/iso_071 and auto_035/iso_063 ) . for completeness , 9 shows the constraint from the iso sample on the lf , where one can see that the bright - end excess is still present . similarly , the constraints on the mf and the stellar mass density derived from the iso sample agree with those based the auto sample to within @xmath163% , and the small differences are mainly caused by the slightly different @xmath61 corrections over the range of interest . from this exercise , we believe that the main results presents in this work are robust . cccccccl iso_085 & 3:32:49.936 @xmath14927:48:18.100 & [email protected] & [email protected] & @[email protected] & c & c & auto_048 ; o12 - 2499448181 + iso_164 & 3:32:41.417 @xmath14927:44:37.831 & [email protected] & [email protected] & [email protected] & c & c & auto_100 + iso_157 & 3:32:42.882 @xmath14927:45:04.268 & [email protected] & [email protected] & [email protected] & b & b & + iso_071 & 3:32:20.981 @xmath14927:48:53.468 & [email protected] & [email protected] & [email protected] & [email protected] & [email protected] & auto_293 ; o12 - 2209848535 + iso_082 & 3:32:14.133 @xmath14927:48:28.911 & [email protected] & [email protected] & [email protected] & @xmath164 & @xmath165 & + iso_011 & 3:32:14.469 @xmath14927:51:48.542 & [email protected] & [email protected] & [email protected] & b & b & + iso_158 & 3:32:47.953 @xmath14927:44:50.436 & [email protected] & [email protected] & [email protected] & @xmath154 & @xmath166 & + iso_063 & 3:32:34.999 @xmath14927:49:21.622 & [email protected] & @xmath167 & [email protected] & [email protected] & [email protected] & auto_035 ; o12 - 2350049216 + iso_017 & 3:32:18.091 @xmath14927:51:18.492 & [email protected] & [email protected] & @[email protected] & b & b & + iso_160 & 3:32:46.111 @xmath14927:44:47.997 & [email protected] & [email protected] & [email protected] & @xmath154 & @xmath168 & + iso_008 & 3:32:16.915 @xmath14927:52:01.878 & [email protected] & @xmath169 & @[email protected] & @xmath154 & @xmath152 & +

in the applications of solar magneto - seismology(sms ) , employing the ratio of the period of the fundamental mode to twice the one of its first overtone , , plays an important role . we examine how field - aligned flows affect the dispersion properties , and hence the period ratios , of standing modes supported by magnetic slabs in the solar atmosphere . we numerically solve the dispersion relations and devise a graphic means to construct standing modes . for coronal slabs , we find that the flow effects are significant , for the fast kink and sausage modes alike . for the kink ones , they may reduce by up to 23% compared with the static case , and the minimum allowed can fall below the lower limit analytically derived for static slabs . for the sausage modes , while introducing the flow reduces by typically% relative to the static case , it significantly increases the threshold aspect ratio only above which standing sausage modes can be supported , meaning that their detectability is restricted to even wider slabs . in the case of photospheric slabs , the flow effect is not as strong . however , standing modes are distinct from the coronal case in that standing kink modes show a that deviates from unity even for a zero - width slab , while standing sausage modes no longer suffer from a threshold aspect ratio . we conclude that transverse structuring in plasma density and flow speed should be considered in seismological applications of multiple periodicities to solar atmospheric structures .

1303.1689

a rich variety of magnetohydrodynamic ( mhd ) waves and oscillations have been detected in the structured solar atmosphere ( e.g. , the reviews by * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . while identifying waves / oscillations was possible with radio and optical instruments prior to the 1990s ( * ? ? ? * and references therein ) , the majority of such observations emerged only after the advent of the solar and heliospheric observatory ( soho ) , the transition region and corona explorer ( trace ) , and later hinode , in conjunction with the availability of ground - based instruments with very high temporal and spatial resolution such as the dutch open telescope ( dot ) and the swedish solar telescope ( sst ) . for instance , imaging and spectroscopic instruments on soho , along with the h@xmath2 measurements with sst and dot have shown that mhd oscillations abound in prominences - cool and dense clouds suspended in the corona ( * ? ? ? * and references therein ) . instruments with high spatial resolution like the solar optical telescope ( sot ) on board hinode have detected transverse oscillatory loop - like structures accompanied by downflows termed `` coronal rains '' @xcite . hinode / sot has also detected transverse oscillations in spicules - chromospheric materials isolated from a coronal environment . using the ultraviolet coronagraph spectrometer ( uvcs ) data , @xcite detected significant periodic oscillations in h i ly@xmath3 intensities in a coronal hole , the quiet sun and a streamer . coronal loops offer by far the most samples of mhd oscillations , among which the majority seems to be standing transverse oscillations triggered by flares ( e.g. , * ? ? ? * ; * ? ? ? identifying the observed oscillations with a particular mhd mode collectively supported by a magnetized structure is physically crucial but proves difficult . this is best illustrated by the substantial controversy associated with interpreting the observed transverse oscillations in terms of alfvn waves ( for the most recent review , see * ? ? ? this happens with both coronal @xcite and spicule measurements @xcite , the interpretation of which as alfvn waves was suggested to be inconsistent with theory by @xcite and later by @xcite . the first unambiguous detection of alfvn waves in the solar atmosphere is due to @xcite where h@xmath3 observations with sst of a group of magnetic bright points are analyzed . less controversial is the identification of fast and slow waves . consider first the coronal case . flare - triggered transverse oscillations supported by loops are usually fast kink modes . on the other hand , fast sausage modes were also inferred albeit much less abundant . as for the slow waves , they are frequently observed and appear both in the form of standing ( * ? ? ? * and references therein ) and as propagating ones ( see * ? ? ? * ; * ? ? ? * ; * ? ? ? * for comprehensive reviews ) . when it comes to atmospheric parts other than the corona , sausage waves are directly observed in the photosphere @xcite and chromosphere @xcite . note that in the fine structures examined in the latter study , these fast sausage waves , manifested as intensity oscillations , appear simultaneously with fast kink waves , manifested in the form of transverse displacements . that periodic longitudinal and transverse oscillatory motions may accompany each other was first shown by for hinode loops . at this point , it should be noted that adopting a helioseismic approach by working with the power spectra of , say , the doppler shift time series , can greatly improve the capability for wave identification , as was advocated by . once identified , the measured properties of waves and oscillations can then be combined with mhd theory to yield parameters of the solar atmosphere that are otherwise difficult to measure . the idea behind this was put forward as early as in the 1970 s following the pioneering work by @xcite and @xcite . along with others , roberts and co - workers ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) expressed the idea in the mathematical form now in routine use . this powerful diagnostic capability , initially termed `` coronal seismology '' , may be properly called `` solar magneto - seismology '' ( sms ) , given that both observations and seismological applications extend well beyond the solar corona ( for recent reviews , see @xcite @xcite , @xcite ; @xcite @xcite ; @xcite @xcite ) . the parameter that is perhaps most frequently inferred is the magnetic field strength . for instance , measuring the periodic flare - generated transverse displacements in several trace loops leads to concrete estimates of the coronal magnetic field strength . this practice was also applied to loops observed by hinode / sot , and recently by the atmospheric imaging assembly ( aia ) on board the solar dynamic observatory ( sdo ) . certainly sms can offer more . as a matter of fact , slow waves running in active region loops measured by the solar terrestrial relations observatory ( stereo ) enabled @xcite to infer loop temperatures via the measured propagation speeds . slow waves measured by hinode also helped derive the effective adiabatic index @xcite . multiple periodicities detected in some oscillating structures are playing an increasingly important role in solar magneto - seismology ( see * ? ? ? * ; * ? ? ? * for recent reviews ) . they may have disparate timescales , differing by nearly a factor of @xmath4 as shown in the light curve data obtained by the lyra irradiance experiment , indicative of the coexistence of slow and fast modes @xcite . this coexistence can help derive the plasma beta and density contrast of the flare loops where the oscillations are found . they may have comparable timescales , favoring an interpretation of the coexistence of a fundamental mode with its overtones . two and three periodicities have been found . one thing peculiar with the multiple periodicities is that the ratio between the period of the fundamental and twice the period of its first overtone , @xmath0 , deviates in general from unity . for instance , the loops labeled c and d in the trace 171 images on 15 april 2001 exhibit a @xmath0 of 0.91 and 0.82 , respectively @xcite . a further study using similar trace 171 loops on 23 nov 1998 yields a @xmath0 of 0.9 . this is not expected for kink modes supported by a longitudinally uniform loop with tiny aspect ratios . hence the deviation of @xmath0 from @xmath5 may signify some longitudinal structuring . actually @xcite come up with a one - to - one look - up diagram where the density scale height along the loop in units of the loop length can be readily deduced once @xmath6 is known . fundamental or global sausage modes with their first overtones were also found . for example , flare - associated quasi - periodic pulsations measured with the nobeyama radioheliograph ( norh ) on 12 jan 2000 contain a global sausage mode with period @xmath7 s , and a higher harmonic with @xmath8 s . moreover , high spatial resolution h@xmath2 images of oscillating cool post - flare loops measured with the solar tower telescope at aryabhatta research institute of observational sciences ( aries ) yield a @xmath9 s and a @xmath10 s @xcite . flows are ubiquitous in the solar atmosphere ( e.g. , * ? ? ? * ) and have been found in oscillating structures . take the corona for instance . the frequently found flow speeds , usually @xmath11 @xmath12 and hence well below the alfvn speed , have only weak effects on the periods of the fundamental kink mode and its overtone supported by thin loops ( e.g. , * ? ? ? therefore it is suggested that the effect of flow may be neglected in the seismological applications of @xmath6 to the corona . however , the flow speeds may not be always small . although less frequent , flow speeds reaching the alfvnic range ( @xmath13 @xmath12 ) have been seen associated with explosive events . taking into account a siphon flow of the order of the alfvn speed may lead to significant revisions to the loop parameters deduced from seismology . for the standing kink mode measured with trace and soho on 15 sept 2001 @xcite , @xcite show that neglecting the flows leads to an underestimation of the loop magnetic field by up to a factor of three ! such a revision is important in its own right , and also important in the context of coronal heating because the energy flux density carried by the transverse waves is proportional to the magnetic field strength . given the importance of multiple periodicities in the context of solar magneto - seismology , and the extent to which a flow may affect the seismological applications , it seems natural to examine further how the flows affect multiple periodicities from the theoretical perspective . in the case of cylindrical geometry , this has been done by @xcite in the thin - tube approximation . here we consider a slab geometry first , and leave the study of coronal cylinders with finite aspect ratio to a future publication . by doing so , we will complement and expand the only study concerning the period ratios of a coronal slab , which shows that the transverse structuring may contribute to the deviation of the period ratio from @xmath5 . this is true even for relatively thin slabs , and is more pronounced for high density contrasts . extending the static slab study by to incorporate flows , we will examine how the flows affect the dispersion properties and hence the period ratios of both standing kink and sausage modes . this paper is organized as follows . in section [ sec_dr_oview ] , we briefly describe the slab dispersion relation . section [ sec_cor_pratio ] deals with coronal slabs by first giving an overview of their dispersion diagrams , then describing a graphical means to derive the period ratios , and also examining in detail how the flow affects the period ratios for the standing kink and sausage modes . the same approach is also applied to isolated photospheric slabs , the results thus derived are given in section [ sec_pho_pratio ] . finally , section [ sec_summary ] summarizes the results , ending with some concluding remarks . consider a slab of half - width @xmath14 with time - independent siphon flows . as illustrated in fig.[fig_slab_illus ] , the slab is infinite in the @xmath15 and @xmath16 directions , but is bordered by two interfaces @xmath17 . let the subscripts @xmath18 and @xmath19 denote the otherwise uniform parameters inside and external to the slab , respectively . the background magnetic fields ( @xmath20 and @xmath21 ) , together with the flow velocities ( @xmath22 and @xmath23 ) , are in the @xmath16direction . let @xmath24 and @xmath25 denote the mass density and thermal pressure . from the force balance condition across the interfaces it follows that @xmath26 where @xmath27 is the adiabatic index , @xmath28 is the adiabatic sound and @xmath29 the alfvn speed . to proceed , it proves convenient to introduce the tube speed @xmath30 where @xmath31 . the dispersion relation ( dr ) for the waves supported by a slab with flow has been examined by a number of authors ( also see @xcite @xcite for the waves supported by cylinders with flows ) . restricting oneself to propagation in the @xmath32 plane , and to linear perturbations , one may adopt the ansatz that any perturbation @xmath33 to the equilibrium @xmath34 is in the form @xmath35\right\ } , \label{eq_wav_ansatz}\end{aligned}\ ] ] where @xmath36 means taking the real part of the function . the phase speed @xmath37 is defined as @xmath38 . of critical importance is the variable @xmath39\left[v_{ai}^2 - \left({\ensuremath{v_{\mathrm{ph}}}}-u_i\right)^2\right ] } { \left(c_i^2 + v_{a i}^2\right ) \left[c_{t i}^2 - \left({\ensuremath{v_{\mathrm{ph}}}}-u_i \right)^2\right ] } , \label{eq_def_m2}\end{aligned}\ ] ] where @xmath31 . to ensure the waves external to the slab are evanescent , one has to require a positive @xmath40 , which translates into @xmath41 where @xmath42 and @xmath43 . on the other hand , the signs of @xmath44 determine the spatial profiles of the perturbations in the transverse ( i.e. , @xmath45 ) direction . for @xmath46 ( @xmath47 ) , the modes are body ( surface ) waves , corresponding to an oscillatory ( a spatially decaying ) @xmath45 dependence inside the slab . evidently , body waves correspond to @xmath48 whereas surface waves take up the rest of the possible @xmath37 . plugging eq.([eq_wav_ansatz ] ) into the linearized , ideal mhd equations , one may derive the dr by requiring the following boundary conditions be satisfied : first , the transverse ( i.e. , @xmath45 ) component of the lagrangian displacement @xmath49 is continuous at @xmath17 . second , so is the total pressure @xmath50 . the dr reads @xcite ( see also @xcite @xcite for its static counterpart ) @xmath51}{\left[v_{a0}^2 - \left({\ensuremath{v_{\mathrm{ph}}}}- u_0\right)^2\right ] } = -\left\{\begin{array}{c } \tanh \\ \coth \end{array } \right\}\left(m_0 d\right ) \label{eq_dr_surface}\end{aligned}\ ] ] for surface waves , and @xmath52}{\left[v_{a0}^2 - \left({\ensuremath{v_{\mathrm{ph}}}}- u_0\right)^2\right ] } = \left\{\begin{array}{c } -\tan \\ \cot \end{array } \right\}\left(n_0 d\right ) \label{eq_dr_body}\end{aligned}\ ] ] for body waves ( @xmath53 ) . furthermore , the upper ( lower ) case is for kink ( sausage ) modes . note that the absolute value of @xmath54 appears to ensure the waves are trapped , i.e. , evanescent outside the slab . it proves necessary to examine also the relative importance of transverse displacement versus the density fluctuation . this is evaluated by @xmath55 , which reads @xmath56 } \frac{\tilde{p}_t}{d\tilde{p}_t / dx } \right |_{x = d } . \label{eq_ratio_rho_xi } \end{aligned}\ ] ] for body waves , @xmath57 inside the slab equals @xmath58 for a kink wave , and @xmath59 for a sausage one , leading to @xmath60 for surface waves , @xmath57 inside the slab is @xmath61 for a kink wave , and @xmath62 for a sausage one , leading to @xmath63 before presenting the solutions to the dispersion relations ( [ eq_dr_surface ] ) and ( [ eq_dr_body ] ) , let us demonstrate the following symmetry properties . 1 . if @xmath64 $ ] represents a solution to the dr , then so does @xmath65 $ ] . this is true because both @xmath44 and @xmath40 depend on @xmath66 only via @xmath67 , and @xmath54 appears in the absolute value bars . therefore changing the sign of @xmath66 changes the sign of @xmath68 , and consequently simultaneously changes the signs of both sides of the drs . the end result is that , the dispersion diagram expressing @xmath37 as a function of @xmath66 is symmetric about the @xmath69 axis and one needs only to consider the 1st and 4th quadrants , as will be done in figs.[fig_cor_dr ] and [ fig_pho_dr ] . for more details , please see appendix . when the external medium is at rest , if @xmath70 $ ] is a solution , then so is @xmath71 $ ] ( see eq.([eq_def_m2 ] ) with @xmath72 ) . it follows that the dispersion diagram for some positive @xmath73 is a reflection about the @xmath74 axis of the diagram for @xmath75 . actually this tendency was clear already in figs.4 and 5 in @xcite , even though the negative flow speed adopted in constructing fig.4 therein is slightly smaller in magnitude than the positive one used in generating fig.5 . more discussions on this are given in the appendix . we stress that if @xmath76 is not zero , then symmetry 2 does not hold . the dispersion behavior , or rather the wave behavior in general , critically depends on the signs of @xmath76 and @xmath73 ( see the recent review by * ? ? ? * ) ( also @xcite ) . let us consider first the coronal case , by which we mean the ordering @xmath77 holds . to be specific , we choose @xmath78 , @xmath79 and @xmath80 . taking the temperature pertaining to trace @xmath81(@xmath82 mk ) as representative , one has @xmath83 @xmath12 , and hence @xmath84 @xmath12 , and @xmath85 @xmath12 , which are very close to the values derived by . the reference value for the external alfvn speed @xmath86 is taken to be @xmath87 , corresponding to a density contrast @xmath88 ( see eq.([eq_rhoe0 ] ) which shows that when @xmath89 , @xmath90 ) . we will also examine the cases where @xmath91 is @xmath92 and @xmath93 , which correspond to a @xmath94 of @xmath95 and @xmath96 , respectively . figure [ fig_cor_dr ] presents the phase speeds @xmath37 as a function of longitudinal wavenumber @xmath66 for a series of @xmath73 . from top to bottom , figs.[fig_cor_dr]a to [ fig_cor_dr]d correspond to an @xmath97 of @xmath98 and @xmath99 , respectively . here @xmath100 expresses the internal flow speed in units of the internal sound speed . the dashed ( solid ) curves are for kink ( sausage ) modes . as shown in fig.[fig_cor_dr]b , the modes are further labeled by combinations of letters b / f+f / s+k / s , representing backward or forward , fast or slow , kink or sausage . while `` fast '' or `` slow '' is determined by the magnitude of the phase speeds , `` backward '' or `` forward '' is determined by the sign of the phase speeds when the flow is absent . the number appended to the letters denotes the order of occurrence . as such , bfk1 represents the first branch of backward fast kink mode , and likewise , ffs2 the second branch of forward fast sausage mode . on the left of each panel , the characteristic speeds external to the slab are given . although nearly indistinguishable , there are two windows where waves are prohibited if their phase speeds are such that @xmath101 . on the right of each panel , characteristic speeds interior to the slab are given . when @xmath102 lies in the intervals @xmath103 , @xmath104 , or @xmath105 , the waves are surfaces waves . otherwise , they are body ones . as was pointed out by @xcite ( hereafter nr95 ) , and also evident from fig.[fig_cor_dr ] , no surface waves are introduced by the flow in the coronal case . the rest of the section complements and expands the study by nr95 . figure [ fig_cor_dr ] shows that the dispersion diagrams demonstrate a clear dependence on the flow speed . with increasing @xmath73 , there is increasingly less symmetry between the lower and upper half of the diagrams , with a perfect symmetry taking place only for a static slab ( fig.[fig_cor_dr]a ) . for the ease of description , let us consider first the slow modes , namely the waves with @xmath37 of the order of the sound speeds . the most obvious variation with varying @xmath73 is that the slow modes are shifted upwards , encompassing always the windows @xmath106 and @xmath107 . this is understandable given that for @xmath108 , @xmath109 where @xmath110 with @xmath111 . in the opposite extreme @xmath112 , one may find @xmath113 where @xmath114 with @xmath111 . in eqs.([eq_cor_slow_smallk ] ) and ( [ eq_cor_slow_bigk ] ) , the plus and minus signs correspond to the upper and lower band , respectively . they also help explain the occurrence of the backward slow waves , labeled bsk and bss , in the upper half of figs.[fig_cor_dr]c and [ fig_cor_dr]d : even though bsk and bss have negative phase speeds in a frame comoving with the internal flow , in the current frame they have positive speeds as long as @xmath115 or @xmath116 . in view of eqs.([eq_hj ] ) and ( [ eq_gj ] ) , there actually exists an infinite number of branches of slow modes . however , in fig.[fig_cor_dr ] it is neither easy to differentiate these branches , or to distinguish between the kink and sausage modes , since @xmath117 differs only slightly from @xmath118 . given that the slow modes are nearly nondispersive , and that the deviation from unity of the period ratio @xmath0 is due entirely to wave dispersion in the current study , we will not examine @xmath0 associated with slow modes . the fast modes are more interesting since they show significant dispersion . with increasing @xmath73 , while the modes in the upper half of the @xmath119 plane become less dispersive , the opposite is true for the ones in the lower - half : the phase speeds of fast modes are bounded in the former by @xmath120 and @xmath86 , and bounded by @xmath121 and @xmath122 in the latter . it can be readily shown that this behavior results from the fact that for @xmath112 , @xmath123 on the other hand , except for the first kink branch , both kink and sausage modes have cutoffs , which are given by @xmath124 where @xmath125}{[(v_{ae}\mp u_0)^2-c_{0}^2][(v_{ae}\mp u_0)^2-v_{a0}^2 ] } } .\end{aligned}\ ] ] for the first kink branch , an analytic approximation to @xmath37 in the slender slab limit @xmath126 can be found , which reads @xmath127 where @xmath128 ^ 2 ( v_{ae}^2-c_{e}^2)}{(v_{ae}^2+c_{e}^2)v_{ae}^2 ( v_{ae}^2-c_{te}^2 ) } , \end{aligned}\ ] ] with @xmath129 and @xmath130 representing the upper and lower branches , respectively . compared with nr95 , the most relevant study to ours , we have extended the expressions pertaining to slow waves in the slender slab limit ( eq.(16 ) in nr95 ) to accommodate negative values of phase speeds ( see eq.([eq_cor_slow_smallk ] ) ) . in addition , we have examined the case of wide slabs ( @xmath112 , see eq.([eq_cor_slow_bigk ] ) ) for slow waves . as for the fast waves , eq.([eq_cor_fast_cutoff ] ) extends eq.(13 ) in nr95 in the sense that only positive phase speeds are examined in nr95 but the phase speed in the present study is allowed to be negative . besides , we have presented the expressions of the dispersion behavior in the slender slab limit for the first fast kink branch ( eq.([eq_cor_fstknk_smallk ] ) ) , and obtained new expressions for both kink and sausage waves at @xmath112 ( eq.([eq_cor_fast_bigk ] ) ) . to begin with , consider a pair of waves corresponding to @xmath131 and @xmath132 , where @xmath133 in an algebraic sense . the combination of the two results in a lagrangian displacement @xmath134 in the form @xmath135 \right\ } \nonumber \\ & + & \mathrm{re}\left\{\tilde{\xi}_{x , r}(x)\exp\left[-i\left(\omega t - k_r z\right)\right ] \right\ } . \label{eq_displacement}\end{aligned}\ ] ] by `` standing '' , we require that @xmath136 and @xmath137 , where @xmath138 and @xmath139 denote the slab ends . putting @xmath140 and @xmath138 in eq.([eq_displacement ] ) , one finds @xmath141 , and hence it makes sense to choose @xmath142 to be real . it then follows that @xmath143 \nonumber \\ & = & -2a_\xi\sin\left(\frac{k_r - k_l}{2}z\right)\sin\left(\omega t-\frac{k_l + k_r}{2}z\right ) . \label{eq_stand_displc}\end{aligned}\ ] ] for @xmath144 to be zero at arbitrary @xmath145 , this requires @xmath146 by convention , @xmath147 corresponds to the fundamental mode , and @xmath148 to its first overtone . equation ( [ eq_k4stand ] ) suggests a graphical means to determine the period ratio @xmath0 for a given aspect ratio @xmath149 . this is illustrated by figure [ fig_sol_proc ] where we use the results computed for @xmath150 as given in fig.[fig_cor_dr]b . here panels ( a ) and ( b ) are for kink and sausage modes , respectively . first , we convert the @xmath119 diagram to an @xmath151 diagram . only the first and second quadrants are relevant , with the former ( latter ) coming from the curves in the upper ( lower ) half - plane of fig.[fig_cor_dr]b . note that instead of combinations of positive @xmath66 with negative @xmath37 ( and hence negative @xmath152 ) , we obtain combinations of negative @xmath66 with positive @xmath152 by reversing the sign of @xmath66 in view of symmetry 1 . this , together with a negative @xmath37 , yields a positive @xmath152 . we then compute @xmath0 for a given @xmath149 by measuring the width between the intersections of a horizontal line with the two curves corresponding to the waves that are to form standing modes . take fig.[fig_sol_proc]a for instance . let @xmath153 and @xmath154 denote the intersections where a horizontal dashed line crosses bfk1 and ffk1 , respectively . the angular frequency @xmath155 for the fundamental mode is then found by requiring @xmath156 , while @xmath157 for its first overtone is found by requiring that @xmath158 . the period ratio is then @xmath159 . in addition , from fig.[fig_sol_proc]b one can readily see that the minimum allowed aspect ratio , namely @xmath160 , is not determined by the difference between the two cutoffs of the two @xmath151 curves involved divided by @xmath161 , but larger than that . we will come back to this point in discussing fig.[fig_cor_prssg_ma ] . furthermore , from symmetry 2 one finds that if @xmath97 is @xmath162 instead , the @xmath163 diagram will be a reflection about the vertical axis of what we have for @xmath150 . it is then clear that the period ratios do not depend on the sign of @xmath73 but only on its magnitude . at this point we note that it is impossible for a kink and a sausage wave to combine to form a standing mode . this is not obvious by examining eq.([eq_k4stand ] ) alone but has to do with the symmetric properties of the eigenfunctions . let @xmath164 and @xmath165 represent , say , a sausage and kink wave , respectively . now that the displacement for kink ( sausage ) modes is an odd ( even ) function of @xmath166 , in light of eq.([eq_stand_displc ] ) one finds that @xmath167 \\ & = & 2 a_\xi \cos\left(\frac{k_r - k_l}{2}z\right)\cos\left(\omega t-\frac{k_l + k_r}{2}z\right ) , \end{aligned}\ ] ] which does not have a permanent node at @xmath138 or @xmath139 when @xmath165 and @xmath164 satisfy eq.([eq_k4stand ] ) , hence violating the requirements for a standing mode . in computing the coronal standing modes , we consider only bfk1 and ffk1 for kink modes , and bfs1 and ffs1 for sausage modes . higher order branches like bfk2 or bfs2 are not considered since they would form standing modes only for relatively thick slabs . take ffk2 and bfk2 for instance . for them to combine to form a standing mode , the slab aspect ratio @xmath149 has to exceed @xmath168 . in the case of ffs1 and bfs2 , @xmath149 has to be larger than @xmath169 . can slow and fast kink ( or sausage ) waves form a standing mode ? for sausage modes like fss or bss to combine with modes like ffs1 or bfs1 , the slab has to be very wide . in the case of fss and bfs1 , the aspect ratio @xmath149 has to exceed @xmath170 . for kink waves , combinations like bfk1 with fsk seem possible but turn out extremely unlikely since the slow mode tends to be dominated by the intensity oscillations rather than the displacement of the slab , whereas the opposite is true for the kink mode . for slender slabs @xmath171 , the backward fast kink wave corresponds to @xmath172 , and @xmath173 , and hence @xmath174 . if one neglects the flow and notes that @xmath175 , then one finds that this ratio is @xmath176 , meaning that fast kink waves are dominated by the displacement rather than the density variation . however , for slow kink waves , by noting that for @xmath177 , @xmath178 , one finds that @xmath179 . with @xmath180 approaching @xmath181 , this will be large , meaning that slow kink waves are dominated by density variations . the end result is that , if a fast kink wave does combine with a slow one to form a standing mode , a transverse displacement on the order of the slab width will cause a relative density variation , and hence a relative intensity variation , that may exceed unity ! this makes such combinations unrealistic . figure [ fig_cor_prknk ] presents the period ratio @xmath0 as a function of the slab aspect ratio @xmath149 for standing fast kink modes . a series of values for the flow speed @xmath73 is adopted , and @xmath73 is measured in units of the internal alfvn speed @xmath182 . common to all the curves is that they start with unity at small @xmath149 , where wave dispersion is negligible , decrease with increasing @xmath149 and attain a minimum before rising again . with increasing @xmath73 , the deviation of @xmath0 from unity is enhanced substantially when compared with a static case @xmath183 . take the minima for instance . while in the static case it reads @xmath184 , when @xmath185 it is significantly reduced to @xmath186 , amounting to a relative difference of 22.8% . one may notice that the former is attained at @xmath187 , while the latter at @xmath188 . however , this is not to say that for smaller aspect ratios the effects of enhanced flow speed is negligible . for instance , for @xmath149 as small as @xmath189 , while @xmath0 reads @xmath190 in the static case , it reads @xmath168 when @xmath191 increases to @xmath192 . from this we conclude that , in agreement with , transverse structuring may contribute to the deviation from unity of @xmath0 even for thin slabs . and we note that , in addition to the density contrast which is the sole contributor to wave dispersion in , the velocity shear also plays a significant role . this has significant bearings on the applications of look - up diagrams expressing the period ratio as a function of longitudinal density stratification like the one constructed by @xcite for coronal loops . while a slab diagram has yet to appear , it seems appropriate to say before using this diagram to deduce the longitudinal stratification , one had better exclude the wave dispersion brought forth by the transverse structuring . this is especially necessary for strong transverse density contrasts , and/or in the presence of substantial flow shears . the effect of flow is further shown in fig.[fig_cor_prknk_ma ] , where ( a ) the minimal period ratio , @xmath193 , as well as ( b ) its location , @xmath194 , are shown as a function of the alfvnic mach number @xmath191 . in addition to the nominal value of @xmath91 being @xmath195 , other ratios of @xmath92 and @xmath93 are also examined and shown in different colors . consider fig.[fig_cor_prknk_ma]a first . one can see that the effect of flow on the period ratios is significant for all the considered @xmath91 , or equivalently , the density contrast . take @xmath196 for instance . while the effect of flow is slightly less pronounced than in the case where @xmath197 , introducing a flow of @xmath198 also rather significantly reduces the period ratio from the static value @xmath199 to @xmath200 , or by 18.3% . likewise , in the case with @xmath201 , a relative reduction of 20.2% is found with @xmath0 being 0.802 for @xmath183 but @xmath202 for @xmath185 . now consider fig.[fig_cor_prknk_ma]b . one may notice that , with increasing density contrast , the aspect ratio at which the minimum period ratio is attained decreases . for a given density contrast , with increasing flow this aspect ratio tends to increase , even though the effect is marginal for @xmath197 and is more pronounced with higher @xmath91 . for instance , for @xmath197 , a value of @xmath203 can be quoted for all the @xmath194 obtained , whereas for @xmath196 , @xmath194 increases substantially from 0.057 in the static case to @xmath204 when @xmath185 . furthermore , the overall tendency for both @xmath193 and @xmath194 to decrease with increasing density contrast agrees closely with fig.10 in . however , while the minimum in the static case can not go below @xmath205 as determined analytically for an epstein density profile and numerically for a step function in , it is not subject to this constraint in a flowing case . now consider standing sausage modes . figure [ fig_cor_prssg ] presents the period ratio @xmath0 as a function of aspect ratio @xmath149 for both the cases where @xmath197 ( solid curves ) and @xmath196 ( dashed curves ) . a series of @xmath73 , once again expressed in terms of the alfvnic mach number @xmath191 , is adopted and indicated by different colors . for the case @xmath197 , one can see that except for the results corresponding to the first three values of @xmath191 , the curves are absent since the range of @xmath149 in which standing sausage modes are allowed is outside the presented @xmath149 range . for the three curves that are present , while one can hardly tell them apart when their overlapping parts are concerned , it is evident that they start with considerably different values of the allowed @xmath149 . with @xmath191 increasing from @xmath18 to @xmath206 to @xmath207 , this cutoff aspect ratio @xmath160 increases from @xmath208 to @xmath209 to @xmath210 . this means that at a given @xmath91 , relative to the static case , slabs with flow can support standing sausage modes only when they are sufficiently wide . the same is also true for other choices of @xmath91 : when @xmath196 , while @xmath160 reads @xmath211 for the static case , it reads @xmath212 when @xmath213 and @xmath214 when @xmath215 . in this case , the effect of flow on the period ratio turns out to be more pronounced than in the case @xmath197 . at @xmath216 , the case with @xmath217 yields a @xmath0 of @xmath218 , which is 5.2% lower than the value @xmath219 obtained in the static case . if comparing the solid and dashed curves , one can see that the period ratio as well as the cutoff aspect ratio depend rather sensitively on @xmath91 , consistent with . here we have shown the new result that , for standing sausage modes , the flows are important in determining the ranges of aspect ratios where standing modes are allowed . for the considered density contrasts the relative changes in the period ratio due to flows are not as significant as in the standing kink cases , though . before proceeding , we note in passing that due to the strong dispersion accompanying sausage waves , it is natural to expect the period ratio @xmath0 to deviate from @xmath5 . as demonstrated by , for static coronal slabs , @xmath0 may reach values as low as @xmath220 when the density contrast increases towards infinity . how low @xmath0 is allowed to go for coronal loops with finite aspect ratios will be explored in a future study that extends the slab computations presented here and in . regardless , it is safe to say that @xmath0 will deviate in general from @xmath5 even in the absence of any longitudinal structuring since sausage waves are also strongly dispersive in a cylindrical geometry . this implies that practices employing standing sausage modes to infer longitudinal density scale heights may be problematic @xcite . figure [ fig_cor_prssg_ma ] further examines the effects of flow speed on the cutoff aspect ratio @xmath160 pertinent to the standing sausage modes . here three cases with different values of @xmath91 are examined , and presented with different colors . besides , the dashed curves represent the distribution with @xmath191 of the lower limit of @xmath160 given by @xmath221 ( see eq.([eq_cor_fast_cutoff ] ) ) . it is evident that with increasing @xmath191 , the deviation of the computed values from this limit increases substantially . note that because @xmath222 , this limit can be satisfactorily approximated by @xmath223 which has only a modest @xmath191 dependence in the parameter range considered . the fact that the computed cutoff aspect ratios have a much stronger @xmath191 dependence results from an increasingly strong asymmetry in the @xmath151 diagram , as discussed regarding fig.[fig_sol_proc]b . from an observational viewpoint , the curves presented in fig.[fig_cor_prssg_ma ] can be tested in a statistical sense if there is a large set of slab - supported standing sausage oscillations . sorting the observations into groups with similar density contrasts , and in any group constructing a parameter space comprising the aspect ratio as well as the alfvnic mach number , then one expects to see a ridge across which detections of such standing modes become drastically less frequent . a few words are in order to justify our examination of standing sausage modes for slabs with a finite width . indeed , loops in the form of bright structures appear to be thin in most ultraviolet images of the solar corona , meaning that thin - tube limits are often valid in many seismological applications . however , coronal loops with finite aspect ratios may exist . for instance , the flaring loop examined in is @xmath224 mm long , and @xmath225 mm in diameter , resulting in an aspect ratio of @xmath226 using our definition . that loop was shown to be wide enough to support a global sausage mode . however , one may ask how typical these `` fat '' structures are , given that their observations seem to be rather sparse . with this caveat in mind , the results found here can nevertheless serve as a reference for our photospheric computations to compare with . now move on to the photospheric case , by which we mean the ordering @xmath227 holds . following nr95 and @xcite , we deal only with an isolated slab immersed in an unmagnetized atmosphere : @xmath228 ( and hence @xmath229 ) . putting some absolute value for @xmath230 as @xmath231 @xmath12 , one finds @xmath232 , @xmath233 @xmath12 , which agree with the photospheric values adopted by nr95 , and fall in category ( ii ) in @xcite . figure [ fig_pho_dr ] presents , in a format similar to fig.[fig_cor_dr ] , the phase speeds @xmath37 as a function of longitudinal wavenumber @xmath66 for a series of @xmath73 . here again @xmath73 is measured in units of the internal sound speed @xmath100 , and kink ( sausage ) modes are given by the dashed ( solid ) curves . labeling different waves is slightly more complicated than in the coronal case , since now in addition to the body waves , surface waves are also permitted and can be further grouped into fast and slow ones @xcite . however , given that body waves can not combine with surface ones to form standing modes ( see section [ sec_sub_comp_pho_pratios ] ) , and propagating body waves show only a modest dispersion , the flow effect turns out to be marginal for pure standing body modes . therefore we leave them out , and label surface waves only , as indicated by fig.[fig_pho_dr]a . with the convention `` b / f+f / s+k / s '' we have , for instance , ffs for forward fast sausage waves , and bsk for backward slow kink waves . figure [ fig_pho_dr ] indicates that the dispersion diagrams possess a stronger dependence on the internal flow than in the coronal case , evidenced by the fact that some propagating windows may disappear . take the body waves for example , for which two bands exist and are shifted upwards with increasing @xmath73 . for forward ( backward ) ones , the phase speeds @xmath37 start with @xmath234 ( @xmath235 ) at @xmath126 in the same manner as described by eq.([eq_cor_slow_smallk ] ) , for both kink and sausage waves alike . with increasing @xmath236 , @xmath37 approaches @xmath237 in the way given by eq.([eq_cor_slow_bigk ] ) , the only exception is the first branch of kink waves : @xmath37 may exceed @xmath238 ( the first dashed curve from top in fig.[fig_pho_dr]a ) or fall below @xmath239 ( the first dashed curves from bottom in all panels ) , thereby rendering these body waves a surface one asymptotically . this phenomenon , termed mode crossing , was also shown in @xcite . when @xmath240 ( fig.[fig_pho_dr]c ) , the upper window is bounded from above by @xmath241 instead of @xmath242 , therefore providing a maximum allowed longitudinal wavenumber beyond which forward slow body kink waves are no longer trapped . when @xmath243 ( fig.[fig_pho_dr]d ) , this upper window disappears altogether . surface waves are examined hereafter . consider first the fast ones . one finds that for @xmath126 , only sausage waves are allowed , kink waves appear only at relatively large @xmath236 as a continuation of the body kink waves , i.e. , via mode crossing . this is readily understandable given that with @xmath108 and @xmath244 , if @xmath245 then the dr for kink waves ( eq.([eq_dr_surface ] ) ) does not allow any solution since its lhs approaches infinity . for the fast sausage wave , one has @xmath246 where @xmath247 . now consider the slow surface waves . one can readily see that for sausage waves @xmath37 starts with @xmath248 , and for small @xmath66 , @xmath249 where @xmath250 . at sufficiently strong internal flow with @xmath251 , the upper branch of slow sausage waves starts with @xmath241 instead , and the behavior of @xmath37 at small @xmath66 is identical to eq.([eq_pho_surf_fstssg_smallk ] ) with the plus sign . on the other hand , for kink waves , @xmath37 at small @xmath66 can be approximated by @xmath252 the study so far conducted in this section differs substantially from nr95 ( their section 3.3 ) , even though both studies are for isolated photospheric slabs . nr95 examined the consequences of a variable @xmath76 while keeping @xmath253 , as opposed to our study where @xmath254 but @xmath73 is allowed to vary . a situation similar to our choice was examined in @xcite albeit a cylindrical geometry is considered there . as have been discussed , when constructing standing modes , we do not employ pure body waves in view of their modest dispersion . once again , by saying it is possible or impossible for a pair of propagating waves to form a standing mode , we mean the standing mode hence constructed corresponds to a realistic density fluctuation @xmath255 in the slender slab limit @xmath108 provided that the waves have no cutoff longitudinal wavenumber . consider kink waves for @xmath108 first . to this end , let us recall that fast kink surface waves do not exist , whereas for slow kink surface waves one finds that @xmath256 by noting that @xmath257 and @xmath245 . on the other hand , the kink body waves in the limit @xmath108 is proportional to @xmath258 which approaches infinity given that @xmath259 . this means that standing kink modes can result from a combination of either pure body or pure surface waves , but not a mixture of both . therefore we are left with only one task of examining pure kink surface modes . standing sausage modes can be considered in a similar manner . once again , they can not be formed by combining a pair of body and surface waves . this is because for body waves , @xmath255 is proportional to @xmath260 and approaches infinity when @xmath236 approaches zero given that @xmath261 . for surface waves , however , one finds that at @xmath262 , for fast waves @xmath263 , whereas for slow waves @xmath264 note that the latter holds since @xmath265 when @xmath266 . nonetheless , different from the coronal case , now in the surface category fast and slow waves are allowed to form standing sausage modes . figure [ fig_pho_prknk ] presents the period ratio @xmath0 as a function of the slab aspect ratio @xmath149 for standing kink surface modes . a series of values for the flow speed @xmath73 is adopted and given by different colors . common to all curves is that they start with @xmath267 at small @xmath149 and decrease monotonically with @xmath149 towards unity . the behavior for slender slabs is in stark contrast to the coronal case where for extremely thin slabs @xmath268 the period ratio approaches unity no matter how strong @xmath73 is . what happens in the photospheric case is , irrespective of @xmath73 , @xmath0 differs substantially from unity even for zero - width slabs ! this is understandable given that for small @xmath236 the surface kink waves have phase speeds @xmath269 , leading to that the angular frequency @xmath270 is proportional to @xmath271 . note that , as illustrated by fig.[fig_sol_proc ] , the period ratios are computed via @xmath159 , where @xmath155 and @xmath157 correspond to the angular frequencies found at the separation @xmath272 being @xmath273 and twice that , respectively . this means , at small @xmath149 , @xmath274 . one can see that the effect of flow on @xmath0 is substantial for @xmath149 only in the range between @xmath189 and @xmath275 . take @xmath276 for example . while @xmath277 in the static case , it reads @xmath278 when @xmath279 , amounting to a reduction of @xmath280 . when @xmath281 , the @xmath282 dependence of phase speed dominates , leading to little difference between different curves . to the contrary , when @xmath283 , all curves show little deviation from unity since the wave dispersion is only modest ( see fig.[fig_pho_dr ] ) . consider now standing sausage surface modes . figure [ fig_pho_prssg ] presents the period ratio @xmath0 as a function of aspect ratio @xmath149 for a series of @xmath73 . distinct from the coronal case , now there is no cutoff in @xmath149 below which no standing modes can be constructed . in addition , the combinations of fast and slow waves are now allowed , resulting in four pairs of possibilities . note that instead of four , there are only two curves in figs.[fig_pho_prssg]b and [ fig_pho_prssg]d , since forward fast waves are absent in the rest of the cases considered . in the parameter range we explored , the strongest @xmath73 dependence of @xmath0 happens in the case where backward fast waves combine with forward slow waves , presented in fig.[fig_pho_prssg]c . the trend of @xmath0 with changing @xmath149 is the same for all curves in this case , decreasing from unity at small @xmath149 to some minimum and then increasing toward unity again . even in this case , the flow effect is only modest , as evidenced by the fact that for @xmath284 , the minimum @xmath0 attains is @xmath285 , only slightly smaller than @xmath286 attained in the static case . directly measuring sausage oscillations with imaging instruments was not possible until very recently @xcite ( hereafter mejm , also see @xcite @xcite , @xcite @xcite ) . by using the rapid oscillations in the solar atmosphere ( rosa ) instrument situated at the dunn solar telescope , mejm identified sausage modes in magnetic pores by examining the phase relations between the pore size and its intensity . a number of periods were present , and the authors suggested that some of the values may be attributed to the overtones of the fundamental mode , with the standing mode set up by reflections between the photosphere and transition region . let @xmath287 ( @xmath288 ) denote the period of the overtones with @xmath289 standing for the fundamental . for now , let us consider only the period @xmath290 s found in the intrinsic mode function ( imf ) c4 therein . choosing @xmath289 to be @xmath291 s pertinent to fast modes , this value would correspond to a @xmath292 in the range @xmath293 $ ] ( @xmath294 $ ] ) provided it corresponds to @xmath148 ( @xmath295 ) . note that the aspect ratio of the structure in question is @xmath296 , since the average pore size is @xmath297 km ( corresponding to an area of @xmath298 km@xmath299 ) and @xmath300 km . at this aspect ratio , the computed @xmath0 as given in our fig.[fig_pho_prssg ] indeed lies in the range that corresponds to @xmath148 , reinforcing the suggestion by mejm that it is the first overtone . nevertheless , choosing @xmath289 to be @xmath301 s pertinent to slow modes , this period would correspond to a @xmath292 in the range @xmath302 $ ] ( @xmath303 $ ] ) provided it corresponds to @xmath148 ( @xmath295 ) . while the former seems inappropriate , the latter is plausible given that the dispersion introduced by transverse structuring in density and magnetic field strength , along with the possible role of velocity shear , reduces @xmath304 from unity for photospheric sausage modes . while this comparison is admittedly inconclusive , the point we want to make is that the transverse structuring does broaden the range of the period ratios , thereby offering more possibilities in interpreting the measured oscillation periods . instead of proceeding further with this discussion , let us note that the slab model may be seen as only a first approximation of solar atmospheric structures , and the oscillations measured by rosa may correspond to propagating rather than standing waves , as was pointed out by mejm . the present study has been motivated by two strands of research in the field of solar magneto - seismology . on the one hand , exploiting the measured multiple periods is playing an increasingly important role in deducing the longitudinal structuring of solar atmospheric structures @xcite on the other hand , it was recently realized that significant flows in coronal structures reaching the alfvnic regime can bring significant revisions to the seismologically deduced physical parameters such as the coronal loop magnetic field strength @xcite . to complement and expand the only theoretical study on multiple periodicities in coronal slabs , we examine the consequences the flows may have . to this end , we numerically solve the dispersion relations for waves supported by slabs incorporating flows , and devise a graphical method to construct standing kink and sausage modes . for coronal slabs , we find that the internal flow has significant effects on the standing modes , for the fast kink and sausage ones alike . for the kink ones , they may reduce the period ratio by up to 23% compared with the static case , and the minimum allowed period ratio may fall below the analytically derived lower limit in the static case . in particular , the reduction due to a finite flow may be significant even for thin slabs . for the sausage modes , while introducing the flow leads to a reduction in the period ratio that is typically @xmath1% relative to the static case , it has significant effects on the threshold aspect ratio only above which standing sausage modes can be supported . at a given density contrast , this threshold may exceed its static counterpart by several times for the parameters considered , thereby limiting the detectability of standing sausage modes to even wider slabs . for the isolated photospheric slabs , we find that the internal flow has only a marginal effect on the period ratios @xmath0 for the surface modes , and even less for body modes . that said , standing modes supported in this case are distinct from the coronal case in a number of aspects . first , standing sausage modes may be supported by slabs with arbitrary aspect ratios and do not suffer from a cutoff @xmath149 any longer . second , for standing kink surface modes , @xmath0 deviates from unity even for a zero - width slab , which originates from the @xmath282 dependence of phase speeds of slow kink surface waves in the slender slab limit . the present study can be extended in several ways . first , to be able to directly compare with coronal loop oscillations , so far the most observed and the best understood , we need to turn to the cylindrical geometry and examine how introducing the flow affects the dispersion behavior and consequently period ratios and cutoff aspect ratios . second , allowing the tube parameters to be time - dependent has been found important as far as the seismological applications of the period ratio are concerned . this is expected to be also true when the flow speeds are dynamic rather than time - stationary . third , concerning the period ratios , one important contributor to its deviation from unity that has received much attention is the longitudinal structuring @xcite , which has yet to be considered in examining the slab period ratios . nonetheless , the main results presented here , which demonstrated the influences of the introduced flow on the standing kink and sausage modes , point to the need for further investigations into such influences . before closing , let us recall that solar magneto - seismology ( sms ) has extended well beyond the territory originally taken up by `` coronal seismology '' . one thing to note in particular , as advocated by @xcite , is to seriously consider applying sms to photospheric structures because there exist ample observations of oscillatory motions on the one hand , relevant theories are available on the other . strictly speaking , our results on the sausage modes are more applicable to the lower parts of the solar atmosphere . in this sense , our computations strengthen the notion raised by @xcite that photospheric studies in the context of sms offer equally rich physics as their coronal counterparts . we thank the anonymous referee for the constructive comments and suggestions which helped improve this paper substantially . this research is supported by the national natural science foundation of china ( 40904047 , 41174154 , and 41274176 ) , the ministry of education of china ( 20110131110058 and ncet-11 - 0305 ) , and by the provincial natural science foundation of shandong via grant jq201212 . * appendix * in this appendix , we examine the two symmetry properties presented in section [ sec_dr_oview ] in some detail . in particular , we show that for the purpose of examining the period ratios of standing modes supported by a flowing structure embedded in a static background , one needs to consider only positive longitudinal wavenumbers @xmath66 and positive internal flow speed @xmath73 . figure [ fig_cor_dr_nonzeroue ] illustrates what happens in a general case where neither @xmath73 nor @xmath76 is zero . it presents the wave phase speed @xmath37 as a function of longitudinal wavenumber @xmath66 , in a format similar to figs.[fig_cor_dr ] and [ fig_pho_dr ] . note that both positive and negative @xmath66-s are considered here . kink and sausage waves are given by the dashed and solid curves . the horizontal and vertical dash - dotted lines in purple separate the graph into four quadrants . the parameters used to construct this graph are identical to those adopted in fig.[fig_cor_dr]b , except that now the external medium is not static , @xmath305 to be specific . this rather arbitrary choice of @xmath76 is to illustrate symmetry property 1 , which in fact holds for arbitrary values of @xmath76 and @xmath73 . graphically , this property translates into the symmetry between the left half - plane ( the 2nd and 3rd quadrants ) and the right one ( the 1st and 4th quadrants ) . in this sense one needs only to consider the 1st and 4th quadrants as long as one is interested only in trapped waves . the second symmetry property is best illustrated if one compares fig.[fig_cor_dr_nonzeroue_pmu0 ] with fig.[fig_cor_dr_zeroue_pmu0 ] , where @xmath37 as a function of @xmath66 is shown for both a positive internal flow ( panel a ) and a negative one ( panel b ) with @xmath306 . the characteristic speeds are identical to those adopted when constructing fig.[fig_cor_dr ] . the difference between figs.[fig_cor_dr_nonzeroue_pmu0 ] and [ fig_cor_dr_zeroue_pmu0 ] is that in the former @xmath76 is not zero ( @xmath76 chosen to be @xmath230 for illustrative purpose ) , whereas in the latter @xmath72 . to label the waves , the same convention as in figs.[fig_cor_dr ] and [ fig_pho_dr ] is adopted . note that in this coronal case no surface modes are introduced by the non - zero external flow , regardless of the sign of the internal flow . the body waves can , as in the case where @xmath254 , be grouped into slow and fast ones . the phase speed for the former is bounded by @xmath248 and @xmath237 , and @xmath37 for the latter is bounded by @xmath307 and @xmath308 . the most important difference , as far as the symmetry properties are concerned , is that fig.[fig_cor_dr_nonzeroue_pmu0 ] lacks a symmetry between the two panels , whereas in fig.[fig_cor_dr_zeroue_pmu0 ] the curves in the 1st ( 4th ) quadrant of panel ( a ) map to those in the 4th ( 1st ) quadrant of panel ( b ) . for instance , the curve labeled bfk1 in fig.[fig_cor_dr_zeroue_pmu0]a corresponds to ffk1 in fig.[fig_cor_dr_zeroue_pmu0]b , and likewise ffk2 in fig.[fig_cor_dr_zeroue_pmu0]a corresponds to bfk2 in fig.[fig_cor_dr_zeroue_pmu0]b . as a matter of fact , this symmetry was present already in figs.4 and 5 in @xcite where wave modes supported by a flowing cylinder is examined . this is not surprising given that the authors also examined the case where the external background is static , and that the @xmath73 they adopted in constructing fig.4 ( @xmath309 ) is close to the one used for constructing fig.5 ( @xmath310 ) . a perfect symmetry would result if the authors chose a pair of @xmath73 of the same magnitude but with opposite signs . the second symmetry , which exists only in the absence of external flows , does not invalidate the concept of resonant flow instability ( rfi ) ( see @xcite @xcite and @xcite @xcite for recent reviews ) . this point can be illustrated if one examines in detail , say , @xcite where the authors address the possible instability a velocity shear between a coronal plume and its environment may introduce . the authors choose to work in the zero - beta limit , and in a frame in which @xmath254 and @xmath73 is negative . when a continuous alfvn speed profile connects the internal and external values , rfi sets in when the `` originally forward - propagating '' ( in the sense that @xmath37 is positive when @xmath311 ) kink waves reach the continuum of the originally backward - propagating alfvn waves . this happens at a @xmath73 smaller in magnitude than required by the non - resonant kelvin - helmholtz instability , the difference between the two thresholds being quite significant for high density contrasts between the internal and external media . in light of the symmetry property we discussed , what fig.3 in @xcite shows for a negative @xmath73 can help one to foresee what happens when @xmath73 is positive . in that case , rfi will also happen when @xmath73 reaches exactly the same magnitude , for at this point the originally backward - propagating kink waves reach the continuum of the originally forward - propagating alfvn waves . extending the present study to address rfi is physically very appealing , but may require quite some effort given the complexity related to solving eigenvalue problems that involve dealing with the dissipative layers at the alfvn resonances and the possible cusp resonances . figure [ fig_omegak_zeroue ] shows the wave angular frequency @xmath152 as a function of @xmath66 , constructed from the @xmath312 diagram presented in fig.[fig_cor_dr_zeroue_pmu0 ] . here figs.[fig_omegak_zeroue]a and [ fig_omegak_zeroue]b correspond to @xmath313 and @xmath314 , respectively . only kink waves are presented , for otherwise the graph will be too crowded . the first thing to note is that if one insists on @xmath152 being positive , one would need the curves in the 3rd quadrant of a @xmath119 diagram where both @xmath66 and @xmath37 are negative . however , from symmetry property 1 illustrated in fig.[fig_cor_dr_nonzeroue ] , one can construct the 3rd quadrant curves using the curves in the 4th quadrant by keeping @xmath37 unchanged while reversing the sign of @xmath66 . second , fig.[fig_omegak_zeroue ] inherits the same symmetry from fig.[fig_cor_dr_zeroue_pmu0 ] . for instance , the branch labeled bfk1 in fig.[fig_omegak_zeroue]a maps to the one labeled ffk1 in fig.[fig_omegak_zeroue]b . likewise , ffk1 in fig.[fig_omegak_zeroue]a corresponds to bfk1 in fig.[fig_omegak_zeroue]b . recall that to compute the period ratios for a given aspect ratio @xmath149 , we simply choose the two curves corresponding to the propagating waves that are to combine to form a standing mode , and then measure @xmath155 and @xmath157 that correspond to the two horizontal cuts resulting in a separation being @xmath273 and twice that , respectively . it follows from this procedure that the period ratio thus calculated would be the same in figs.[fig_omegak_zeroue]a and [ fig_omegak_zeroue]b . evidently , if we choose to work in a frame where the external medium is not static , then the symmetry between figs.[fig_omegak_zeroue]a and [ fig_omegak_zeroue]b will be broken , meaning that the period ratio for the same @xmath149 derived for a positive @xmath73 will differ from that derived for a negative one , even if the magnitudes of @xmath73 are equal . this means , at any given non - zero @xmath76 , for one to examine the effects of @xmath73 on the period ratios one would have to first calculate a full range of different @xmath73 , both positive and negative , then construct the @xmath163 diagram , and finally measure @xmath155 and @xmath157 . this certainly merits a detailed discussion , but is beyond the scope of the present study .

numerical quantum transport calculations are commonly based on a tight - binding formulation . a wide class of quantum transport algorithms requires the tight - binding hamiltonian to be in the form of a block - tridiagonal matrix . here , we develop a matrix reordering algorithm based on graph partitioning techniques that yields the optimal block - tridiagonal form for quantum transport . the reordered hamiltonian can lead to significant performance gains in transport calculations , and allows to apply conventional two - terminal algorithms to arbitrary complex geometries , including multi - terminal structures . the block - tridiagonalization algorithm can thus be the foundation for a generic quantum transport code , applicable to arbitrary tight - binding systems . we demonstrate the power of this approach by applying the block - tridiagonalization algorithm together with the recursive green s function algorithm to various examples of mesoscopic transport in two - dimensional electron gases in semiconductors and graphene . , coherent quantum transport , recursive green s function technique , block - tridiagonal matrices , matrix reordering , graph partitioning 72.10.bg , 02.70.-c , 02.10.ox 05c50 , 05c78

0806.2739

if the dimensions of a device become smaller than the phase coherence length @xmath0 of charge carriers , classical transport theories are not valid any more . instead , carrier dynamics is now governed by quantum mechanics , and the wave - like nature of particles becomes important . in general , the conductance / resistance of such a device does not follow ohm s law . in the regime of coherent quantum transport , the landauer - bttiker formalism @xcite relates the conductance @xmath1 of a device to the total transmission probability @xmath2 of charge carriers through the device , @xmath3 where @xmath4 is the transmission amplitude between different states with transverse quantum numbers @xmath5 and @xmath6 in the left and right lead , respectively . a state with a given transverse quantum number @xmath5 is also called _ channel _ @xmath5 . the problem of calculating the conductance is thus reduced to calculating scattering eigenfunctions @xmath7 for a given energy @xmath8 : @xmath9 where @xmath10 is the hamiltonian of the system . alternatively , the transmission probability can be extracted from the retarded green s function @xmath11 that obeys the equation of motion @xmath12 the fisher - lee relation @xcite then allows to calculate the transmission ( @xmath4 ) and reflection ( @xmath13 ) amplitudes from @xmath14 . in its simplest form , the fisher - lee relation reads @xmath15 and @xmath16 where @xmath17 is the velocity of channel @xmath5 and the integration runs over the cross - section @xmath18 ( @xmath19 ) of the left ( right ) lead . the landauer - bttiker formalism can also deal with multi - terminal systems , but is restricted to linear response , i.e. small bias voltages . in the general case including external bias , the conductance can be calculated using the non - equilibrium green s function formalism ( see , e.g. @xcite ) . except for particularly simple examples , solving eqs . ( [ eq : scattering_eigenstate ] ) and ( [ eq : retarded_greenfunction ] ) exactly is not possible , and therefore a numerical computation is often the method of choice . instead of solving directly a differential equation with its continuous degrees of freedom , such as the schrdinger equation , numerical computations are usually only attempted within a discrete basis set . the differential equation is then replaced by a set of linear equations , and the hamiltonian @xmath10 can be written as a matrix . very often , only few of the matrix elements @xmath20 are nonzero . such _ tight - binding _ representations of the hamiltonian are ubiquitous in quantum transport calculations and can arise from finite differences @xcite , from the finite element method @xcite , from atomic orbitals in empirical tight - binding @xcite or kohn - sham orbitals within density functional theory @xcite . when describing transport , the systems under consideration are open and thus extend to infinity . as a consequence , the tight - binding matrix @xmath10 is infinite - dimensional . however , the conductance calculation can be reduced to a finite problem by partitioning the system into a finite scattering region attached to leads that extend to infinity , as schematically depicted in fig . [ fig : grids](a ) . for the case of two - terminals , the matrix @xmath10 can be written as @xmath21 where @xmath22 is the ( infinite ) hamiltonian of the left ( right ) lead , @xmath23 is the hamiltonian of the scattering region and of finite size . the matrices @xmath24 and @xmath25 represent the coupling between the scattering region and the left and right lead , respectively . in order to reduce the problem size , it is useful to introduce the retarded self - energy @xmath26 , where @xmath27 is the surface green s function of the left ( right ) lead , i.e. the value of the green s function at the interface of the lead disconnected from the scattering region . then , the green s function @xmath28 of the scattering region can be calculated as @xcite @xmath29 reminiscent of eq . ( [ eq : retarded_greenfunction ] ) but with an effective hamiltonian @xmath30 of finite size . this treatment is easily extended to multi - terminal systems . note that it suffices to know the surface green s function of the ( semi-)infinite leads , as in a tight - binding hamiltonian the matrices @xmath31 and @xmath32 have only few nonzero entries . for simple systems , the surface green s function can be calculated analytically @xcite , whereas in more complex situations it must be computed numerically , either by iteration @xcite , or by semi - analytical formulas @xcite . the original infinite - dimensional problem has thus been reduced to a finite size matrix problem that can , in principle , be solved straight - forwardly on a computer . however , for any but rather small problems , the computational task of the direct inversion in eq . ( [ directinversion ] ) is prohibitive . therefore , for two - terminal transport , many algorithms make use of the _ sparsity _ of the hamiltonian matrix in tight - binding representation - in particular that this matrix can be written in block - tridiagonal form : @xmath33 where the index @xmath34 ( @xmath35 ) denotes the blocks in the left ( right ) lead , @xmath36 the blocks within the scattering region , and @xmath37 ( @xmath38 ) the first block in the left ( right ) lead . such a form arises , for example , naturally in the method of finite differences , when grid points are grouped into vertical slices according to their @xmath39-coordinates , as shown in fig . [ fig : grids](b ) , but also applies to any other sparse tight - binding hamiltonian . the block - tridiagonal form of the hamiltonian is the foundation of several quantum transport algorithms for two - terminal systems . the transfer matrix approach applies naturally to block - tridiagonal hamiltonians , but becomes unstable for larger systems . however , a stabilized version has been developed by usuki _ et al . _ @xcite . in the decimation technique @xcite , the hamiltonian of the scattering region is replaced by an effective hamiltonian between the two leads by eliminating internal degrees of freedom . the contact block reduction method @xcite calculates the full green s function of the system using a limited set of eigenstates . the recursive green s function ( rgf ) technique @xcite uses dyson s equation to build up the system s green s function block by block . it has also been adapted to hall geometries with four terminals @xcite and to calculate non - equilibrium densities @xcite . furthermore , the rgf algorithm has been formulated to be suitable for parallel computing @xcite . contains all blocks @xmath40 . ( b ) the green s function @xmath41 is obtained by adding another matrix block . ] of course , there are also other transport techniques not directly based on the block - tridiagonal form of the hamiltonian matrix , such as extracting the green s function from wave packet dynamics @xcite . still , such algorithms are not as widely used as the large class of algorithms , that are directly based on the block - tridiagonal form of the hamiltonian . in order to illustrate the typical computational tasks of this class of algorithms , we briefly explain , as a representative example , the rgf algorithm . the rgf technique is based on dyson s equation @xmath42 ( see , e.g @xcite ) , where @xmath14 denotes the green s function of the perturbed system , @xmath43 that of the unperturbed system and @xmath44 the perturbation . using this equation , the system is built up block by block , as depicted in fig . [ fig : rgf ] . let @xmath45 denote the green s function for the system containing all blocks @xmath40 . then , at energy @xmath8 , the green s function @xmath41 is related to @xmath45 by @xmath46 and @xmath47 starting from @xmath48 , the surface green s function of the right lead , @xmath49 slices are added recursively , until @xmath50 has been calculated . the blocks of the green s function of the full system necessary for transport are then given by @xmath51 and @xmath52 where @xmath53 is the surface green s function of the left lead . @xmath54 and @xmath55 are sufficient to calculate transmission and reflection probabilities via the fisher - lee relation , eqs . and . each step of the algorithm performs inversions and matrix multiplications with matrices of size @xmath56 . since the computational complexity of matrix inversion and multiplications scales as @xmath57 , the complexity of the rgf algorithm is @xmath58 . thus , it scales linearly with the `` length '' n , and cubically with the `` width '' @xmath56 of the system . this scaling also applies to most of the other transport algorithms mentioned above . while for particular cases general transport algorithms , such as the rgf algorithm , can not compete with more specialized algorithms , such as the modular recursive green s function technique @xcite that is optimized for special geometries , they are very versatile and easily adapted to two - terminal geometries provided that the leads are arranged collinearly . amongst other things , this restriction will be lifted by the approach presented in this work . although the block - tridiagonal structure of @xmath10 , eq . ( [ eq : blocktridiagonal ] ) , that arises naturally in many problems appears to have a small `` width '' and thus seems to be quite suitable for transport algorithms , optimizing the block - tridiagonal structure by reordering the matrix @xmath10 may lead to a significant speed - up in the conventional two - terminal algorithms , as we show below . furthermore , such a reordering allows for the application of the established two - terminal algorithms to more complex geometries , such as non - collinear leads or multi - terminal structures , that would otherwise need the development of specialized algorithms . below , we develop a matrix reordering algorithm based on graph partitioning techniques that brings an arbitrary matrix @xmath10 into a block - tridiagonal form optimized for transport calculations . to this end , the paper is organized as follows . in section [ section : algorithm ] we formulate the matrix reordering problem in the language of graph theory and develop the reordering algorithm . in section [ section : examples ] we apply this algorithm to various examples and investigate its performance and the performance of the rgf algorithm for the reordered hamiltonian @xmath10 . we conclude in section [ section : conclusions ] . as shown in the introduction , the typical runtime of transport algorithms , is proportional to @xmath59 , does depend on the particular block - tridiagonal structure of @xmath10 . therefore , the runtime of these algorithms can be improved in principle by conveniently reordering @xmath10 with a permutation @xmath60 , @xmath61 in order to quantify how a typical transport algorithm performs for a given matrix structure , we define a weight @xmath62 associated with a matrix @xmath10 as @xmath63 where @xmath56 is the size of block @xmath64 . optimizing the matrix for transport algorithms is then equivalent to minimizing the weight @xmath62 . since @xmath65 , where @xmath66 is the total number of grid points , @xmath62 is minimal , if all @xmath56 are equal , @xmath67 therefore , a matrix tends to have small weight , if the number @xmath49 of blocks is large , and all blocks are equally sized . the reordering problem of the matrix @xmath10 is thus summarized as follows : [ matrix_reordering_problem ] * matrix reordering problem : * find a reordered matrix @xmath68 such , that 1 . @xmath69 and @xmath70 are blocks given by the left and right leads ( as required by transport algorithms ) 2 . @xmath68 is block - tridiagonal ( @xmath71 , iff @xmath72 ) , 3 . the number @xmath49 of blocks is as large as possible , and all blocks are equally sized . in principle , this constrained optimization problem could be solved by generic optimization algorithms , such as _ simulated annealing_. however , for larger problems the optimization could take much more time than the actual transport calculation , rendering the optimization process useless . it is therefore necessary to use heuristics especially designed for the problem at hand . to this end , we formulate the matrix reordering problem in the language of graph theory . a _ graph _ @xmath73 is an ordered pair @xmath74 , where @xmath75 is a set of _ vertices _ @xmath76 and @xmath77 a set of ordered pairs of vertices @xmath78 . such a pair is called an _ edge_. a graph is called _ undirected _ , if for every edge @xmath79 also @xmath80 . two vertices @xmath81 and @xmath82 are called _ adjacent _ , if @xmath79 . in order to simplify the notation , we will also consider a vertex @xmath76 to be adjacent to itself . there is a natural one - to - one correspondence between graphs and the structure of sparse matrices . for a given @xmath83 matrix @xmath10 , we define a graph @xmath84 with @xmath85 and @xmath86 iff the entry @xmath87 . a graph thus stores information about the _ structure _ of a matrix , i.e. which entries are nonzero . it does not contain any information about the values of the respective entries , although these may be stored easily along with the graph . however , for the formulation of the quantum transport algorithms , only the block - tridiagonal form , i.e. the structure of the matrix , is relevant . hermitian matrices , that are considered in quantum transport , have a symmetric structure of zero and nonzero entries , and therefore the corresponding graphs are undirected . a graph can be depicted by drawing dots for each vertex @xmath76 , and lines connecting these dots for every edge @xmath88 , as shown in fig . [ fig : graphs](a ) . it should be noted that a graphical representation of a tight - binding grid , such as shown in fig . [ fig : graphs](b ) , can be directly interpreted as a representation of a graph and the corresponding matrix structure . in terms of graph theory , matrix reordering corresponds to renumbering the vertices of a graph . since we are only interested in reordering the matrix in terms of matrix blocks ( the order within a block should not matter too much ) , we define a _ partitioning _ of @xmath73 as a set @xmath89 of disjoint subsets @xmath90 such that @xmath91 and @xmath92 for @xmath93 . using these concepts , we can now reformulate the original matrix reordering problem into a graph partitioning problem : [ graph_part_problem ] * graph partitioning problem : * find a partitioning @xmath94 of @xmath73 such that : 1 . [ req1 ] @xmath95 and @xmath96 contain the vertices belonging to left and right leads , 2 . [ req2 ] 1 . vertices in @xmath95 and @xmath96 are only connected to vertices in @xmath97 and @xmath98 , respectively , 2 . for @xmath99 , there are edges between @xmath100 and @xmath101 iff @xmath72 , 3 . [ req3 ] the number @xmath102 of sets @xmath100 is as large as possible , and all sets @xmath100 have the same cardinality @xmath103 . a partitioning with all @xmath103 equally sized is called _ balanced_. a partitioning obeying requirement [ graph_part_problem].[req2 ] is called a _ level set _ with _ levels _ @xmath100 @xcite . level sets appear commonly as an intermediate step in algorithms for bandwidth reduction of matrices @xcite . these algorithms seek to find a level set of minimal width , i. e. @xmath104 as small as possible which is equivalent to requirement [ graph_part_problem].[req3 ] . the main difference between our graph partitioning problem and the bandwidth reduction problem is requirement [ graph_part_problem].[req1 ] : in the graph partitioning problem , @xmath95 and @xmath98 are determined by the problem at hand , while in the bandwidth reduction problem these can be chosen freely . due to this difference , bandwidth reduction algorithms can be applied successfully to our graph partitioning problem only for special cases , as we show below . the term _ graph partitioning _ usually refers to the general problem of finding a balanced partitioning @xmath89 of a graph and has many applications in various fields such as very - large - scale integration ( vlsi ) design @xcite , sparse matrix reorderings for lu or cholesky decompositions @xcite , or block ordering of sparse matrices for parallel computation @xcite . in particular , the latter examples also include block - tridiagonal orderings @xcite . however , as these reorderings are geared towards parallel computation , they obtain a fixed number @xmath49 of sets @xmath100 given by the number of processors of a parallel computer , whereas in our block - tridiagonal reordering the number @xmath49 should be as large as possible . in addition to that , the constraints on the blocks @xmath95 and @xmath96 ( requirement [ graph_part_problem].[req1 ] ) are again not present there . as we can not directly employ existing techniques to solve the graph partitioning problem , we will develop an algorithm combining ideas from both bandwidth reduction and graph partitioning techniques in the subsequent sections : concepts from bandwidth reduction are employed to construct a level set which is then balanced using concepts from graph partitioning . a breadth - first - search ( bfs ) @xcite on a graph immediately yields a level set @xcite . in our particular example , the level set is constructed as follows : [ bfsalgo ] level set construction by breadth - first - search . 1 . start from @xmath105 . then , @xmath106 , as the first level is given by the constraints of requirement ( [ req1 ] ) . 2 . if there is a vertex in @xmath100 that is adjacent to a vertex in @xmath96 , assign all the remaining unassigned vertices into @xmath100 and end the algorithm . all vertices adjacent to @xmath100 that are not contained in the previous levels @xmath107 are assigned to @xmath108 . 4 . continue at step b with @xmath109 . note that the sets @xmath89 form a level set by construction a set @xmath100 may only have vertices adjacent to @xmath110 and @xmath108 . the construction by bfs not only obtains the number of levels @xmath102 for a particular realization , but yields a more general information : the number of levels @xmath102 in the level set constructed by algorithm [ bfsalgo ] is the maximum number of levels compatible with the constraints on the initial and final level @xmath95 and @xmath96 for a graph @xmath73 . this can be seen from the fact that a bfs finds the shortest path in the graph between the initial sets @xmath95 and @xmath96 , @xmath111 where @xmath112 and @xmath113 . any vertex on this shortest path can be uniquely assigned to a single level @xmath100 and it would not be compatible with a larger number of levels than @xmath102 . algorithm [ bfsalgo ] not only yields the maximum number of levels : all vertices contained in the first @xmath5 levels of the bfs must be contained in the first @xmath5 levels of _ any _ other level set . [ lemma_minimal_set ] let @xmath114 be a level set constructed by algorithm [ bfsalgo ] , and @xmath115 another level set consistent with the requirements of problem [ graph_part_problem ] with @xmath116 . then @xmath117 for @xmath118 . the statement is proved by induction . it is true trivially for @xmath119 ( because of requirement [ req1 ] in problem [ graph_part_problem ] ) and for @xmath120 ( then the levels cover the whole graph ) . suppose now that the statement holds for @xmath121 . note that for the proof it suffices to show that @xmath122 . consider now the set of all vertices adjacent to @xmath123 , @xmath124 . by construction , @xmath125 . since @xmath126 and @xmath127 is a level set , all vertices adjacent to @xmath123 must be contained in the set of vertices including the next level , i.e. @xmath128 . but then also @xmath122 , which concludes the proof . . different levels are shown in alternating shades of grey . ] thus , the vertices contained in the first @xmath5 levels of the bfs form a minimal set of vertices needed to construct @xmath5 levels . however , this also implies that the last level which then covers the remaining vertices of the graph , may contain many more vertices than the average , leading to an unbalanced level set . this is not surprising , since the algorithm does not explicitly consider balancing and only local information is used , i.e. whether a vertex is adjacent to a level or not . an example for this imbalance is shown in fig . [ example_bfs ] , the bfs construction yields a very large last level . note that throughout the manuscript we visualize the graph theoretical concepts using examples of graphs obtained from discretizing a two - dimensional structure . however , the ideas and algorithms presented here apply to any graph and are not limited to graphs with coordinate information . two - dimensional graphs have the advantage of being visualized easily . in particular , the bfs search has an intuitive physical analog : wave front propagation of elementary waves emanating from the vertices of the initial level @xmath95 . the problem that a bfs does not yield a balanced partitioning was also noted in the theory of bandwidth reduction . the gibbs - poole - stockmeyer ( gps ) algorithm tries to overcome this deficiency by constructing a level set through the combination of two bfs searches starting from the initial and the final levels . however there the initial and final levels are sought to be furthest apart , contrary to our problem . in general , the gps construction only yields a balanced level set if the initial and final level are close to furthest apart , as we will show in sec . [ section : examples ] . in order to obtain a balanced partitioning , graph partitioning algorithms commonly perform a recursive bisection , i.e. successively bisect the graph and the resulting parts until the desired number of parts is obtained @xcite . this approach has the advantage of reducing the partitioning problem to a simpler one , i.e. bisection . furthermore , if the resulting parts of every bisection are equally sized , the overall partitioning will be balanced . in addition , bisection is inherently a global approach , as the whole graph must be considered for splitting the system into two equally sized parts . thus , it can be expected to yield better results than a local approach , such as bfs . we intent to construct a level set with @xmath102 levels , where @xmath102 is the maximum number of levels as determined by algorithm [ bfsalgo ] . to this end we start from an initial partitioning @xmath129 , where @xmath95 and @xmath96 contain the vertices of the leads ( requirement [ graph_part_problem].[req1 ] ) , and @xmath97 all other vertices . the level set is then obtained by applying the bisection algorithm recursively to @xmath97 and the resulting subsets , until @xmath49 levels are obtained , as shown schematically in fig . [ fig : recursive_bisection_example ] . here bisection means splitting a set @xmath100 into two sets , @xmath130 and @xmath131 , such that @xmath132 and @xmath133 . in oder to be applicable to the graph partitioning problem [ graph_part_problem ] , the bisection must comply with certain requirements : [ bisection_prob ] the bisection algorithm must be 1 . [ bisect_req1 ] compatible with a level set with @xmath102 levels . [ bisect_req2 ] balanced . [ bisect_req3 ] performed such that subsequent bisections may lead to a balanced level set . requirement [ bisection_prob].[bisect_req3 ] is formulated rather vaguely : usually there are many different choices how to perform a bisection . a particular choice will influence the subsequent bisections ( for a similar problem in graph partitioning see @xcite ) , and thus the bisection algorithm must in principle take all following bisection steps into account . since an exact solution to that problem seems computationally intractable , we will resort to heuristics there . we start explaining how to comply with requirements [ bisection_prob].[bisect_req1 ] and [ bisection_prob].[bisect_req2 ] . in the following we assume that @xmath134 , as @xmath135 are trivial cases . then the initial partitioning @xmath129 forms a level set , and so will the final result of the recursive bisection , if the result of every intermediate bisection yields a level set . for this , consider a set @xmath100 with vertices adjacent to the sets @xmath136 and @xmath137 , where `` left''(``right '' ) is defined as being closer to @xmath95 ( @xmath96 ) . then the sets resulting from the bisection , @xmath130 and @xmath131 may only have vertices adjacent to @xmath136,@xmath131 and @xmath130,@xmath137 , respectively . apart from the condition of forming a level set , requirement [ bisection_prob].[bisect_req1 ] also dictates the total number of levels . due to the nature of the recursive bisection , the number of final levels contained in an intermediate step is always well - defined . if a set @xmath100 contains @xmath138 levels , then @xmath130 and @xmath131 must contain @xmath139 and @xmath140 levels , respectively . here , @xmath141 denotes rounding off to the next smallest integer . the bisection is thus balanced , if @xmath142 note that @xmath138 can take any value , and usually is not a power of two . from lemma [ lemma_minimal_set ] we know that the minimum set of vertices necessary to form @xmath5 levels is given by a bfs up to level @xmath5 . let @xmath143 ( @xmath144 ) denote the set of vertices found by a bfs starting from @xmath136 ( @xmath137 ) up to level @xmath145 ( @xmath146 ) . then , for any bisection complying with requirement [ bisection_prob].[bisect_req1 ] , @xmath147 and @xmath148 . these vertices are uniquely assigned to @xmath130 and @xmath131 and are consequently marked as _ locked _ , i.e. later operations may not change this assignment . an example for the vertices found in a bfs is shown in fig . [ fig : bisection_explanation](a ) . note that in the initial bisection , @xmath149 , @xmath150 , @xmath151 , and @xmath152 . ( black ) and @xmath144 ( dark grey ) that must be contained in @xmath130 and @xmath131 , respectively . the remaining vertices ( light grey ) can be assigned to either set . ( b ) and ( c ) : examples illustrating the difference between cut edges and cut nets . the number of cut edges is 5 in both ( b ) and ( c ) , while the number of cut nets ( boundary vertices ) is 10 in ( b ) and 9 in ( c ) . ] the remaining unassigned vertices can be assigned to either set , and the bisection will still be compatible with a level set containing @xmath102 vertices . thus for complying with requirement [ bisection_prob].[bisect_req2 ] , any prescription obeying the balancing criterion may be used . we choose to distribute the remaining vertices by continuing the bfs from @xmath136 and @xmath137 and assigning vertices to @xmath130 and @xmath131 depending on their distance to the left or right neighboring set , while additionally obeying the balancing criterion . this approach assigning vertices to levels according to their distance from the initial and final set is rather intuitive and probably the procedure that would be used if the level set were to be constructed `` by hand '' . this procedure may lead to reasonable level sets , however in general , additional optimization on the sets @xmath130 and @xmath131 is needed , as discussed below . if this optimization is used , it can also be useful to distribute the unassigned vertices randomly , as this may help avoiding local minima . as mentioned above , there is a lot of arbitrariness in distributing the unassigned vertices into @xmath130 and @xmath131 . however , the particular choice of the bisection will influence whether a later bisection is balanced or not : if @xmath153 contains more vertices than the balance criterion , the bisection _ can not _ be balanced . obviously , the bfs that constructs @xmath153 depends on the details of the set @xmath100 and thus on the details of the previous bisection step . in order to formulate a criterion that may resolve the above mentioned arbitrariness and help to find a balanced level set , it is useful to consider the matrix representation of the graph @xmath73 . bisecting a graph means ordering the corresponding matrix into two blocks that are connected by an off - diagonal matrix @xmath154 : @xmath155 this off - diagonal matrix will be unchanged by further bisections and thus determines the minimum level width that can be achieved . therefore , the size of the off - diagonal matrix @xmath154 should be minimized . in a bisection , an edge @xmath156 is said to be _ cut _ , if @xmath81 and @xmath82 belong to different sets , i.e. @xmath157 and @xmath158 or vice versa . the entries of @xmath154 correspond to edges cut by the bisection , and minimizing the number of entries in @xmath154 corresponds to minimizing the number of edges cut by the bisection ( _ min - cut criterion _ ) . this criterion is often used in reordering matrices for parallel processing , where the off - diagonal matrix size determines the amount of communication between processors . however , the number of entries in @xmath154 is not directly related to the size of the matrix , as has been noted in the graph partitioning problem for parallel computing @xcite . instead , the size of the off - diagonal matrix is given by the number of _ surface _ vertices , i.e. the number of vertices that have cut edges . for this , we define a _ net _ of a vertex @xmath76 in a graph @xmath84 as @xcite @xmath159 note that @xmath160 , as @xmath76 is adjacent to itself . a net is said to be cut by a bisection , if any two vertices @xmath161 are contained in different sets @xmath130 and @xmath131 . then , the number of surface vertices and thus the size of the off - diagonal matrix @xmath154 is given by the number of cut nets . thus , minimizing the number of cut nets ( _ min - net - cut criterion _ ) corresponds to minimizing the the number of surface vertices , and thus to minimizing the size of the off - diagonal matrix @xmath154 . furthermore , since the vertices in @xmath162 are determined by a bfs emanating from the surface vertices , minimizing the number of cut nets will usually also lead to a smaller number of vertices in @xmath162 , leaving more freedom towards achieving a balanced bisection . figs . [ fig : bisection_explanation](b ) and ( c ) show a comparison of the min - cut and min - net - cut criterion for simple examples . in practice , when minimizing the number of cut nets , we also use the min - cut criterion to break ties between different bisections with the same number of cut nets ( _ min - net - cut - min - cut criterion _ ) in order to avoid wide local minima , that occur frequently in the min - net - cut problem . both the min - cut and min - net - cut bisection problem have been shown to be @xmath163-hard @xcite . therefore , only heuristics are available to solve them . these heuristics start from an initial ( balanced ) bisection , such as constructed by the steps outlined above , and improve upon this initial bisection . here , we choose to use the fiduccia - mattheyses ( fm ) algorithm @xcite , as it is readily available for min - cut and min - net - cut bisection . in fact , min - net - cut bisection is a _ hypergraph _ partitioning problem , and the fm algorithm was originally designed for hypergraph partitioning . furthermore , the fm algorithm can naturally deal with locked vertices that may not be moved between sets , is reasonable fast and its underlying concepts are easy to understand . the fm heuristic is a _ pass - based _ technique , i.e. it is applied repeatedly to the problem ( several _ passes _ are performed ) , iteratively improving the bisection . more detailed information about the fundamentals of the fiduccia - mattheyses algorithm are given in appendix [ sec : fm_explanation ] . we now summarize the steps outlined above and formulate an algorithm for bisection : [ bisection_algo ] bisection of set @xmath100 containing @xmath138 levels , with left ( right ) neighboring set @xmath136 ( @xmath137 ) . 1 . stop , if @xmath164 . 2 . do a bfs starting from @xmath136 up to level @xmath139 and a bfs starting from @xmath137 up to level @xmath165 . the vertices found by the bfs are assigned to @xmath130 and @xmath131 , respectively , and are marked as locked . 3 . distribute the remaining unassigned vertices taking into account the balance criterion . the vertices may be assigned according to either one of the following prescriptions : 1 . continue the bfss from step b and assign vertices to @xmath130 , if they are first reached by the bfs from @xmath136 , and to @xmath131 , if they are first reached by the bfs from @xmath137 . if a set has reached the size given by the balance criterion , assign all remaining vertices to the other set . 2 . distribute the unassigned vertices randomly to @xmath130 and @xmath131 . if a set has reached the size given by the balance criterion , assign all remaining vertices to the other set . 4 . optimize the sets @xmath130 and @xmath131 by changing the assignment of unlocked vertices according to some minimization criterion . in particular , the following optimizations may be performed : 1 . no optimization . min - cut optimization using the fm algorithm . min - net - cut optimization using the fm algorithm . min - net - cut - min - cut optimization using the fm algorithm . recursive application of the bisection algorithm [ bisection_algo ] then leads to an algorithm for constructing a level set complying with the requirements of the graph partitioning problem [ graph_part_problem ] , and thus an algorithm for block - tridiagonalizing a matrix . [ graph_part_algo ] block - tridiagonalization of matrix @xmath10 1 . construct the graph @xmath84 corresponding to the matrix @xmath10 , and the sets @xmath95 and @xmath96 corresponding to the leads . 2 . use algorithm [ bfsalgo ] to determine the maximum number of levels @xmath102 . if @xmath166 , stop . 3 . construct @xmath167 , containing @xmath49 levels . 4 . apply the bisection algorithm [ bisection_algo ] to @xmath97 and then recursively on the resulting subsets . do not further apply if a set only contains one level . it should be emphasized , that the block - tridiagonalization does not require any other input than the graph structure . in principle , the number of fm passes may affect the result . however , from experience , this number can be chosen as a fixed value , e.g. 10 fm passes , for all situations @xcite . thus , the block - tridiagonalization algorithm can serve as a black box . ) , and application of the block - tridiagonalization algorithm [ graph_part_algo ] with distribution of vertices by bfs ( algorithm [ bisection_algo ] , step c.(a ) ) ( b ) without further optimization , ( c ) with min - cut optimization , ( d ) with min - net - cut optimization . ] in fig . [ fig : algo_application ] we show for examples of level sets arising from the natural ordering of grid points ( fig . [ fig : algo_application](a ) , _ natural level set _ ) and from the block - tridiagonalization algorithm developed in this work ( fig . [ fig : algo_application](b)(d ) ) for the case of a disk - type geometry . the level set in fig . [ fig : algo_application](b ) arises from recursive bisection , where the vertices were distributed according to a bfs without any optimization . the resulting level set strongly resembles the natural level set . this is due to the highly symmetric structure and the fact that vertices are assigned to levels according to their distance from the leads only small deviations are present due to the balance criterion . when the bisection is optimized according to the min - cut criterion , fig . [ fig : algo_application](c ) , the resulting level set changes little , as the min - cut criterion favors horizontal and vertical cuts for a square lattice , as presented in the example . in contrast , min - net - cut optimization ( fig . [ fig : algo_application](d ) ) yields a new , non - trivial level set that has less symmetry than the underlying structure . note that the minimization of surface vertices leads to levels in the form of `` droplets '' , analogous to surface tension in liquids . in fact , we will show in sec . [ section : examples ] that min - net - cut optimization usually leads to level sets and thus block - tridiagonal orderings that are superior to those arising from other methods . in particular , they are better than the natural level sets , leading to a significant speed - up of transport algorithms , as demonstrated in sec . [ sec : examples_mesoscopics ] . in addition to that , the reordering algorithms allow one to use conventional two - terminal transport algorithms also for more complicated , even multi - terminal structures ( see secs . [ sec : examples_mesoscopics ] and [ sec : examples_multiterminal ] ) . we conclude the theoretical considerations with an analysis of the computational complexity of algorithms [ bisection_algo ] and [ graph_part_algo ] . the bisection algorithm involves a bfs search on @xmath100 , which scales linearly with the number of edges within @xmath100 , and thus has complexity @xmath168 , where @xmath169 is the set of edges within @xmath100 . in addition to that , a single optimization pass of the fm algorithm scales also as @xmath168 @xcite . usually , a constant number of passes independent of the size of the graph is enough to obtain converged results , and therefore the optimization process of several fm passes is also considered to scale as @xmath168 . thus , the full bisection algorithm also has complexity @xmath168 . usually , the number of edges per vertex is approximately homogeneous throughout the graph . since the recursive bisection is a divide - and - conquer approach , the computational complexity of the full block - tridiagonalization algorithm is then @xmath170 @xcite . in typical graphs arising from physics problems , the number of edges per vertex is a constant , the computational complexity can also be written as @xmath171 , where @xmath172 is the number of vertices in @xmath44 , or the size of the matrix @xmath10 . in contrast , many quantum transport algorithms , such as the recursive green s function technique , scale as @xmath173 in the optimal case of @xmath49 equally sized matrix blocks ( levels ) of size @xmath174 . often , the number of blocks ( levels ) @xmath175 . typically , to name a few examples , @xmath176 in one - dimensional chains , @xmath177 in two dimensions , and the transport calculation scales as @xmath178 . thus , except for the case of a linear chain , where @xmath179 and matrix reordering is pointless anyways , the block - tridiagonalization algorithm always scales more favorably than the quantum transport algorithms . this scaling implies that the overhead of the matrix reordering in the transport calculation will become more negligible , the larger the system size . we now evaluate the performance of the block - tridiagonalization algorithm using representative examples from mesoscopic physics . the schrdinger equation for the two - dimensional electron gas ( 2deg ) is usually transformed into a tight - binding problem by the method of finite differences @xcite , where the continuous differential equation is replaced by a set of linear equations involving only the values of the wave function on discrete grid points . commonly , these points are arranged in a regular , square grid . this grid , together with the shape of the particular structure under consideration then defines the structure of the hamilton matrix and the corresponding graph . denotes a length characterizing the extent for the different structures . ] the representative examples considered here are shown in fig . [ fig : examples_2deg ] : the circle ( fig . [ fig : examples_2deg](a ) ) and the asymmetric sinai billiard ( fig . [ fig : examples_2deg](b ) ) that are examples of integrable and chaotic billiards in quantum chaos , the ring ( fig . [ fig : examples_2deg](c ) ) that may exhibit various interference physics , and the circular cavity with leads that are not parallel ( fig . [ fig : examples_2deg](d ) ) as an example of a structure that does not have an intuitive , natural block - tridiagonal ordering . for all these structures , we introduce a length scale @xmath180 , given by the outer radius of the circular structures and the side length of the square structure , characterizing the maximum extent . the fineness of the grid , and thus the size of the corresponding graph will be measured in number of grid points per length @xmath180 . we now apply the block - tridiagonalization algorithm using the various optimization criteria discussed in the previous section , and compare the resulting orderings with the natural ordering and the ordering generated by the gps algorithm . the weights @xmath62 , eq . , of the different orderings are given in table [ tab : weights ] . > m0.22||>m0.16|>m0.16|>m0.16|>m0.16 & circular billiard&asymmetric sinai billiard&ring&cavity with perp . leads + natural block- + tridiagonal ordering & @xmath181&@xmath182&@xmath183&@xmath184 + gibbs - poole - stockmeyer&@xmath185&@xmath186&@xmath187&@xmath188 + distribution by bfs , no optimization&@xmath189&@xmath190&@xmath191&@xmath192 + distribution by bfs , min - cut&@xmath189&@xmath193&@xmath191&@xmath194 + random distribution , min - cut&@xmath195&@xmath196&@xmath191&@xmath197 + distribution by bfs , min - net - cut&@xmath189&@xmath198&@xmath191&@xmath199 + random distribution , min - net - cut&@xmath200&@xmath201&@xmath202&@xmath199 + distribution by bfs , min - net - cut - min - cut&@xmath203&@xmath204&@xmath205&@xmath206 + random distribution , min - net - cut - min - cut&@xmath207&@xmath208&@xmath202&@xmath209 + the initial distributions for the bisection algorithm are done in two different ways : the vertices are distributed both in an ordered way by bfs and randomly . the outcome after the optimization however is always similar for both types of initial distributions which indicates that the resulting weights are close to the global minimum and not stuck in a local minimum . note that we use twice as many fm passes for a random initial distribution than for an initial distribution by bfs , as convergence is usually slower for a random initial distribution . in all examples , the min - net - cut criterion yields orderings with the best weights , as expected from the considerations of the previous section . based on the weight , orderings according to this criterion are expected to give the best performance in transport calculations such as the rgf algorithm . note that the min - net - cut - min - cut ordering is on average closest to the best ordering . the min - net - cut ordering sometimes suffers from slow convergence , when the algorithm must traverse a wide local minimum . the additional min - cut criterion helps to break ties and thus avoids these wide local minima . except for the ring , where all algorithms perform well , the gps algorithm yields weights that are even larger than the weight of the natural ordering . as discussed above , the gps algorithms performs well , if both leads are furthest apart in terms of the graph . in the case of the ring , this is approximately fulfilled . in the general case , when the leads are at arbitrary positions , the gps algorithm usually produces some very large levels . as the level size enters cubically in the @xmath62 , this results in a prohibitively large weight . the gps algorithm thus can not be used as a generic reordering algorithm for quantum transport according to problem [ graph_part_problem ] . in summary , the block - tridiagonalization algorithm [ graph_part_algo ] in the combination of initial distribution by bfs and min - net - cut - min - cut optimization yields the best reorderings with respect to the weight @xmath62 . experience shows that usually 10 fm passes are enough for optimizing a bisection . as a consequence , we will use this combination exclusively in the rest of this work . as a function of the level ( matrix block ) index @xmath210 for the natural level set ( dashed line ) and the min - net - cut - min - cut reordering ( solid line ) , shown for ( a ) the circle billiard , ( b ) the asymmetric sinai billiard , ( c ) the ring , and ( d ) the circular cavity with perpendicular leads . note that for ( d ) , there is no natural ordering . in all examples , there were 400 grid points per length @xmath211 . ] the weight @xmath62 of a matrix is a global measure of the quality of a ordering . additional insight can be gained from the distribution of the sizes @xmath56 of the matrix blocks / levels . in fig . [ fig : blocksizes ] we show this distribution before and after reordering . for the natural ordering of the finite difference grids , the number of matrix blocks is determined by the number of lattice points along the @xmath39-coordinate direction ( see fig . [ fig : grids](b ) ) . in contrast , the number of matrix blocks after reordering is given by the length of the shortest path between the two leads , in terms of the corresponding graph . in the case of the circle billiard , fig . [ fig : blocksizes](a ) , the number of matrix blocks is the same for the natural ordering and the reordered matrix , as the shortest path between the leads is simply a straight line along the @xmath39-coordinate direction . the improvements in the weight originate only from balancing the matrix block sizes : while the matrix block sizes vary for the natural ordering the lateral size changes along the @xmath39-direction the reordered matrix has equally sized matrix blocks . for this particular example , the result of the block - tridiagonalization algorithm is optimal , as it yields the best solution with respect to the requirements set forth in problems [ matrix_reordering_problem ] and [ graph_part_problem ] . note that in general it is not always possible to find a perfectly balanced partitioning , but the circle billiard is such an example . in contrast , in the case of the asymmetric sinai billiard and the ring the number of matrix blocks generated by the block - tridiagonalization algorithm is larger than in the natural ordering ( see figs . [ fig : blocksizes](b ) and ( c ) , respectively ) . in both cases , the obstacle within the scattering region increases the length of the shortest path connecting the two leads . in both examples , this increase in the number of matrix blocks leads to a significantly decreased weight @xmath62 with respect to the natural ordering , although the partitioning is only approximately balanced . for instance , in the particular case of the ring , the number of matrix blocks after reordering is approximately given by the number of lattice points around half of the circumference . the reordered ring thus has a weight very similar to a straight wire with a width twice as large as the width of one arm of the ring , and a length given by half of the ring circumference . for the cavity with perpendicular leads , there is no natural ordering , and a specialized transport algorithm would be required . the reordering creates a matrix with approximately balanced block sizes , and allows the direct application of conventional algorithms . the weight @xmath62 was introduced as a theoretical concept in order to simulate the computational complexity of a transport calculation . after discussing the influence of the reordering on this theoretical concept , we now demonstrate how the reordering increases the performance of an actual quantum transport calculation . to this end we use a straight - forward implementation of the well - established recursive green s function algorithm for two terminals , as described in ref . the necessary linear algebra operations are performed using the atlas implementation of lapack and blas @xcite , optimized for specific processors . it should be emphasized that the code that does the actual transport calculation such as calculation of the green s function and evaluation of the fisher - lee relation is the same for all examples considered here , including the non - trivial cavity with perpendicular leads . the abstraction necessary for the reordering , i.e. the graph structure and the corresponding level set , allows for a generic computational code applicable to any tight - binding model . , eq . , through the reordering as a function of the grid size for the circular billiard , the asymmetric sinai billiard , and the ring , respectively . @xmath212 is shown excluding ( @xmath213 ) and including ( @xmath214 ) the overhead of matrix reordering . the estimate for @xmath212 from the weights @xmath62 of the different orderings is shown as a dashed line . ( d ) : fraction of time @xmath215 , eq . , used for reordering the matrix as a function of the grid size . data is shown for the circular billiard ( @xmath213 ) , the asymmetric sinai billiard ( @xmath214 ) , the ring ( @xmath216 ) , and the circular cavity with perpendicular leads ( @xmath217 ) . the benchmarks were run on pentium 4 processor with 2.8 ghz and 2 gbs of memory . ] we measure the performance gain through matrix reordering as @xmath218 note that during a real calculation , the conductance is usually not only calculated once , but repeatedly as a function of some parameters , such as fermi energy or magnetic field . thus , the respective quantum transport algorithm is executed repeatedly , too . in contrast , the block - tridiagonalization has to be carried out again _ only _ when the structure of the matrix and thus the corresponding graph changes . for the examples considered here this would correspond to changing the grid spacing or the shape of the structure . in such a case , the overhead of matrix reordering must be taken into account for @xmath212 . this overhead can be quantified as @xmath219 in a typical calculation however , the matrix structure given by the underlying tight - binding grid does not change , and the matrix reordering must be carried out only once . in this common situation , the overhead of matrix reordering is negligible . for example , any change of physical parameters such as fermi energy , magnetic field or disorder averages does not change the matrix structure . in fig . [ fig : reordering_speedup ] we show the performance gain through matrix reordering , @xmath212 , as a function of grid size for the circle billiard , the asymmetric sinai billiard , and the ring ( figs . [ fig : reordering_speedup](a)(c ) , respectively ) . we include both measurements excluding and including the overhead of matrix reordering , as discussed above . remember that in the case of the cavity with perpendicular leads , fig . [ fig : examples_2deg](d ) , there is no natural ordering and thus a performance comparison is not possible . in fact for this system , only matrix reordering makes a transport calculation possible in the first place . we find that block - tridiagonalization always increases the algorithmic performance in the typical situation , when the overhead of matrix reordering can be neglected . however , even if the reordering overhead is taken into account , we see a significant performance gain except for small systems but there the total computing time is very short anyway . in fact , as the system sizes increases , the overhead of reordering becomes negligible , as predicted from the analysis of the computational complexity , and the performance gains including and excluding the reordering overhead converge . this can also be seen in fig . [ fig : reordering_speedup](d ) , where we show the reordering overhead @xmath215 as a function of system size . especially for large systems , the total computing time can become very long , and any performance gain is beneficial . reordering leads to significant performance gains up to a factor of 3 in the case of the ring . the performance gain @xmath212 can also be estimated from the weights @xmath62 of the original matrix ( the natural ordering ) and the reordered matrix , shown as the dashed line in figs . [ fig : reordering_speedup](a)(c ) . the actual , measured performance gain approaches this theoretical value , as the system size increases . note that we do not fully reach the theoretically predicted performance gain in the case of the ring . on modern computer architectures , computing time does not only depend on the number of arithmetic operations @xcite , and thus the weight @xmath62 overestimates the performance gain , though the performance still improves significantly . , for the circular billiard ( @xmath213 ) , the asymmetric sinai billiard ( @xmath214 ) , the ring ( @xmath216 ) , and the circular cavity with perpendicular leads ( @xmath217 ) . the solid line is a fit to the predicted scaling of the computational complexity , @xmath220 . ] finally , we demonstrate the @xmath171 scaling of the reordering algorithm . [ fig : reordering_scaling ] shows the computing times of the block - tridiagonalization algorithm as a function of matrix / graph size @xmath49 for the geometries considered in this section . for all systems , the computing times scale according to the prediction from the complexity analysis in sec . [ sec : complexity_analysis ] , as apparent from the fit @xmath221 . note that for large @xmath172 , @xmath171 scaling is practically indistinguishable from @xmath222 , as can also be seen in fig . [ fig : reordering_scaling ] . in the examples of this section , we considered the pedagogic case of charge transport on a square , finite difference grid . the approach presented here can however immediately applied to more complex situations , such as spin transport , as reviewed in ref . in addition , extending the transport calculation to a different grid is straightforward , as any tight - binding grid can be encoded into a graph , and the block - tridiagonalization algorithm has already been applied to the case of the hexagonal grid of graphene @xcite ( for a review on graphene see @xcite ) . a further example of this versatility is shown in the next section , where we apply the block - tridiagonalization algorithm to solve multi - terminal structures involving different tight - binding models . in the previous section , we demonstrated that matrix reordering increases the performance of quantum transport algorithms for two - terminal structures and additionally makes it possible to apply these conventional algorithms to non - trivial structures . whereas there is a great variety of quantum transport algorithms for systems with two leads , there are only few algorithms that are suitable for multi - terminal structures , and most of these are restricted to rather specific geometries ( e.g. ref . only recently algorithms have been developed that claim to be applicable to any multi - terminal structure . the _ knitting algorithm _ of ref . @xcite is a variant of the rgf algorithm where the system is built up adding every lattice point individually , instead of adding whole blocks of lattice points at a time . therefore , instead of a matrix multiplication , the central computational step is an exterior product of vectors . unfortunately , this implies that the knitting algorithm can not use highly optimized matrix multiplication routines ( level 3 blas operations ) , that are usually much more efficient than their vector counterparts ( level 2 blas operations ) , as discussed in ref . another multi - terminal transport algorithm presented recently @xcite , is based on the transfer matrix approach . however , it requires the hamiltonian to be in a specific block - tridiagonal form , and the corresponding level set is set up manually . here we show how to employ the block - tridiagonalization algorithm in order to apply the well - established two - terminal quantum transport algorithms to an arbitrary multi - terminal system . the basic idea is sketched in fig . [ fig : multiterminal_virtualleads](a ) : combining several ( real ) leads into only two virtual leads the multi - terminal problem is reduced to an equivalent two - terminal problem . after reordering , the resulting problem can then be solved by conventional two - terminal algorithms . note that in this approach the number of matrix blocks is given by the shortest path between leads in two different virtual leads . if all leads are very close together , this may lead to only few , large blocks in the reordered matrix and respectively levels in the graph partitioning , leading to a very large weight @xmath62 . in such a case it is advisable to collect all leads into a single virtual lead . the second virtual lead is then formed by a vertex in the graph , that is furthest away from all leads as depicted in fig . [ fig : multiterminal_virtualleads](b ) . such a vertex can be found by a bfs search originating from all leads . thereby the number of matrix blocks / levels is maximized . in fact , this approach yields a block - tridiagonal matrix structure as required by the algorithm of ref . @xcite . is shown as a function of @xmath223 , where @xmath224 is the width of the hall bar and @xmath225 the cyclotron radius in a magnetic field @xmath226 . note that @xmath227 . the dotted lines indicate the quantized values of the hall resistance , @xmath228 , where @xmath5 is a positive integer . ] we now demonstrate these strategies on the example of the quantum hall effect ( qhe ) in a 2deg formed in a semiconductor heterostructure @xcite and in graphene @xcite . for this we use a four - terminal hall bar geometry as sketched in fig . [ fig : multiterminal_virtualleads](a ) , on top of a square lattice ( finite difference approximation to 2deg ) and a hexagonal lattice . again , it should be emphasized that the code of the actual transport calculation is the same as employed in the two - terminal examples of the previous section . the results of the calculation are shown in fig . [ fig : example_halleffect ] , where the integer qhe of the 2deg and the odd - integer qhe of graphene are clearly visible . the methods outlined above make it possible to calculate quantum transport in _ any _ system described by a tight - binding hamiltonian . this generality is one of the main advantages gained by using the matrix reordering . however , generality also implies that it is difficult to make use of properties of specific systems , such as symmetries , in order to speed up calculations . special algorithms developed specifically for a certain system however can , and will usually be faster than a generic approach at the cost of additional development time . in the case of the hall geometry in a 2deg , such a special algorithm was presented by baranger et al . @xcite , and we have implemented a variant of it . comparing the computing times for the hall bar geometry in a 2deg , we find that the special algorithm is only a factor of @xmath229 faster than our generic approach . although such a performance comparison may depend crucially on the details of the system under consideration , experience shows that the use of the generic approach often does not come with a big performance penalty . we have developed a block - tridiagonalization algorithm based on graph partitioning techniques that can serve as a preconditioning step for a wide class of quantum transport algorithms . the algorithm can be applied to any hamilton matrix originating from an arbitrary tight - binding formulation and brings this matrix into a form that is more suitable for many two - terminal quantum transport algorithms , such as the widely used recursive green s function algorithm . the advantages of this reordering are twofold : first , the reordering can speed up the transport calculation significantly . second , it allows for applying conventional two - terminal algorithms to non - trivial geometries including non - collinear leads and multi - terminal systems . the block - tridiagonalization algorithm scales as @xmath171 , where @xmath172 is the size of the hamilton matrix , and thus induces only little additional overhead . we have demonstrated the performance of the matrix reordering on representative examples , including transport in 2degs and graphene . the block - tridiagonalization algorithm can operate as a black box and serve as the foundation of a generic transport code that can be applied to arbitrary tight - binding systems . such a generic transport code is desirable , as it minimizes development time and increases code quality , as only few basic transport routines are necessary , that can be tested thoroughly . we acknowledge financial support from dfg within grk638 and sfb689 . the fiduccia - mattheyses algorithm was originally developed for hypergraph partitioning @xcite . a hypergraph @xmath230 is an ordered pair @xmath231 , where @xmath75 is a set of vertices , and @xmath232 a set of _ nets _ @xmath233 ( also called _ hyperedges _ ) between them . a net @xmath233 is a set of vertices , i.e. @xmath234 . an undirected graph is a special realization of a hypergraph , where every net contains exactly two vertices . thus , any algorithm for hypergraph partitioning can also be applied to an undirected graph . during the fm bisection , we have to consider the graph structure arising from the hamiltonian matrix in order to minimize the number of cut edges ( min - cut ) , whereas for minimizing the number of surface vertices , i.e. the number of cut nets ( min - net - cut ) , the hypergraph structure arising from all nets @xmath235 as defined in eq . , @xmath236 , is essential . for min - net - cut - min - cut optimization , we have to consider both structures simultaneously . a schematic representation of a graph and the corresponding hypergraph structure is shown in fig . [ fig : example_graph_hypergraph ] . the fm algorithm is based on the concept of _ gain_. the gain of a vertex in an existing bisection is defined as the change in weight , i.e. the number of cut edges or nets , that occurs when this vertex is move to the other part . this gain can also be negative , if such a move increases the number of cut edges or nets . the basic idea of the fm algorithm is to swap vertices with the highest gain between parts , while obeying some balance criterion . the fact that the highest gain can be negative , helps the fm algorithm to escape local minima . after moving , the respective vertex is locked in order to avoid an infinite loop , where a single vertex might be swapped back and forth repeatedly . the fm pass ends , when all ( free ) vertices have been moved , and the best bisection encountered during the pass is returned as result . further passes can then successively improve on this bisection . p. havu , v. havu , m. j. puska , r. m. nieminen , nonequilibrium electron transport in two - dimensional nanostructures modeled using green s functions and the finite - element method , phys . rev . b 69 ( 11 ) ( 2004 ) 115325 . s. sanvito , c. j. lambert , j. h. jefferson , a. m. bratkovsky , general green39s - function formalism for transport calculations with @xmath237 hamiltonians and giant magnetoresistance in co- and ni - based magnetic multilayers , phys . b 59 ( 18 ) ( 1999 ) 1193611948 . m. luisier , a. schenk , w. fichtner , g. klimeck , atomistic simulation of nanowires in the @xmath238 tight - binding formalism : from boundary conditions to strain calculations , phys . rev . b 74 ( 20 ) ( 2006 ) 205323 . a. di carlo , a. pecchia , l. latessa , t. frauenheim , g. seifert , tight - binding dft for molecular electronics ( gdftb ) , in : g. cuniberti , g. fagas , k. richter ( eds . ) , introducing molecular electronics , springer , berlin , heidelberg , 2006 , pp . 153184 . t. usuki , m. saito , m. takatsu , r. a. kiehl , n. yokoyama , numerical analysis of ballistic - electron transport in magnetic fields by using a quantum point contact and a quantum wire , phys . rev . b 52 ( 11 ) ( 1995 ) 82448255 . c. m. fiduccia , r. m. mattheyses , a linear - time heuristic for improving network partitions , in : dac 82 : proceedings of the 19th conference on design automation , ieee press , piscataway , nj , usa , 1982 , pp . 175181 . m. wimmer , m. scheid , k. richter , spin - polarized quantum transport in mesoscopic conductors : computational concepts and physical phenomena , to appear in the encyclopedia of complexity and system science , arxiv:0803.3705v1 ( 2008 ) . k. s. novoselov , a. k. geim , s. v. morozov , d. jiang , m. i. katsnelson , i. v. grigorieva , s. v. dubonos , a. a. firsov , two - dimensional gas of massless dirac fermions in graphene , nature 438 ( 7065 ) ( 2005 ) 197200 .

the mathematical description of the quantum harmonic oscillator is essentially based on the gaussian function . in the case of a quantum oscillator with finite - dimensional hilbert space , the position space consists in a finite number of points , and the ground state can be regarded as a finite version of the gaussian function . conversely , each finite gaussian can be used in order to define some finite oscillators . we investigate certain finite gaussians as concern their fourier transform , wigner function and associated oscillators .

1512.02589

in quantum mechanics and quantum information there exists an increasing interest in quantum systems with finite - dimensional hilbert space . the finite - dimensional versions of the mathematical objects used in the continuous case are expected to play an important role . in section 2 , we review some fundamental mathematical tools available in a finite - dimensional hilbert space for the description of quantum systems . particularly , we present some results concerning the kravchuk functions in a version suitable for our purpose . in the case of a quantum system with finite - dimensional hilbert space there are several candidates for the role of gaussian function . in section 3 we investigate five of them as concern their finite fourier transform and the corresponding wigner function . in section 4 , each of these finite gaussians is used in order to generate a finite tight frame representing a finite counterpart of the canonical system of coherent states . in the first part of section 5 , we review some known one - dimensional finite oscillators . more examples can be found in @xcite . in the second part of section 5 , based on the considered finite gaussians and the finite frame quantization , we define some new finite oscillators . in section 6 , we investigate the occurrence of revivals . some explicit gaussians and oscillators are presented in section 7 . our aim is to bring a contribution to the mathematical formalism available for the description of the quantum systems with finite - dimensional hilbert space . let @xmath0 and @xmath1 . any @xmath2-dimensional hilbert space is isomorphic to @xmath3 , and can be regarded as the space of all the complex functions defined on a set with @xmath2 elements . we consider @xmath4 with the scalar product defined as @xmath5 in order to define certain mathematical objects , we have to embed @xmath3 in the space of all the functions of the form @xmath6 by identifying it either with the space @xmath7 of periodic functions of period @xmath2 or with the space @xmath8\!=\!\ { \ , \psi \!:\!\mathbb{z}\!\longrightarrow \!\mathbb{c}\ \ , |\ \ , \psi(n)\!=\!0\ \ \mbox{for}\ \ n\!\not\in \!\{-j , -j\!+\!1,\ , ... \ , , j\!-\!1 , j\}\ , \}\ ] ] of all the functions null outside @xmath9 . here @xmath10 is the ring of integers modulo @xmath2 . depending on the used identification , a function @xmath11 is extended up to a periodic function with period @xmath2 or up to a function null outside @xmath9 . conversely , any function @xmath12 from @xmath13 or @xmath14 $ ] coresponds uniquely to a function @xmath11 , which can be obtained by taking the restriction to @xmath9 . the scalar product in all the three spaces @xmath3 , @xmath13 and @xmath14 $ ] is defined by the same formula , namely ( [ scalprod ] ) . the hilbert spaces @xmath3 , @xmath13 and @xmath14 $ ] are isomorphic , but they offer distinct formal advantages . the evolution of a quantum particle along an axis is described by the hilbert space @xmath15 with the scalar product defined as @xmath16 similarly , the evolution of a quantum particle along @xmath9 is usually described by using one of the hilbert spaces @xmath3 , @xmath13 and @xmath14 $ ] . + the _ fourier transform _ @xmath17 $ ] , where @xmath18(k)=\frac{1}{\sqrt{d}}\sum_{n =- j}^j { \rm e}^{-\frac{2\pi { \rm i}}{d}kn}\ , \psi(n),\ ] ] is a unitary transformation . its inverse is the adjoint transformation @xmath19(k)=\frac{1}{\sqrt{d}}\sum_{n =- j}^j { \rm e}^{\frac{2\pi { \rm i}}{d}kn}\ , \psi(n).\ ] ] we have @xmath20(k)\!=\!\psi(-k)$ ] , and @xmath21\!=\!f[\psi]$ ] for any even function @xmath22 . the fourier transform @xmath23 corresponding to @xmath24 is also defined by ( [ deff ] ) . the functions @xmath25 , defined by the relation @xmath26 0 & \mbox{for } & n\!\neq \!k , \end{array } \right.\ ] ] form an orthonormal basis in @xmath3 and @xmath14 $ ] , called the canonical or computational basis . the corresponding basis in @xmath13 , written as @xmath27 , is defined by the relation @xmath28 0 & \mbox{for } & n\!\neq \!k\ ( { \rm modulo}\ , d ) . \end{array } \right.\ ] ] by writing @xmath29 instead of @xmath30 , we have @xmath31 and the resolution @xmath32 of the identity operator @xmath33 . the functions @xmath34\}_{k =- j}^j$ ] form a complementary orthonormal basis . by writing @xmath35 instead of @xmath36 $ ] , we have @xmath37 and @xmath38 as concern the fourier transform , we have @xmath39 f^2\!=\!\!\sum\limits_{n =- j}^j\!\ ! |j ;- n\rangle \langle j;n|\!=\!(f^+)^2,\qquad & f^2\!=\!\!\sum\limits_{n =- j}^j\!\ ! |j ;- n\rangle\!\rangle\langle\ ! \langle j;n|\!=\!(f^+)^2 . \end{array}\ ] ] the self - adjoint operators @xmath40 , where @xmath41 admit the spectral decompositions @xmath42 the schwinger s unitary operators @xcite @xmath43 b={\rm e}^{\frac{2\pi { \rm i}}{d}q}=\sum\limits_{n =- j}^j{\rm e}^{\frac{2\pi { \rm i}}{d}n}|j;n\rangle \langle j;n| \end{array}\ ] ] can be simpler analyzed by using @xmath44 instead of @xmath3 . they satisfy the relations @xcite @xmath45 a^\alpha|j;k\rangle\!\rangle = { \rm e}^{-\frac{2\pi { \rm i}}{d}k\alpha}|j;k\rangle\!\rangle , \qquad & b^\beta |j;k\rangle\!\rangle = |j;k\!+\!\beta\rangle\!\rangle , \\[2 mm ] ( a^\alpha \psi ) ( n)=\psi ( n\!-\!\alpha ) , & ( b^\beta \psi ) ( n)={\rm e}^{\frac{2\pi { \rm i}}{d}n\beta}\psi ( n ) , \\[2 mm ] a^d = b^d=\mathbb{i } , & a^\alpha b^\beta = { \rm e}^{-\frac{2\pi { \rm i}}{d}\alpha \beta } b^\beta a^\alpha , \end{array}\ ] ] for any @xmath46 and @xmath47 . the @xmath48 operators @xmath49 satisfying the relation @xmath50 define a projective representation of a finite version of the heisenberg - weyl group . the convolution of two functions @xmath51 is the function @xmath52 , @xmath53 and we have @xmath54(k ) & \!\!\!\!=\frac{1}{\sqrt{d}}\sum\limits_{n , m =- j}^j{\rm e}^{-\frac{2 \pi { \rm i}}{d}kn}\ , \varphi(m)\ , \psi(n - m)\\[2 mm ] & \!\!\!\!=\frac{1}{\sqrt{d}}\sum\limits_{m,\ell =- j}^j{\rm e}^{-\frac{2 \pi { \rm i}}{d}k(m+\ell)}\ , \varphi(m)\ , \psi(\ell ) , \end{array}\ ] ] that is , @xmath55=\sqrt{d}\ , f[\varphi]\ , f[\psi].\ ] ] the _ discrete wigner function _ @xmath56 , @xmath57 corresponding to a function @xmath58 satisfies the relation @xmath59(m\!-\!k)\ , \overline{f[\psi](m\!+\!k)}\ ] ] which implies that @xmath60 } ( n , m)\!=\!\frac{1}{d}\sum\limits_{k =- j}^j\mathrm{e}^{-\frac{4\pi \mathrm{i}}{d}nk}\ , \psi(-m\!+\!k)\ , \overline{\psi(-m\!-\!k)}.\ ] ] particularly , in the case of an even function @xmath22 we have @xmath61 } ( n , m)=w_{\!\psi } ( m ,- n).\ ] ] by admitting for gamma function that @xmath62 we get an extended definition for the binomial coefficients @xmath63 0 & \mbox{for } & n\!\in \!\mathbb{z}\backslash\ { 0,1,2 , ... , k\}. \end{array}\right.\ ] ] particularly , the function @xmath64 belongs to the space @xmath65 $ ] . for any @xmath66 , the coefficient of @xmath67 in @xmath68 is the polynomial of degree @xmath69 @xmath70 called _ kravchuk polynomial_. we have @xmath71 , and @xmath72 the first three polynomials are : @xmath73 , @xmath74 , @xmath75 . + * theorem 1*. _ the polynomial functions @xmath76 , @xmath77\ ] ] satisfy the relation + @xmath78 for any @xmath79 . _ + _ proof_. direct consequence of the polynomial relation @xmath80 \mbox{}\qquad = \frac{1}{2^{2j}}\sum\limits_{n =- j}^j c_{2j}^{j+n}\ , \sum\limits_{m =- j}^{j}k_k(n)\ , x^{j+m}\ , \sum\limits_{\ell =- j}^{j}k_\ell ( n)\ , x^{j+\ell } \\[2 mm ] \mbox{}\qquad = \frac{1}{2^{2j}}\sum\limits_{n =- j}^j c_{2j}^{j+n}\,(1\!-\!x)^{j+n}(1\!+\!x)^{j - n}(1\!-\!y)^{j+n}(1\!+\!y)^{j - n } \\[2 mm ] \mbox{}\qquad = \frac{1}{2^{2j}}[(1\!-\!x)(1\!-\!y)+(1\!+\!x)(1\!+\!y)]^{2j}= ( 1+xy)^{2j } \\[2 mm ] \mbox{}\qquad = \sum\limits_{m =- j}^j c_{2j}^{j+m}x^{j+m } y^{j+m}.\qquad \opensquare \end{array}\ ] ] the _ kravchuk functions _ @xmath81 , @xmath82 satisfying the relation @xmath83 , form an orthonormal basis in @xmath44 : @xmath84 * theorem 2*. _ the _ kravchuk functions _ @xmath85 satisfy the relation @xmath86 and particularly , we have _ @xmath87 _ proof_. it is known that the hypergeometric function @xmath88 where @xmath89 , satisfies the relation @xcite @xmath90 since @xmath91 , the direct consequence of the previous formula @xmath92 can be written as @xmath93 in @xmath65 $ ] , by admitting that @xmath94 we have @xmath95 * theorem 3*. _ in @xmath65 $ ] , the kravchuk function @xmath96 satisfies the relation _ _ proof_. by differentiating ( [ defkpoly ] ) we get the relation @xmath98 that is , the polynomial equality @xmath99 leading to @xmath100 the corresponding recurrence relation for the kravchuk functions is @xmath101 the operators @xmath102\longrightarrow \ell^2[-j , j]$ ] defined as @xcite @xmath103 j_+|j;m\rangle = \sqrt{(j\!-\!m)(j\!+\!m\!+\!1)}\ |j;m\!+\!1\rangle \\[1 mm ] j_-|j;m\rangle = \sqrt{(j\!+\!m)(j\!-\!m\!+\!1)}\ |j;m\!-\!1\rangle \end{array}\ ] ] satisfy the relations @xmath104=\pm j_{\pm},\qquad [ j_-,j_+]=-2j_z\ ] ] and define an irreducible representation of the group @xmath105 . the operators @xmath106 satisfying the relations @xmath107={\rm i}j_z,\qquad [ j_y , j_z]={\rm i}j_x,\qquad [ j_z , j_x]={\rm i}j_y,\ ] ] define an irreducible representation of the rotation group @xmath108 . the operator corresponding to the rotation of angle @xmath109 around a unit vector @xmath110 is @xmath111 . by using the relation @xmath112 the equality ( [ reck2 ] ) can be written in the form @xmath113 equivalent to @xmath114 the _ kravchuk transform _ @xmath115\longrightarrow \ell^2[-j , j]$ ] , @xmath116 is a unitary transform such that @xmath117 . it also satisfies the relations @xmath118 we have @xmath119 but @xmath120 is not the only unitary operator satisfying this relation . the generalized kravchuk transform @xmath121\longrightarrow \ell^2[-j , j]$ ] , @xmath122 satisfies for any real numbers @xmath123 the relation @xmath124 a finite system @xmath125 formed by @xmath126 non - null vectors from @xmath3 is a _ frame _ if there exist two constants @xmath127 such that @xmath128 by using the frame operator @xmath129 this relation can be written in the form @xmath130 that is @xmath131 a frame with @xmath132 , called a _ tight frame _ , satisfies the relation @xmath133 by denoting @xmath134 and @xmath135 , the resolution of the identity ( [ frame - id ] ) becomes @xmath136 we have @xmath137 particularly , if all @xmath138 are equal , then @xmath139 each frame @xmath140 satisfying ( [ frame - idd ] ) defines an embedding of @xmath3 into a larger space @xmath141 . in the hilbert space @xmath141 , the vectors @xmath142 defined by using an orthonormal basis @xmath143 in @xmath141 , form an orthonormal system . + our hilbert space @xmath3 can be identified with the subspace @xmath144 by using the isometry @xmath145 after identification , @xmath146 satisfies the relation @xmath147 where @xmath148 is the orthogonal projector @xmath149 thus , the frame @xmath140 is the normalized projection of the orthonormal basis @xmath143 . + by using the frame @xmath140 , we can associate to a function @xmath150 the operator @xmath151 and to a linear operator @xmath152 the function @xmath153 some useful mathematical objects can be defined by using the quantization @xmath154 and the dequantization @xmath155 procedures . the probability distribution @xmath156 corresponding to @xmath157 and the discrete wigner function @xmath158 defined by ( [ wg ] ) , in the case @xmath159.,title="fig : " ] the probability distribution @xmath156 corresponding to @xmath157 and the discrete wigner function @xmath158 defined by ( [ wg ] ) , in the case @xmath159.,title="fig : " ] the gaussian functions of the form @xmath160 , where @xmath161 and @xmath162 are parameters , play an important role in quantum mechanics , signal processing and mathematics . they belong to the hilbert space @xmath163 of the square integrable functions , and the corresponding norm can be evaluated exactly @xmath164 the direct / inverse fourier transform of a function @xmath165 is the function @xmath166:\mathbb{r}\longrightarrow \mathbb{c},\qquad \mathcal{f}^{\pm 1}[f](p)=\frac{1}{\sqrt{2\pi } } \int_{-\infty } ^\infty { \rm e}^{\mp { \rm i}pq}\ , f(q)\ , dq\ ] ] and the corresponding wigner function is the function @xmath167 the fourier transform of a gaussian function is also a gaussian function @xmath168=\frac{1}{\sqrt{\kappa } } \ , g^{(1/\kappa ) } \end{array}\ ] ] and the corresponding wigner function is a product of gaussian functions @xmath169 in this section , our aim is to investigate the even functions ( see fig . 1 - 4 ) @xmath170 @xmath171 @xmath172 @xmath173 @xmath174 @xmath175 where @xmath176 is a parameter . we call _ finite gaussians _ the functions of the form @xmath177 the probability distribution @xmath178 corresponding to @xmath179 and the discrete wigner function @xmath180 defined by ( [ wgp ] ) , in the case @xmath159.,title="fig : " ] the probability distribution @xmath178 corresponding to @xmath179 and the discrete wigner function @xmath180 defined by ( [ wgp ] ) , in the case @xmath159.,title="fig : " ] * theorem 4*. _ the fourier transform of a finite gaussian is a finite gaussian : _ @xmath181=\frac{1}{\sqrt{\kappa}}\ , \mathfrak{g}_1^{(1/\kappa ) } , \end{array}\ ] ] @xmath182\!=\!\frac{1}{\sqrt{\kappa}}\ , \mathfrak{g}_3^{(1/\kappa)}\quad \quad and \qquad \qquad f[\mathfrak{g}_3^{(\kappa)}]=\frac{1}{\sqrt{\kappa}}\ , \mathfrak{g}_2^{(1/\kappa ) } , \end{array}\ ] ] @xmath183=\mathfrak{g}_5\qquad \quad\quad and \qquad \qquad \quad f[\mathfrak{g}_5]=\mathfrak{g}_4 . \end{array}\ ] ] _ proof_. a ) a proof can be found in @xcite . + b ) the periodic function @xmath184 @xmath185 with period @xmath2 admits the fourier expansion + @xmath186 with + @xmath187 & = \frac{1}{d}\sum\limits_{\alpha = -\infty } ^\infty \int _ 0^d\mathrm{e}^{-\frac{2\pi \mathrm{i}}{d}\ell x } \mathrm{e}^{-\frac{\kappa}{2}\left(\sqrt{\frac{2\pi } { d}}\ , \left(\left(\alpha + \frac{1}{2}\right)d + x\right)\right)^2 } dx . \\[-1 mm ] \end{array}\ ] ] by denoting @xmath188 we get @xcite @xmath189 & = \frac{(-1)^\ell } { \sqrt{2\pi d}}\sum\limits_{\alpha = -\infty } ^\infty \int_{(\alpha -1/2)\sqrt{2\pi d}}^{(\alpha + 1/2)\sqrt{2\pi d}}\mathrm{e}^{-\mathrm{i}\ell t\sqrt{\frac{2\pi } { d } } } \mathrm{e}^{-\frac{\kappa}{2}t^2}\ , dt\\[4 mm ] & = \frac{(-1)^\ell } { \sqrt{2\pi d}}\int_{-\infty } ^{\infty}\mathrm{e}^{-\mathrm{i}\ell t\sqrt{\frac{2\pi } { d } } } \mathrm{e}^{-\frac{\kappa}{2 } t^2}\ , dt=\frac{(-1)^\ell } { \sqrt{\kappa d}}\ , \mathrm{e}^{-\frac{\pi}{\kappa d}\ell ^2 } \end{array}\ ] ] whence @xmath190 particularly , we have @xcite @xmath191 & \!\!\!\!=\frac{1}{\sqrt{\kappa d}}\sum\limits_{n =- s}^{s}\sum\limits_{\alpha = -\infty } ^\infty \mathrm{e}^{\frac{2\pi \mathrm{i}}{d}k(\alpha d+n ) } \,(-1)^{(\alpha d+n ) } \mathrm{e}^{-\frac{\pi } { \kappa d}(\alpha d+n ) ^2}\ , \\[4 mm ] & \!\!\!\!=\!\frac{1}{\sqrt{d}}\sum\limits_{n =- s}^{s}\!\!\mathrm{e}^{\frac{2\pi \mathrm{i}}{d}kn } \frac{(-1)^n}{\sqrt{\kappa } } \!\!\sum\limits_{\alpha = -\infty } ^\infty \!\!(-1)^\alpha \ , \mathrm{e}^{-\frac{\pi } { \kappa d}(\alpha d+n ) ^2 } \!\!=\!\frac{1}{\sqrt{\kappa } } f[\mathfrak{g}_3^{(1/\kappa)}](k ) . \end{array}\ ] ] c ) we have @xmath192(k ) & \!\!\!\!=\frac{1}{\sqrt{d}}\frac{1}{2^{2j}}\sum\limits_{n =- j}^j \frac{(2j)!}{(j\!-\!n)!\ , ( j\!+\!n ) ! } { \rm e}^{-\frac{2\pi { \rm i}}{d}kn}\\[3 mm ] & \!\!\!\!=\frac{1}{\sqrt{d}}\frac{1}{2^{2j } } \left({\rm e}^{\frac{\pi { \rm i}}{d}kj}\!+\!{\rm e}^{-\frac{\pi { \rm i}}{d}k}\right)^{2j}=\frac{1}{\sqrt{d}}\ , \cos^{2j}\frac{k\pi } { d}=\mathfrak{g}_5(k).\qquad \opensquare \end{array}\ ] ] in this paper , we are mainly interested in the normalized gaussians ( see fig . 1 - 4 ) @xmath193 satisfying the relations @xmath194\!=\!\mathfrak{g}_1,\quad f[\mathfrak{g}_2]\!=\!\mathfrak{g}_3,\quad f[\mathfrak{g}_3]\!=\!\mathfrak{g}_2,\quad f[\mathfrak{g}_4]\!=\!\mathfrak{g}_5,\quad f[\mathfrak{g}_5]\!=\!\mathfrak{g}_4.\ ] ] due to the periodicity , the distribution of probability @xmath178 corresponding to @xmath195 has the shape of a gaussian function ( see fig . [ gpw ] ) . in this case the maximum value is taken in the neighbouring points @xmath196 and @xmath197 . the function @xmath198 takes positive as well as negative values , but the corresponding distribution of probability @xmath199 has the shape of a gaussian function ( see fig . [ gmw ] ) . the probability distribution @xmath199 corresponding to @xmath200 and the probability distribution @xmath201 corresponding to @xmath202 , in the case @xmath159.,title="fig : " ] the probability distribution @xmath199 corresponding to @xmath200 and the probability distribution @xmath201 corresponding to @xmath202 , in the case @xmath159.,title="fig : " ] the jacobi theta function @xcite @xmath203 has several remarkable properties among which we mention : @xmath204 @xmath205 and @xcite @xmath206 by using ( [ theta3prop ] ) we get @xmath207 that is , the formula @xmath208 ruzzi s relation ( [ ruzzi ] ) is equivalent to our formula ( [ fgk ] ) . by using the jacobi function @xmath209 and ( [ theta3prop ] ) we get the relation @xmath210 & \!\!\!\!=\ ! \sqrt{\kappa d}\ \mathrm{e}^{-\frac{\kappa \pi } { d } \left(n+\frac{d}{2}\right)^2 } \theta _ 3\left ( -{\rm i}\kappa \left(n+\frac{d}{2}\right ) , { \rm i}\kappa d\right)\\[2 mm ] & \!\!\!\!=\sqrt{\kappa d}\sum\limits_{\alpha = -\infty } ^\infty \mathrm{e}^{-\frac{\kappa \pi } { d } \left(\left(\alpha + \frac{1}{2}\right)d + n\right)^2 } \end{array}\ ] ] equivalent to @xmath211 in a similar way , by using the jacobi function @xmath212 and ( [ theta3prop ] ) we obtain the formula @xmath213 & \!\!\!\!=\ ! \sqrt{\kappa d}\ \mathrm{e}^{-\frac{\pi}{4\kappa d}+\frac{\pi { \rm i}}{d}n}\ \mathrm{e}^{-\frac{\kappa \pi } { d } \left(n+\frac{\rm i}{2\kappa}\right)^2 } \theta _ 3\left ( -{\rm i}\kappa n+\frac{1}{2 } , { \rm i}\kappa d\right)\\[2 mm ] & \!\!\!\!=\sqrt{\kappa d}\sum\limits_{\alpha = -\infty}^{\infty}(-1)^\alpha { \rm e}^{-\frac{\kappa\pi}{d}(\alpha d+n)^2 } \end{array}\ ] ] equivalent to @xmath214 by writing @xmath215 ^ 2},\\[3 mm ] \mathfrak{g}_3^{(\kappa)}(2m)\!=\!\sum\limits_{\alpha = -\infty}^{\infty}(-1)^{2\alpha } { \rm e}^{-\frac{\kappa\pi}{d}(2\alpha d+2m)^2}\!+\!\sum\limits_{\alpha = -\infty}^{\infty}(-1)^{2\alpha+1 } { \rm e}^{-\frac{\kappa\pi}{d}[(2\alpha+1 ) d+2m]^2 } , \end{array}\ ] ] we get the relations @xmath216 \mathfrak{g}_3^{(\kappa)}(2m)=\mathfrak{g}_1^{(4\kappa)}(m)-\mathfrak{g}_2^{(4\kappa)}(m ) . \end{array}\ ] ] if we separate an absolutely convergent series as @xmath217 and use the substitutions @xmath218 and @xmath219 we obtain @xmath220 the probability distribution @xmath221 corresponding to @xmath222 and the discrete wigner function @xmath223 , in the case @xmath159 . , title="fig : " ] the probability distribution @xmath221 corresponding to @xmath222 and the discrete wigner function @xmath223 , in the case @xmath159 . , title="fig : " ] * theorem 5*. _ the wigner functions corresponding to @xmath224 , @xmath225 and @xmath226 are sums + of products of finite gaussians ( see fig . 1 - 2 ) : _ @xmath227 & \ + \frac{1}{\sqrt{2\kappa d}}\ , \mathfrak{g}_2^{(2\kappa)}(n)\ , \mathfrak{g}_1^{(2/\kappa ) } ( m)\!-\!\frac{1}{\sqrt{2\kappa d}}\ , \mathfrak{g}_2^{(2\kappa)}(n)\,\mathfrak{g}_2^{(2/\kappa ) } ( m ) , \end{array}\ ] ] @xmath228 & \ + \frac{1}{\sqrt{2\kappa d}}\ , \mathfrak{g}_2^{(2\kappa)}(n)\ , \mathfrak{g}_1^{(2/\kappa ) } ( m)\!+\!\frac{1}{\sqrt{2\kappa d}}\ , \mathfrak{g}_2^{(2\kappa)}(n)\,\mathfrak{g}_2^{(2/\kappa ) } ( m ) , \end{array}\ ] ] @xmath229 & \ -\frac{1}{\sqrt{2\kappa d}}\ , \mathfrak{g}_2^{(2\kappa)}(n)\ , \mathfrak{g}_1^{(2/\kappa ) } ( m)\!+\!\frac{1}{\sqrt{2\kappa d}}\ , \mathfrak{g}_2^{(2\kappa)}(n)\,\mathfrak{g}_2^{(2/\kappa ) } ( m ) . \end{array}\ ] ] * proof*. a ) a proof can be found in @xcite . + b ) by using ( [ fgk ] ) , ( [ fgkpm ] ) , ( [ g2n ] ) and ( [ doublesum ] ) we get @xmath230 & \ , \ + \frac{1}{d}\sum\limits_{k =- s}^s\mathrm{e}^{\frac{4\pi \mathrm{i}}{d}mk}\ , \sum\limits_{\mu , \eta = -\infty } ^\infty \mathrm{e}^{-\kappa \frac{\pi } { d } ( ( \mu + \eta + \frac{3}{2 } ) d + n - k)^2}\ \mathrm{e}^{-\kappa \frac{\pi } { d } ( ( \mu -\eta + \frac{1}{2 } ) d + n+k)^2}\\[4 mm ] & \!\!\!\!=\frac{1}{d}\sum\limits_{k =- s}^s\mathrm{e}^{\frac{4\pi \mathrm{i}}{d}mk}\ , \sum\limits_{\mu , \eta = -\infty } ^\infty \mathrm{e}^{-2\kappa \frac{\pi } { d } ( ( \mu + \frac{1}{2})d + n)^2}\ \mathrm{e}^{-2\kappa \frac{\pi } { d } ( \eta d -k)^2}\\[4 mm ] & \ , \ + \frac{1}{d}\sum\limits_{k =- s}^s\mathrm{e}^{\frac{4\pi \mathrm{i}}{d}mk}\ , \sum\limits_{\mu , \eta = -\infty } ^\infty \mathrm{e}^{-2\kappa \frac{\pi } { d } \left(\mu d + n\right)^2}\ \mathrm{e}^{-2\kappa \frac{\pi } { d } \left(\left(\eta + \frac{1}{2}\right ) d -k\right)^2}\\[4 mm ] & \!\!\!\!=\frac{1}{\sqrt{d}}\ , \ , \mathfrak{g}_2^{(2\kappa)}(n)\ , \ , f[\mathfrak{g}_1^{(2\kappa)}](2 m ) + \frac{1}{\sqrt{d}}\ , \ , \mathfrak{g}_1^{(2\kappa)}(n)\,\ , f[\mathfrak{g}_2^{(2\kappa)}](2m)\\[4 mm ] & \!\!\!\!=\!\frac{1}{\sqrt{2\kappa d}}\ , \ , \mathfrak{g}_2^{(2\kappa)}(n ) \left ( \mathfrak{g}_1^{(2/\kappa)}(m)\!+\!\mathfrak{g}_2^{(2/\kappa)}(m)\right)\!+\ ! \frac{1}{\sqrt{2\kappa d}}\ , \mathfrak{g}_1^{(2\kappa)}(n)\!\left ( \mathfrak{g}_1^{2/\kappa ) } ( m)\!-\!\mathfrak{g}_2^{(2/\kappa)}(m)\right).\\[4 mm ] \end{array}\ ] ] c ) since @xmath231 $ ] and @xmath232 is an even function , from ( [ fwigner ] ) we get @xmath233}(n , m ) = \frac{1}{\kappa}\ , w_{\mathfrak{g}_2^{(1/\kappa)}}(m ,- n ) . \qquad \opensquare\ ] ] the wigner function @xmath234 is presented in fig . [ bw ] , and @xmath235 . by using the relation ( [ doublesum ] ) , one obtains @xmath236 for @xmath2 large enough , we have @xmath237 and @xmath238 . in this case @xmath239 if @xmath22 is an even function , then @xmath240 particularly , we have @xmath241 \mathfrak{g}_1^{(2)}(0)\right)^2\!+\ ! \sqrt{\frac{d}{2}}\left(\mathfrak{g}_2^{(2)}(0)\right)^2 = ||\mathfrak{g}_3^{(1)}||^2 . \end{array}\ ] ] the binomial coefficients @xmath242 satisfy the relation @xmath243 obtained by evaluating the coefficient of @xmath244 from both sides of the equality @xmath245 since @xmath246 is a unitary transformation , this means that @xmath37 the standard coherent states are @xmath247 where @xmath248 is the displacement operator and @xmath249 is the normalized gaussian @xmath250{\pi}}\ , { \rm e}^{-\frac{1}{2}q^2}. \end{array}\ ] ] the states @xmath251 satisfy the relation @xmath252 and the resolution of the identity @xmath253 to each function @xmath254 with convergent integral we associate the linear operator @xmath255 this procedure , called _ coherent state quantization _ , allows us to define new operators or to obtain integral representations for known operators @xcite . for example , we have @xmath256 by using the finite displacement operators @xmath257 and a normalized finite gaussian @xmath258 we define in @xmath13 the states @xmath259 they satisfy the relation @xmath260 and the resolution of the identity @xmath261 by using the frame @xmath262 we can associate to each function @xmath263 the linear operator @xmath264 , @xmath265 this procedure , we call _ finite frame quantization _ , allows us to define some new useful linear operators @xcite . the fourier transform @xmath266 $ ] , where @xmath267(p)=\frac{1}{\sqrt{2\pi } } \int_{-\infty } ^{\infty } { \rm e}^{-{\rm i}pq}\psi ( q)\ , { \rm d}q,\ ] ] satisfies the relation @xmath268(p)=\frac{-{\rm i}}{\sqrt{2\pi } } \int_{-\infty } ^{\infty } { \rm e}^{-{\rm i}pq}\psi ' ( q)\ , { \rm d}q = p\ , \mathcal{f}[\psi ] ( p)\ ] ] which can be written as @xmath269 by using these formulas , the hamiltonian of the quantum harmonic oscillator @xmath270 can be re - written as @xmath271 the self - adjoint operator @xmath272 where @xmath273(k)\!=\!\frac{1}{\sqrt{d}}\sum_{n =- s}^s{\rm e}^{-\frac{2\pi { \rm i}}{d}kn}\ , \psi ( n),\ ] ] is the hamiltonian of a finite - dimensional version of the quantum harmonic oscillator . it is fourier invariant , that is , we have @xmath274 . the finite - difference operator @xmath275\ ] ] is a finite - dimensional version of the differential operator @xmath276 . therefore @xmath277 can be regarded as the hamiltonian of a finite oscillator . it is fourier invariant , and one can prove that the eigenspaces corresponding to its eigenvalues are one - dimensional . from the relation @xmath278 it follows that any eigenfunction of @xmath279 is at the same time an eigenfunction of @xmath246 . the normalized eigenfunctions @xmath280 , @xmath281 , ... , @xmath282 of @xmath279 , considered in the increasing order of the number of sign alternations , satisfy the relation @xcite @xmath283 and are called _ harper functions_. they form an orthonormal basis in @xmath284 , and @xmath285 is the starting point for the definition @xcite @xmath286 of the _ fractional fourier transform_. if @xmath287 is in a small enough neighbourhood of @xmath288 , then @xmath289 are deformed versions of the finite oscillators @xmath290 and @xmath279 . atakishiyev and wolf proved @xcite that @xmath291 can be regarded as the hamiltonian of a finite oscillator with coordinate @xmath292 and momentum @xmath293 . one can remark the analogy existing between the formulas @xmath294 a^+=\frac{1}{\sqrt{2}}\left(\hat{q}\!-\!{\rm i}\hat{p } \right)\\[1 mm ] [ h , a]=-a\\[1 mm ] [ h , a^+]=a^+\\[1 mm ] h = a^+a+\frac{1}{2 } \end{array } \qquad \mbox{and}\qquad \begin{array}{l } j_-=j_x-{\rm i}\ , j_y\\[1 mm ] j_+=j_x+{\rm i}\ , j_y\\[1 mm ] [ h _ { { } _ { \rm kravchuk}},j_-]=-j_-\\[1 mm ] [ h _ { { } _ { \rm kravchuk}},j_+]=j_+\\[1 mm ] h _ { { } _ { \rm kravchuk}}=\frac{1}{2}(j_+j_-\!-\!j_-j_+)\!+\!j\!+\!\frac{1}{2}. \end{array}\ ] ] from the relation @xmath295 it follows that the eigenvalues of @xmath296 are @xmath297 , @xmath288 , @xmath298 , ... , @xmath299 and the corresponding eigenstates are @xmath300 , @xmath301 , ... , @xmath302 , @xmath303 . the relation @xmath304 presents the expansion of @xmath305 in terms of the coordinate eigenbasis @xmath306 . the ground state of @xmath290 ( top left ) , @xmath279 ( top center ) , @xmath307 ( top right ) , @xmath308 ( bottom left ) , @xmath309 ( bottom center ) and @xmath310 ( bottom right ) in the case @xmath159 . , title="fig : " ] the ground state of @xmath290 ( top left ) , @xmath279 ( top center ) , @xmath307 ( top right ) , @xmath308 ( bottom left ) , @xmath309 ( bottom center ) and @xmath310 ( bottom right ) in the case @xmath159 . , title="fig : " ] the ground state of @xmath290 ( top left ) , @xmath279 ( top center ) , @xmath307 ( top right ) , @xmath308 ( bottom left ) , @xmath309 ( bottom center ) and @xmath310 ( bottom right ) in the case @xmath159 . , title="fig : " ] + the ground state of @xmath290 ( top left ) , @xmath279 ( top center ) , @xmath307 ( top right ) , @xmath308 ( bottom left ) , @xmath309 ( bottom center ) and @xmath310 ( bottom right ) in the case @xmath159 . , title="fig : " ] the ground state of @xmath290 ( top left ) , @xmath279 ( top center ) , @xmath307 ( top right ) , @xmath308 ( bottom left ) , @xmath309 ( bottom center ) and @xmath310 ( bottom right ) in the case @xmath159 . , title="fig : " ] the ground state of @xmath290 ( top left ) , @xmath279 ( top center ) , @xmath307 ( top right ) , @xmath308 ( bottom left ) , @xmath309 ( bottom center ) and @xmath310 ( bottom right ) in the case @xmath159 . , title="fig : " ] the hamiltonian of the harmonic oscillator admits the integral representation @xcite @xmath311 since the frame @xmath262 is a finite counterpart of the standard system of coherent states @xmath312 , the linear operator @xmath313 is the hamiltonian of a finite oscillator , for any @xmath314 . from the equality @xmath315 it follows the relation @xmath316 leading to @xmath317 and @xmath318 ( 120,55)(-15,5 ) ( 10,5)(0,1)55 ( 100,5)(0,1)55 ( 9.7,10)(1,0)0.7 ( 9.7,20)(1,0)0.7 ( 9.7,30)(1,0)0.7 ( 9.7,40)(1,0)0.7 ( 9.7,50)(1,0)0.7 ( 99.7,10)(1,0)0.7 ( 99.7,30)(1,0)0.7 ( 99.7,50)(1,0)0.7 ( 15,11.19)(1,0)15 ( 15,13.58)(1,0)15 ( 15,15.96)(1,0)15 ( 15,18.35)(1,0)15 ( 15,20.74)(1,0)15 ( 15,23.13)(1,0)15 ( 15,25.50)(1,0)15 ( 15,27.94)(1,0)15 ( 15,30.13)(1,0)15 ( 15,33.06)(1,0)15 ( 15,34.08)(1,0)15 ( 15,37.88)(1,0)15 ( 15,39.80)(1,0)15 ( 15,44.48)(1,0)15 ( 15,54.12)(1,0)15 ( 35,11.88)(1,0)15 ( 35,14.27)(1,0)15 ( 35,16.66)(1,0)15 ( 35,19.04)(1,0)15 ( 35,21.41)(1,0)15 ( 35,23.75)(1,0)15 ( 35,26.00)(1,0)15 ( 35,28.27)(1,0)15 ( 35,29.80)(1,0)15 ( 35,32.35)(1,0)15 ( 35,32.73)(1,0)15 ( 35,35.66)(1,0)15 ( 35,38.46)(1,0)15 ( 35,42.80)(1,0)15 ( 35,49.35)(1,0)15 ( 55,11.88)(1,0)15 ( 55,14.27)(1,0)15 ( 55,16.66)(1,0)15 ( 55,19.04)(1,0)15 ( 55,21.41)(1,0)15 ( 55,23.75)(1,0)15 ( 55,26.02)(1,0)15 ( 55,28.16)(1,0)15 ( 55,30.07)(1,0)15 ( 55,31.69)(1,0)15 ( 55,33.34)(1,0)15 ( 55,35.49)(1,0)15 ( 55,38.50)(1,0)15 ( 55,42.80)(1,0)15 ( 55,49.35)(1,0)15 ( 75,11.99)(1,0)15 ( 75,14.38)(1,0)15 ( 75,16.76)(1,0)15 ( 75,19.14)(1,0)15 ( 75,21.49)(1,0)15 ( 75,23.81)(1,0)15 ( 75,26.01)(1,0)15 ( 75,28.25)(1,0)15 ( 75,29.62)(1,0)15 ( 75,32,29)(1,0)15 ( 75,32.67)(1,0)15 ( 75,35.71)(1,0)15 ( 75,38.40)(1,0)15 ( 75,42.73)(1,0)15 ( 75,49.18)(1,0)15 ( 105,10.42)(1,0)15 ( 105,11.72)(1,0)15 ( 105,13.75)(1,0)15 ( 105,16.61)(1,0)15 ( 105,19.86)(1,0)15 ( 105,23.81)(1,0)15 ( 105,27.73)(1,0)15 ( 105,32.09)(1,0)15 ( 105,36.01)(1,0)15 ( 105,40.0)(1,0)15 ( 105,43.28)(1,0)15 ( 105,46.18)(1,0)15 ( 105,48.24)(1,0)15 ( 105,49.56)(1,0)15 ( 8,9)@xmath319 ( 7,19)@xmath320 ( 7,29)@xmath321 ( 7,39)@xmath322 ( 7,49)@xmath323 ( 98,9)@xmath319 ( 98,29)@xmath324 ( 98,49)@xmath325 ( 18,6)@xmath326 ( 42,6)@xmath327 ( 62,6)@xmath328 ( 82,6)@xmath329 ( 108,6)@xmath330 the normalized kravchuk polynomials @xmath331 can be obtained by orthonormalizing @xmath332 with respect to the scalar product @xmath333 and the kravchuk functions can be defined as @xmath334 let @xmath335 and let @xmath336 , @xmath337 , ... , @xmath338 be the polynomials obtained by orthonormalizing @xmath332 with respect to the scalar product @xmath339 by using the gramm - schmidt method of orthonormalization . the functions @xmath340 represent a finite counterpart of the hermite - gauss functions @xmath341 , and @xmath342 a finite counterpart of the relation @xmath343 satisfied by the hamiltonian of the harmonic oscillator . therefore , the operator @xmath344 can be regarded as the hamiltonian of a finite oscillator with the ground state @xmath258 . rrrrrrr @xmath345 & @xmath346 & @xmath347 & @xmath348 & @xmath349 & @xmath350 & @xmath351 + @xmath352 & @xmath353 & @xmath354 & @xmath355 & @xmath297 & @xmath356 & @xmath297 + @xmath357 & @xmath358 & @xmath357 & @xmath359 & @xmath356 & @xmath357 & @xmath354 + @xmath288 & @xmath353 & @xmath356 & @xmath355 & @xmath297 & @xmath354 & @xmath297 + when the dimension @xmath2 becomes larger and larger , the eigenvalues of the considered finite oscillators have the tendency to become equidistant ( see fig . [ eigval ] ) . on the other hand , the existence of some equidistant energy levels implies the existence of revivals . if the hamiltonian @xmath360 of a finite oscillator admits @xmath361 eigenvalues @xmath362 , @xmath363 , ... , @xmath364 such that @xmath365 then , by denoting with @xmath366 , @xmath367 , ... , @xmath368 the corresponding eigenvectors , we have @xmath369 & \!\!\!={\rm e}^{-{\rm i}t\varepsilon _ 1}\left(\alpha _ 1 |\psi _ 1\rangle + \alpha _ 2 { \rm e}^{-{\rm i}t(\varepsilon _ 2-\varepsilon _ 1 ) } |\psi _ 2\rangle + ... + \alpha _ k { \rm e}^{-{\rm i}t(\varepsilon _ k-\varepsilon _ 1 ) } |\psi _ k\rangle \right)\\[2 mm ] & \!\!\!={\rm e}^{-{\rm i}t\varepsilon _ 1}\left(\alpha _ 1 |\psi _ 1\rangle \!+\!\alpha _ 2 { \rm e}^{-{\rm i}t(\varepsilon _ 2-\varepsilon _ 1 ) } |\psi _ \!+ ... +\!\alpha _ k { \rm e}^{-{\rm i}t(k-1)(\varepsilon _ 2-\varepsilon _ 1 ) } |\psi _ k\rangle \right ) \end{array}\ ] ] for any state @xmath370 of the form @xmath371 particularly , up to a phase factor , we have the equality @xmath372 which shows that the time evolution is periodic , and the number @xmath373 is a period for any coefficients @xmath374 , @xmath375 , ... , @xmath376 . the fourier transform has in the computational basis @xmath377 the matrix @xmath378 and its eigenvalues are 1 , @xmath379 , -1 . it admits the spectral decomposition @xmath380 where @xmath381 , @xmath382 , @xmath383 are the normalized eigenvectors presented in table [ tabfk ] . + the finite gaussians @xmath384 , @xmath385 , @xmath386 , obtained by using the relations @xmath387=\mathfrak{g}_1,\qquad f[\mathfrak{g}_2\!+\!\mathfrak{g}_3]=\mathfrak{g}_2\!+\!\mathfrak{g}_3,\qquad f[\mathfrak{g}_2\!-\!\mathfrak{g}_3]=-(\mathfrak{g}_2\!-\!\mathfrak{g}_3 ) \end{array}\ ] ] are presented in table [ tabg ] . the kravchuk transform has the matrix @xmath388 and we have @xmath389 , @xmath390 , @xmath391 , @xmath392 j_x=\frac{1}{2}\left ( \begin{array}{ccc } 0 \ & \sqrt{2 } \ & 0\\ \sqrt{2 } & 0 & \sqrt{2}\\ 0 & \sqrt{2 } & 0 \end{array}\right)=kj_zk^+=-|\mathfrak{k}_1\rangle \langle \mathfrak{k}_1|+|\mathfrak{k}_{-1}\rangle \langle \mathfrak{k}_{-1}| . \end{array}\ ] ] the eigenvectors of the fourier transform are at the same time eigenvectors of the fourier invariant hamiltonians @xmath290 , @xmath279 and @xmath393 . more exactly , we have @xmath394 h_{{}_{\rm harper}}=\frac{1}{2}\left(1\!-\!\frac{1}{\sqrt{3}}\right ) |\mathfrak{f}_{-1}\rangle \langle \mathfrak{f}_{-1}|\!+\!3\ , |\mathfrak{f}_{0}\rangle \langle \mathfrak{f}_{0}|+\frac{1}{2}\left(1\!+\!\frac{1}{\sqrt{3}}\right ) |\mathfrak{f}_{1}\rangle \langle \mathfrak{f}_{1}|\\[3 mm ] h_1=\frac{1}{2}\left(1\!-\!\frac{1}{2\sqrt{3}}\right)\ |\mathfrak{f}_{-1}\rangle \langle \mathfrak{f}_{-1}|\!+\!\frac{1}{4}\left(3\!+\!\frac{1}{\sqrt{3}}\right)\ , |\mathfrak{f}_{0}\rangle \langle \mathfrak{f}_{0}|+\frac{3}{4}\ |\mathfrak{f}_{1}\rangle \langle \mathfrak{f}_{1}| . \end{array}\ ] ] these operators and the functions @xmath395 , @xmath396 , @xmath397 may be useful in the study of qutrits . rrrrrr @xmath345 & @xmath384 & @xmath385 & @xmath386 & @xmath398 & @xmath399 + @xmath352 & @xmath353 & @xmath400 & @xmath401 & @xmath402 & @xmath403 + @xmath357 & @xmath358 & @xmath404 & @xmath405 & @xmath406 & @xmath407 + @xmath288 & @xmath353 & @xmath400 & @xmath401 & @xmath402 & @xmath403 + there exists a close relation between the finite gaussians and finite oscillators . we have used certain finite gaussians in order to define some finite oscillators . conversely , the ground state of a finite oscillator ( see fig . [ ground ] ) can be regarded as a finite version of the gaussian function . in the case of the fourier invariant hamiltonians @xmath290 , @xmath279 and @xmath393 , the ground state is an eigenfunction of @xmath246 . the normalized eigenfunction @xmath408 of @xmath246 can also be regarded as a finite version of the gaussian function @xmath409{\pi}}\ , { \rm e}^{-\frac{1}{2}q^2}$ ] . 99 atakishyiev n m pogosyan gand wolf k 2005 finite models of the oscillator _ phys . part . nucl . _ * 36 * 521 - 55 atakishyiev n m and wolf k b 1997 fractional fourier - kravchuk transform _ j. opt . am . a _ * 14 * , 1467 - 77 barker l , candan , hakiolu t , kutay m a and ozaktas h m 2000 the discrete harmonic oscillator , harper s equation , and the discrete fractional fourier transform _ j. phys . a : math . gen . _ * 33 * 2209 - 22 candan , kutay m a and ozaktas h m 2000 the discrete fractional fourier transform _ ieee trans . signal process . _ * 48 * , 1329 cotfas n and dragoman d 2012 properties of finite gaussians and the discrete - continuous transition _ j. phys . a : math . and theor . _ * 45 * 425305 cotfas n and dragoman d 2013 finite oscillator obtained through finite frame quantization _ j. phys . a : math . and theor . _ * 46 * 355301 cotfas n , gazeau j p and vourdas a 2011 finite - dimensional hilbert space and frame quantization _ j. phys . theor . _ * 44 * 175303 gazeau j - p 2009 _ coherent states in quantum physics _ ( berlin : wiley - vch ) hakiolu t and wolf k b 2000 the canonical kravchuk basis for discrete quantum mechanics _ j. phys . a : math . gen . _ * 33 * 3313 - 23 magnus w , oberhettinger f and soni r p1966 _ formulas and theorems for the special functions of mathematical physics _ ( new - york : springer - verlag ) marchiolli m a and ruzzi m 2012 theoretical formulation of finite - dimensional discrete phase spaces : i. algebraic structures and uncertainty principles _ ann . phys . _ * 327 * 1538 - 1461 mehta m l 1987 eigenvalues and eigenvectors of the finite fourier transform _ j. math . _ * 28 * 781 nikiforov a f , suslov s k and uvarov v b 1991 _ classical orthogonal polynomials of a discrete variable _ ( berlin : springer ) perelomov a m 1986 _ generalized coherent states and their applications _ ( berlin : springer ) ruzzi m 2006 jacobi @xmath410-functions and discrete fourier transform _ j. math . phys . _ * 47 * 063507 schwinger j 1960 unitary operator bases _ proc . ( usa ) _ * 46 * 570 vilenkin n j and klimyk a u 1992 _ special functions and integral transforms _ ( dordrecht : kluwer academic ) vourdas a 2004 quantum systems with finite hilbert space _ rep . phys . _ * 67 * 267 - 320 wolf k b and kr " otzsch g 2007 geometry and dynamics in the fractional discrete fourier transform _ j. opt . * 24 * 651 - 8 whittaker e t and watson g n 2000 _ cambridge mathematical library : a course of modern analysis _ ( cambridge : cambridge university press )

water molecules inside g - protein coupled receptor have recently been spotlighted in a series of crystal structures . to decipher the dynamics and functional roles of internal waters in gpcr activity , we studied a adenosine receptor usingsec - molecular dynamics simulations . our study finds that the amount of water flux across the transmembrane ( tm ) domain varies depending on the receptor state , and that the water molecules of the tm channel in the active state flow three times slower than those in the inactive state . depending on the location in solvent - protein interface as well as the receptor state , the average residence time of water in each residue varies from psec to nsec . especially , water molecules , exhibiting ultraslow relaxation ( nsec ) in the active state , are found around the microswitch residues that are considered activity hotspots for gpcr function . a continuous allosteric network spanning the tm domain , arising from water - mediated contacts , is unique in the active state , underscoring the importance of slow waters in the gpcr activation .

1608.01473

the activation of gpcrs is associated with a conformational change of the receptor modulated by an agonist binding . as suggested by the x - ray crystal structure of @xmath3-adrenergic receptor , the outward tilting of tm5 and tm6 helices that accompanies a helix rotation of the cytoplasmic part of tm5-icl3-tm6 is one of the key structural changes that facilitate g - protein recruitment @xcite . more microscopically , orchestrated dynamics among a common set of highly conserved fingerprint residues , dubbed as _ microswitches _ , of the class a gpcr family is critical for the regulation of gpcr activity @xcite . these microswitches , including three ubiquitous sequence motifs , dry , cwxp , and npxxy ( where ` x ' denotes any amino acid residue ) , adopt distinct rotameric configurations in their side chains , which are sensitive to the type of a cognate ligand bound to the orthosteric site @xcite . in the agonist - bound state , the dynamic correlations among microswitches allow the receptor to adopt and maintain the active conformation , which facilitate accommodation of g - protein @xcite . besides the allosteric network defined by the inter - residue contacts @xcite , water molecules can also contribute to the formation of the intra - receptor signaling network . given that gpcrs are the key therapeutic targets that sense and process external stimuli with an exquisite precision , it is critical to understand the roles played by water as well as the residue dynamics in gpcr activation . as is well appreciated by many studies , water is essential for structure , dynamics , and function of biomolecules @xcite . water molecules that form hydration shell on protein surfaces @xcite or a lipid - mediated effect of water via hydrophobic matching to the protein surface @xcite have long been studied . especially , possible functional roles of internal water molecules found inside tm domain in gpcr activation have recently been discussed @xcite . recently , md simulations studies have analyzed water in various type of gpcrs , including a@xmath0ar @xcite , @xmath3-adrenergic receptor @xcite , rhodopsin @xcite , dopaminergic receptor @xcite , and @xmath1-opioid receptor @xcite ; yet the foci of these studies were mostly on mapping the locations of internal waters and metal ions inside tm domain , or on the local dynamics of water specific to the ligand binding cleft in the process of ligand - receptor recognition @xcite . _ dynamical _ properties of internal water over the entire architecture of gpcr have not fully been discussed . while x - ray crystal structures provide a glimpse of ordered water molecules interacting with the interior of gpcrs @xcite , these structural waters are not completely static but have finite lifetimes . furthermore , the roles of mobile water with a relatively faster relaxation kinetics remain unknown . thus , it is timely to examine the dynamics of water in gpcr and ask the functional roles of water in conjunction with the local dynamics of amino acid residues . here , we explored water dynamics in a@xmath0ar by conducting @xmath1sec all - atom md simulations . in order to assess the role of water dynamics in gpcr function , we strategically evaluated various quantities . we first calculated the water capacity in the tm region , and mapped the probability density of water for each receptor state . explicit calculation of water flux through the tm region has revealed that the water flux in the agonist ( uk432097)-bound active form is three times smaller than that in the apo or antagonist ( zm241385)-bound form . water streams are divided into multiple pathways , some of which form one - dimensional water wire " @xcite . the receptor surfaces , mapped with the water relaxation time , indicate that the water dynamics is heterogeneous in space . waters in the extra- and intra - cellular domains move fast ( @xmath4 nsec ) . especially in the agonist - bound form , waters trapped around the microswitches display ultra - slow relaxation ( @xmath5 nsec ) , coating the polar and charged surface of the tm channel . our study finally shows that water - mediated inter - residue interaction network , formed along microswitches in tm7 , reinforces and extends the range of allosteric interface in the active state , aiding robust activation of gpcrs . * preparation of the receptor structures complexed with ligands . * the x - ray crystal structures of a@xmath6ar bound with an agonist or antagonist were modeled based on the structures from the protein data bank ( pdb ) @xcite . we used the x - ray crystal structures with pdb entries of 3qak @xcite for agonist - bound and 3eml @xcite for antagonist - bound model , and used uk432097 and zm241385 for agonist and antagonist ligands , respectively . since some of the loop regions are not resolved in the crystal structures , we performed homology modeling using the modeller program implemented in discovery studio v.3.1 to prepare the full - length a@xmath6ar models including all the loop regions . the x - ray crystal structures with pdb entries of 2ydv @xcite and 3pwh @xcite were used to model the loop regions missing in 3qak and 3eml , and the conserved disulfide bridges connecting the loops , i.e. , c71-c159 , c74-c146 , c77-c166 , and c259-c262 , were retained . these models were optimized by simulated annealing and selected based on the dope score . the final structures were obtained by energy - minimization using the conjugate gradient method and the generalized born with simple switching ( gbsw ) implicit solvent model . although an antagonist - bound engineered crystal structure of a@xmath0ar complexed with apocytochrome b@xmath7ril replacing intracellular loop 3 ( icl3 ) ( pdb entry : 4eiy @xcite ) , whose structure was fully determined with no missing region including extracellular ( ec ) and intracellular ( ic ) loops , became available after we started this research , the extent of structural overlap between the crystal structure and our modeled structure is as high as @xmath8-rmsd @xmath9 . furthermore , as the memory of initial configuration , especially the initial configuration of flexible " loops , would be erased after some time ( at most a few tens of nsec ) , and an identical force field " is used for simulation , the conclusion of this study would not be altered even when a more precise and accurate structure is used as a starting conformation for the simulation run . for the apo form , we used the protein structure , minimized after removing the antagonist ligand from the original structure ( 3eml ) as a starting conformation of our md simulations . + * molecular dynamics simulations . * to construct the explicit membrane system , the transmembrane region of a@xmath6ar was predicted based on the orientations of proteins in membranes ( opm ) database and the hydrophobicity of the protein . using solvate 1.0 ( http://www.mpibpc.mpg.de/home/grubmueller/solvate ) , we first solvated the receptor structure with tip3p water molecules and removed the water molecules outside the receptor in the tm region . then , the 1-palmitoyl-2-oleoylphosphatidylcholine ( popc ) lipid bilayer ( @xmath10 @xmath11 wide ) was placed around the tm region of a@xmath6ar . after aligning the initially solvated receptor structure and the bilayer system , we removed the lipid and water molecules overlapping within the distance of 0.5 from the receptor structure . the prepared system was solvated with the water box , covering all the previous output molecules , and neutralized with k@xmath12/cl@xmath13 ions to make 150 mm salt concentration . our system contains 68,799 ( apo , 13,549 water + 173 lipid ) , 70,052 ( antagonist - bound , 13,551 water + 182 lipid ) , and 69,449 ( agonist - bound , 13,552 water + 177 lipid ) atoms in @xmath14 @xmath15 periodic box . md simulations were performed using namd package v.2.8 @xcite with charmm22/cmap force field @xcite . the topology and parameter files for the ligands were generated by swissparam web server which provide topologies and parameters for ligand , based on merck molecular force field ( mmff ) , compatible with the charmm force field @xcite . the simulations were then conducted with the following steps : ( i ) energy - minimization for 2,000 steps using conjugate gradient method in the order of membrane , water molecules , protein structure , and the whole system ; ( ii ) gradual heating from 0 k to 300 k using a 0.01 k interval at each step ; ( iii ) 50 nsec pre - equilibration with nvt ensemble before the production runs ; and ( iv ) 1.2 @xmath1sec production runs with npt ensemble ( @xmath16 k , @xmath17 atm ) . pressure was maintained at @xmath18 atm using the nos - hoover langevin piston method with an oscillation period of 200 fsec and a decay time of 50 fsec . the temperature was controlled by applying langevin forces to all heavy atoms with a damping constant of 1 psec@xmath19 . non - bonded interactions were truncated with a 12.0 cutoff and a 10.0 switching distance , and the particle mesh ewald summation with a grid spacing 0.91 was used to handle electrostatic interactions . we used a uniform integration time step of 1 fsec , and during the whole production runs , no specific constraints were applied to the system except for restraining the hydrogen atoms . settle algorithm was used to keep water molecules rigid , and shake to restrain the bond lengths of hydrogens other than water molecules . the atomic coordinates of the simulated system were used every 50 psec for water flux analysis and 400 psec for other calculations . importantly , in the production runs it takes more than 200 300 nsec to fully equilibrate the ( receptor+lipid bilayer ) system in npt ensemble . water molecules fill the tm channel gradually ( figure [ average_water_density ] ) . while the protein rmsd gets to its steady state at @xmath20 nsec ( figure s1a ) , the area per lipid or the bilayer thickness is still under relaxation in the first 200 300 nsec ( see figure s1b and s1c ) . in the steady state , the area per lipid ( apl ) for popc molecule is @xmath21 @xmath22 @xmath11 and bilayer thickness is @xmath21 40 , which compare well with the experimental values at @xmath23 ( apl @xmath24 @xmath11 , thickness @xmath25 ) as well as with other simulation studies @xcite . after the first @xmath26 nsec simulation , we restrained the xy - plane area of the simulation box by using the option of ` useconstantarea ' in namd package . as a result , apl and thickness of lipid bilayer are stably maintained for @xmath27 nsec . in our study , the production runs of the first 0 400 nsec , which could usually be regarded a long simulation time , are still in the process of nonequilibrium relaxation . thus , we conducted various analyses in this study using the data from @xmath28 nsec after the system had reached the steady state . + * time correlation function to probe water unbinding kinetics . * to quantitatively probe the dynamics of water near the receptor structure , we calculated the auto - correlation function @xcite : @xmath29 the function @xmath30 describes whether a contact is formed between an atom in the receptor and a water molecule at time @xmath31 . @xmath32 for a water molecule within 4 from any heavy atom of a residue @xmath33 at time @xmath31 ; otherwise , @xmath34 . @xmath35 denotes a time average along the trajectory and an ensemble average over the number of water molecules ( @xmath36 ) associated with the residue @xmath33 . since @xmath37 , @xmath38 is equivalent to the survival probability of a bound water at time @xmath31 . thus , the average lifetime of water is obtained using @xmath39 . + * betweenness centrality . * simplifying a structure of complex system into a graph , represented with vertices and edges , can be used as powerful means to extract key structural features of the system @xcite . the graph theoretical analysis can be performed for protein structures @xcite , so as to judge the importance of each vertex that defines the graph . for instance , a residue in contact with many other surrounding residues , which in the language of graph theory is called a vertex with high degree centrality , could be deemed important for characterizing a main feature of the given structure . however , in order to address the issue of allostery , the betweenness centrality can be used for quantifying the extent to which a node ( @xmath40 ) has control over the transmission of information between nodes in the network @xcite . the betweenness centrality is defined as @xmath41 with @xmath42 . @xmath43 is the number of paths with the shortest distance linking the nodes @xmath44 and @xmath31 , and @xmath45 is the number of minimal paths linking the nodes @xmath44 and @xmath31 via the node @xmath40 . to build a network ( graph ) , the residue - residue contact was defined ( i ) for water - free contact using the threshold distance of 4 between any two heavy atoms of two residues , and ( ii ) for water - mediated contacts between two residues if a water oxygen is shared within 3.5 by any two heavy atoms of two residues or if the two residues are within 4 threshold distance as in ( i ) . to quantify the importance of the @xmath40-th residue in mediating signal transmission , we calculated two distinct betweenness centralities ; ( i ) @xmath46 for water - free and ( ii ) @xmath47 for water - mediated inter - residue network of gpcr . @xmath46 and @xmath47 calculated for each snapshot of molecular structure were averaged over the ensemble of structures obtained along the md trajectories . to calculate @xmath48 , we employed brandes algorithm @xcite , which substantially reduces the computational cost of eq.[cb ] . at each 3d grid point @xmath49 ( @xmath50 with the grid cell volume of @xmath51 1 @xmath52 1 @xmath52 1 ) calculated for the apo , agonist - bound , and antagonist - bound states based on the final 200 nsec of simulation . two dimensional slices of @xmath53 were calculated using volmap plugin in vmd . [ average_water_density ] ] * water capacity in the tm region . * our simulations show that it takes more than 200 300 nsec for water molecules to fill the empty tm channel and to reach the steady state ( figure [ average_water_density]a , see also figure s1 for the time evolutions of the receptor rmsd , area per popc lipid , and membrane thickness ) . in the steady state , on average 121@xmath5420 , 69@xmath5418 , and 63@xmath5416 water molecules are found in the interior ( between the phosphate atoms of upper and lower leaflets ) of the apo , agonist - bound active , and antagonist - bound inactive states , respectively ( figure [ average_water_density]a ) . as depicted in the probability density map of water ( figure [ average_water_density]b ) , the apo form , due to the water - filled ligand binding cleft , contains twice the volume of the water compared with the other ligand - bound states . in the ligand - bound states , a large volume of water has to be discharged from the vestibule . the expanded g - protein binding cleft is also seen in the agonist - bound active state ( figure [ average_water_density]b , middle ) . a water cluster , whose functional role will be further discussed in the discussion , is identified in the midst of tm domain ( see the white arrow in figure [ average_water_density]b , middle ) . + ic : solid lines , ic@xmath55ec : dashed lines ) . water fluxes ( @xmath56 ) through the receptor , calculated using the slope from linear fits to the data over the interval @xmath57 ns , are @xmath58 ( antagonist ) , @xmath59 ( apo ) , and @xmath60 @xmath1sec@xmath19 ( agonist ) . [ flux ] ] * water flux through the tm pore . * although the average number of water in the tm channel remains constant in the steady state ( figure [ average_water_density]a , @xmath61 ns ) , this does not imply that water is static inside the channel . the number of waters in and out of the channel are balanced in the steady state . in order to compare the water flux between the different receptor states , we traced the coordinate of every water molecule along the axis ( @xmath62-axis ) perpendicular to the lipid bilayer plane and counted the number of water molecules that traverse through the channel from ec to ic or from ic to ec domain ( figure [ flux ] ) . in the early stage of simulations , non - steady state fluxes are observed in the apo and antagonist - bound forms . thus , we excluded the first 400 nsec from the analysis . for @xmath63 ns , the water fluxes in the two directions satisfy @xmath64 in all the receptor states . overall , the water flux of the antagonist - bound state is three fold greater ( @xmath65 @xmath1sec@xmath19 ) than that of the agonist - bound state ( @xmath66 @xmath1sec@xmath19 ) . the water flux of the apo state lies in between ( @xmath67 @xmath1sec@xmath19 ) . in fact , the lowest water flux in the agonist - bound form has its molecular origin . below we will map each receptor state by probing the relaxation kinetics of water around each residue , which will help elucidate the role of water in receptor activation as well as the molecular origin of differential water fluxes . + * relaxation kinetics of water . * we mapped the water dynamics on the receptor surfaces by computing time correlation functions of water from each residue ( @xmath38 , see materials and methods ) ( figure [ corr_tau]a - b ) . water in the vicinity of the ic or ec loops , exposed to the bulk , is expected to have a shorter lifetime . in the tm region , in contrast , dynamics of water is much slower ( figure [ corr_tau]c ) , displaying a broad spectrum of relaxation times depending on ( i ) the receptor state and ( ii ) their locations ( figure [ corr_tau]d - f ) . the average lifetimes of water are 14.5 nsec , 26.3 nsec , and 7.1 nsec for the apo , agonist , and antagonist - bound states , respectively . the water molecules , hydrating the tm1 , tm6 , and tm7 helices in the agonist - bound form , reside longer than 100 nsec especially around the microswitches ( marked with the cyan circles at the top of figure [ corr_tau]d - f ) . this long residence time of water in the interior of tm domain ( @xmath68 nsec ) is noteworthy , given that a typical water lifetime on biomolecular surface probed by spin - label measurement is @xmath69 psec @xcite . similar conclusions as the above analysis calculating relaxation times were drawn by calculating the fano factor involving water number fluctuation around each residue ( see si text and figure s2 ) . * detailed look into water flux across the tm domain . * in order to glean the molecular origin of receptor state - dependent water flux , we first visualize the geometry of water channel . the radius of water channel , calculated using the program hole @xcite ( see figure [ water_flux_analysis]a ) , reveals that the geometrical bottleneck ( @xmath70 ) in the midst of tm domain ( @xmath71 ) ( figure [ water_flux_analysis]a , see the black arrow ) is formed around w246@xmath72 and y288@xmath73 in the agonist - bound active form . this is compatible with our finding that the water flux is substantially reduced in the agonist - bound form ( figure [ flux ] ) . w246@xmath72 , a key microswitch that senses an agonist and relays its signal to other microswitches @xcite , is located deep inside the orthosteric binding vestibule , regulates the entry of water from the ec domain ; whereas y288@xmath73 is located at the lower part of tm channel , regulating the entry of water from the ic domain . although the geometry of channel in each receptor state provides a glimpse of pipeline across the tm region , the actual water flux through the channel is not fully explained by the radii of the channel alone . for example , the apo state , overall , has greater radii along the channel and does not have a particularly more restrictive geometrical bottleneck than those in the antagonist - bound state ( figure [ water_flux_analysis]a ) ; yet the flux is smaller than the one observed in the antagonist - bound state . depending on the extent of hydrophobicity or electrostatic nature of residues comprising each region of the channel surface , stochastic wetting - dewetting transition @xcite can occur along the channel . furthermore , a stable water cluster in the channel , not exchanging water molecules with the surroundings , would impede the water dynamics through the channel . explicit calculations of water occupancy along the @xmath62-axis @xcite visualize how water molecules actually fill the tm channel at time @xmath31 ( figure [ water_flux_analysis]b ) . as expected , in the apo state ( figure [ water_flux_analysis]b , top ) the ec region ( @xmath74 ) is always filled with a high density of water since the empty ligand binding pocket is accessible from the bulk ; however , ic region ( @xmath75 ) remains dry " , indicating that the entry of water through the ic region is blocked . this dry zone corresponds to the second hydrophobic layer ( hl2 ) around npxxy motif above y288@xmath73 , illustrated by yuan _ _ @xcite . in the agonist - bound form , another water - free layer is observed right above the hl2 ( @xmath76 ) ( figure [ water_flux_analysis]b , middle ) . this is the region below which the water cluster is formed . on the other hand , the first hydrophobic layer ( hl1 ) @xcite , corresponding to another dry zone between the orthosteric and the allosteric sites @xcite(@xmath71 ) is observed in the inactive state . notably , despite the hl1 , the water occupancy map of the antagonist - bound form ( figure [ water_flux_analysis]b , bottom ) finds multiple instances ( @xmath77 ns ) that both hl1 and hl2 are filled with water bridging across the entire tm region . this accounts for the greater water flux in the inactive state . the location of the hydrophobic layer and the receptor state - dependent water flux are affected by the alignment of polar residues along the channel . the tm domain is mostly composed of non - polar hydrophobic residues , but there are polar / charged amino acids buried inside the tm domain as well . the streams of water molecules are found along an y " shaped array of these polar and charged residues that bridge through the tm domain ( figure s3 ) . this array of polar / charged residues corresponds to the buried ionizable networks in gpcrs which have recently been underscored @xcite . our study reveals that these networks shape the passages of water molecules through the tm domain . a misalignment of polar residues in the agonist - bound state is led to dehydration of the ic zone around @xmath78 , which gives rise to the hl2 ( figure s3 left ) . the rotameric state of y288@xmath73 sidechain in the antagonist - bound form enables a single file of water molecules constituting a water wire to flow across the hl2 ( figure s3 right , si movie m3 . see also figure s4 ) . the formation of the y - shaped bridge made of polar residues including the npxxy motif is the molecular origin underlying the hydrophobic gating " @xcite that regulates the water flux through the ic region of a@xmath0ar . + * microscopic origin of receptor state - dependent water flux . * to further glean the microscopic underpinnings of the receptor state - dependent water flux ( figure [ flux ] ) , configurations of microswitches at three key locations in the tm domain are probed ( figure [ three_switches ] ) : ( i ) the ionic lock ( r102@xmath79-e228@xmath80 ) ( figure [ three_switches]a , top panel ) , the hallmark of the inactive state of gpcrs , is intact in the antagonist - bound form , maintaining the inter - residue distance @xmath81 . in the apo form , the ionic - lock repeatedly disrupts and rebinds , suggestive of the receptor s basal activity @xcite . in the agonist - bound form , the ionic lock is completely disrupted ; ( ii ) the distance between r102@xmath79 and y288@xmath73 depends on the receptor state ( figure [ three_switches]a , the second panel from the top ) , and importantly y288@xmath73 gates the entry of water from the ic region . in the active form , r102@xmath79 released from the influence of e228@xmath80 can interact with and stabilize the side chain orientation of y288@xmath73 , resulting in blocking the passage of water stream as well as misaligning the bridge made of npxxy motif ( see supporting movie m2 , figure [ three_switches]b , figure s3 ) ; ( iii ) the side chain of w246@xmath72 residue gates the entry of water from the ec domain . in the active form , w246@xmath72 blocks the water passage from the ec region , but it allows water to flow more freely in the inactive form . in the apo form , the rotamer angle of w246@xmath72 undergoes sharp transitions ( @xmath82 in figure [ three_switches]a , black trace ) multiple times during the simulation ( @xmath83 ns ) , displaying correlations with the ionic - lock ( @xmath84 in figure [ three_switches]a , black line ) and with the increased level of water flux ( notice the sudden increase of the flux at @xmath85 nsec from ec to ic ( black trace in solid line in figure [ flux ] ) ) . for example , when the ionic - lock was stabilized at @xmath86 nsec in the apo form ( figure [ three_switches]a , top panel , black trace ) , the rotameric angle of w246@xmath72 also displayed a sharp change from @xmath87 to @xmath88 ( cyan arrow in the @xmath82 plot of figure [ three_switches]a ) . this is the moment when @xmath89 in the apo form has increased ( @xmath90 nsec in figure [ flux ] ) . the movies ( si movies m1 , m2 , and m3 ) from simulations of the water dynamics across tm region provide nice visualizations of receptor state - dependent water flux , which is recapitulated in the cartoons in figure [ water_flux_map ] . in the apo form , the water molecules freely navigate the wide volume of the empty ligand binding cleft in the ec domain , but a further penetration across the tm region is regulated by the narrow channel gated by w246@xmath72 . when the agonist or antagonist occupies the binding cleft , however , water flux is divided into the major ( solid lines in figure [ water_flux_map ] ) and minor streams ( dashed lines in figure [ water_flux_map ] ) ; the major stream is formed between tm1 , 2 , and 7 , and the minor stream is formed between tm3 , 5 , and 6 . in the agonist - bound form , the w246@xmath72 blocks the minor stream and the water flow along the major stream is tightly regulated by the several other microswitch residues ( n24@xmath91 , d52@xmath92 , n280@xmath93 , s281@xmath94 , and n284@xmath95 ) , which creates the stable water cluster . in case of the antagonist - bound form , no water cluster is observed ; w246@xmath72 gate is open and lets water molecules flow in and out of the tm channel . the water flux from the ic region is regulated by the side chain configuration of y288@xmath73 . in the active state whose ionic - lock is disrupted , r102@xmath79 interacts with y288@xmath73 , which in turn blocks the passage of water flux in the ic domain . + * allosteric interface reinforced by water - mediated interactions . * in order to underscore the contribution of water - mediated contacts between tm residues to the receptor s allosteric signaling and function we conducted a graph theoretical analysis on the ensemble of gpcr structures . as shown by our previous study @xcite the microswitches of gpcrs in general are identified by graph theoretical analysis using betweenness centrality to be the key sites for intra - molecular orthosteric ( allosteric ) signal transmission ( see methods ) . allosteric interface can be visualized by highlighting those allosteric hotspots @xcite . while waters inside channel are generally dynamic , some water molecules , especially around microswitches in the active state , display slow relaxation kinetics and even can be trapped for an extended amount of time ( @xmath96 ns ) ( see si text , figure s5 and figure s6 ) . as long as water dynamics is sufficiently slow , stable water - mediated contacts can be made between two residues that are not in direct contact . defining a water - mediated contact when two residues share a water oxygen within 3.5 from any heavy atom in each residue , we constructed a water - mediated residue interaction network for a given structure . using an ensemble of structure obtained from simulations , we calculated an average betweenness centrality ( @xmath48 ) at the @xmath40-th residue ( see si text , figure [ allosteric_interface]a ) . at present , there are many ways to consider the protein allostery ; some of them consider thermodynamic aspect of protein allostery @xcite , and others focus more on identification of allosteric hotspot of a given structure @xcite . the graph theoretical method @xcite can also be employed to identify an allosteric hotspot of a given network structure , and in this method a residue with high @xmath97 value corresponds to a site important for allosteric signal transmission @xcite . in the agonist - bound active state , the @xmath48 values calculated with ( @xmath98 ) and without ( @xmath99 ) water - mediated contacts ( see figure s7 ) show clear differences ( @xmath100 ) along the microswitches in tm7 ( figure [ allosteric_interface]b , c ) , which is in accord with our findings that a number of slow water molecules stably coordinating with microswitches along the water channels are present in the active state . highlighted in figure [ allosteric_interface]c with magenta surface is the allosteric interface in the active state reinforced by water - mediated interactions ( @xmath101 ) . the interface mainly formed from along the tm7 helix , reaches r102@xmath79 in the tm3 helix via y288@xmath102 and a231@xmath103 , spanning the whole tm domain . we surmise that this wide - spread interface across the tm domain enables a robust long - range signal transmission " ( compare the map of agonist - bound active state with those for apo and antagonist - bound inactive states in figure [ allosteric_interface]c ) . notably , recent calculation of energy tranport in homodimeric hemoglobin also underscores the importance of interface water cluster , substantiating our proposal of ultra - slow water mediated allosteric signaling @xcite . systematic analyses of gpcr structures reveal the network of inter - tm contacts mediated by microswitches @xcite . network analysis also put forward that these microswitches act as the hub of the intra - receptor signaling network of a@xmath0ar @xcite , and analyses of md simulation trajectories confirmed that dynamics of microswitches occurs in concert @xcite . here , we extended our analysis to the dynamics of water molecules traversing through the tm channel and investigated their role in allosteric ( orthosteric ) signaling . as is well appreciated , water , an essential component of living systems , provides driving force for self - assemblies of biomolecules and enhances their conformational fluctuations @xcite . without water , biomolecules can not function @xcite . guided by the array of polar residues , water permeates inside the mostly dry and hydrophobic pore of gpcrs . in the antagonist - bound inactive state , this array of polar residues is connected continuously from the ec to ic domain . in accord with yuan _ et al . _ @xcite , our simulations confirm the presence of the two hydrophobic layers , hl1 and hl2 ; however , these hydrophobic layers are not static , but highly dynamic . our study also confirms the yuan _ s finding @xcite that the water flux along the tm channel is regulated by the rotameric states of several microswitches ( figure [ water_flux_map ] ) . among them , two key microswitches , w246@xmath72 and y288@xmath73 act together to form an and - gate " controlling the water flow . there are also other md simulation studies on gpcrs ( e.g. rhodopsin @xcite ) which propose that a continuous stream of internal water is important for gpcr activation . by counting the number of water molecules inside tm domain , leioatts _ et al . _ @xcite showed that upon an elongation of retinal an influx of water increases inside the tm domain rhodopsin . they found that the increase of hydration level was significant in the complex - counterion simulation of rhodopsin , but not in the dark - state . their simulation results on water hydration in the dark - state of rhodopsin differ from a@xmath6ar in the inactive state in that only 20 30 water molecules are allowed inside the tm domain during the 1.6 @xmath1sec simulation time @xcite . however , the hydration level in the complex - counterion form is consistent with our study on the active form of a@xmath6ar in terms of the number of water molecules that fills the tm channel . in our study , both active and inactive states of a@xmath0ar could accommodate approximately an equivalent amount of water molecules , 60 100 , at steady state ( figure [ average_water_density]a ) , but the water flux in the inactive form was found greater than that in the inactive form by three times ( figure [ flux ] ) . here , it is crucial to distinguish the flux of water across tm domain " ( figure [ flux ] ) from the number of water inside tm domain " ( figure [ average_water_density ] ) . as described in _ results _ , the water flux across tm domain was calculated by tracing the individual water molecule and counting the number of waters that enter and exit from one side of the membrane to the other . it is not merely the increase of tm water with time . while the number of water molecule inside the dark - state of rhodopsin is smaller than that in the inactive state of a@xmath6ar , the water flux in the dark - state of rhodopsin could also be large . our study puts more emphasis on the dynamic aspect of water molecules across tm domain in the _ steady state _ , which leads us to further ask the questions of which residues are hydrated by slow water and how those slow waters contribute to the water - contact mediated allosteric network . our explicit calculations of water fluxes indicate that the continuous water stream can be formed in all three states ( i.e. , @xmath104 in figure [ flux ] ) , but with differing degrees . thus , we want to argue that what is more relevant for gpcr activation is the water - mediated contacts among the key allosteric residues than the existence of continuous water stream . it is easier to make a water - mediated contact if a water molecule is slower . in the agonist - bound active state , water molecules inside the tm channel are almost stagnant , displaying minimal flux ( figure [ flux ] ) ; they can stably hydrate the microswitches mainly along the tm7 helix . we show that water - mediated residue network extends from extracellular domain to the intracellular part of tm6 helix via tm3 helix ( figure [ corr_tau ] and figure [ allosteric_interface ] ) . the waters around tm microswitches , some of which constitute the water cluster , stabilize the relative orientation and distance between tm helices by bridging them together ( figure [ allosteric_interface ] and figure s7 ) . our study highlights the interactions of internal water with microswitches , which contribute to extending and reinforcing the allosteric interface of gpcrs ( figure [ allosteric_interface ] ) . our study puts forward that these interactions are especially critical for the functional fidelity of the gpcr activity . this work was supported by the grant from the national leading research laboratory ( nlrl ) program ( 2011 - 0028885 ) funded by the ministry of science , ict & future planning and the national research foundation of korea ( to s.c . ) , and by the rp - grant 2015 funded by ewha womans university ( to s.c . and y.l . ) . we thank kias and kisti supercomputing center for providing computing resources . 77 [ 1]`#1 ` rosenbaum , d. , s. rasmussen , and b. kobilka , 2009 . . _ nature _ 459:356363 . rasmussen , s. , b. devree , y. zou , a. kruse , k. chung , t. kobilka , f. thian , p. chae , e. pardon , d. calinski , j. mathiesen , s. shah , j. lyons , m. caffrey , s. gellman , j. steyaert , g. skinoitis , w. weis , r. sunahara , and b. kobilka , 2011 . . _ nature _ 477:549555 . lebon , g. , t. warne , p. edwards , k. bennett , c. langmead , a. leslie , and c. tate , 2011 . . _ nature _ 474:521525 . nygaard , r. , t. frimurer , b. holst , m. rosenkilde , and t. schwartz , 2009 . ligand binding and micro - switches in 7tm receptor structures . _ trends pharmacol . _ 30:249259 . katritch , v. , v. cherezov , and r. c. stevens , 2013 . . _ 53:531556 . lee , y. , s. choi , and c. hyeon , 2015 . . _ plos comp . _ 11:e1004044 . lee , y. , s. choi , and c. hyeon , 2014 . . _ proteins : struct _ 82:727743 . tarek , m. , and d. j. tobias , 2000 . . _ biophys . 79:32443257 . ball , p. , 2008 . water as an active constituent in cell biology . _ chem . rev . _ 108:74108 . frauenfelder , h. , g. chen , j. berendzen , p. w. fenimore , h. jansson , b. h. mcmahon , i. r. stroe , j. swenson , and r. d. young , 2009 . a unified model of protein dynamics . u. s. a. _ 106:51295134 . tsai , a. m. , d. a.neumann , and l. n. bell , 2000 . . _ biophys . 79:27282732 . pal , s. k. , j. peon , and a. h. zewail , 2002 . biological water at the protein surface : dynamical solvation probed directly with femtosecond resolution u. s. a. _ 99:17631768 . jungwirth , p. , 2015 . biological water or rather water in biology ? _ j. phys . _ 6:24492450 . mallikarjunaiah , k. , a. leftin , j. j. kinnun , m. j. justice , a. l. rogozea , h. i. petrache , and m. f. brown , 2011 . solid - state 2 h nmr shows equivalence of dehydration and osmotic pressures in lipid membrane deformation . _ biophys . 100:98107 . yuan , s. , s. filipek , k. palczewski , and h. vogel , 2014 . . _ 5:4733 . yuan , s. , h. vogel , and s. filipek , 2013 . . chemie international ed . _ 52:1011210115 . sun , x. , h. gren , and y. tu , 2014 . functional water molecules in rhodopsin activation . _ j. phys . b _ 118:1086310873 . burg , j. s. , j. r. ingram , a. venkatakrishnan , k. m. jude , a. dukkipati , e. n. feinberg , a. angelini , d. waghray , r. o. dror , h. l. ploegh , and k. c. garcia , 2015 . . _ science _ 347:11131117 . leioatts , n. , b. mertz , k. martnez - mayorga , t. d. romo , m. c. pitman , s. e. feller , a. grossfield , and m. f. brown , 2014 . retinal ligand mobility explains internal hydration and reconciles active rhodopsin structures . _ biochemistry _ 53:376385 . lee , j. y. , and e. lyman , 2012 . . _ biophys . j. _ 102:21142120 . yuan , s. , z. hu , s. filipek , and h. vogel , 2015 . . _ angewandte chemie _ 127:566569 . sabbadin , d. , a. ciancetta , and s. moro , 2014 . . _ j. chem . inf . model . _ 54:28462855 . bai , q. , h. prez - snchez , y. zhang , y. shao , d. shi , h. liu , and x. yao , 2014 . . _ 16:1587415885 . grossfield , a. , m. c. pitman , s. e. feller , o. soubias , and k. gawrisch , 2008 . . _ 381:478486 . jardn - valadez , e. , a .- n . bondar , and d. j. tobias , 2010 . coupling of retinal , protein , and water dynamics in squid rhodopsin . _ biophys . j. _ 99:22002207 . selent , j. , f. sanz , m. pastor , and g. de fabritils , 2010 . . _ plos comp . _ 6:e1000884 . liu , w. , e. chun , a. thompson , p. chubukov , f. xu , v. katritch , g. han , c. roth , l. heitman , a. ijzerman , v. cherezov , and r. stevens , 2012 . . _ science _ 337:232236 . venkatakrishnan , a. , x. deupi , g. lebon , c. g. tate , g. f. schertler , and m. m. babu , 2013 . . _ nature _ 494:185194 . raghavender , u. s. , s. aravinda , n. shamala , r. rai , and p. balaram , 2009 . characterization of water wires inside hydrophobic tubular peptide structures . _ 131:1513015132 . reddy , g. , j. e. straub , and d. thirumalai , 2010 . dry amyloid fibril assembly in a yeast prion peptide is mediated by long - lived structures containing water wires . u. s. a. _ 107:2145921464 . thirumalai , d. , g. reddy , and j. e. straub , 2011 . role of water in protein aggregation and amyloid polymorphism . _ acc . chem . res . _ 45:8392 . xu , f. , h. wu , v. katritch , g. han , k. jacobson , z. gao , v. cherezov , and r. stevens , 2011 . . _ science _ 332:322 . jaakola , v. , m. griffith , m. hanson , v. cherezov , e. chien , j. lane , a. ijzerman , and r. stevens , 2008 . . _ science _ 322:12111217 . dor , a. s. , n. robertson , j. c. errey , i. ng , k. hollenstein , b. tehan , e. hurrell , k. bennett , m. congreve , f. magnani , c. g. tate , m. weir , and f. h. marshall , 2011 . . _ structure _ 19:12831293 . phillips , j. c. , r. braun , w. wang , j. gumbart , e. tajkhorshid , e. villa , c. chipot , r. d. skeel , l. kale , and k. schulten , 2005 . . _ j. comp . chem . _ 26:17811802 . buck , m. , s. bouguet - bonnet , r. w. pastor , and a. d. mackerell , 2006 . . _ biophys . j. _ 90:l36l38 . zoete , v. , m. a. cuendet , a. grosdidier , and o. michielin , 2011 . swissparam : a fast force field generation tool for small organic molecules . _ j. comput . _ 32:23592368 . kuerka , n. , m .- nieh , and j. katsaras , 2011 . fluid phase lipid areas and bilayer thicknesses of commonly used phosphatidylcholines as a function of temperature . _ biochimica et biophysica acta ( bba)-biomembranes _ 1808:27612771 . tsai , h .- h . g. , j .- b . huang , and r. juwita , 2013 . a molecular dynamics study of the structural and dynamical properties of putative arsenic substituted lipid bilayers . _ 14:77027715 . luzar , a. , and d. chandler , 1996 . hydrogen - bond kinetics in liquid water . _ nature _ 379:5557 . yoon , j. , j .- c . lin , c. hyeon , and d. thirumalai , 2014 . . _ j. phys . b _ 118:79107919 . freeman , l. , 1979 . centrality in social networks conceptual clarification . _ social networks _ 1:215239 . strogatz , s. , 2001 . exploring complex networks . _ nature _ 410:268276 . albert , r. , h. jeong , and a. barabsi , 2000 . error and attack tolerance of complex networks . _ nature _ 406:378382 . newman , m. , 2005 . a measure of betweenness centrality based on random walks . _ social networks _ 27:3954 . greene , l. , and v. higman , 2003 . uncovering network systems within protein structures . _ 334:781791 . amitai , g. , a. shemesh , e. sitbon , m. shklar , d. netanely , i. venger , and s. pietrokovski , 2004 . network analysis of protein structures identifies functional residues . _ _ 344:11351146 . del sol , a. , h. fujihashi , d. amoros , and r. nussinov , 2006 . residues crucial for maintaining short paths in network communication mediate signaling in proteins . _ _ 2:2006.0019 . bagler , g. , and s. sinha , 2007 . assortative mixing in protein contact networks and protein folding kinetics . _ bioinformatics _ 23:17601767 . vendruscolo , m. , n. dokholyan , e. paci , and m. karplus , 2002 . small - world view of the amino acids that play a key role in protein folding . e. _ 65:061910 . dokholyan , n. , l. li , f. ding , and e. shakhnovich , 2002 . topological determinants of protein folding . u. s. a. _ 99:8637 . brandes , u. , 2001 . a faster algorithm for betweenness centrality . _ j. math . _ 25:163177 . franck , j. m. , m. sokolovski , n. kessler , e. matalon , m. gordon - grossman , s .- han , d. goldfarb , and a. horovitz , 2014 . probing water density and dynamics in the chaperonin groel cavity . _ 136:93969403 . franck , j. m. , y. ding , k. stone , p. z. qin , and s. han , 2015 . anomalously rapid hydration water diffusion dynamics near dna surfaces . _ 137:1201312023 . smart , o. s. , j. g. neduvelil , x. wang , b. wallace , and m. s. sansom , 1996 . hole : a program for the analysis of the pore dimensions of ion channel structural models . _ 14:354360 . li , j. , a. l. jonsson , t. beuming , j. c. shelley , and g. a. voth , 2013 . . soc . _ 135:87498759 . aryal , p. , f. abd - wahab , g. bucci , m. s. sansom , and s. j. tucker , 2014 . . _ nature communications _ 5:4377 . anishkin , a. , and s. sukharev , 2004 . water dynamics and dewetting transitions in the small mechanosensitive channel mscs . _ biophys . j. _ 86:28832895 . isom , d. g. , and h. g. dohlman , 2015 . . u. s. a. _ 112:57025707 . aryal , p. , m. s. sansom , and s. j. tucker , 2015 . hydrophobic gating in ion channels . _ 427:121130 . powell , m. r. , l. cleary , m. davenport , k. j. shea , and z. s. siwy , 2011 . electric - field - induced wetting and dewetting in single hydrophobic nanopores . _ nature nanotechnology _ 6:798802 . kobilka , b. k. , and x. deupi , 2007 . . _ trends pharmaco . _ 28:397406 . j. , h. v. , j. o. wrabl , and h. n. motlagh , 2012 . structural and energetic basis of allostery . _ 41:585609 . motlagh , h. n. , j. o. wrabl , j. li , and v. j. hilser , 2014 . the ensemble nature of allostery . _ nature _ 508:331339 . lockless , s. w. , and r. ranganathan , 1999 . evolutionarily conserved pathways of energetic connectivity in protein families . _ science _ 286:295299 . halabi , n. , o. rivoire , s. leibler , and r. ranganathan , 2009 . protein sectors : evolutionary units of three - dimensional structure . _ cell _ 138:774786 . zheng , w. , b. r. brooks , s. doniach , and d. thirumalai , 2005 . network of dynamically important residues in the open / closed transition in polymerases is strongly conserved . _ structure _ 13:565577 . ribeiro , a. a. , and v. ortiz , 2014 . determination of signaling pathways in proteins through network theory : importance of the topology . _ j. chem . _ 10:17621769 . feher , v. a. , j. d. durrant , a. t. van wart , and r. e. amaro , 2014 . computational approaches to mapping allosteric pathways . _ 25:98103 . di paola , l. , and a. giuliani , 2015 . protein contact network topology : a natural language for allostery . _ 31:4348 . leitner , d. m. , 2016 . . _ j. phys . b _ 120:40194027 . yoon , j. , d. thirumalai , and c. hyeon , 2013 . . _ 135:1211212121 . rat , v. , r. dunn , m. ferrand , j. l. finney , r. m. daniel , and j. c. smith , 2000 . solvent dependence of dynamic transitions in protein solutions . u. s. a. _ 97:99619966 . fitter , j. , r. lechner , g. buldt , and n. dencher , 1996 . internal molecular motions of bacteriorhodopsin : hydration - induced flexibility studied by quasielastic incoherent neutron scattering using oriented purple membranes . _ proc . u. s. a. _ 93:76007605 . li , t. , a. a. hassanali , y .- t . kao , d. zhong , and s. j. singer , 2007 . hydration dynamics and time scales of coupled water - protein fluctuations . _ 129:33763382 . wood , k. , m. plazanet , f. gabel , b. kessler , d. oesterhelt , d. tobias , g. zaccai , and m. weik , 2007 . coupling of protein and hydration - water dynamics in biological membranes . natl . acad . u. s. a. _ 104:1804918054 . fogarty , a. c. , and d. laage , 2014 . water dynamics in protein hydration shells : the molecular origins of the dynamical perturbation . _ j. phys . chem . b. _ 118:77157729 . * details of multiexponential fit of the auto - correlation function of water in fig . * @xmath38s of water around the residues given in fig . 3b are fitted to tri - expoential function as follows : @xmath105 ; @xmath106 ; @xmath107 ; @xmath108 ; @xmath109 ; @xmath110 , and the average relaxation time of water is obtained by using @xmath111 . + * number fluctuation of water molecules on the receptor surfaces . * water undergoes sharp transition from the bulk to interface . heterogeneity of biomolecular surfaces and conformational dynamics give rise to a number of different classes of interfacial water @xcite . tempo - spatial variation in the water number and its fluctuation on the receptor surfaces can provide glimpses of the water motion at the interfaces . some regions of the receptor are persistently occupied by water molecules with small number fluctation , and others with large fluctuation . to quantify the water fluctuation , we counted the number of water molecules as a function of time , @xmath112 , around each residue , and calculated their average ( @xmath113 ) and variance ( @xmath114 ) ( fig . [ fluctuation_fano ] ) . as expected , both @xmath113 and @xmath115 are larger around the loop regions ( icls and ecls ) than near the tm helices ( see fig . [ fluctuation_fano ] ) . taking the ratio between the two numbers , @xmath113 and @xmath115 , i.e. , @xmath116 ( fano factor ) , one can appraise the influence of the receptor surface on water number fluctuation ( fig . [ fluctuation_fano ] ) . in calculation , we counted the number of water molecules ( @xmath117 ) within 4 from any heavy atom of each amino acid residue . using our 1.2 @xmath1sec simulations , we calculated the average number of water ( @xmath113 ) , variance ( @xmath118 ) , and the ratio between the twos , i.e. , fano factor @xmath116 . residues with suppressed water fluctuations ( @xmath119 ) are commonly found in the icls and ecls of all the three receptor states , and especially in the central zone of tm domain of apo and agonist - bound forms . notably , the region where the water cluster is detected ( fig . 1b , middle ) is surrounded by the residues with @xmath119 ( fig . [ fluctuation_fano]b , black arrows in the bottom panel ) . most of the residues with enhanced water fluctuations ( @xmath120 ) are found at the interface between the receptor and lipid bilayer , especially in tm1 , 2 , 5 , and helix 8 . the apo and agonist - bound forms have residues in the tm region with @xmath121 , suggestive of strong attractive interactions from tm residues , whereas the tm region of the antagonist - bound form ( fig . [ fluctuation_fano]c ) is devoid of such residues . + * internal waters trapped inside the receptor . * among the water molecules that show slow dynamics , some molecules are trapped inside the receptor and tightly bound to specific sites . tracing the position of the entire water molecules along the trajectories , we systematically identified the trapped water molecules . we divided the full trajectory of 1.2 @xmath1s to six intervals ( i.e. , every 200 ns ) and calculated the rmsd of each water oxygen relative to the position in the final snapshot of each interval . since the average rmsd of trapped water is smaller than that of untrapped free water ( @xmath122 66 ) , it is not difficult to identify the trapped water molecules . the apo form has two such positions ( @xmath123 and @xmath124 in fig . [ trapped]a ) . in @xmath123 , a water molecule is trapped via tight h - bondings with w129 and s47 during the entire simulation time ( 1.2 @xmath1s ) ( fig . [ trapped_dist]a ) . in case of @xmath124 , maintaining the h - bonds with y197 , l95 , and a99 , three water molecules alternate to be trapped ( see the three plateau regions in the graph plotting the distances between @xmath124 and three water molecules in fig . [ trapped]a ) . that is , one water molecule in the site is displaced by another water ( supporting movie m4 ) . among the residues interacting with the trapped waters , w129@xmath125 and y197@xmath126 residues are the microswitches . in the agonist - bound form , trapped waters were observed in three positions ( fig . [ trapped]b ) : the water molecules bound in @xmath127 form the h - bonding interactions with n24@xmath91 , v282 , and f286 , connecting the tm1 and tm7 helices . in fact , @xmath128 site corresponds to the position where the water cluster with high density is identified in the water density map ( fig . 1b , middle ) ; the water in @xmath129 interacts with c245 , l249 , and a273 , mediating the interaction between the tm6 and tm7 ; two water molecules are bound simultaneously to @xmath130 , maintaining the h - bonds with t279 , n280@xmath93 , v283 , and n284@xmath95 . in case of @xmath129 and @xmath130 , several water molecules compete to bind at the sites ( figs . [ trapped_dist]b - c ) . along with these trapped waters , the water cluster is observed in the pocket formed by tm1 , tm2 , tm3 , and tm7 , especially interacting with n24@xmath91 , d52@xmath92 , n280@xmath93 , and s281@xmath94 . these water molecules constitutes an extended allosteric interface between tm helices , helping the receptor activation . in the antagonist - bound form , trapped water molecules are observed in two sites ( fig . [ trapped]c ) . one ( @xmath131 ) is near the dry motif , which forms the ionic - lock ( i.e. , salt - bridge between r102@xmath79 and e228 observed in the inactive state ) . the water molecules in @xmath131 do not form a direct interaction with dry motif but make a tight h - bonding network with t117 and r120 . the residue r120 also makes the h - bonds with d101 and y112 . the other site @xmath132 is identified among the microswitch residues in tm1 , tm2 , and tm7 . the water molecule in this site makes the h - bonds with n24@xmath91 , s281@xmath94 , and y288@xmath73 , and in the antagonist - bound state , n24@xmath91 interacts with d52@xmath92 in tm2 via h - bonds . in each receptor state , water molecules help extending interactions of the residues belonging to different tm helices , thus stabilizing the interhelical configurations of gpcrs . + * formation of water wires . * a closer look at the movies of water dynamics through the tm domain ( si movies m1 , m2 , and m3 ) reveals the presence of water wires " in a few locations . especially in the antagonist - bound inactive state , water molecules are aligned into one dimensional arrays and move in concert , stabilized by the successive h - bonds @xcite . macro - dipole moment , @xmath133 , where @xmath134 denotes the index of atoms ( oxygen or hydrogen ) comprising water molecules and @xmath135 is the total number of atoms , was calculated for the waters in the entire tm domain , and to demonstrate the presence of water wires quantitatively we calculated _ regional _ macrodipole moment , @xmath136 where @xmath137 is the number of atoms composing water molecules in the region @xmath138 , b , or c specified in figure s5 . excluding the early stage of simulation ( @xmath139 ns ) when the water flux is not yet in the steady state , we find that @xmath140 debye and occasionally reaches @xmath141 debye ( figure s5 ) . in the regions above and below w246@xmath72 , where water wires are clearly observed , the regional macrodipole moments are @xmath142 , @xmath143 , and @xmath144 debye . it is noteworthy that the orientation of the regional macro - dipole moment of the water wire in the region a is opposite to those in b and c. + * supporting movie m1 . * water dynamics in the _ apo _ form during the time interval @xmath145 ns . all the water oxygens are shown with small spheres in different colors . the key micro - switch residues are depicted using the stick representation marked with their residue numbers . + * supporting movie m2 . * water dynamics in the _ agonist_-bound form during the time interval @xmath146 ns . the details of the representation are identical with movie m1 . + * supporting movie m3 . * water dynamics in the _ antagonist_-bound form during the time interval @xmath146 ns . the details of the representation are identical with movie m1 . + * supporting movie m4 . * dynamics of two water molecules in the apo form around the @xmath147 for the time interval @xmath148 ns . the receptor structure is represented as white ribbon , and two water molecules that compete around @xmath147 are depicted using spheres in cyan and green . , width=220 ]

the electrical hall conductivities of any anisotropic interacting system with reflection symmetry obey . in contrast , we show that the analogous relation between the transverse thermoelectric peltier coefficients , , does not generally hold in the same system . this fact may be traced to interaction contributions to the heat current operator and the mixed nature of the thermoelectric response functions . remarkably , however , it appears that emergence of quasiparticles at low temperatures forces . this suggests that quasiparticle - free groundstates ( so - called non - fermi liquids ) may be detected by examining the relationship between and in the presence of reflection symmetry and microscopic anisotropy . these conclusions are based on the following results : ( i ) the relation between the peltier coefficients is exact for elastically scattered noninteracting particles ; ( ii ) it holds approximately within boltzmann theory for interacting particles when elastic scattering dominates over inelastic processes . in a disordered fermi liquid the latter lead to deviations that vanish as . ( iii ) we calculate the thermoelectric response in a model of weakly - coupled spin - gapped luttinger liquids and obtain strong breakdown of antisymmetry between the off - diagonal components of . we also find that the nernst signal in this model is enhanced by interactions and can change sign as function of magnetic field and temperature .

1609.02176

typically , an electronic system sustains average charge and heat current densities , @xmath6 , @xmath7 , when subjected to a uniform temperature gradient , @xmath8 , and constant electric field , @xmath9 . its linear thermoelectric response is described by [ eq : linres ] ( c j^e + j^h + ) = ( cc & + & + ) ( c + -t + ) , where @xmath10 is the conductivity tensor , @xmath11 and @xmath12 are the peltier tensors , and @xmath13 is the thermal conductivity tensor . in noninteracting systems , the electrical and heat - current operators are simply related to each other , giving rise to relations between @xmath10 , @xmath13 and @xmath11 . these relations continue to hold in fermi liquids , up to asymptotically vanishing corrections . an example is the wiedemann - franz law , @xmath14 , ( we use throughout @xmath15 . @xmath16 is the electron charge ) , whose breakdown has been interpreted as a signature of physics beyond the fermi liquid framework @xcite . another is the exact relation for noninteracting electrons @xcite between @xmath11 at a given temperature @xmath17 and chemical potential @xmath18 , and @xmath10 of the same system at zero temperature and shifted chemical potential [ eq : mott ] ( t , ) = _ - ^ d ( t=0,+ ) , where @xmath19 is the fermi function . this formula hence implies that in the absence of interactions , @xmath11 shares the same symmetry properties as @xmath10 . a similar conclusion is reached by solving the boltzmann equation within an energy - dependent relaxation - time approximation @xcite . owing to the pioneering works of onsager @xcite and subsequently of kubo @xcite it is well known that various linear - response transport coefficients are related via the time reversal symmetry of microscopic dynamics . consequently , one finds on general grounds that in the presence of a magnetic field @xmath20 , @xmath21 and @xmath22 , where @xmath23 . in turn , it is straightforward to show that even for an anisotropic system , as long as it is invariant under reflections , say with respect to the @xmath24 axis , @xmath25 . the above discussion implies that under similar conditions one also finds @xmath26 , provided that the system is noninteracting or considered within approximated boltzmann transport theory . a natural question then arises : is the relation @xmath26 valid beyond the limits of these two conditions ? beside its intrinsic theoretical appeal , this issue is also important for identifying non - fermi liquid behavior in the thermoelectric properties of correlated electronic systems . one such property is the nernst signal , defined by the off - diagonal elements @xmath27 and @xmath28 of the thermopower tensor @xmath29 . the latter relates the measured electric field to an applied temperature gradient , @xmath30 , in the presence of a magnetic field @xmath31 and in the absence of an electrical current . the dependence of @xmath32 on both the resistivity tensor @xmath33 and @xmath11 means that generally @xmath34 only for isotropic systems . therefore , the symmetry properties of @xmath32 do not carry direct information about interaction effects without independent knowledge of @xmath35 . however , such information may be gleaned from discrepancies between the measured nernst signal and the predictions of boltzmann transport theory . while this theory accounts for the observed data in a number of materials @xcite it underestimates the effect by orders of magnitude in several quasi - one - dimensional conductors @xcite . the nernst effect is also a sensitive probe of superconducting fluctuations , which contribute positively to the signal @xcite , in contrast to quasiparticles of various ordered normal states whose contribution is often of a negative sign @xcite . a positive nernst effect has been measured over a wide range above the critical temperature , @xmath36 , in a series of superconductors including the cuprates @xcite , as well as amorphous films of nb@xmath37si@xmath38 and ino@xmath39 @xcite . while the fluctuation contribution in the cuprates emerges from a high - temperature negative quasiparticle signal , the latter dominates the nernst effect down to , and even below , @xmath36 in other compounds such as the pnictides @xcite . it is therefore interesting to investigate the interplay between these opposing contributions in systems which exhibit concomitant strong fluctuations towards competing orders including superconductivity . motivated by the aforementioned issues we study in sec . [ sec : alpha ] the symmetry properties of @xmath11 within a generic model of interacting electrons . we begin by considering the thermoelectric linear response using the kubo formula . we show that the close relation which exists between the electrical and heat current operators in the noninteracting limit naturally leads , in the presence of reflection symmetry , to @xmath26 . however , contrary to the corresponding relation for the hall conductivities the property @xmath26 is not protected by reflection and time - reversal symmetries , and we demonstrate its explicit violation in the exactly solvable problem of two harmonically interacting electrons in a magnetic field . having established this point of principle we move on to consider the issue using boltzmann transport theory for the interacting system . we show that @xmath26 is obtained within the relaxation - time approximation of this theory , or more generally whenever inelastic processes can be neglected . since this is the case in a disordered fermi liquid at low temperatures we conclude that violation of the above relation under the specified conditions is a telltale sign of interactions beyond the fermi liquid framework . in sec . [ sec : nernst ] we consider a non - fermi liquid model of weakly coupled luttinger chains in the presence of a spin gap . we show that the antisymmetry of the off - diagonal elements of @xmath11 is indeed violated . furthermore , we calculate the nernst signal and show that interactions can lead to its substantial enhancement in such low dimensional systems . this may bare relevance to understanding the large signal observed experimentally in the quasi - one - dimensional materials . finally , we also find that the sign of the effect in the spin gapped system changes from negative to positive as the temperature is lowered and the magnetic field increased . we interpret this behavior as being due to the stronger superconducting fluctuations induced by the spin gap . various technical aspects of our study are relegated to the appendices . we consider interacting spinless fermions in a two - dimensional system of area @xmath40 , which includes mass anisotropy and coupling to static electromagnetic potentials . the system is described by the hamiltonian @xmath41 , with [ eq:2dhdensity ] ( ) & = & ^-e ( ) ( ) + & & + d^2r u(-)^()( ) ( ) , where @xmath42 , summation over repeated greek indices , which take the values @xmath43 , is implied , and the interaction is assumed to obey @xmath44 . a route for calculating the thermoelectric coefficients was laid out by luttinger @xcite , who argued that in the long - wavelength , low - frequency limit the linear response to a temperature variation @xmath45 is the same as the response to a fictitious gravitational field @xmath46 . an extension of luttinger s results to the case with a magnetic field was given by oji and streda @xcite . the gravitational field enters the calculation in two ways : first , it couples to the unperturbed density @xmath47 of @xmath48 , such that the latter reads @xmath49 . secondly , the unperturbed current density operators @xmath50 , @xmath51 are themselves modified , with @xmath50 becoming @xmath52 , see appendix [ app : currents ] . consequently , [ eq : kubo1 ] _ ij&= & + & & _ ij^(1 ) + _ ij^(2 ) . henceforth , latin indices , which take the values @xmath43 , are not summed over , and @xmath53 , where @xmath54 , @xmath55 . the contribution @xmath56 is analogous to the diamagnetic term in the electrical conductivity . for a spatially constant temperature gradient one finds , ( see appendix [ app : kubo ] ) [ eq : alpha2 ] _ ij^(2)=- d^2r j^e_i ( * r * ) r_j _ k = ^ijzm_z , where @xmath57 is the @xmath58 component of the orbital magnetization . the importance of this contribution and its origin in the redistribution of the equilibrium magnetization currents which flow in the system , has been extensively discussed by cooper , halperin and ruzin @xcite . here , we note that its appearance is a direct consequence of the kubo formalism . whereas @xmath56 is clearly antisymmetric in @xmath59 and @xmath60 , the other contribution ( see appendix [ app : kubo ] ) [ eq : alpha1 ] _ i j^(1)=_0 , expressed in terms of the retarded correlation function @xmath61 of the averaged electrical and heat current densities , is generally not . the transformation properties of the correlation functions are discussed in appendix [ app : timerev ] . under spatial reflection , when such a transformation is a symmetry of the system , they imply that the diagonal elements of @xmath62 are even functions of the magnetic field @xmath63 , while the off - diagonal elements are odd . since also @xmath64 one finds that [ eq : refsymmalpha ] _ ij(b ) = \ { cc _ ij(-b ) & i = j + -_ij(-b ) & ij + . , with similar relations for @xmath65 and @xmath13 . concomitantly , the transformation of @xmath62 under time reversal and the expressions for @xmath66 and @xmath67 , eqs ( [ eq : alphatilde1],[eq : alphatilde2 ] ) , lead to the conclusion [ eq : timerevsymmalpha ] t_ij(b ) = _ ji(-b ) . hence , combining property ( [ eq : refsymmalpha ] ) , when applied to @xmath68 , with eq . ( [ eq : timerevsymmalpha ] ) yields the relation @xmath69 between the off diagonal elements of the peltier tensors . however , symmetry considerations do not imply a similar relation between the elements of @xmath11 , per se . this stands in contrast to @xmath10 ( and @xmath13 ) , whose elements are related by time reversal symmetry via @xmath21 , thereby implying @xmath25 for the hall conductivity of a reflection symmetric system . notwithstanding , noninteracting electrons constitute an exception to the above statement . for this case it is sufficient to consider the most general hamiltonian @xmath70 of a single particle , whose position operator we denote by @xmath71 . in first quantization , @xmath72 , where the curly brackets denote the anti - commutator . using the continuity equation , @xmath73 , to identify the energy current density @xmath74 , one finds for @xmath75 [ eq : jqni ] * j*^h=-\{h-,*j*^e}. as a result , the correlation functions appearing in @xmath76 transform in the same way as the @xmath77 correlation functions determining @xmath10 . specifically , @xmath78 , implying together with eq . ( [ eq : timerevsymmalpha ] ) that @xmath79 . this , in turn , when combined with reflection symmetry , gives @xmath26 . however , we reiterate that such a behavior is not guaranteed in the presence of interactions . let us note in passing that when @xmath80 the above discussion implies that for a generic interacting system with no reflection symmetry @xmath81 @xcite . in this case it is impossible to make @xmath5 purely diagonal by choosing suitably aligned principle axes . such an `` anomalous '' peltier effect is different from the hall conductivity under the same conditions , which can always be made to vanish , and is necessarily a consequence of interactions , since in their absence @xmath82 . we now proceed to demonstrate the explicit violation of @xmath26 in an exactly solvable example . consider two interacting particles in a magnetic field , whose hamiltonian [ eq:2ph ] h&=&h_0+u(_1-_2)=_i=1,2(m_x v_i , x^2+m_y v_i , y^2 ) + & & + m_x_x^2(x_1-x_2)^2+m_y_y^2(y_1-y_2)^2 , is reflection symmetric , but anisotropic due to the rotation asymmetry of the mass tensor and harmonic interaction . the latter is characterized by the frequencies @xmath83 , which together with the cyclotron frequency , @xmath84 , set the energy scales of the problem . we work in the symmetric gauge for which the velocity operators take the form [ eq : velop ] v_x&=&(p_x - y ) , + v_y&=&(p_y+x ) . the above hamiltonian does not include a boundary potential , which is responsible for generating equilibrium edge currents and magnetization . however , in a system much larger than the magnetic lengths @xmath85 it has a negligible effect on the current correlation functions in the bulk , which are our main point of interest . transforming to the center of mass coordinates , @xmath86 , and relative coordinates @xmath87 , separates the hamiltonian into two commuting sectors @xmath88 , with [ eq : hsplit ] h_cm&= & , + h_r&= & + & & + . defining the complex coordinate @xmath89 and the operators [ eq : a1def ] a_1&=&+2 ^ 1/2 , + [ eq : a2def ] a_2&=&+2 ^ 1/2 , satisfying @xmath90=[a_2,a_2^\dagger]=1 $ ] and @xmath91=[a_1,a_2^\dagger]=0 $ ] , leads to the familiar diagonalized form of @xmath92 [ eq : diaghcm ] h_cm=_c(a_1^a_1 + ) . the relative hamiltonian can be diagonalized via a series of canonical transformations that are detailed in appendix [ app:2pdiag ] . the result is [ eq : diaghrmain ] h_r=_1(d_1^d_1+)+_2(d_2^d_2 + ) , where @xmath93=[d_2,d_2^\dagger]=1 $ ] and @xmath94=[d_1,d_2^\dagger]=0 $ ] , and the frequencies @xmath95 are given in eq . ( [ eq : omega12def ] ) . the energy eigenstates @xmath96 are therefore characterized by the eigenvalues of @xmath97 , @xmath98 , @xmath99 and @xmath100 , respectively , with energies @xmath101 . the fermionic statistics forces odd @xmath102 , see appendix [ app:2pdiag ] . the first quantized form of eq . ( [ eq : jedef ] ) , @xmath103 , where @xmath104 , leads to the averaged electrical current density @xmath105 with [ eq : jex2p ] j^e_x&=&-i(a_1 ^ -a_1 ) , + j^e_y&=&-(a_1^+a_1 ) . an explicit calculation then readily confirms that the electrical current correlation functions satisfy @xmath106= -\tr\left[e^{-\beta h}j^e_y(t)j^e_x(0)\right]$ ] , as required for @xmath0 . it follows from the results of appendix [ app : currents ] that the averaged energy current density takes the form [ eq : jedef2p ] ^e&=&_i=1,2v_i + & & + ( _ 1-_2)+h.c . , where the commutativity of @xmath6 with @xmath107 has been used . we are interested in the correlation functions [ eq : jecorfun2p ] & & = _ n , n,ne^(it-)e_n , n - it e_n,n + & & n , n|j^e_x|n,nn,n|j^e_y|n , n , and @xmath108 $ ] , relevant to @xmath2 and @xmath3 . we therefore require only the piece in @xmath109 which is diagonal in @xmath110 . calculation reveals that the corresponding piece in @xmath111 may be expressed as @xmath112 , with [ eq : jexdiag ] i_x&= & + & & ( d_1^d_1+d_1 d_1^ ) + & + & + & & ( d_2^d_2+d_2 d_2^)-h , where the parameters @xmath113 , @xmath114 and @xmath115 are given in appendix [ app:2pdiag ] . at the same time the corresponding piece in @xmath116 reads @xmath117 , with [ eq : jeydiag ] i_y&= & + & & ( d_1^d_1+d_1 d_1^ ) + & + & + & & ( d_2^d_2+d_2 d_2^)-h . since @xmath118=[i_y , h]=0 $ ] , the same argument presented following eq . ( [ eq : jqni ] ) would imply that @xmath119= -\tr\left[e^{-\beta h}j^e_y(t)j^e_x(0)\right]$ ] , provided that @xmath120 . however , this condition is fulfilled only when @xmath121 , leading to @xmath122 and @xmath123 . hence , we conclude that @xmath124 except when the system is isotropic ( @xmath125 and @xmath121 ) , or when the anisotropy in the interaction matches the mass anisotropy ( @xmath126 and @xmath121 ) , in which case it may be removed by coordinate rescaling . space . ] let us next apply the boltzmann equation to the transport of spinless electrons in a two - dimensional system subjected to a perpendicular magnetic field @xmath127 . this approach is appropriate on time and length scales much larger than the corresponding scales characterizing the scattering events . consequently , the effects of scattering are captured by a local collision integral . close to equilibrium , the distribution function can be written as @xmath128 , with @xmath129 and @xmath130 . as a result , the collision integral takes the form @xmath131 , where the kernel @xmath132 depends on the equilibrium transition rates @xcite , and the integral @xmath133 extends over the reciprocal unit cell spanned by the vectors @xmath134 , see fig . [ fig : bo ] . to linear order in the applied homogeneous electric field and thermal gradient the boltzmann equation reads @xcite [ eq : linbol ] b_g_+_ i_,g_ = _ , where we have assumed that the energy spectrum consists of a single band and defined the differential operator [ eq : bk ] b_&=&- ( _ ) _ + & = & ( - ) . solving eq . ( [ eq : linbol ] ) yields [ eq : solgk ] g_&=&__0i_,_0 ^ -1__0 + & + & _ _ 0i_,_0 ^ -1_n=1^(-1)^n_m=1^n _ _ mb__m-1i__m-1,_m^-1 + & & _ _ n , where @xmath135 . since the electrical and heat current densities are given by @xmath136 , and @xmath137 it follows that [ eq : alphabol1 ] ( c t_ij + _ ij + ) & = & e___0 v_i , i_,_0 ^ -1v_j,_0 ( c _ _ 0- + _ - + ) + & + & e___0 v_i , i_,_0 ^ -1_n=1^(-1)^n _ m=1^n__m + & & b__m-1i__m-1,_m^-1 v_j,_n ( c _ _ n- + _ - + ) . the above result obeys the onsager relation ( [ eq : timerevsymmalpha ] ) at @xmath80 , as can be readily verified by using the symmetry @xmath132 and exchanging @xmath138 in the first line of eq . ( [ eq : alphabol1 ] ) . to demonstrate that the onsager relation continues to hold for @xmath139 we integrate by parts the integrals in the second line , use the symmetry of the collision kernel and exchange @xmath140 , @xmath141 for @xmath142 . this brings the expression back to itself up to @xmath143 , and @xmath144 in the last parenthesis . accordingly , the desired relation is established , provided that the contribution from the boundary terms , incurred during the integration by part , vanishes . on general grounds , @xmath145 and @xmath146 , for any reciprocal vector @xmath147 . we find that the boundary contribution vanishes if @xmath148 also respects the lattice periodicity , , @xmath149 . under such conditions the integrand is invariant under translation by a reciprocal wave - vector and for every contribution from an end point @xmath150 $ ] there exists an opposite contribution from an end point at @xmath150+{\bf k}_2 $ ] or @xmath150-{\bf k}_1 $ ] , see fig . [ fig : bo ] . a similar argument works for the other end points . the preceding analysis shows that @xmath151 , and therefore @xmath152 in reflection symmetric systems , only if @xmath153 in eq . ( [ eq : alphabol1 ] ) . this condition is fulfilled whenever @xmath154 is proportional to @xmath155 , as is the case for elastic impurity scattering , or within the relaxation time approximation where @xmath156 . an important case of interest is the disordered fermi liquid which includes both elastic impurity scattering and inelastic processes due to electron - electron interactions . while the elastic piece in @xmath154 is temperature independent , the inelastic channel contribution to @xmath154 scales as @xmath157 in three dimensions @xcite . therefore , at low temperatures the physics is dominated by the former , @xmath158 , and the relation @xmath152 holds up to corrections of order @xmath4 . in the following section we turn our attention to the behavior of @xmath11 in a system which is manifestly a non - fermi liquid we consider a model of a two - dimensional array of @xmath159 one - dimensional chains extending along the @xmath160 direction from @xmath161 to @xmath162 and separated by a distance @xmath163 in the @xmath24 direction , with both @xmath164 . the chains are immersed in a magnetic field @xmath127 , which is generated by the vector potential @xmath165 . the spinfull electrons that populate the system interact via an attractive contact interaction , which opens a gap in the spin sector of each chain @xcite . this gap is assumed to be much larger than any remaining energy scale in the problem , such as the temperature and inter - chain couplings . owing to the spin gap , single - particle tunneling between the chains is irrelevant . in contrast , the superconducting and @xmath166 charge - density wave ( cdw ) susceptibilities of the chains are enhanced and the inter - chain josephson and cdw couplings are important @xcite . in order to have a non - trivial transverse thermoelectric response one needs to include the josephson tunneling . we will neglect the cdw coupling , whose main effect is to compete against the superconducting ordering tendency of the system , since we are interested in the case where the latter prevails . consequently , we study the following bosonized form of @xmath167 , where [ eq:1dh0 ] h_0&=&_j=1^n_cdx , + [ eq:1dhj ] h_&=&-_j=2^n_cdx . here @xmath168 and @xmath169 are the velocity and luttinger parameter of the charge sector , respectively . @xmath170 is the josephson energy per unit length , and @xmath171 is a wavevector associated with the magnetic field . ( [ eq:1dhj ] ) shows that the field adds an oscillatory phase to the pair hopping term , thus rendering it irrelevant in the renormalization group sense . however , a second order term in @xmath170 is relevant for @xmath172 and induces a crossover to a strong coupling regime at @xmath173 , where @xmath174 is the short distance cutoff of the theory @xcite . therefore , the perturbative treatment of @xmath170 , which we employ below , is valid only for @xmath175 . the current density operators may be deduced from the continuity equations for the conserved quantities . for the average current densities we obtain [ eq : jex1d ] j^e_x&=&-_j=1^n_cdx_x_j , + [ eq : jey1d ] j^e_y&=&-_j=2^n_cdx , + [ eq : jhx1d ] j^h_x&=&-_j=1^n_cdx\{_x_j,_x_j } , + [ eq : jhy1d ] j^h_y&=&-_j=2^n_cdx + & & \{_x_j+_x_j-1 , } , where @xmath176 . note that in the luttiner model ( [ eq:1dh0 ] ) the energy is measured relative to the chemical potential and therefore @xmath51 is calculated from the continuity equation for the hamiltonian density . using the above expressions and eq . ( [ eq : alpha1 ] ) we compute @xmath177 to second order in @xmath170 , see appendix [ app:1dalpha ] for details . the result [ eq : alpha1yx1d ] _ yx^(1)=_0- , is expressed in terms of the function [ eq : cqomega ] c(q,)&=&()()^2 - 2/k + & & b + & & b , where @xmath178 is the beta function , and @xmath179 is the thermal length . appendix [ app:1dalpha ] also contains the computation of @xmath57 , which , together with eq . ( [ eq : alpha2 ] ) , leads to [ eq : alpha2yx1d ] _ yx^(2)=- . the final result for @xmath3 may be cast into a scaling form [ eq : f1scaling ] _ yx = e()^2()^4 - 2/k , where the functions @xmath180 originate from @xmath181 , respectively . both @xmath182 and @xmath183 scale as @xmath160 for @xmath184 , and decay as @xmath185 for @xmath186 , due to the rapid oscillations in the josephson coupling , see fig . [ fig : alphascaling ] . while @xmath187 is always positive , consistent with a diamagnetic response ( @xmath188 ) , the sign of @xmath189 changes as function of @xmath190 . at weak fields and high temperatures the two contributions add up , leading to a positive @xmath3 , which behaves as @xmath191 . on the other hand , at large magnetic fields and low temperatures they tend to cancel each other leaving a total negative @xmath3 , which varies according to @xmath192 . the sign of @xmath3 in this regime is the one expected from superconducting fluctuations . in contrast , we show in appendix [ app:1dalpha ] that @xmath193 is smaller by a factor @xmath194 than @xmath177 , and hence negligible in the thermodynamic limit . this is a consequence of the fact that in the clean model considered here @xmath195=0 $ ] , up to corrections from boundary terms . as a result the retarded @xmath196 correlation function which determine @xmath193 vanishes identically . this demonstrates that in the inherently interacting problem studied here , @xmath197 . we expect that upon breaking the conservation of @xmath198 , , by introducing disorder into the chains , @xmath193 will no longer vanish . nevertheless , its magnitude will be proportional to the disorder strength and will not match that of @xmath177 . let us comment that a model for two superconducting wires , similar to @xmath199 defined by eqs . ( [ eq:1dh0 ] ) and ( [ eq:1dhj ] ) , was considered in ref . . however , unlike the present study each wire was assumed to be in equilibrium , described by a density matrix @xmath200 with a different temperature , while the josephson coupling was turned on adiabatically . consequently , it was found that @xmath201 . upon including a term which breaks the linear dispersion and characterized by a dimensionless curvature @xmath202 , this result changed to @xmath203 , where @xmath204 . , shown here for @xmath205 . the inset depicts the sign change of @xmath206 , and thus of @xmath3 for large @xmath190 . ] for a particle - hole and reflection symmetric model , such as the one considered here , the relation between the peltier coefficients and the thermopower is considerably simplified . under particle - hole transformation @xmath207 and @xmath208 . therefore , in the symmetric case , where @xmath209 , we conclude that @xmath210 and @xmath211 . when combined with eq . ( [ eq : refsymmalpha ] ) due to reflection symmetry , it leads to the result @xmath212 . in turn , one finds for the nernst signals [ eq : nernstsignals ] s_xy= , -s_yx=- . for a quasi - one - dimensional system embedded in a magnetic field and possessing galilean invariance along the chains one finds @xmath213 , where @xmath214 is the curvature of the free chain spectrum @xcite . in our linearized model @xmath215 diverges and as a result @xmath216 . to calculate @xmath217 we apply eq . ( [ eq : pichi ] ) ( with @xmath218 ) and find to second order in @xmath170 [ eq : sigmayy ] _ yy&=&_0 - 2ide^2 ^ 2 + & = & e^2()^2()^3 - 2/kf_(bl_t ) . the conductivity scaling function @xmath219 is depicted in the inset of fig , [ fig : nernst ] . it tends to a constant at small @xmath160 and decays as @xmath220 for large @xmath160 . when combined with the behavior of @xmath3 this results in a nernst signal along the @xmath24 direction that is negative and scales according to @xmath221 for low fields and large temperatures ( @xmath222 ) . as the field is increased and the temperature lowered the nernst signal turns positive and eventually , when @xmath223 , follows @xmath224 , see fig , [ fig : nernst ] . the resulting scale for the nernst signal , @xmath225 , is very large . for typical values relevant for the quasi - one - dimensional conductors , @xmath226 ms@xmath227 , @xmath228 nm , and @xmath229 k , the nernst signal is of order @xmath230 mv k@xmath227 , to be compared with values of order 0.1 mv k@xmath227 , measured in ( tmtsf)@xmath231clo@xmath232 @xcite . the nernst coefficient , @xmath233 , calculated for low fields where @xmath234 is linear in @xmath20 , is also large . for the above parameters we find @xmath235 @xmath18v k@xmath227 t@xmath227 , while @xmath236 measured in ( tmtsf)@xmath231clo@xmath232 is of order 10 @xmath18v k@xmath227 t@xmath227 @xcite . this is in contrast to @xmath237 @xmath18v k@xmath227 t@xmath227 calculated using boltzmann theory for a similar band structure @xcite . for the case @xmath205 . the inset depicts the scaling function of the @xmath217 . ] the transformation properties of a system under spatial reflections , time reversal and charge conjugation relate many of its transport coefficients . here we showed that the frequently used relation @xmath1 does not belong to this category . rather , its validity requires the additional condition of no interactions between the electrons , either directly or via mediators such as phonons . nevertheless , it becomes a good approximation whenever the interacting system can be considered to comprise of weakly and locally interacting particles , , a fermi liquid . in that sense , the above relation is similar to the wiedemann - franz law . they both reflect an underlying assumption that heat transfer is restricted to convection by motion of the charge carriers . the violation of @xmath1 in a reflection symmetric system is therefore a clear sign that energy is also transported via interactions between the particles , or in the extreme limit that the concept of a quasiparticle fails . thus , it would be interesting to follow the relation between @xmath2 and @xmath3 as function of temperature . if , for example , @xmath238 is observed at high temperatures but approaches @xmath239 below a characteristic temperature , @xmath240 , this would mean one of the following : ( i ) @xmath240 indicates a nematic transition inside a non - fermi liquid state , , a breaking of the @xmath241 rotation symmetry around the @xmath58 axis . ( ii ) the system is anisotropic and breaks the reflection symmetry about the @xmath160 and @xmath24 directions below @xmath240 . ( iii ) the system is a fermi - liquid and breaks both reflection symmetry and @xmath241 rotation symmetry at low temperatures . ( iv ) the system is anisotropic but reflection symmetric and non - fermi liquid behavior onsets at the temperature scale @xmath240 . the pseudogap regime of the high - temperature superconductors , with its tendencies to develop various ordered states , seems to be a good candidate for such an experiment . by studying the nernst effect in an interacting quasi - one - dimensional model with strong superconducting fluctuations we were able to demonstrate that the effect is much stronger than in two - dimensional models considered using boltzmann transport theory . this finding points to the importance of interactions and low dimensionality in establishing a large nernst signal , and may bare relevance to experiments done on quasi - one - dimensional materials . we would like to thank steve kivelson , eun - ah kim and kamran behnia for helpful discussions . this research was supported by the united states - israel binational science foundation ( grant no . 2014265 ) and by the israel science foundation ( grant no . here we obtain the electrical and heat current density operators of model ( [ eq:2dhdensity ] ) . to begin with , the continuity equation for the charge density @xmath242 in the presence of a gravitational field [ eq : contcharge ] _ ^e_=id^2r , is satisfied by the electrical current density operator [ eq : jedef ] j^e_j()=^()d_j()+h.c .. the heat current density operator @xmath243 is related to the energy current density operator @xmath244 , which in turn is to be determined by the continuity equation for the energy density [ eq : contenergy ] _ ^e_=i&=&_\{^ ( ) } + & & + d^2r\{^()()+()}+h.c .. here , in order to avoid a surface term which arises in the derivation , we have assumed that no charge current is flowing out of the system , , @xmath245 , with @xmath246 the normal to the system s boundary . to make progress we need to integrate eq . ( [ eq : contenergy ] ) with the appropriate boundary conditions . to this end , we assume that the system is thermally isolated in the sense @xmath247 . both conditions on the currents are fulfilled if @xmath248 . denoting the second line in eq . ( [ eq : contenergy ] ) by @xmath249 we further assume that its contribution to @xmath244 is irrotational and hence can be expressed as @xmath250 , where @xmath251 . it follows from the divergence theorem that a solution to this equation exists only if @xmath252 , which holds true in our case . consequently , we find [ eq : jedef ] ^e_j()&=&^ ( ) + & & + d^2r d^2r _ j^()()+h.c . , where @xmath253 is the green s function of the laplace equation with neumannn boundary conditions , satisfying @xmath254 and @xmath255 . for a rectangular domain @xmath256 it is given by g(,)=_m , n=0^ , where the term @xmath257 is excluded from the sum and the laplacian eigenfuncions and eigenvalues are given by u_mn()&= & ( ) ( ) , + _ mn&=&-()^2-()^2 , with @xmath258/2}$ ] . subsequently , it follows from [ eq : intg ] d^2r _ j g(,)=-r_j , that the average current density [ eq : avejop ] j^e = d^2r ^e ( ) , is readily obtained from eq . ( [ eq : jedef ] ) ( up to the factor @xmath259 ) by integrating the first line over @xmath260 and replacing @xmath261 $ ] with @xmath262 in the second . alternatively , it can also be expressed as [ eq : intje ] j^e_j&=&-d^2r ^_t ( ) + & & -d^2r d^2r(r_j - r_j)u(- ) + & & \{^()()_t ( ) -^()^( ) ( ) } + & & + h.c .. consider a time independent @xmath263 ( with @xmath264=0 $ ] ) , perturbed by @xmath265 , where @xmath266 is an external field coupled to a conserved charge @xmath267 , satisfying @xmath268 . to linear order in @xmath269 an observable @xmath270 , with @xmath271 , acquires the expectation value @xcite [ eq : kuboeq1 ] o(t)_k+k = o(t)_k - d^2r ( , t)_,o(,+i ) . here we have assumed that no @xmath272 flows out of the system . , @xmath273 , denoted by @xmath274 the change in the form of @xmath275 in the presence of @xmath276 , and [ eq : pi ] _ , o(,)=_0^dt_0^de^it(,-t - i)o(0)_k . using a lehmann representation in terms of @xmath263 eigenstates , @xmath277 , we can write the latter as [ eq : pileh ] _ , o(,+i)&=&_m , n e^-_n + & & _ 0^de^(_n-_m ) + & = & _ m , n + & & + & & n|()|mm|o|n , where the limit @xmath278 is introduced in order to recover the correct result of the @xmath279 integration in the case @xmath280 , and is to be taken first , followed by the limit @xmath281 . on the other hand , consider the imaginary - time correlation function [ eq : chileh ] _ o,(,i_n)&=&-_0^de^i_n o(-i)(,0 ) + & = & _ m , n + & & , where @xmath282 is a bosonic matsubara frequency , and the limit @xmath278 takes care of the case @xmath283 and @xmath280 . from eqs . ( [ eq : pileh],[eq : chileh ] ) it then follows that [ eq : pichi ] _ , o(,+i)= , where @xmath284 has been analytically continued via @xmath285 to yield the retarded correlation function . the above results applies to the calculation of @xmath5 given the identification @xmath286 , @xmath287 , @xmath288 , and @xmath289 . this in turn leads , together with the definition [ eq : defavechi ] _ o , j()=d^2r _ o , ( , ) , to eq . ( [ eq : alpha1 ] ) . from eq . ( [ eq : jedef ] ) it follows that @xmath290 , with the consequent contribution to @xmath11 [ eq : alpha2ex1 ] ^(2)_ij = -d^2 j^e_i ( ) r_j _ k. to relate it to the @xmath58 component of the orbital magnetization , @xmath57 , note that [ eq : magrel1 ] 0&=&d^2r _ t^e()r_i r_j _ k= -d^2r ^e()r_i r_j _ k + & = & d^2r j_i^e()r_j + d^2r j_j^e()r_i , where the first equality is a result of @xmath291\}=0 $ ] , and the third a result of our assumption @xmath245 . eq . ( [ eq : magrel1 ] ) , together with the definition [ eq : mzdef ] m_z = d^2r _ k , allow us to express @xmath292 by eq . ( [ eq : alpha2 ] ) . next , let us discuss the calculation of @xmath68 using the kubo formula . since the calculation is done for finite @xmath293 , which is taken to zero only at the end , one needs to determine the appropriate form of @xmath244 in the presence of a time - varying electric field . to this end , we split the scalar potential into @xmath294 , such that the driving electric field is given by @xmath295 , while @xmath296 ( ) describes the constant background potential , due to the ions for example . we denote by @xmath297 the current density that is given by eq . ( [ eq : jedef ] ) with time - dependent electromagnetic potentials , and note that it satisfies @xmath298 $ ] . consequently one finds , [ eq : cont - timevar ] _ t[-^e]&=&i[h,]+_t - i-_t^e + & = & -(^e-^e)-(_t^e+^e ) + & & -(+_ta / c)^e + & = & -(^e-^e)+^e , which is to be interpreted as a continuity equation , with a source term due to joule heating , for the energy density @xmath299 , and current @xmath300 @xcite . these @xmath301 and @xmath244 are also both gauge invariant , with @xmath109 obtained from eq . ( [ eq : intje ] ) via the substitution @xmath302 . introducing the electric field in the gauge @xmath303 and applying eq . ( [ eq : kuboeq1 ] ) with @xmath304 , @xmath305 , @xmath306 , and @xmath218 leads to [ eq : alphatilde1 ] _ i j^(1)=_0 a . the above form of the heat current does not change in the presence of @xmath307 . however , in the limit @xmath308 , the system relaxes to a state which is close to local ( but not global ) thermodynamic equilibrium , for which @xmath309 becomes a part of the @xmath310 and @xmath311 a part @xmath51 . as a result , an additional contribution to @xmath312 appears , and is given by [ eq : alphatilde2 ] _ i j^(2)=- d^2r j^e_i ( * r * ) r_j _ k = ^ijzm_z . the existence of the magnetization term can be traced back to the assumed local thermodynamic equilibrium , which implies the relation @xmath313 for an infinitesimal heat change . this , when divided by @xmath314 and combined with faraday s law @xmath315 , gives eq . ( [ eq : alphatilde2 ] ) . finally , we demonstrate that @xmath316 is gauge invariant . to this end , we employ the conventional form of the kubo formula @xcite , which for the gauge @xmath303 reads [ eq : kubo - scalargauge ] _ i^h(,t)_k+k&= & i_-^t dt d^2r e^-i(+i)t + & & _ k , where we used the fact that in this gauge @xmath317 . alternatively , one can use the gauge @xmath318 , where the first piece is responsible for the magnetic field while the electric field is introduced via @xmath319 . the kubo formula then becomes [ eq : kubo - timedepgauge ] & & _ i^h(,t)_k+k=_i^h(,t)_k + & & + _ -^t dt d^2r _ k , + where @xmath263 includes @xmath320 but not @xmath321 . to proceed , we note that [ eq : sidecalc ] d^2r ^e & = & d - d^2r ( ) ^e + & = & d^2r _ t^e , where in going from the first to the second line we assumed that the surface integral of @xmath322 vanishes . plugging eq . ( [ eq : sidecalc ] ) into eq . ( [ eq : kubo - timedepgauge ] ) and integrating by parts over @xmath323 yields eq . ( [ eq : kubo - scalargauge ] ) and a boundary term [ eq : boundaryterm ] & & d^2r _ k + = & & -a^e_(t)[d_i]^d _ + & & + ^ + & & + d^2r d^2r _ i + & & ^( ) ( ) ( )+_i,^e _ k + & & + h.c . , which exactly cancels @xmath324 , as can be checked using eqs . ( [ eq : jedef],[eq : jedef ] ) . we comment that by applying the considerations outlined in appendix [ app : timerev ] it can be shown that @xmath324 vanishes in the presence of reflection symmetry . we are interested in the case where the potential appearing in the hamiltonian density ( [ eq:2dhdensity ] ) is invariant under reflection about the @xmath24 axis [ eq : parcond1 ] ( x , y)=(l_x - x , y)(x , y ) , and similarly the magnetic field satisfies @xmath325 . the latter condition in obeyed provided that [ eq : parcond2 ] a_x(x , y)=a_x(x , y)+_x f(x , y ) , + a_y(x , y)=-a_y(x , y)-_y f(x , y ) . it is then straightforward to check that @xmath326 under the reflection transformation [ eq : partrans ] ( x , y)^=(x , y)e^i(e / c)f(x , y ) . as a result , any two bosonic hermitian operators @xmath327 , transforming according to [ eq : paroptrans ] o_1,2(b)^=^_1,2o_1,2(-b ) , with @xmath328 , satisfy [ eq : parcorrtrans ] o_1(b ) o_2(b ) _ k(b)&= & /z_k(-b ) + & = & ^_1^_2o_1(-b ) o_2(-b ) _ k(-b ) . consequently , the imaginary - time correlation function obey [ eq : parfreqcorr ] & & _ o_1,o_2(i_n;b ) + & & = -_0^de^i_n o_1(-i;b ) o_2(0;b ) _ k(b ) + & & = ^_1^_2_o_1,o_2(i_n ;- b ) . note that for both electrical and heat currents @xmath329 . ( [ eq : parfreqcorr ] ) also holds for the disorder averaged correlation function @xmath330 , even when condition ( [ eq : parcond1 ] ) is not fulfilled , as long as the disorder distribution obeys @xmath331=p[\phi(\tilde x , y)]$ ] . under time reversal @xmath332 . provided that [ eq : timerevoptrans ] o_1,2(b)^-1=^_1,2o_1,2(-b ) , with @xmath333 , and using @xmath334 @xcite , where @xmath335 is the time reversed state , one finds [ eq : timerevcorrtrans ] & & o_1(-i;b ) o_2(0;b ) _ k(b ) + & & = /z_k(-b ) . + hence , [ eq : timerevfreq ] _ o_1,o_2(i_n;b)=^_1^_2_o_2,o_1(i_n ;- b ) . for both electrical and heat current densities the relative two - particle hamiltonian in eq . ( [ eq : hsplit ] ) is expressed in terms of the operators [ eq : bdef1 ] b_1&= & ( + l_x ) , + [ eq : bdef2 ] b_2&= & ( + l_y ) , satisfying @xmath338=[b_2,b_2^\dagger]=1 $ ] , @xmath339=[b_1,b_2^\dagger]=0 $ ] , as [ eq : hrdiag1 ] h_r&=&(b_1^b_1+b_2^b_2 + 1 ) -i(b_1^b_2-b_2^b_1 ) + & & - , where [ eq : omegadef ] & = & , + [ eq : gammadef ] & = & . it may be decoupled into two independent pieces via the canonical transformation [ eq : cdef ] ( c b_1 + b_2 + ) = ( cc i & + - & -i + ) ( c c_1 + c_2 + ) , with [ eq : phidef ] 2=- , which leads to [ eq : hdiag2 ] h_r&= & ( cc c_1^ & c_1 + ) ( cc _ - & + & _ - + ) ( c c_1 + c_1^ + ) + & + & ( cc c_2^ & c_2 + ) ( cc _ + & - + - & _ + + ) ( c c_2 + c_2^ + ) , where [ eq : defomegapm ] _ = 0 . finally , we employ the bogoliubov transformation [ eq : ddef ] ( c c_1,2 + c_1,2^ + ) = ( cc _ 1,2 & _ 1,2 + _ 1,2 & _ 1,2 + ) ( c d_1,2 + d_1,2^ + ) , with [ eq : theta12def ] 2_1,2= , to bring @xmath337 into the diagonalized form [ eq : diaghr ] h_r=_1(d_1^d_1+)+_2(d_2^d_2 + ) , where @xmath93=[d_2,d_2^\dagger]=1 $ ] , @xmath94=[d_1,d_2^\dagger]=0 $ ] , and the eigenfrequencies are given by [ eq : omega12def ] _ 1,2=2 . the center of mass part of the two particle eigenstate is obviously symmetric under particle exchange . to maintain antisymmetry of the full state we require that the relative part would be antisymmetric . one can check that the wavefunction of the ground state , @xmath340 , of @xmath337 is proportional to @xmath341 $ ] , with @xmath342 constants , and hence symmetric . from eqs . ( [ eq : bdef1],[eq : bdef2 ] ) it follows that @xmath343 under particle exchange , and the linearity of the ensuing transformations means that @xmath344 share this property . thus , the allowed @xmath345 are those for which @xmath102 is odd . according to eq . ( [ eq : alpha1 ] ) , @xmath189 is determined from the correlation function @xmath346 , which we evaluate perturbatively in @xmath347 . one can readily verify that the zeroth order term vanishes in the limit @xmath348 , and the lowest non - vanishing contribution is [ eq : chiyx1d1 ] _ j_y^e , j_x^h(i_n)=_0^dd e^i_nt_j_y^e()j_x^h(0)h_( ) _ 0 , where here @xmath349 and @xmath350 is the imaginary time ordering operator . using expressions ( [ eq : jey1d ] ) and ( [ eq : jhx1d ] ) of the current densities and the averages @xcite [ eq : f1 ] f_1(x,)&=&k^-1t_^2_0 + & = & kt_^2_0 + & = & \{()^2 } , + + [ eq : f2 ] f_2(x,)&=&t_(x,)(0,0)_0 + & = & \{. + & & .- } , where @xmath351 , and @xmath179 , we obtain [ eq : chiyx1d2 ] _ j_y^e , j_x^h(i_n)&=&_0^dddx dx e^i_n + & & c(x - x,- ) + & & + & & , with [ eq : cdef ] c(x,)=e^-2k^-1f_1(x , ) . the parity of the functions @xmath352 and @xmath353 leads , after defining @xmath354 , to [ eq : chiyx1d3 ] _ j_y^e , j_x^h(i_n)&=&_0^dddx dr e^i_n + & & c(r,-)(br ) + & & . for @xmath355 $ ] the integral over @xmath160 can be evaluated with the result [ eq : chiyx1d4 ] _ j_y^e , j_x^h(i_n)&=&_0^dddr e^i_n(br ) + & & c(r,- ) . integrating by parts , we find that the @xmath356 term vanishes . finally , integration by parts over @xmath357 and a change of variables to @xmath358 , gives [ rq : chiyx1dfinal ] _ j_y^e , j_x^h(i_n)= , where @xmath359 , and [ eq : cqomegadef ] c(q , i_n)=_-^dx_0^d e^i(_n - qx)c(x , ) . we are now left with the task of calculating [ eq : cqomegaint1 ] c(q , i_n)&=&()^2/k_-^dx _ 0^dy e^i(_n y- q x ) + & & ( ^2 x+^2 y)^-1/k , with @xmath360 and @xmath361 . clearly , this function is symmetric under @xmath362 , so in the following we assume @xmath363 . next , a change in the integration variable to @xmath364 rotates the integral onto the segment @xmath365 , see fig . [ fig : ccontour ] . applying cauchy s theorem , while noting the branch cuts at @xmath366 and @xmath367 , we may trade @xmath365 by the contour @xmath368 . finally , we use the invariance of the integrand under @xmath369 to shift @xmath370 below the real axis and obtain [ eq : cqomegaint2 ] c(q , i_n)&=&i()^2/k_-^dx _ -^0 dz e^_n z- iq x + & & \{^-1/k . + & & .-^-1/k } , where @xmath371 is a positive infinitesimal . taking @xmath372 and using [ eq : sinhiden ] & & ^1/k + & & = \ { . , + we arrive at [ eq : cqomegaint3 ] c(q , i_n)&=&2()^2/k ( ) + & & _ 0^dz _ -z^z dx . by changing variables to @xmath373 the remaining integrals can be evaluated for @xmath169 . the result , after analytically continuing @xmath285 , is given by eq . ( [ eq : cqomega ] ) . within a perturbative treatment of @xmath170 the leading contribution to @xmath374 is determined by [ eq : chixy1d1 ] _ j_x^e , j_y^h(i_n)=_0^dd e^i_nt_j_x^e()j_y^h(0)h_( ) _ 0 . concentrating on the spatial integrals which appear in this contribution , we find that it is proportional to [ eq : chixy1dint ] & & _ j=1^n_c_j=2^n_c_-l/2^l/2 dx dx dx + & & c(x-x,-)\{_j , j^2_x f_2(x - x,- ) . + & & + ( /k)(_j , j-_j , j-1 ) _ x f_2(x-x,- ) + & & } . clearly , the sum over @xmath375 of the last two lines vanishes . the sums and @xmath160 integral over the remaining part give @xmath376 . it follows from eq . ( [ eq : f2 ] ) that this term is appreciable only for @xmath377 within a distance of order @xmath378 from the the edges at @xmath379 . consequently , we conclude that @xmath374 is smaller by a factor @xmath194 than the corresponding @xmath189 . the magnetization can be computed from the thermodynamic relation [ eq : magterno ] m_z=-()_,t , where @xmath113 is the grand canonical potential . to second order in @xmath170 we obtain [ eq : omegaexp ] & = & _ 0-_0^dt_h_()h_(0)_0 + & = & _ 0-_0^ddx c(x,)(bx ) , with @xmath380 $ ] , from which it follows , using eqs . ( [ eq : magterno ] ) and ( [ eq : cqomegadef ] ) , that [ eq : magres ] m_z=- . the same result is also obtained from the definition of @xmath57 in terms of currents , eq . ( [ eq : mzdef ] ) . to see this we note that eqs . ( [ eq:1dhj ] ) and ( [ eq : jey1d ] ) imply [ eq : magcurrenty ] d^2 r xj^e_y_h&=&-2ed _ h + & = & ed _ 0^dt_h_()h_(0)_0 + & = & cm_z . furthermore , explicit calculation reveals that to order @xmath381 [ eq : magcurrentx ] d^2 r yj^e_x_h&= & - _ j=1^n_c_j=2^n_cj(_j , j-_j , j-1 ) + & & _ 0^dd dx dx dx + & & c(x-x,-)_x f_1(x - x,- ) . + a naive evaluation of the integral , disregarding the finite size of the system , would yield zero owing to the fact that the integrand is odd in @xmath382 and @xmath383 . however , a more careful analysis leads to a different conclusion . first , the sums add up to @xmath384 . secondly , eq . ( [ eq : f1 ] ) implies that to a good approximation @xmath385/\pi$ ] , as long as @xmath377 is situated more than @xmath378 away from the edges . using this we obtain [ eq : magcurrentx2 ] d^2 r yj^e_x_h&= & _ 0^ddx c(x , ) x ( bx ) + & = & c =- cm_z , as required . g. catelani and i. l. aleiner , `` interaction corrections to thermal transport coefficients in disordered metals : the quantum kinetic equation approach '' , zh . fiz . * 127 * , 372 ( 2005 ) [ sov . jetp * 100 * , 331 ( 2005 ) ] . z. a. xu , n.p . ong , y. wang , t. kakeshita , and s. uchida , `` vortex - like excitations and the onset of superconducting phase fluctuation in underdoped la@xmath391sr@xmath39cuo@xmath232 '' , nature ( london ) * 406 * , 486 ( 2000 ) . y. wang , z. a. xu , t. kakeshita , s. uchida , s. ono , y. ando , and n. p. ong , `` onset of the vortexlike nernst signal above @xmath36 in la@xmath391sr@xmath39cuo@xmath232 and bi@xmath231sr@xmath392lacuo@xmath389 '' , , 224519 ( 2001 ) . o. cyr - choini@xmath393re , r. daou , f. lalibert@xmath393 , d. leboeuf , n. doiron - leyraud , j. chang , j .- q . yan , j .- cheng , j .- s . zhou , j. b. goodenough , s. pyon , t. takayama , h. takagi , y. tanaka , and l. taillefer , `` enhancement of the nernst effect by stripe order in a high-@xmath36 superconductor '' , nature ( london ) * 458 * , 743 ( 2009 ) . r. daou , j. chang , d. leboeuf , o. cyr - choini@xmath393re , f. lalibert@xmath393 , n. doiron - leyraud , b. j. ramshaw , r. liang , d. a. bonn , w. n. hardy , and l. taillefer , `` broken rotational symmetry in the pseudogap phase of a high-@xmath36 superconductor '' , nature ( london ) * 463 * , 519 ( 2010 ) . j. chang , n. doiron - leyraud , f. lalibert@xmath393 , r. daou , d. leboeuf , b. j. ramshaw , r. liang , d. a. bonn , w. n. hardy , c. proust , i. sheikin , k. behnia , and l. taillefer , `` nernst effect in the cuprate superconductor yba@xmath231cu@xmath394o@xmath395 : broken rotational and translational symmetries '' , , 014507 ( 2011 ) . j. chang , n. doiron - leyraud , o. cyr - choini@xmath393re , g. grissonnanche , f. lalibert@xmath393 , h. hassinger , j. ph . reid , r. daou , s. pyon , t. takayama , h. takagi , and l. taillefer , `` decrease of upper critical field with underdoping in cuprate superconductors '' , nat . * 8 * , 751 ( 2012 ) . a. pourret , h. aubin , j. lesueur , c. a. marrache - kikuchi , l. berg@xmath396 , l. dumoulin , and k. behnia , `` observation of the nernst signal generated by fluctuating cooper pairs '' , nature physics * 2 * , 683 ( 2006 ) . z. w. zhu , z. a. xu , x. lin , g. h.cao , c. m. feng , g. f. chen , z. li , j. l. luo , and . n. l. wang , `` nernst effect of a new iron - based superconductor lao@xmath397f@xmath39feas '' , new j. phys . * 10 * , 063021 ( 2008 ) . m. matusiak , z. bukowski , and j. karpinski , `` nernst effect in single crystals of the pnictide superconductor cafe@xmath398co@xmath399as@xmath231 and parent compound cafe@xmath231as@xmath231 '' , , 020510(r ) ( 2010 ) .

we have monitored the cassiopeia dwarf galaxy ( ic 10 ) in a series of 10 _ chandra _ acis - s observations to capture its variable and transient x - ray source population , which is expected to be dominated by high mass x - ray binaries ( hmxbs ) . we present a sample of 21 x - ray sources that are variable between observations at the 3 level , from a catalog of 110 unique point sources . we find 4 transients ( flux variability ratio greater than 10 ) and a further 8 objects with ratio 5 . the observations span years 2003 - 2010 and reach a limiting luminosity oferg s , providing sensitivity to x - ray binaries in ic 10 as well as flare stars in the foreground milky way . the nature of the variable sources is investigated from light - curves , x - ray spectra , energy quantiles , and optical counterparts . the purpose of this study is to discover the composition of the x - ray binary population in a young starburst environment . ic 10 provides a sharp contrast in stellar population age ( my ) when compared to the magellanic clouds ( 40 - 200 my ) where most of the known hmxbs reside . we find 10 strong hmxb candidates , 2 probable background active galactic nuclei , 4 foreground flare - stars or active binaries , and 5 not yet classifiable sources . complete classification of the sample requires optical spectroscopy for radial velocity analysis and deeper x - ray observations to obtain higher s / n spectra and search for pulsations . a catalog and supporting dataset are provided .

1611.08611

the dwarf - irregular galaxy ic 10 in cassiopeia presents one of the most active starbursts in the local group and a unique environment to study the immediate relics of the most massive stars : black holes and neutron stars in high mass x - ray binaries ( hmxbs ) . motivated by the discovery of the black hole hmxb ic 10 x-1 @xcite and the very large amplitude transient ic 10 x-2 @xcite we conducted a 7-shot monitoring campaign with _ chandra _ during 2009 - 10 . combined with deep observations taken in 2003 and 2006 , we were able to span timescales ranging up to 7 years reaching a limiting sensitivity of @xmath2erg s@xmath3 . this cadence and sensitivity , coupled with the angular resolution of _ chandra _ , enable light - curves to be constructed for individual point sources . ic 10 hosts a young ( @xmath56 @xmath6 yr ) stellar population @xcite accompanied by the highest known space - density of wolf rayet ( wr ) stars @xcite , which are the evolved helium cores of stars with initial stellar masses in excess of 20 @xmath7 . the star formation rate ( sfr ) is reported to be as high as 0.5 @xmath7@xmath8 @xcite , which when normalized by the mass of the galaxy ( 2 @xmath9 @xmath7 ; @xcite ) yields 2.5 @xmath10 @xmath7@xmath8 @xmath7@xmath3 which is among the highest specific sfrs in the local universe exceeding the values for the milky way ( mw ) , small magellanic cloud ( smc ) , and large magellanic cloud ( lmc ) . depending on the indicators and methods used , there is an order of magnitude dispersion among sfr values for ic 10 @xcite . the chaotic structure of ic 10 is probably part of the picture and indeed much of the galaxy resembles an ob association , but on a kiloparsec scale @xcite . ic 10 is replete with a complex structure of @xmath11-bright and radio - bright bubbles , filaments , and hii regions , all signaling its turbulent evolution under the influence of the kinetic energy being injected by winds and supernovae . the similarities between ic10 and the smc are striking ; both are gas - rich dwarf irregulars and both are thought to have undergone tidal disruption in the recent past @xcite leading to triggered star formation activity . however , two critical differences exist between these galaxies that make ic 10 an exciting new laboratory for stellar astrophysics . the first and most important contrast is the time since star formation began . this is @xmath56 @xmath6 yr for ic 10 compared to 40 - 200 @xmath6 yr for the various distinct age - segregated populations identified in the smc @xcite , where the hmxbs are found in the 40 - 70 myr sub - population @xcite . the second contrast is in metallicity which for ic 10 is @xmath12 @xcite , intermediate between the smc and the mw . complete censuses of x - ray binary populations in local group galaxies are a powerful approach to discover the underlying physics of stellar evolution , for example the effect of metallicity . the magellanic clouds have historically served this purpose , but gaining proper control of secular differences between galaxies requires new independent laboratories such as ic 10 to be added . important large samples have been constructed by mining the _ chandra _ archive , ( for example @xcite ) and by concerted multi - year efforts to intensively observe nearby galaxies ( e.g. , @xcite ) . an x - ray study of point sources in ic 10 was conducted by @xcite using single epoch data from _ xmm - newton _ and _ chandra_. among their important findings was a population of point sources significantly above the expected background density and spatially concentrated within the optical outline of the galaxy . the combined x - ray spectrum of these point - sources was consistent with an absorbed power - law with parameters characteristic of hmxbs . key physics goals to be addressed by ic 10 s x - ray binary ( xrb ) sample include : constructing the x - ray luminosity function ( xlf ) in order to compare it with the sfr history ; discovering what type of xrbs are formed at the earliest times ; and breaking the degeneracy between compact - object mass and accretion - mechanism that currently bedevils attempts to understand the nature of both ultraluminous x - ray ( ulx ) sources and the xlf - sfr relation . the relative proximity of ic10 is crucial because the individual xrbs can be studied in detail once they are identified . in the larger samples of galaxies , the questions are addressed on a purely statistical basis . identifying optical counterparts to x - ray sources at the 660 kpc distance of ic 10 is a daunting prospect since it lies at low galactic latitude and it is viewed by us through the outer portion of the galactic plane . since most xrbs are highly variable , a monitoring campaign offers an alternative path to discovering the nature of these individual sources . this paper begins with a uniform reduction / analysis of the _ chandra _ monitoring and archival observations and visualizations of the ccd data ( [ sect : obs ] ) followed by the construction of the source catalog ( [ sect : catalog ] ) and the generation of x - ray light curves ( [ sect : lc ] ) . we outline our spectral analysis using quantiles and ccd pulse - height spectra in [ sect : quantile ] , we briefly describe how optical counterparts were identified ( [ sect : optical ] ) , and then we proceed to [ sect : discussion ] , where we look at each variable source in turn in order to summarize the available information and to discuss its likely nature . finally , in [ sect : conclusions ] , we present an overview of the results and we lay out the direction of future progress . a monitoring series of 7@xmath1315 ksec _ chandra _ observations , spaced at roughly 6-week intervals was obtained during 2009 - 2010 . we used the advanced ccd imaging spectrometer ( acis ) , positioning the galaxy ic 10 on the back - illuminated chip s3 . the observing cadence was designed to match the native variability timescales of hmxbs which tend to be transients with typical outbursts lasting for 1 - 3 weeks repeating on intervals of months . the individual exposure times were calculated to reach a limiting luminosity of 10@xmath14 erg s@xmath3 ( assuming 5 net acis - s counts at the high resolution mirror assembly ( hrma ) aim - point for a detection threshold ) . this sensitivity would reach the typical luminosity of hmxbs ( both persistent supergiant hmxbs and be - hmxbs in outburst ) . a pair of very deep acis - s3 observations ( 2@xmath1345 ksec ) made in november 2006 provide a reference dataset for improved source positions and spectral information . the original @xcite _ chandra _ ( acis - s3 , 1/2 subarray ) observation of 30 ksec made in 2003 was included in our analysis to provide additional time baseline for the variability study and to make use of their published source identifications . the complete listing of 10 observation identifiers ( obsids ) comprising the dataset is summarized in table [ tab : dataset ] . all of the data were processed in a uniform way following the standard _ ciao _ threads , and the prescription of @xcite for consistent selection of energy bands b ( 0.3 - 8 kev ) , s ( 0.3 - 1.5 kev ) , and h ( 2.5 - 8 kev)and extraction regions . llllllll 52710.7 & 12 mar 2003 & 03953(a ) & 28.9 & 00:20:25 & 59:16:55 & 339.27 & 31 + 54041.8 & 2 nov 2006 & 07082 & 40.1 & 00:20:04 & 59:16:45 & 223.70 & 48 + 54044.2 & 5 nov 2006 & 08458 & 40.5 & 00:20:04 & 59:16:45 & 223.70 & 41 + 55140.7 & 5 nov 2009 & 11080 & 14.6 & 00:20:17 & 59:17:56 & 226.53 & 19 + 55190.2 & 25 dec 2009 & 11081 & 8.1 & 00:20:19 & 59:18:02 & 286.15 & 24 + 55238.5 & 11 feb 2010 & 11082 & 14.7 & 00:20:23 & 59:17:10 & 320.56 & 24 + 55290.6 & 4 apr 2010 & 11083 & 14.7 & 00:20:34 & 59:19:01 & 10.32 & 25 + 55337.8 & 21 may 2010 & 11084 & 14.2 & 00:20:25 & 59:20:16 & 67.89 & 27 + 55397.5 & 20 jul 2010 & 11085 & 14.5 & 00:20:11 & 59:19:13 & 121.25 & 22 + 55444.6 & 5 sep 2010 & 11086 & 14.7 & 00:20:15 & 59:18:11 & 157.71 & 27 + ( b ) & 2 - 5 nov 2009 & 57082 & 80.6 & - & - & - & 63 + images for visual inspection were generated using both gaussian - smoothed and adaptively - smoothed b ( 0.3 - 8 kev ) , s ( 0.3 - 1.5 kev ) , and h ( 2.5 - 8 kev ) band data . smoothing is necessary to create an image from sparse data . the adaptive smoothing maps created with _ ciao smooth _ for the b - band images were applied to make the s & h band images , and all 3 bands used the exposure maps generated for 1.5 kev photons ( roughly the peak of acis sensitivity ) . to illustrate our search for transient sources we present all 10 b - band images in figure [ fig : cxo_images ] which shows a number of prominent transient events and also illustrates the effects of shifting spacecraft roll - angle and aim - point . the individual variable sources can be identified with the aid of figure [ fig : cxo_image ] which is a representative color image constructed using the visualization and analysis program _ saoimage ds9 _ from the deep merged dataset 57082 ( table [ tab : dataset ] ) . for each _ chandra _ observation we ran _ ciao _ _ wavdetect _ on each of the three energy bands ( s , b , h ) . _ wavdetect _ was run with the default significance threshold of @xmath15 at wavelet scales of 1 , 2 , 4 , 8 , and 16 pixels . each detection is given its own source @xmath16 , and the detections in each band are cross - matched by position ( 95% uncertainty radius computed by the formula of @xcite ) to produce a list of unique sources within the observation , linked by an identifier called @xmath17 . the list derived from the 90 ksec stacked @xmath18 57082 was bore - sighted against the optical catalog of @xcite . source lists from each @xmath18 were then corrected to the same reference frame , using the coordinates of ic 10 x-1 ( by far the brightest source ) . we then compiled the final point - source catalog from the entire set of observations by using the tool _ classifytable _ in the _ starbase _ relational database package . _ classifytable _ is used to group together sources that lie within a certain distance ( 5 ) of each other , these matches are then filtered using the 95% uncertainty radii . all detections belonging to a unique source are linked by the identifier _ srcnumber_. coordinates and error radius are inherited from the @xmath16 with the best " detection ( most counts , smallest error radius ) . the number of unique x - ray point sources detected in each observation is listed in the final column of table [ tab : dataset ] . next , the unique source list for each observation was bore - sighted to the unique source list obtained from the deep merged dataset of obsids 07082 & 08485 ( which had identical pointing ) . this reference data set had itself already been bore - sighted to the optical catalog of @xcite . the bore - sighting procedure between x - ray observations used only the brightest source ( ic 10 x-1 ) due to its very small statistical positional uncertainty arising from its large count rate . once bore - sighting was complete , we constructed a master catalog using a 5 arcsec matching radius around each source ( @xmath17 ) . finally we applied a filter to eliminate false matches : requiring the matching @xmath17 coordinates to lie within the quadrature sum of their 95% error radii and an additional 0.75 " arcsec accounting for statistical and systematic errors . the resulting master catalog contains 110 unique point sources , many of which have multiple detections . all detections of the same source share the same _ srcnumber _ identifier and a _ chandra _ source name based on the usual convention ( cxou j hhmmss@xmath19ddmmss ) taking the coordinates from the @xmath16 with the smallest error circle . the _ chandra _ observations all have slightly different fields of view as a consequence of the operational requirements of the spacecraft . a certain pitch and roll angle are required to maintain the proper orientation with respect to the sun . we worked with the mission planners to adjust the exact pointing for each observation once the pitch / roll combination was known . this enabled us to keep most of the same point sources within the acis - s3 ccd during each observation . the master catalog was processed with a _ unix _ script that called _ ciao _ to query the exposure map for the coordinates of every source in each observation to determine whether the source was actually observed with acis - s3 or not . if the source was observed _ and not detected _ by _ wavdetect _ , then a new row was added to the master catalog with the modified julian date ( mjd ) of the observation , the value @xmath20 = null " and an estimated upper - limit for the broad - band count rate . this upper limit was computed as the equivalent count - rate that would generate 5 net counts at the hrma aim - point according to pimms ( portable multi - mission simulatorbin / tools / w3pimms / w3pimms.pl ] ) for an absorbed power - law spectrum ( @xmath21 @xmath22 ) divided by the normalized exposure - map value at the source location . the simulations of @xcite indicate this is a @xmath2375% detection probability for the off - axis distances encountered within the s3 chip , at low background rates . if the source was found to lie outside the s3 chip , then no modifications were made to the table . the master source catalog is maintained as a single table , combining all detections and upper limits , from which light curves are extracted as described in the next section . positional matching to the catalog of @xcite was performed using our and their 95% error radii in quadrature ; the resulting ids are provided in table [ tab : variables ] . to facilitate observations at other wavelengths , and cross - correlation with other databases we provide our complete point - source catalog as supplemental data in the online journal . light curves for all of the 110 unique sources were constructed by processing the master catalog with a _ unix _ script which computed the following parameters for each group of rows bearing the same @xmath24 identifier : number of occasions observed ( @xmath25 ) , number of detections in the broad - band ( @xmath26 ) , minimum count rate ( @xmath27 ) , maximum count rate ( @xmath28 ) , average count rate ( @xmath29 ) ; root - mean - square @xmath30 ; variability range ( @xmath31 = @xmath28 - @xmath27 ) ; variability ratio ( @xmath32 ) ; relative variability ( @xmath33 ) ; and total number of net broad - band counts ( @xmath34 ) obtained by summing the values from each detection . in computing the above parameters we included all b - band detections and b - band upper limits , except that @xmath35 ( and its associated error bar ) are required for a positive detection . the complete set of parameters characterizing each source are provided in the online journal as supplemental data accompanying table [ tab : variables ] , which shows only the variable objects , whose selection is described below . light curves were plotted for all x - ray sources in the master catalog ; selected examples are shown in figures [ fig : lcs_gt10 ] , [ fig : lcs_gt5 ] , and [ fig : lcs_lt5 ] . the plots display detections in all 3 energy bands and upper limits for observations with no b - band detection . all 110 light curves were visually examined alongside the _ acis _ images to verify the detections and to note any abnormalities such as sources falling very close to the edge of a ccd or having a close neighbor . in this way , we flagged 8 sources as probably false " ( objects detected only once at the extreme edge of the ccd ) and 4 sources as being duplicates ( multiple detections of a real source being falsely identified as separate sources when detected in subsequent observations ) . these eventualities are expected , and the number of cases was within the expected range . _ wavdetect _ expects to generate one false source per ccd ( i.e. @xmath23 10 in total ) and faint off - axis sources are subject to larger positional errors . we note that most of the sources were detected in at least two independent observations . specifically , 43 sources were detected more than once in the b - band , 36 only once in b - band ( some of which were also detected in a separate observation in s or h only ) , and 25 were only ever detected in the s or h band . the probability of a source with multiple detections being false is extremely small given the low surface density of sources . the light - curve parameters described above ( and given in table [ tab : variables ] ) were used to identify variable and transient sources by sorting on the @xmath36 and @xmath37 parameters . assuming the usual statistical definitions , a light curve with @xmath38 is variable at the 99% confidence level . there are 21 such objects , listed in table [ tab : variables ] ranked in descending order by @xmath38 . our criterion for detecting a variable or transient source is at least one b - band detection with count rate at least @xmath39 above the lowest upper limit ; thus , only real sources are included in table [ tab : variables ] . complete data for each variable source are provided in tabular form as `` data - behind - the figure '' accompanying this paper in the online journal . lllllllllllll 46 & 11c & 10 & 2 & 1.61 & 0.1 & 11.75 & 11.65 & 117.5 & 18.2 & 370 & 3.35 & t , x + 90 & - & 9 & 1 & 0.45 & 0.1 & 1.83 & 1.73 & 18.3 & 4.7 & 26 & 1.64 & t , s + 28 & 24c & 8 & 3 & 0.73 & 0.18 & 2.86 & 2.67 & 15.6 & 9.2 & 130 & 1.13 & t , s + 50 & 17c , 46x & 8 & 6 & 0.58 & 0.1 & 1.24 & 1.14 & 12.4 & 5.7 & 88 & 2.07 & t , x ? + 13 & - & 7 & 2 & 0.39 & 0.1 & 0.96 & 0.86 & 9.6 & 3.1 & 19 & 1.69 & tc , s ? + 30 & 19c & 10 & 4 & 0.36 & 0.1 & 0.92 & 0.82 & 9.2 & 3.1 & 33 & 1.97 & tc , x ? + 41 & - & 10 & 5 & 0.63 & 0.2 & 1.68 & 1.48 & 8.4 & 7.4 & 154 & 2.53 & tc , x + 43 & 24x & 8 & 7 & 1.11 & 0.3 & 2.45 & 2.15 & 8.2 & 5.1 & 170 & 2.06 & v , x + 20 & 18c , 39x & 10 & 10 & 105.8 & 23.7 & 179.87 & 156.17 & 7.6 & 55.6 & 22632 & 1.75 & v , x + 53 & - & 8 & 1 & 0.47 & 0.1 & 0.66 & 0.56 & 6.6 & 3.7 & 27 & 1.38 & tc , u + 5 & 14c , 35x & 8 & 8 & 1.55 & 0.54 & 3.17 & 2.63 & 5.9 & 6.7 & 292 & 0.89 & v , u + 29 & 4c & 10 & 1 & 0.31 & 0.1 & 0.51 & 0.41 & 5.1 & 3.6 & 22 & 3.60 & tc , x + 23 & 17x & 9 & 7 & 0.98 & 0.33 & 1.62 & 1.29 & 4.9 & 4.9 & 203 & 1.86 & v , u + 52 & - & 9 & 2 & 0.53 & 0.3 & 1.21 & 0.91 & 4.0 & 4.7 & 64 & 1.24 & v , s + 14 & 3c , 25x & 8 & 7 & 0.76 & 0.39 & 1.68 & 1.29 & 4.3 & 3.3 & 117 & 3.64 & v , x + 42 & 12c , 33x & 10 & 9 & 0.73 & 0.4 & 1.44 & 1.04 & 3.6 & 3.3 & 144 & 2.70 & v , x / a ? + 33 & 16x & 9 & 7 & 0.87 & 0.4 & 1.38 & 0.98 & 3.4 & 3.0 & 132 & 2.55 & v , x / a ? + 1 & 3c , 36x & 5 & 4 & 1.24 & 0.68 & 1.89 & 1.21 & 2.8 & 3.0 & 70 & 1.76 & v , x + 12 & 15c , 36x & 9 & 9 & 4.96 & 2.84 & 7.93 & 5.1 & 2.8 & 8.1 & 1137 & 1.69 & v , u + 18 & 8c & 10 & 10 & 1.99 & 1.18 & 2.77 & 1.59 & 2.3 & 3.6 & 435 & 1.12 & v , u + 8 & 7c , 28x & 8 & 8 & 4.22 & 3.1 & 5.8 & 2.7 & 1.9 & 4.5 & 863 & 1.83 & v , x + given the large distance to ic 10 we are able to extract spectra for relatively few sources . consequently the median energy for the combined events is reported in table [ tab : variables ] as an indication of spectral hardness . we plotted a hardness intensity diagram for all 110 point sources in figure [ fig : hardint ] , using cumulative events as a proxy for intensity , and the median event energy for the hardness axis . this parameter space serves to illustrate some key properties of the objects in our catalog . figure [ fig : hardint ] shows that large amplitude variability is seen at all intensity levels , that most bright sources are moderately variable ( @xmath40 ) , and that below 100 counts our selection method ( section [ sect : lc ] ) is sensitive only to large amplitude variables . thus we are not picking up random fluctuations in faint sources as variables . figure [ fig : hardint ] also shows that the distribution of median event energy is bimodal , with a hard / soft split at 1.4 kev . this was revealed by a kernel density estimator ( _ density _ in the _ r _ package , with bandwidth = 0.1 kev ) which we ran both with and without weighting by the errors . weighting increased the strength of the bimodality , as seen by comparing the solid and dashed lines in the right - hand panel of figure 6 . we performed a bootstrap resampling ( 10@xmath41 trials ) using the error - weighted values to generate 95% confidence intervals for the density plot . the _ kmeans _ clustering test identifies a similar hard - soft split , which can be attributed to different extinction values for ic 10 vs foreground galactic objects . energy quantiles @xcite were computed from the event lists for each unique source in the master catalog . quantiles were computed for each observation and also for the combined event list , which provides the maximum number of counts and hence delivers better s / n values . these values and the accumulated number of events in the source extraction regions is listed in the supplementary table in the online journal . in our analysis we use two of the quantiles defined by @xcite : the slope quantile @xmath42 which is related to the steepness of the spectrum , and the curvature quantile @xmath43 which is constructed from the ratio of the quartiles @xmath44 . full details are available in @xcite and in the source code .. the @xmath42 , @xmath43 parameter space shown in figure [ fig : quantiles ] was designed to explore in a single analysis , the spectral shape of low count sources that differ widely in hardness . panel a shows spectral model grids generated for the back - illuminated _ acis - s3 _ detector for two classes of sources : power law ( photon index @xmath45 ) and thermal bremsstrahlung ( @xmath46 ) , both of which with absorption varying between @xmath47@xmath22 . these models are intended to represent accretion - powered and stellar coronal sources , respectively . the quantile values for the variable sources selected in table [ tab : variables ] are plotted in panel b , with labels to facilitate discussion . error bars are in panel c , and those sources with optical counterparts ( table [ tab : optical ] ) are highlighted in panel d. pulse - height spectra were extracted for all sources having greater than 200 net counts in the master catalog , yielding 7 objects ( there are 13 sources with @xmath1 100 net counts ) . in this article , we include only those appearing in table [ tab : variables ] after excluding ic 10 x-1 and x-2 ( which have been reported elsewhere ) , which leaves 5 objects for the analysis in this paper . for every such object , we used the _ ciao _ script _ specextract _ to generate source and background spectra and response files ( rmf , arf ) for each observation ; and also for the merged event list containing all observations in which the source was detected . the resulting spectra are plotted in figure [ fig : spectra ] . the spectra were examined in xspec and were fitted by the following models ( combined with the _ phabs _ absorption model ) : power - law ( pl ) , raymond - smith ( rs ) , mekal ( m ) , black - body ( bb ) , and thermal bremmstrahlung ( tb ) . these models were selected as appropriate for x - ray spectra of accretion , magnetic flare , and thermal origin . in cases where the absorption was unconstrained by the model we froze it at 5@xmath48 @xmath22 for the column density toward ic 10 . the resulting model parameters and goodness of fit statistics are presented in table [ tab : xrayspec ] . given the low s / n ratios , only single component fits were considered . we caution that both quantiles ( in common with all flux ratio - based color systems ) and single - component spectral fits can be non - unique . for example spuriously low @xmath49 values can occur when soft components remain unaccounted for @xcite . this is a motivation to obtain deeper datasets where such components can be resolved . cccccc source 5 + pl & 0.842 @xmath19 0.3 & - & 7.5 @xmath19 2.2 & 2.0 @xmath19 0.9 & 1.167/74 + rs & 0.442 @xmath19 0.1 & 0.28 @xmath19 0.06 & - & 2.1 @xmath19 1.2 & 0.596/74 + m & 0.000 @xmath19 0.1 & 0.55 @xmath19 0.08 & - & 0.21 @xmath19 0.09 & 0.631/74 + bb & 0.006 @xmath19 0.1 & 2.8 @xmath19 0.02 & - & 7.6 @xmath19 0.8 & 1.829/74 + tb & 0.78 @xmath19 0.5 & 0.13 @xmath19 0.04 & - & 0.02 @xmath19 0.001 & 0.934/74 + source 8 + pl & 0.565 @xmath19 0.09 & - & 1.6 @xmath19 0.2 & 1.3 @xmath19 0.2 & 1.020/53 + rs & 0.483 @xmath19 0.07 & 10.2 @xmath19 4.2 & - & 4.2 @xmath19 0.3 & 1.011/53 + m & 0.482 @xmath19 0.07 & 10.4 @xmath19 4.4 & - & 4.2 @xmath19 0.2 & 1.017/53 + bb & 0.051 @xmath19 0.06 & 0.92 @xmath19 0.005 & - & 0.061 @xmath19 0.004 & 1.049/53 + tb & 0.49 @xmath19 0.07 & 9.4 @xmath19 3.8 & - & 1.5 @xmath19 0.1 & 1.001/53 + source 12 + pl & 0.73 @xmath19 0.08 & - & 2.2 @xmath19 0.1 & 2.9 @xmath19 0.4 & 0.835/64 + rs & 0.57 @xmath19 0.06 & 4.1 @xmath19 0.5 & - & 5.8 @xmath19 0.4 & 0.866/64 + m & 0.57 @xmath19 0.06 & 4.1 @xmath19 0.6 & - & 5.7 @xmath19 0.4 & 0.883/64 + bb & 0.19 @xmath19 0.05 & 0.71 @xmath19 0.03 & - & 0.065 @xmath19 0.3 & 1.180/64 + tb & 0.58 @xmath19 0.05 & 3.6 @xmath19 0.7 & - & 2.5 @xmath19 0.3 & 0.849/64 + source 18 + pl & 0.831 @xmath19 0.2 & - & 6.2 @xmath19 2.1 & 0.41 @xmath19 0.2 & 0.993/46 + rs & 0.072 @xmath19 0.03 & 6.2 @xmath19 1.8 & - & 0.16 @xmath19 0.09 & 1.013/46 + m & 0.623 @xmath19 0.2 & 0.56 @xmath19 0.2 & - & 0.19 @xmath19 0.1 & 0.990/46 + bb & 0.000 @xmath19 0.04 & 0.32 @xmath19 0.05 & - & 0.001 @xmath19 0.0004 & 1.004/46 + tb & 0.453 @xmath19 0.1 & 0.39 @xmath19 0.1 & - & 1.17 @xmath19 0.9 & 0.9934/46 + source 23 + pl & 2.45 @xmath19 0.9 & - & 5.8 @xmath19 0.6 & 5.4 @xmath19 1.2 & 0.803/71 + rs & 3.31 @xmath19 0.5 & 0.21 @xmath19 0.06 & - & 0.002 @xmath19 0.0005 & 0.805/71 + m & 0.834 @xmath19 0.3 & 1.06 @xmath19 0.9 & - & 0.95 @xmath19 0.6 & 0.834/71 + bb & 0.001 @xmath19 0.002 & 3.31 @xmath19 0.9 & - & 0.15 @xmath19 0.08 & 0.827/71 + tb & 1.02 @xmath19 0.8 & 1.7 @xmath19 1.2 & - & 0.42 @xmath19 0.2 & 0.810/71 + a list of positional counterparts from @xcite was generated by the procedure described in @xcite . a limiting visual magnitude of @xmath50 was used for sources without counterparts , this being the faintest magnitude of any object in the catalog . the distance independent x - ray to optical flux ratios were found using @xmath51 where the observed flux @xmath52 was computed from the net count rate assuming an absorbed power - law with a photon index of 1.5 and @xmath53 @xmath22 . the values in table [ tab : optical ] are the maximum measured values , remembering that @xmath52 is variable . in the following analysis , sources with optical counterparts believed to lie in ic 10 have their apparent magnitudes corrected for the foreground extinction assuming a value of @xmath54= 2.85 and distance modulus @xmath55 . narrow - band @xmath11 and @xmath56 images from the gemini observatory were examined in order to look for sources of balmer - line emission as described in @xcite . preliminary spectroscopic followup was reported by @xcite , although spectral type and radial velocity ( rv ) are only available for 3 of the variable sources . four sources have variability @xmath57 and @xmath58 , these are src 46 ( ic 10 x-2 ) , src 90 , src 28 , & src 50 ( listed in order of decreasing amplitude ) . the light curves for these sources appear in figure [ fig : lcs_gt10 ] . there are a further 2 sources ( src 53 and src 29 ) that we include as transients as each was detected in a single observation , with a broad - band count rate exceeding the lowest upper limit by three @xmath0 ( i.e. @xmath59 and @xmath60 , but not reaching @xmath57 ) listed in table [ tab : variables ] . both were in the acis - s3 field of view 8 - 10 times , so that the null detections and corresponding upper limits place meaningful limits on the duty - cycle . lightcurves for these two objects are include in figure [ fig : lcs_gt5 ] . only one transient , src 46 , which is ic 10 x-2 , @xcite provided enough counts to produce pulse - height spectra and appears to be an x - ray binary with a supergiant be companion and similarities to supergiant fast x - ray transients ( sfxts ) . for reference we include its lightcurve in figure [ fig : lcs_gt10 ] and it appears on the quantile diagram ( figure [ fig : quantiles ] ) at ( -0.18,1.17 ) which is consistent with a very hard power - law spectrum . the case of ic 10 x-2 highlights the need for optical spectra , since its photometric colors are unremarkable unless one knows the appropriate correction , and its nature as an extragalactic hmxb required the unambiguous doppler - shift to place it firmly in ic 10 . src 28 was observed 8 times yielding 3 b - band ( and one s - band only ) detections with a variability ratio of 15.6 , a 9.3@xmath0 variation above the most constraining upper limit . quantile values for the 4 observations were all consistent with the merged value . lying at [ -0.9,1.1 ] in figure [ fig : quantiles ] the source is consistent with a soft power law ( @xmath61 ) or a @xmath23few kev thermal plasma . in both cases the absorption is small ( @xmath62@xmath22 ) suggesting a foreground galactic location for the object ( subject to the caveats in section [ sect : quantile ] . ) there are two optical counterparts ( table [ tab : optical ] from @xcite ) and the source is not in the field of the gemini images , so no followup spectrum was obtained . only one of these stars is consistent with the x - ray detection having the smallest error circle . both stars have color indices suggestive of spectral type m. taking the most likely counterpart , whose b - v is already at the un - reddened value for m0v , we calculate distance modulus @xmath63 . the distance is then @xmath23900 pc , yielding peak @xmath64erg s@xmath3 , in the range of isolated dme stars , which is supported by the distance independent ratio @xmath65=-0.53 being lower than seen for active binaries . note that in the absence of a spectral type or rv value , ic 10 membership can not be entirely ruled out . if located at 660 kpc , and de - reddened by @xmath66=2.85 the star would have an extreme absolute magnitude @xmath67 , which favors the galactic dme interpretation . src 29 was detected in _ obsid 08458 _ with 22 net counts which placed it a factor of 5 ( or 3.6@xmath0 ) above its most sensitive non - detection . the detection is in the b and h ( and not s ) bands . the quantile value [ -0.13,1.2 ] lies in the hard power law region ( @xmath68 ) of figure [ fig : quantiles ] on the contour for @xmath69 @xmath22 . there are 2 optical counterparts in m07 having near identical photometry ( @xmath70 21.7 , @xmath71 ) , gemini @xmath11 line and continuum images indicate that two stars are blended in the @xcite images , with one catalog position on top of the brighter star , and the other midway between the two . the source lies in a region containing clumpy diffuse @xmath11 emission , but neither star is a strong @xmath11 point - source . a gmos spectrum of the brighter star ( which is near the center of the x - ray error circle ) shows a strong continuum without obviously identifiable features other than telluric lines . this rules out a foreground m - dwarf and suggests a reddened early - type star in ic 10 . applying the distance modulus and extinction correction assumed throughout this paper gives @xmath72 , and @xmath73 . the star is also blue in u - b and v - r color indices compared to the other stars in table [ tab : optical ] , and once de - reddened looks like a b spectral type . this object is likely an hmxb in ic 10 , based on x - ray and optical properties . src 50 was observed 8 times , yielding 6 b - band detections and two upper limits . the most constraining upper limit comes from one of the pair of 45 ksec exposures made in nov 2006 providing the evidence for a large variability ratio ( @xmath74=12 , @xmath36=5.7 ) . the quantile values varied with time , the spectrum being harder in _ obsids 11081,11082,11084_. the merged value plotted in figure [ fig : quantiles ] lies at [ -0.52,1.05 ] which is dominated by photons from its softer apparition in 2006 . the model grid indicates a power law ( @xmath75 ) spectrum with typical ic 10 absorption ( @xmath23 5@xmath48@xmath22 ) . there is an optical counterpart in @xcite but it lies outside the gemini field and no followup spectrum has been obtained . if the object is in ic 10 , then its de - reddened photometry gives @xmath76 , @xmath77 . src 53 was detected only in _ obsid 08458 _ with 27 net counts , at a count - rate a factor of 6.6 ( 3.7@xmath0 ) above the most constraining non - detection _ obsid 07082 _ which was of equal exposure - time ( 45 ksec ) and pointing , and occurred only a day before . a point - source is clearly visible in the image of _ obsid 08458 _ and absent in _ obsid 07082_. the detection was made in the s and b bands only . the very rapid variability and soft spectrum implied by the s - band detection are indicative of a magnetic stellar flare but not conclusive . the quantile value [ -0.8,0.5 ] lies outside our model grids and has large errors so we can say little about its spectral shape . there is an optical counterpart ( v=21.4 , table [ tab : optical ] ) but it lies outside the gemini field and no followup spectrum has yet been obtained . src 90 was covered by the acis - s field of view 9 times , but was detected only once , in _ obsid 11082 _ in the s , b & h bands . it is clearly visible as a point source in the adaptively smoothed image . the observed count - rate was a factor of 18 ( 4.7@xmath0 ) above the most constraining upper limit . the energy quantile ( figure [ fig : quantiles ] ) lies at [ -0.67,1.23 ] within the power law model grid at a location consistent with @xmath78 , @xmath79 @xmath22 . there is an optical counterpart ( [ tab : optical ] ) which we identify as an @xmath11 emission line star in gemini narrow - band imaging . we obtained a gemini gmos optical spectrum of this star @xcite , which shows a red continuum with molecular bands characteristic of spectral type m. prominent lines of @xmath11 , @xmath80 in emission enable us to measure the radial velocity rv=-165 km / s , rv=-171 km / s for the respective lines . thus the star belongs to the milky way since the mean rv of ic10 is -340 km / s . photometry from @xcite ( v=21.348 , b - v=1.8 ) indicates the color - index is consistent with early m. for example spectral type m0ve ( @xmath81 would require only e(b - v)@xmath230.3 , @xmath82 . the distance modulus is then @xmath83 , for a distance of @xmath232 kpc , implying @xmath84erg s@xmath3 . the distance - independent flux ratio during the flare reaches @xmath65= + 0.3 . stellar flares in dme stars reach @xmath85 , with the by dra binary type being particularly luminous @xcite . the luminosity , quantile , and duty - cycle point to a hard and energetic event on a magnetically active m - dwarf . such interlopers are frequent bi - catch of x - ray surveys , and we have seen one in the foreground of the smc @xcite . many of the brighter sources in our study show significant variability ( @xmath38 ) as expected given that astrophysical x - ray emission mechanisms are dynamic in nature . the light curves of these candidates are presented in figure [ fig : lcs_gt5 ] for those varying by a factor of 5 - 10 , and in figure [ fig : lcs_lt5 ] for those not exceeding a factor of 5 . all of these objects have sufficient counts in the merged event - list to provide energy quantiles with reasonable s / n . in figure [ fig : quantiles ] we plot the quantiles for all unique objects in the master catalog having at least 10 counts in their merged event list , with the candidate variable sources picked out in larger points . a group of 10 candidate variables clusters along the @xmath49 contour corresponding to @xmath86@xmath22within the power law model grid region indices in the range @xmath87 . based on this evidence they are consistent with being hmxbs in ic10 . in the following section we present individual analyses for the remaining variability candidates listed in table [ tab : variables ] , beginning with the five candidate variables to exceed 200 net counts . for these objects we performed spectral fitting in xspec testing pl , bb , tb , nei models combined with the photo - absorption model _ phabs_. the best fit for each of these sources is displayed in table [ tab : xrayspec ] . source 1 was observed in half of the acis - s observations and is detected in 4 of the 5 . the are enough counts for the quantile diagram to place it firmly in the xrb locus . the optical counterpart @xmath88 if adjusted for distance ( @xmath55 ) and reddening ( @xmath89 ) gives @xmath90 which implies an early b spectral type . thus the object is an hmxb , subject to optical spectral confirmation of the precise subtype . the light - curve for src 5 ( figure [ fig : lcs_gt5 ] ) shows a large amplitude variable with a 100% duty cycle , being detected in all observations for which it was in the fov . it also shows that virtually all the counts are in the soft band . the hardness - intensity plot ( figure [ fig : hardint ] ) shows src 5 is at once the softest variable , and one of the brightest objects in the survey , pointing to a foreground object . quantile analysis suggests a thermal spectrum with median energy of just 0.5 kev that did not vary between detections . the @xmath91 net - count spectrum was best fit by a black - body model ( fig . [ fig : spectra ] , table [ tab : xrayspec ] ) . a power - law model can provide an acceptable fit but the spectral index of 7.7 implies the cool thermal plasma is more physically motivated . no optical counterpart is found in @xcite and the source lies outside the gemini field . src 8 is a very low amplitude variable with 100% duty cycle in the chandra survey . the summed 863 net - count spectrum was fit with a power - law model of @xmath92 , and typical ic 10 absorption ( fig . [ fig : spectra ] , table [ tab : xrayspec ] ) . there is an optical counterpart in @xcite ( lies outside the gemini field ) for which we can estimate @xmath93 , @xmath94 which would fit a supergiant or be - x interpretation . src 12 is a persistent source with 100% duty cycle and a lightcurve that suggests very slow x - ray variability . the source has the second highest number of total x - ray counts ( @xmath11000 ) in the survey , so spectral fitting was feasible . the quantiles lie at the upper end of the cluster of points in figure [ fig : quantiles ] , and accordingly a powerlaw fit to the spectrum yields @xmath95 powerlaw with typical absorption for the ic 10 field ( figure [ fig : spectra ] , table [ tab : xrayspec ] ) . turning to physically motivated models , acceptable fits are obtained for the thermal plasma models ( tb , rs , mekal ) , all giving temperatures consistent with @xmath96 , and @xmath97@xmath22 similar to the powerlaw fit . a blackbody fit yields a significantly lower column density and temperature ( kt@xmath981 kev ) . the lack of an optical counterpart in m07 provides the following constraints on distance - independent flux ratio ( @xmath65 @xmath99.9 ) and absolute magnitude ( @xmath100 ) if located in ic 10 . the object could be an agn on the basis of the spectral index , large @xmath65 and slow variability . a foreground dme star seems unlikely due to the lack of flares and persistently high @xmath65 . an xrb can not be ruled out . given the persistent nature of this source it will be possible to obtain much higher s / n x - ray spectra and timing in the future . a transient x - ray source appearing to exhibit rapid variability , src 13 was detected ( at a very low flux ) in only one of the deep ( 45 ksec ) observations separated by 2 days in 2006 . subsequently the source flared up in a single one of the 5 monitoring observations of 2009/10 which are spaced approx . 6 week apart . these two independent positive detections confirm the reality of the object . the limiting fluxes for those four nulls are at or above the flux detected in the deep 2006 exposure . both positive detections were dominated by soft - band counts , with no detection in the hard band . the quantile value is consistent with a plasma temperature of @xmath101 , or a @xmath102 powerlaw . there is an optical counterpart within the 95% confidence radius . the photometry ( table [ tab : optical ] : @xmath103 ) would be consistent with an un - reddened m0v star at a distance of 2 kpc , or a hotter spectral type at larger distance . it is impossible to make the star closer since the observed color index imposes the limit of @xmath104 . if located in ic 10 we obtain @xmath105=-6.4 . an optical spectrum is needed to resolve the nature of this source . src 14 has an interesting lightcurve that contains an outburst of duration @xmath232 months during 2010 . quantile analysis reveals a persistently hard source , having the highest median photon energy ( e50 = 3.64 kev ) of any object in table [ tab : variables ] , and lying in the high - absorption hard power - law region of figure [ fig : quantiles ] . the lack of an optical counterpart down to a limiting magnitude of @xmath10623.8 provides us with a distance independent flux ratio @xmath65@xmath107 + 1.24 for the brightest recorded x - ray flux . if located in ic10 , the absolute magnitude would be @xmath108 if we assume the typical extinction correction . a more luminous star would require more extinction to remain hidden . spectral type b0v has @xmath109 so the data support a be - xrb . if a dme star located in the milky way , then the spectral type would have to be late m ( for m0v the luminosity distance implied by @xmath50 is greater than 9 kpc ) and the x - ray quantiles are too persistently hard . in summary the long duration outburst , persistently hard spectrum , and high @xmath65 point to an xrb . src 18 is a low amplitude ( less than a factor of 2 ) variable with 100% duty cycle in the chandra survey . the total 435 net - count spectrum is soft , equivalent to powerlaw @xmath110 or cool thermal plasma for any reasonable @xmath49 ( fig . [ fig : spectra ] , table [ tab : xrayspec ] ) . there is no optical counterpart in m07 , providing a lower limit on @xmath65 @xmath1 + 1.12 . the x - ray position lies at the center of a faint compact @xmath11 shell identified in narrow - band gemini images . a gmos slit placed across the ring did not attain sufficient s / n to detect any lines . src 23 is a variable source with a high duty cycle ( 7/10 ) and no optical counterpart in m07 which puts a lower limit on @xmath65 @xmath1 + 1.2 . with 203 net counts summed over the 6 observations where it was detected we obtained an acceptable x - ray spectral fit to the thermal bremsstrahlung spectral model with @xmath111=1.7 kev , @xmath112 @xmath22 ( fig . [ fig : spectra ] , table [ tab : xrayspec ] ) . due to the low number of counts all simple models obtained similar @xmath113 values , with the powerlaw model confirming a soft spectrum ( @xmath110 ) and the blackbody fit unable to constrain the @xmath49 to a significantly non - zero value . indicative of a foreground galactic location . transient source with duty cycle @xmath114 and no optical counterpart , if located in ic 10 the limiting magnitude @xmath115 implies @xmath65@xmath116 subject to a typical @xmath49 . the quantile value is also consistent with an hmxb . another high duty - cycle , hard , variable source without optical counterpart . as with src 42 , the object could be an hmxb or agn with its high @[email protected] , and limiting @xmath115 implying @xmath117 depending on absorption . hard transient source with duty cycle @xmath114 and no optical counterpart , the lightcurve is reminiscent of a be - xrb . if located in ic 10 the limiting magnitude @xmath115 implies @xmath65@xmath118 subject to a typical @xmath49 . the quantile value is also consistent with an hmxb . this is a persistent variable source with a hard x - ray spectrum ( median energy = 2.7 kev ) , and no optical counterpart in m07 . table [ tab : variables ] reports 9 broad - band detections out of 10 observations , however the x - ray lightcurve ( fig [ fig : lcs_lt5 ] ) shows a hard - band only detection in that observation ( final point in 2010 ) , so the source has a 100% duty cycle . there are not enough counts to analyze the spectrum , however the quantile diagram fig [ fig : quantiles ] shows src 42 ( the point at @xmath119 ) is consistent with a heavily absorbed powerlaw @xmath120@xmath22 . the limiting magnitude of @xmath121 constrains the optical counterpart to @xmath122 for typical ic10 absorption , however the likely much higher @xmath49 could hide an o star . the flux ratio @xmath65@xmath1 + 1.2 together with the hard spectrum and variability make an hmxb the most likely interpretation , although it could conceivably be a background agn . variable source on all timescales probed , and seen in all observations except for the 2003 exposure ( which places a limit on its quiescent flux ) , although since it appears as [ w05]24x we know the source was seen by _ xmm - newton _ in 2003 . the lack of an optical counterpart limiting magnitude @xmath115 implies @xmath65@xmath123 if located in ic10 . the energy quantile is hard and lies in the hmxb region . this object is a large amplitude variable with 3 detections out of 9 observations . two of these detections are mere days apart , and show a factor of 5 change in flux ( perhaps a flare ) , with the fainter point only seen due to the long observation ( 45 ksec ) , while a 3rd and final detection was seen in the soft - band ( only ) in a single one of the 2009/10 monitoring series . there is a bright optical counterpart in m07 , which at @xmath124=17.3 , @xmath125=+1.6 is almost certainly a galactic star ( the alternative being an extreme luminosity @xmath105=-9.6 hyper - giant if located in ic 10 ) . assuming spectral type m0 ( based on the color index ) the distance modulus is @xmath126 giving a distance of 450 pc . the distance independent flux ratio @xmath65= -1.5 supports this interpretation but an optical spectrum is needed to confirm it . lllllllllll 1 & j002008.68 + 591340.7 & 0.15 & 0.43 & 22.384 & 0.700 & -0.087 & 0.762 & 0.855 & 0.77 & n + 8 & j002012.68 + 591501.4 & 0.1 & 0.32 & 22.974 & 1.242 & & 0.735 & 1.144 & 1.45 & n + 13 & j001958.69 + 591527.7 & 0.37 & 0.53 & 20.473 & 1.525 & & 0.932 & 0.962 & -0.33 & n + 20 & j002029.08 + 591651.7 & 0.30 & 0.28 & 21.722 & 0.905 & -1.094 & 0.137 & 0.754 & 2.43 & + 20 & j002029.12 + 591651.8 & 0.17 & 0.28 & 22.478 & 0.017 & -0.916 & 0.851 & 0.805 & 2.74 & y + 28 & j002046.55 + 591731.7 & 0.61 & 0.53 & 18.793 & 1.521 & 1.327 & 0.957 & 0.958 & -0.53 & + 28 & j002046.36 + 591732.4 & 0.55 & 0.80 & 20.199 & 1.786 & 99.999 & 1.191 & 1.395 & -0.53 & n + 29 & j002009.14 + 591738.3 & 0.32 & 0.39 & 21.777 & 1.059 & -0.820 & 0.536 & 1.052 & -0.09 & n + 29 & j002009.15 + 591738.4 & 0.37 & 0.39 & 21.732 & 0.932 & -0.588 & 0.494 & 1.049 & -0.1 & n + 46 & j002020.94 + 591759.3 & 0.28 & 0.31 & 19.954 & 1.211 & -0.259 & 1.043 & 0.820 & 0.54 & y + 50 & j002047.28 + 591935.7 & 0.89 & 0.66 & 22.539 & 1.823 & & 1.292 & 1.564 & 0.6 & + 52 & j002033.29 + 592135.4 & 1.16 & 0.75 & 17.277 & 1.551 & 1.365 & 0.946 & & -1.5 & + 53 & j002013.70 + 592236.1 & 0.49 & 0.96 & 21.404 & 1.872 & & 1.354 & 1.603 & -0.2 & + 90 & j002028.42 + 592024.0 & 0.34 & 0.54 & 21.348 & 1.805 & & 1.294 & 1.579 & 0.37 & y + 18 & * & & & & & & & & & y + from a total of 110 unique point sources detected in 10 _ chandra _ observations ( years 2003 - 2010 ) , we found 21 that are variable at the 3@xmath0 level . the x - ray light curves exhibit a range of qualitatively different behaviors , including transients , large amplitude variables , and persistent sources with duty cycles ranging up to 100% . analysis in the x - ray quantile diagram reveals a split between soft thermal sources and harder power - law like spectra that cluster along the model grid contour consistent with the average column density towards ic 10 . while this separation is not a complete diagnostic of source classes , it supports to a large extent the hypothesis of @xcite that ic 10 hosts numerous hmxbs . our variability study , better positions , and the increased spectral s / n ratios made possible by analyzing a cumulative 225 ksec of chandra exposure ( vs the previous 30 ksec ) have provided a richer and more detailed view of this population . in this paper , we were able to confidently classify most of the variable sources and to usefully constrain the nature of the remaining sources . deeper observations are needed to increase the number of individual sources for which x - ray spectral fits can be obtained . variability is a characteristic of all known astrophysical x - ray production mechanisms , and the fact that the majority of the catalog sources do not make the 3@xmath0 cut is due to their faintness ( and consequent difficulty in establishing variability ) of most objects encountered . the hmxb candidates presented in this paper are all characterized by power - law spectra , strong x - ray variability , and either the presence of an optical counterpart consistent with an o / b star in ic10 ; or the lack of an optical counterpart to a limiting magnitude of @xmath127 which still allows for the existence of an early type star ( b0v or fainter , @xmath100 for minimal ic 10 extinction @xmath89 ) . substantially higher extinction values do exist in parts of ic 10 as there is a high degree of structure in the interstellar medium ( ism ) of that galaxy . in the former category , we find src 1 , src 8 , and src 29 , in addition to the well - studied src 20 = ic 10 x-1 and src 46 = ic 10 x-2 , for a total of 5 hmxbs with optical counterparts , only one of which has a known orbital period and three have no optical spectrum . of the sample in this paper , only 12 x - ray sources have positional optical counterparts in @xcite , among which we identify 4 sources with @xmath11 line emission . optical spectra have been obtained for just 4 counterparts to x - ray variables . these were the known bh+wr binary ic 10 x-1 ; the previously reported hmxb ic 10 x-2 ; a reddened early - type star associated with the hard absorbed transient source # 29 ; and a foreground dme star associated with the soft transient source # 90 . given that highly luminous stars exist in ic 10 , spectroscopic followup is required to positively identify the majority of the optical counterparts to these sources . the photometric colors of the counterparts are expected to be somewhat degenerate between extinction and spectral type due to the large line - of - sight depth through the foreground galactic plane . future efforts must be directed at obtaining optical spectra of the counterparts which can be efficiently done with multi - object spectrographs on 8-m telescopes thanks to the compact size of ic10 . positive spectroscopic identification of the x - ray binaries requires spectral types , confirmed ic 10 membership via radial velocity , and ultimately orbital elements from rv and/or photometry . the distance and extinction values to ic 10 indicate that only the upper - end of the spectral sequence can be resolved from ground - based telescopes , however this is not an insurmountable problem since hmxbs are formed from precisely these stars . this paper reports objects with x - ray properties similar to known hmxbs , both with and without candidate optical counterparts . no transients were seen at super - eddington luminosity , thus no candidate bh x - ray novae occurred during any of the observations . transient x - ray sources are conventionally identified by a variability ratio in the @xmath110 range and transient candidates in the 5 - 10 range ( e.g. @xcite ) . we are aware of powerful and innovative approaches that rely on bayesian analysis ( e.g. @xcite ) to obtain the probability that the variability ratio for an individual source lies within the range of interest . these techniques reduce bias and hence offer two advantages : ( a ) enhanced sensitivity to extremely faint transients even in the case of a very small number of observations , ( b ) reduced tendency to falsely identify positive photon - counting fluctuations as variability . our selection criteria are based on the conventional statistical prescription for comparing data points . in particular the 3@xmath0 cut is conservative ( 99% significance ) and avoids sweeping up low probability transient candidates . with 110 point sources detected , the expectation is of order 1 false variable candidate out of the 21 variables identified . admittedly , the cost of this conservative approach is reduced sensitivity to fainter transients . there are 36 single - detection b - band sources , only 2 of which made it into the sample presented here . as we showed in figure [ fig : hardint ] all of those 34 other objects are very faint , therefore they are not hmxbs in outburst . as to the nature of the new hmxbs reported here , we can already assert that they are a different population than that in the magellanic clouds . the young age of the underlying stellar population and the elevated production of massive stars and stellar remnants noted by other authors ( e.g. @xcite ) have implications for the expected x - ray binary population . the smc with its episodic starburst history contains at least 100 known or suspected hmxbs , the vast majority of which are be+ns systems , and all earlier than b3 @xcite . it was first proposed by @xcite that the narrow spectral type range seen in be - hmxb is an evolutionary signature . this idea is supported by the groundbreaking work of @xcite , who showed the smc be - hmxb are associated with distinct stellar populations of age 40 - 70 myr and @xcite who found a similar age - association for hmxbs in ngc 300 and ngc 2403 . in ic 10 there has not been sufficient time to form these objects , therefore we expect to see completely different hmxb species . following this logic , we assume that the be phenomenon is not yet widespread in ic 10 , hence accretion - powered x - rays must involve mass donors with stronger winds and/or smaller orbital separation in order to deliver the requisite mass - transfer rate . supergiants and wr stars are the obvious candidates , although we find only one x - ray source ( x-1 ) coincident with the wr catalog of @xcite ; weak - lined wr stars could be present but as - yet unidentified . some very young hmxbs have recently been reported : @xcite found an age of @xmath45 myr for sn 2010da ( an hmxb in ngc 300 ) ; in own galaxy , cir x-1 has been revealed as the youngest known related example at @xmath44600 yr @xcite . however , there is currently debate about the nature of the optical companion in sn 2010da ( lbv , yellow supergiant , etc . ) , and the companion in cir x-1 has proven infamously difficult to classify . @xcite proposed that the recent sn in cir x-1 has shock - heated the envelope of the companion , puffing it up , which can reconcile the strangely red color for an hmxb . consequently we are wary of excluding red objects from the ic 10 sample and we must obtain spectra of all counterparts . in summary , our _ chandra _ monitoring study has found a heterogeneous population of exotic hmxbs in ic 10 . we are witnessing a starburst at an early stage , even perhaps before its peak @xcite , so these unfamiliar systems will play an important role in understanding the age / metallicity dependence of massive binaries , the x - ray binary luminosity function , and ulx sources . we are most grateful to the anonymous referee for their constructive improvements to the manuscript . this project was made possible by the support of sao grant nas8 - 03060 and the physics department of university of massachusetts , lowell . sl also thanks the gemini observatory for its support . gemini is operated by the association of universities for research in astronomy , inc . , on behalf of the international gemini partnership of argentina , australia , brazil , canada , chile , the united kingdom , and the united states of america .

in 1976 thurston associated to amanifold a marked polytope in which measures the minimal complexity of surfaces representing homology classes and determines all fibered classes in . recently the first and the last author associated to a presentation with two generators and one relator a marked polytope in and showed that it determines the bieri neumann strebel invariant of . in this paper , we show that if the fundamental group of a 3manifold admits such a presentation , then the corresponding marked polytopes in agree .

1507.05660

throughout this paper all @xmath0manifolds are compact , connected and orientable . suppose @xmath1 is a 3manifold . in 1976 thurston @xcite introduced a seminorm @xmath7 on @xmath8 , henceforth referred to as the _ thurston norm _ , which is a natural measure of the complexity of surfaces dual to integral classes . a class @xmath9 is _ fibered _ if @xmath10 can be represented by a non - degenerate closed 1-form . if @xmath10 is integral , then @xmath10 is fibered if and only if it is induced by a surface bundle @xmath11 . we refer to [ section : thurstonnorm ] for details . thurston @xcite showed that the information on the thurston seminorm and the fibered classes can be encapsulated in terms of a _ marked polytope_. a marked polytope is a polytope in a vector space together with a ( possibly empty ) set of marked vertices . in order to state thurston s result precisely we need one more definition . given a polytope in a vector space @xmath12 we say that a homomorphism @xmath13 _ pairs maximally with the vertex @xmath14 _ if @xmath15 for all other vertices @xmath16 . in this language , the main result of thurston @xcite can be stated as follows . [ thm : thurston ] let @xmath1 be a 3manifold . there exists a unique symmetric marked polytope @xmath17 in @xmath18 with vertices in @xmath19 such that for any @xmath20 we have @xmath21 and such that @xmath10 is fibered if and only if it pairs maximally with a marked vertex of @xmath17 . subsequently , by a @xmath22presentation we mean a group presentation with precisely two generators and one non - empty relator . a @xmath22presentation is _ cyclically reduced _ if the relator is a cyclically reduced word . recently the first and the third author @xcite associated to a cyclically reduced @xmath22presentation @xmath23 a marked polytope @xmath24 in @xmath5 . now we outline the definition of @xmath24 in the case that @xmath25 . a different ( but equivalent ) definition is given in section [ section : defpolytopepi ] , as well as a definition for cyclically reduced @xmath22presentations @xmath4 with @xmath26 . identify @xmath27 with @xmath28 such that @xmath29 corresponds to @xmath30 and @xmath31 corresponds to @xmath32 . then the relator @xmath33 determines a discrete walk on the integer lattice in @xmath34 and the marked polytope @xmath24 is obtained from the convex hull of the trace of this walk as follows : start at the origin and walk across @xmath28 reading the word @xmath33 from the left . take the convex hull @xmath35 of the set of all lattice points reached by the walk . mark precisely those vertices of @xmath35 which the walk passes through exactly once . now consider the unit squares that are completely contained in @xmath35 and touch a vertex of @xmath35 . mark a midpoint of a square precisely when one ( and hence all ) vertices of @xmath35 incident with the square are marked . the set of vertices of @xmath24 is the set of midpoints of all of these squares , and a vertex of @xmath24 is marked precisely when it is a marked midpoint of a square . in figure [ fig : dunfield - intro ] we sketch the construction of @xmath24 for the presentation @xmath36 where @xmath37 this example is due to dunfield @xcite and presents the fundamental group of the exterior of the 2component link in @xmath38 shown in figure [ fig : dunfield ] ( see section [ section : dunfield ] ) . given two polytopes @xmath39 and @xmath40 in a vector space @xmath41 we write @xmath42 if the polytopes @xmath39 and @xmath40 differ by a translation , i.e. if there exists @xmath43 with @xmath44 . the following is the main theorem of this paper . theoremmainthm [ mainthm ] let @xmath1 be an irreducible 3manifold that admits a cyclically reduced @xmath22presentation @xmath45 . then @xmath46 the proof of theorem [ mainthm ] relies on the virtually special theorem of agol @xcite , liu @xcite , przytycki - wise @xcite and wise @xcite , which we recall in section [ section : special ] . it also hinges on the following general result , which is of independent interest . theorempinresg [ thm : pinresg ] let @xmath1 be an irreducible 3manifold with empty or toroidal boundary . if @xmath1 is not a closed graph manifold , then @xmath47 is residually a torsion - free and elementary amenable group . the proof of theorem [ thm : pinresg ] uses the virtually special theorem and builds on work of linnell schick @xcite . it is proved in section [ section:3mfdgroups ] , where we also give several consequences . + we give a brief outline of the proof of theorem [ mainthm ] . the starting point is an alternative definition of the marked polytope @xmath48 using fox derivatives @xcite ( see section [ section : pipolytope ] ) . this definition is less pictorial , but it allows us to relate the polytope @xmath48 to the chain complex of the universal cover of the 2complex @xmath49 associated to the presentation @xmath4 . this makes it possible to study the ` size ' of @xmath48 using twisted reidemeister torsions corresponding to finite - dimensional complex representations and corresponding to skew fields of @xmath49 @xcite . since @xmath49 is simple homotopy equivalent to @xmath50 these twisted reidemeister torsions agree with the twisted reidemeister torsions of @xmath1 . in the following we denote @xmath51 and @xmath52 the polytopes @xmath17 and @xmath24 without the markings . given two polytopes @xmath39 and @xmath40 in a vector space @xmath12 we write @xmath53 if there exists @xmath43 with @xmath54 . the proof of theorem [ mainthm ] now breaks up into three parts : 1 . we first show that @xmath55 . put differently , we show that @xmath52 is ` big enough ' to contain @xmath56 . this is achieved with the main theorem of @xcite , which states that twisted reidemeister torsions corresponding to finite - dimensional complex representations detect the thurston norm of @xmath1 . this relies on the virtually special theorem . see section [ section : thurstonsmaller ] . next we show the reverse inclusion @xmath57 . this means that @xmath52 is ` not bigger than necessary ' . at this stage it is crucial that @xmath33 is cyclically reduced . by @xcite this implies that the summands in the fox derivative @xmath58 are pairwise distinct elements in the group ring @xmath59 $ ] . using theorem [ thm : pinresg ] and the non - commutative reidemeister torsions of @xcite we show that indeed @xmath60 . see section [ section : thurstonlarger ] . finally we need to show that the markings of @xmath17 and @xmath24 agree . this follows immediately from ( * ? ? ? * theorem 1.1 ) and ( * ? ? ? * theorem e ) . see section [ section : proofmainthm ] . the paper is concluded with a conjecture and a question in section [ section : questions ] . throughout this paper , all groups are finitely generated , all vector spaces finite dimensional , and all 3manifolds are compact , connected and orientable . the first author is very grateful for being hosted by the university of sydney in march 2013 and again in march 2014 . the first and the second author also wish to thank the mathematische forschungsinstitut oberwolfach for hosting them in january 2015 . the first author was supported by the sfb 1085 ` higher invariants ' , funded by the deutsche forschungsgemeinschaft ( dfg ) . the third author is partially supported under the australian research council s discovery funding scheme ( project number dp140100158 ) . the authors thank nathan dunfield for very insightful conversations . let @xmath12 be a real vector space and let @xmath61 be a finite ( possibly empty ) subset . denote @xmath62 the _ polytope spanned by @xmath63_. a _ polytope _ in @xmath12 is a subset of the form @xmath64 for some finite subset @xmath63 of @xmath12 . for any polytope @xmath39 there exists a unique smallest subset @xmath65 such that @xmath39 is the polytope spanned by @xmath66 . the elements of @xmath66 are the _ vertices _ of @xmath39 . note @xmath67 is a vertex if and only if there exists a homomorphism @xmath68 such that @xmath69 for every @xmath70 . let @xmath12 be a real vector space and let @xmath39 and @xmath40 be two polytopes in @xmath12 . the _ minkowski sum of @xmath39 and @xmath40 _ is @xmath71 it is straightforward to see that @xmath72 is again a polytope . furthermore , for each vertex @xmath73 of @xmath72 there exists a unique vertex @xmath14 of @xmath39 and a unique vertex @xmath74 of @xmath40 such that @xmath75 . conversely , for each vertex @xmath14 of @xmath39 there exists a ( not necessarily unique ) vertex @xmath74 of @xmath40 such that @xmath76 is a vertex of @xmath72 . if @xmath77 and @xmath78 are polytopes with @xmath79 then we write @xmath80 . we have @xmath81 in particular @xmath82 is well - defined . there is a natural scaling operation on polytopes : @xmath83 where @xmath84 is a polytope and @xmath85 if @xmath86 , then the minkowski sum of @xmath87 copies of @xmath39 equals @xmath88 . let @xmath89 be a non - empty convex set in the real vector space @xmath12 . given @xmath13 we define the _ thickness of @xmath89 in the @xmath10-direction _ by @xmath90 it is straightforward to see that the function @xmath91 is a seminorm . conversely , a seminorm @xmath92 defines the convex set @xmath93 note that @xmath94 is _ symmetric _ since @xmath95 implies @xmath96 . for any seminorm @xmath97 we have @xmath98 . on the other hand , if @xmath89 is a non - empty convex set , then @xmath99 equals the symmetrization @xmath100 of @xmath89 , @xmath101 finally , given a convex set @xmath89 in @xmath12 the _ dual _ of @xmath89 is @xmath102 let @xmath12 be a real vector space . a _ marked polytope _ @xmath103 in @xmath12 is a polytope @xmath39 and a ( possibly empty ) subset @xmath104 of @xmath105 the elements of @xmath104 are the _ marked vertices _ , the elements of @xmath106 are the _ unmarked vertices _ and @xmath39 is the _ underlying polytope _ of @xmath103 . if @xmath107 and @xmath108 are two marked polytopes , then the _ minkowski sum of @xmath103 and @xmath109 _ has underlying polytope the minkowski sum of the underlying polytopes and set of marked vertices precisely those that are sums of marked vertices : @xmath110 the marked polytope @xmath107 is _ symmetric _ if the underlying polytope @xmath39 is symmetric and @xmath111 let @xmath1 be a 3manifold . for each @xmath112 there is a properly embedded oriented surface @xmath113 , such that @xmath114\in h_2(n , \partial n ; \mathbb{z})$ ] is the poincar dual to @xmath115 letting @xmath116 , where @xmath117 are the connected components of @xmath118 the _ thurston norm _ of @xmath112 is @xmath119 = \phi \big\}.\ ] ] the class @xmath9 is called _ fibered _ if it can be represented by a non - degenerate closed 1-form . by @xcite an integral class @xmath120 is fibered if and only if there exists a fibration @xmath121 such that @xmath122 . thurston @xcite showed that @xmath7 extends to a seminorm @xmath7 on @xmath3 and that the dual @xmath123 to the unit norm ball @xmath124 of the seminorm @xmath7 is a polytope @xmath56 with vertices in @xmath125 . furthermore , thurston showed that we can turn @xmath56 into a marked polytope @xmath126 which has the property that @xmath127 is fibered if and only if it pairs maximally with a marked vertex . let @xmath128 be a group . throughout this paper , given @xmath129 $ ] and @xmath130 we denote @xmath131 the @xmath132coefficient of @xmath133 . let @xmath134 be the canonical map . we write @xmath135 and we view @xmath136 as a subset of @xmath12 . with this convention the above map @xmath137 gives rise to a map @xmath138 . given @xmath129 $ ] we refer to @xmath139 as the _ polytope of @xmath133_. we will now associate to @xmath140 a marking . in order to do this we need a few more definitions . for @xmath43 we refer to @xmath141 as the _ @xmath14component of @xmath133_. we say that an element @xmath142 $ ] is a monomial if it is of the form @xmath143 for some @xmath130 . a vertex @xmath14 of @xmath140 is _ marked _ precisely when the @xmath14-component of @xmath133 is a monomial . we then refer to the polytope @xmath140 together with the set of all marked vertices as the _ marked polytope @xmath144 of @xmath133_. the proof of ( * ? ? ? * lemma 3.2 ) applies with the above definitions , to give : [ lem : productadd ] let @xmath128 be a group and let @xmath145 $ ] . then the following hold : if for every vertex @xmath14 of @xmath140 the @xmath14-component @xmath146 $ ] is not a zero divisor , then @xmath147 . if each vertex of @xmath144 is marked , then @xmath148 . [ section : defpolytopepi ] let @xmath149 be the free group with generators @xmath29 and @xmath31 . following @xcite we denote by @xmath150\to \z[f]$ ] the _ fox derivative with respect to @xmath29 _ , i.e. the unique @xmath151linear map such that @xmath152 for all @xmath153 . we similarly define the fox derivative with respect to @xmath31 , and often write @xmath154 in @xcite we proved the following proposition . [ prop : foxpolytope ] let @xmath155 be a @xmath22presentation with @xmath25 . then there exists a marked polytope @xmath103 , unique up to translation , such that @xmath156 denote @xmath24 the marked polytope of proposition [ prop : foxpolytope ] . up to translation it is a well - defined invariant of the presentation , and it is shown in @xcite that this definition is equivalent to the one sketched in the introduction . a @xmath22presentation @xmath155 is _ simple _ if @xmath157 , @xmath29 defines a generator of @xmath158 and @xmath31 represents the trivial element in @xmath158 . in @xcite we showed that given a simple @xmath22presentation @xmath155 there exists a marked polytope @xmath24 , unique up to translation , such that @xmath159 it was shown in @xcite that there is a canonical way to associate to any @xmath22presentation @xmath155 with @xmath157 a simple presentation @xmath160 representing the same group . we then define @xmath161 . manifolds having fundamental group with a @xmath22presentation are described in section [ sec : examples ] . the only specific result needed to develop our theory is the following , which follows from work of epstein @xcite . [ thm : ep61 ] let @xmath1 be an irreducible ( compact , connected and orientable ) 3manifold such that @xmath163 admits a @xmath22presentation . then the boundary of @xmath1 consists of one or two tori . it follows from ( * ? ? ? * lemma 1.7 ) that @xmath4 has deficiency 1 , and from ( * ? ? ? * section 3 ) that the fundamental group of a closed irreducible 3manifold has deficiency zero . whence @xmath1 has non - empty boundary , and ( * ? ? ? * lemma 2.2 ) implies that @xmath164 the 3ball is the only irreducible 3manifold with a spherical boundary component due to the poincar conjecture . hence no boundary component of @xmath1 is a sphere since @xmath165 . since @xmath1 ( and hence each of its boundary components ) is orientable , we now have @xmath166 and every boundary component is a torus . a standard half - lives - half - dies argument shows that @xmath167 . since @xmath168 we deduce that @xmath169 either consists of one or two tori . following agol @xcite we say that a group is _ residually finite rationally solvable rfrs _ if there exists a sequence @xmath170 of subgroups of @xmath4 such that @xmath171 and such that for any @xmath172 the following hold : the subgroup @xmath173 is normal and finite index in @xmath4 , @xmath174 , and the natural surjection @xmath175 factors through @xmath176 . as usual , given a property of groups or spaces we say this property is satisfied _ virtually _ if a finite - index subgroup ( not necessarily normal ) or a finite - index cover ( not necessarily regular ) has the property . the following theorem is now a variation on the virtually special theorem . [ thm : virtuallyrfrs ] let @xmath1 be an irreducible 3manifold with empty or toroidal boundary . if @xmath1 is not a closed graph manifold , then @xmath47 is virtually rfrs . the theorem was proved by agol @xcite for all closed hyperbolic 3manifolds , by wise @xcite for all hyperbolic 3manifolds with boundary , by liu @xcite and przytycki wise @xcite for all graph manifolds with boundary and by przytycki wise @xcite for all 3manifolds with a non - trivial jsj - decomposition that has at least one hyperbolic jsj - component . we refer to @xcite for precise references . in the following , given a 3manifold @xmath1 we say that @xmath9 is _ quasi - fibered _ if it is a limit of fibered classes in @xmath3 . the following theorem is an immediate consequence of theorem [ thm : virtuallyrfrs ] and agol s virtual fibering theorem ( * ? ? ? * theorem 5.1 ) ( see also ( * ? ? ? * theorem 5.1 ) for an exposition ) . [ thm : virtuallyfib][thm : virtualfib ] let @xmath1 be an irreducible 3manifold with empty or toroidal boundary . if @xmath1 is not a closed graph manifold , then for every @xmath177 there exists a finite - index cover @xmath178 such that @xmath179 is quasi - fibered . if we apply the theorem to the zero class we get in particular the following corollary . [ cor : virtuallyfib ] an irreducible 3manifold with empty or toroidal boundary is virtually fibered unless it is a closed graph manifold . we start with several definitions , most of which are standard . let @xmath39 be a class of groups . the group @xmath4 is _ residually @xmath39 _ if for every non - trivial @xmath180 , there exists a homomorphism @xmath181 to a group in @xmath182 in @xmath39 such that @xmath183 . the group @xmath4 is _ fully residually @xmath39 _ if for every finite subset @xmath184 , there exists a epimorphism @xmath185 to a group in @xmath182 in @xmath39 such that @xmath186 for all @xmath187 . the group @xmath4 has the _ @xmath39factorization property _ if for every epimorphism @xmath185 onto a finite group @xmath128 there exists an epimorphism @xmath188 to a group @xmath182 in @xmath39 such that @xmath189 factors through @xmath190 . we are mostly interested in the following classes of groups . the class @xmath191 of _ elementary amenable groups _ is the smallest class of groups that contains all abelian and all finite groups and that is closed under extensions and directed unions . we denote by @xmath192 the class of all groups that are torsion - free and elementary amenable . it is clear that @xmath192 is closed under taking finite direct products . using corollary [ cor : virtuallyfib ] and work of linnell schick @xcite we will prove the following theorem : [ thm : pi1nfactors ] let @xmath1 be an irreducible 3manifold with empty or toroidal boundary . if @xmath1 is not a closed graph manifold , then @xmath47 has the @xmath192-factorization property . the question of to what degree this statement holds for closed graph manifolds is discussed in section [ section : closedgraphmanifolds ] . we postpone the proof of the theorem to section [ section : proofpi1 ] , and point out several consequences , including the following , which was stated in the introduction . let @xmath39 be any class of groups . if a group @xmath4 is residually finite and has the @xmath39factorization property , then @xmath128 is also residually @xmath39 . the statement of the theorem now follows from theorem [ thm : pi1nfactors ] and the fact that 3manifold groups are residually finite @xcite . [ cor : groupring ] let @xmath4 be the fundamental group of an irreducible 3manifold that has empty or toroidal boundary and is not a closed graph manifold . for every non - zero element @xmath193,$ ] there exists a homomorphism @xmath194 such that @xmath195 $ ] . we write @xmath196 , where @xmath197 and @xmath198 are pairwise distinct . by theorem [ thm : pinresg ] the group @xmath4 is residually @xmath192 . since @xmath192 is closed under taking finite direct products , @xmath4 is also fully residually @xmath192 . we can thus find a homomorphism @xmath181 to a group @xmath199 such that all @xmath200 and all products @xmath201 @xmath202 are non - trivial . whence @xmath203 $ ] is non - zero . the following lemma is probably well - known to the experts . [ lem : tf ] let @xmath204 be a surface group @xmath205i.e . the fundamental group of a compact orientable surface@xmath206 and let @xmath207 be a normal subgroup . then @xmath208 $ ] is torsion free . let @xmath209 $ ] be a non trivial element . we pick a representative for @xmath132 in @xmath204 which by slight abuse of notation we also denote by @xmath132 . we denote by @xmath210 the subgroup of @xmath204 generated by @xmath132 and @xmath211 . it suffices to prove the following claim . the group @xmath212 $ ] is torsion free . we consider the following short exact sequence @xmath213/[r , r]\to s/[r , r]\to s/[s , s]\to 0.\ ] ] since @xmath211 and @xmath210 are either surface groups or free we deduce that @xmath214=h_1(s;\z)$ ] and @xmath215=h_1(r;\z)$ ] are torsion free . the group @xmath216 is generated by one element , which implies that @xmath216 is cyclic , in particular abelian . it follows that @xmath217\subset r$ ] . we thus see that @xmath217/[r , r]$ ] is a subgroup of @xmath215 $ ] . so the groups on the left and on the right of the above short exact sequence are torsion - free . it follows that @xmath212 $ ] is torsion free . [ thm : ext ] if @xmath218 is an exact sequence with @xmath204 a surface group and @xmath219 , then @xmath4 has the @xmath192-factorization property . let @xmath220 be a map to a finite group . let @xmath221 . by lemma [ lem : tf ] the group @xmath222 $ ] is torsion - free . furthermore it is elementary amenable by the exact sequence @xmath223 \rightarrow e / [ r , r ] \rightarrow e / r \rightarrow 1.\ ] ] now @xmath224 factors through @xmath225,$ ] and this is in @xmath192 due to the sequence @xmath226 \to \pi/ [ r , r ] \to m \to 1.\hfill\qedhere\ ] ] the _ profinite completion _ of the group @xmath4 is denoted @xmath227 see ( * ? ? ? * section 3.2 ) for a definition and its main properties . following serre ( * ? ? ? * d.2.6 exercise 2 ) we say that a group @xmath4 is _ good _ if the natural morphism @xmath228 is an isomorphism for any finite abelian group @xmath229 with a @xmath4-action . [ thm : extendfactorizingproperty ] let @xmath4 be a finitely generated torsion - free group that has a finite - dimensional classifying space and which is good . if @xmath4 admits a finite index subgroup @xmath182 which has the @xmath192-factorization property , then @xmath4 also has the @xmath192-factorization property . in the proof of the theorem we will on several occasions use the following standard notation : if @xmath182 is a subgroup of @xmath4 , then @xmath230 . note that @xmath231 is always a normal subgroup of @xmath4 , and if @xmath182 is of finite - index , then @xmath231 is of finite - index . we also note that the methods of the proof build heavily on the work of linnell and schick @xcite . let @xmath185 be a homomorphism to a finite group . we denote by @xmath232 the intersection of @xmath233 and @xmath231 . the subgroup @xmath234 is of finite index in @xmath4 and is clearly contained in @xmath235 . it is straightforward to see that @xmath234 also has the @xmath192-factorization property . we write @xmath236 . first suppose that @xmath63 is a @xmath237-group . it suffices to show there is a subgroup @xmath238 such that the map @xmath239 factors through @xmath240 and @xmath240 is in @xmath192 . if no such @xmath241 exists , then since @xmath234 has the @xmath192-factorization property , there is a nontrivial subgroup @xmath242 of @xmath63 that splits in the induced sequence of profinite completions @xmath243 see ( * ? ? ? * lemmas 6.7 , 6.8 , 6.9 ) . however , putting the following two observations together shows that this is not possible . 1 . the cohomology @xmath244 is nonzero in infinitely many dimensions . 2 . by ( * ? ? ? * exercise 1 , 2.6 ) any finite - index subgroup @xmath245 ( such as @xmath234 or the preimage of @xmath242 under @xmath246 ) of @xmath4 is also good and has a finite - dimensional classifying space . this implies that @xmath247 is nonzero in only finitely many dimensions . for the general case , we use a trick from @xcite . for each sylow @xmath237-subgroup @xmath210 of @xmath63 , consider the exact sequence @xmath248 , where @xmath249 is the preimage of @xmath210 . by the above , we get a collection of subgroups @xmath250 such that the quotients @xmath251 are torsion - free elementary amenable . let @xmath252 . since @xmath253 is a finite extension of @xmath254 , elementary amenability follows again from ( * ? ? ? * lemma 4.11 ) . it remains to show that @xmath253 is torsion - free . suppose that @xmath253 has a non - trivial torsion element @xmath255 . by raising @xmath255 to some power we get an element @xmath256 that is @xmath237-torsion for some prime @xmath237 . we have an exact sequence @xmath257 where @xmath258 and @xmath259 are torsion - free by lemma 4.11 in @xcite . since @xmath260 is torsion - free , @xmath256 would map to some sylow @xmath237-subgroup , in which case @xmath261 , which is torsion - free . therefore , @xmath253 is torsion - free . now we are finally in a position to prove theorem [ thm : pi1nfactors ] . let @xmath1 be an irreducible 3manifold that has empty or toroidal boundary and that is not a closed graph manifold . according to corollary [ cor : virtuallyfib ] , @xmath1 has a finite cover @xmath262 that is fibered . the fundamental group of @xmath262 is a semidirect product of @xmath151 with a surface group , and hence lemma [ lem : tf ] and proposition [ thm : ext ] imply @xmath263 has the @xmath192-factorization property . it follows from ( * ? ? ? * exercise 2b ) pg . 16 ) that @xmath263 is good . by ( * exercise 1 pg . 16 ) the group @xmath47 is also good . it is well - known , see e.g. ( * ? ? ? * ( a1 ) ) , that @xmath1 is aspherical and that in particular @xmath47 is torsion - free . thus we can apply theorem [ thm : extendfactorizingproperty ] to @xmath47 and the finite - index subgroup @xmath263 , giving the desired result that @xmath47 has the @xmath192-factorization property . it is natural to ask for which closed graph manifolds the conclusions of theorem [ thm : pi1nfactors ] and its corollaries hold . it follows from the work of liu @xcite that the conclusion of the theorem also holds for closed non - positively curved graph manifolds . the question of which closed graph manifolds are non - positively curved was treated in detail by buyalo and svetlov @xcite . in the following we give a short list of examples of graph manifolds that are not non - positively curved : spherical 3-manifolds , sol- and nil - manifolds , seifert fibered 3-manifolds that are finitely covered by a non - trivial @xmath264-bundles over a closed surface it is clear that the statements do not hold for spherical 3-manifolds with non - trivial fundamental group . the following lemma takes care of the second case . the fundamental groups of sol- and nil - manifolds are @xmath192 , in particular they have the @xmath192-factorization property . sol- and nil - manifolds are finitely covered by torus - bundles over @xmath264 . hence their fundamental groups are elementary amenable , but the fundamental groups are also torsion - free , so they are @xmath192 . let @xmath1 be a seifert fibered space with infinite fundamental group . then @xmath47 has the @xmath192-factorization property . since we will not make use of this lemma we only sketch the proof . the manifold @xmath1 is finitely covered by an @xmath264-bundle over a surface . by theorem [ thm : extendfactorizingproperty ] we can thus without loss of generality assume that @xmath1 is an @xmath264-bundle over a surface @xmath149 . since @xmath265 is infinite there exists a short exact sequence @xmath266 where the subgroup @xmath267 is generated by the @xmath264-fiber . by proposition [ thm : ext ] the group @xmath268 has the @xmath192-factorization property . denote by @xmath269 the euler number of the @xmath264-bundle over @xmath149 and denote by @xmath262 the total space of the @xmath264-bundle over the torus with euler number @xmath269 . then there exists a fiber - preserving map from @xmath270 to @xmath262 . since @xmath263 is @xmath192 we found a homomorphism from @xmath47 to a @xmath192 group which is injective on @xmath267 . now it is straightforward to see that @xmath47 has the @xmath192-factorization property . the above discussion shows that the fundamental groups of many closed graph manifolds have the @xmath192-factorization property . nonetheless we expect that there are many closed graph manifolds whose fundamental groups do not have the @xmath192-factorization property . the goal of this section is to prove the following proposition . [ mainprop1 ] let @xmath45 be a cyclically reduced @xmath22presentation for the fundamental group of an irreducible 3manifold @xmath1 . then @xmath271 the main ingredient in the proof will be the fact that twisted reidemeister torsions corresponding to finite - dimensional complex representations detect the thurston norm of 3manifolds . let @xmath4 be a group , let @xmath272 be a representation and let @xmath273 be a homomorphism to a free abelian group . we denote by @xmath274 the quotient field of the group ring @xmath275 $ ] . the homomorphisms @xmath189 and @xmath137 give rise to the representation @xmath276 that we refer to as the tensor product of @xmath189 and @xmath137 . this representation extends to a ring homomorphism @xmath59\to m(k\times k,\c(h)),$ ] which we also denote by @xmath277 . let @xmath49 be a finite @xmath278complex , @xmath279 , and denote by @xmath280 the universal cover of @xmath49 . the action of @xmath4 via deck transformations on @xmath280 equips the chain complex @xmath281 with the structure of a chain complex of @xmath59$]left modules . let @xmath272 be a representation . we denote by @xmath282 the obvious projection map . using the representation @xmath277 we can now view @xmath283 as a right @xmath59$]-module , where the action is given by right multiplication on row - vectors . we consider the chain complex @xmath284 } c_*(\ti{x};\z)\ ] ] of @xmath274modules . for each cell in @xmath49 pick a lift to a cell in @xmath285 . we denote by @xmath286 the standard basis for @xmath283 . the tensor products of the lifts of the cells and the vectors @xmath287 turns @xmath288 into a chain complex of based @xmath274vector spaces . if the chain complex @xmath288 is not acyclic , then we define the corresponding twisted reidemeister torsion @xmath289 to be zero . otherwise we denote by @xmath290 the torsion of the based chain complex @xmath288 . we refer to @xcite for the definition of the torsion of a based chain complex . standard arguments show that @xmath290 is well - defined up to multiplication by an element of the form @xmath291 where @xmath292 and @xmath293 . the indeterminacy arises from the fact that we had to choose lifts and an ordering of the cells . suppose @xmath1 is a 3manifold and let @xmath294 be a representation . choose a @xmath278structure @xmath49 for @xmath1 and define @xmath295 . it is well - known , see e.g. @xcite , that this definition does not depend on the choice of the @xmath278structure . as above , suppose @xmath1 is a 3manifold and @xmath294 a representation . if @xmath296 is zero , then we define @xmath297 . otherwise we write @xmath298 with @xmath299 $ ] . if the minkowski difference @xmath300 exists ( and by @xcite this is the case if @xmath301 ) , then we define @xmath302 and otherwise define @xmath303 . [ prop : tapcontainedinmpi ] let @xmath304 be a @xmath22presentation for the fundamental group of an irreducible 3manifold @xmath1 . then for any representation we have @xmath305 in the proof of the proposition we will need one more definition and one more lemma . let @xmath4 be a group , @xmath306,$ ] @xmath307 be a representation , and @xmath308 be the canonical epimorphism . then @xmath309 $ ] and we write @xmath310 [ lem : detcontained ] let @xmath4 be a group , @xmath306 $ ] and @xmath307 be a representation . then @xmath311 we write @xmath312 with @xmath313 and @xmath314 . we consider @xmath315 put differently , @xmath210 is the set of all elements in @xmath275 $ ] with support a subset of @xmath316 . for every @xmath317 we have @xmath318 . this implies that if @xmath319 are elements in @xmath210 , then @xmath320 we write @xmath321 . each entry of @xmath322 lies in @xmath210 . it follows from the laplace formula that @xmath322 is a sum of products of the form @xmath323 where each @xmath324 lies in @xmath210 . by the above we have @xmath325 . the definitions imply that if @xmath326 $ ] are such that @xmath327 and @xmath328 are contained in a polytope @xmath40 , then we have also have @xmath329 . hence @xmath330 . we again denote by @xmath331 the canonical epimorphism . note that @xmath332 or @xmath333 . without loss of generality we may assume @xmath333 . theorem [ thm : ep61 ] shows that @xmath1 has non - trivial toroidal boundary . it thus follows from ( * ? ? ? * theorem a ) , see also @xcite , that @xmath334 by lemma [ lem : detcontained ] we have @xmath335 . since @xmath333 we know that @xmath336 and @xmath337 are the two , distinct , vertices of @xmath338 . also , we have @xmath339 and it is straightforward to see that this polytope equals @xmath338 . combining these results we obtain @xmath340 proposition [ mainprop1 ] is an immediate consequence of theorem [ thm : ep61 ] , proposition [ prop : tapcontainedinmpi ] and the second statement of the following proposition . [ prop : tapdetecttn ] let @xmath1 be a 3manifold with empty or toroidal boundary and @xmath341 be a unitary representation . then @xmath342 furthermore , if @xmath1 is irreducible , then there exists a unitary representation @xmath341 such that @xmath343 let @xmath1 be a 3manifold with empty or toroidal boundary . we write @xmath344 . let @xmath345 be a unitary representation . if @xmath346 , then there is nothing to show . so suppose that @xmath347 . in ( * ? * theorem 1.1 ) and ( * ? ? ? * theorem 3.1 ) it was shown that for any @xmath348 we have @xmath349 it follows from the definitions and the discussion in section [ section : convex ] that @xmath350 . since @xmath189 is a unitary representation it follows from ( * theorem 1.2 ) that @xmath351 . it thus follows that indeed @xmath352 . if @xmath1 is not a closed graph manifold , then , building on theorem [ thm : virtualfib ] , it was shown in ( * ? ? ? * corollary 5.10 ) that there exists a unitary representation @xmath345 such that @xmath353 for every @xmath9 . the same argument as above then implies that @xmath354 . if @xmath1 is a closed graph manifold , then the same statement holds by @xcite . the goal of this section is to prove the following proposition , and to complete the proof of the main theorem . [ mainprop2 ] let @xmath45 be a cyclically reduced @xmath22presentation for the fundamental group of an irreducible 3manifold @xmath1 . then @xmath355 in the proof of proposition [ mainprop1 ] we used twisted reidemeister torsions corresponding to finite - dimensional complex representations . in the proof of proposition [ mainprop2 ] we use a different , but related , object , namely non - commutative reidemeister torsions . in this context they were first studied in @xcite . let @xmath356 . it follows from ( * ? ? ? * theorem 1.4 ) that the group ring @xmath357 $ ] is a domain . since @xmath182 is amenable it follows from ( * ? ? ? * corollary 6.3 ) that @xmath357 $ ] satisfies the ore condition . this means that for any two non - zero elements @xmath358 $ ] there exist non - zero elements @xmath359 $ ] such that @xmath360 . by * section 4.4 ) this implies that @xmath357 $ ] has a classical fraction field , referred to as the ore localization of @xmath357 $ ] , which we denote by @xmath361 . let @xmath362 be a homomorphism . for every non - zero @xmath363 $ ] we define @xmath364 we extend this to all of @xmath357 $ ] by letting @xmath365 . since @xmath357 $ ] has no nontrivial zero divisors it follows that for @xmath359 $ ] we have @xmath366 . given @xmath367 we also define @xmath368 it is straightforward to see that this is indeed well - defined . let @xmath49 be a finite @xmath278complex with @xmath369 , and denote @xmath285 the universal cover of @xmath49 . as in section [ section : twitorsion ] we view @xmath370 as a chain complex of left @xmath371$]-modules . now let @xmath372 be a homomorphism , and consider the chain complex of left @xmath373-modules @xmath374}c_*(\wti{x}),\ ] ] where @xmath128 acts on @xmath361 on the right via the homomorphism @xmath375 . if @xmath376 is not acyclic , define the corresponding reidemeister torsion @xmath377 to be zero . otherwise choose an ordering of the cells of @xmath49 and for each cell in @xmath49 pick a lift to @xmath285 . this turns @xmath376 into a chain complex of based @xmath361 left - modules and we define @xmath378 to be the reidemeister torsion of the based chain complex @xmath376 . here @xmath379 is the abelianization of the direct limit @xmath380 of the general linear groups over @xmath361 ( see @xcite for details ) . we write @xmath381 and denote by @xmath382 the abelianization of the multiplicative group @xmath383 . the dieudonn determinant , see @xcite , gives rise to an isomorphism @xmath384 which we will use to identify these two groups . the invariant @xmath385 is well - defined up to multiplication by an element of the form @xmath386 where @xmath387 . furthermore , it does not depend on the homeomorphism type of @xmath49 . we refer to @xcite for details . it follows from @xmath388 for @xmath389 that @xmath390 descends to a homomorphism @xmath391 . in particular @xmath392 is defined . let @xmath1 be an irreducible 3manifold and suppose @xmath45 is a cyclically reduced @xmath22presentation of its fundamental group . without loss of generality we may assume that @xmath29 represents a non - zero element in @xmath393 . we need to show that @xmath57 . we call @xmath394 _ generic _ if there are vertices @xmath14 and @xmath74 of @xmath395 such that @xmath10 pairs maximally with @xmath14 and @xmath10 pairs minimally with @xmath74 . for any generic epimorphism @xmath396 we have @xmath397 we denote by @xmath14 and @xmath74 the ( necessarily unique ) vertices of @xmath395 such that @xmath10 pairs maximally with @xmath14 and minimally with @xmath74 . by corollary [ cor : groupring ] and theorem [ thm : ep61 ] there exists a homomorphism @xmath398 such that @xmath399 . in particular , @xmath400 and @xmath401 . denote @xmath273 the canonical epimorphism . after possibly replacing @xmath189 by @xmath402 we can and will assume that @xmath137 factors through @xmath189 . in particular @xmath10 factors through @xmath189 and @xmath403 is a non - trivial element in @xmath182 . we denote by @xmath49 the @xmath278complex corresponding to the presentation @xmath4 with one 0cell , two 1cells corresponding to the generators @xmath29 and @xmath31 and one 2cell corresponding to the relator @xmath33 . as in @xcite we have @xmath404 . we then have @xmath405 here the first two equalities follows from the definitions and the choice of @xmath14 and @xmath74 . the fifth equality is ( * ? ? ? * theorem 2.1 ) and the last inequality is given by ( * ? ? ? * theorem 1.2 ) ( see also @xcite ) . this concludes the proof of the claim . + it is straightforward to see that the non - generic elements in @xmath406 correspond to a union of proper subspaces of @xmath406 . by continuity and linearity of seminorms we see that the inequality @xmath407 holds in fact for all @xmath394 . it follows from the definitions and the discussion in section [ section : convex ] that @xmath408 . [ [ section : proofmainthm ] ] for the reader s convenience we recall the statement of theorem [ mainthm ] . it follows from propositions [ mainprop1 ] and [ mainprop2 ] that @xmath55 and @xmath409 . by the symmetry of the thurston norm we also have @xmath410 and this implies @xmath411 . the fact that the markings agree is an immediate consequence of ( * ? ? ? * theorem 1.1 ) and ( * ? ? ? * theorem e ) . currently there is no geometric characterization of those 3manifolds whose fundamental group may be presented using only two generators and one relator . waldhausen s question @xcite of whether the rank of the fundamental group equals the heegaard genus gives the conjectural picture that all of these manifolds have tunnel - number one . li @xcite gives examples of 3manifolds whose rank is strictly smaller than the genus , including closed manifolds , manifolds with boundary , hyperbolic manifolds , and manifolds with non - trivial jsj decomposition . see also related work of boileau , weidmann and zieschang @xcite . however , waldhausen s question remains open for hyperbolic 3manifolds of rank 2 and for knot complements in @xmath38 . a _ tunnel - number one _ 3manifold is a 3manifold obtained by attaching a 2handle to a 3dimensional 1handlebody of genus two . the fundamental group has a presentation with two generators from the handlebody and one relator corresponding to the attaching circle of the 2handle . theorem [ mainthm ] allows us to compute the unit ball of the thurston norm with ease , whilst other methods , such as normal surface theory @xcite have limited scope ( cf . moreover , with theorem [ mainthm ] one can easily construct examples with prescribed combinatorics or geometry of the unit ball . brown s algorithm is an essential ingredient in dunfield and d. thurston s proof @xcite that the probability of a tunnel - number one manifold fibring over the circle is zero . this can be paraphrased as : the probability that the unit ball has a non - empty set of marked vertices is zero . interesting applications of theorem [ mainthm ] combined with the methods of @xcite would be further predictions about the unit ball of a random tunnel - number one manifold . norwood @xcite showed that if the complement of a given knot in @xmath38 has fundamental group generated by two elements , then the knot is prime . the complements of _ tunnel - number one _ knots or links in @xmath38 are tunnel - number one manifolds . this includes the 2bridge knots and links , but johnson @xcite showed that there are hyperbolic tunnel - number one knots with arbitrarily high bridge number . there is a complete classification of all tunnel - number one satellite knots by morimoto and sakuma @xcite , and morimoto @xcite also showed that a composite link has tunnel - number one if and only if it is a connected sum of a 2bridge knot and the hopf link . 3.8 cm [ fig : dunfield ] 11 cm we conclude this section with an explicit calculation for the link @xmath245 shown in figure [ dunfield ] , which was studied by dunfield @xcite . write @xmath413 and denote @xmath414 the link group . then @xmath4 has the presentation @xmath415 where a meridian for the unknotted component is given by @xmath416 and a meridian for the other component is given by @xmath417 . theorem [ mainthm ] implies that @xmath418 . we use the map induced by @xmath419 and @xmath420 to identify @xmath421 with @xmath28 . a straightforward calculation shows that @xmath422 is the polytope with vertices @xmath423 and @xmath424 shown in figure [ fig : pp(r_x ) ] . here @xmath425 and @xmath426 are opposite vertices of @xmath422 and @xmath427 and @xmath428 are opposite vertices of @xmath422 . subtracting the underlying polytope of @xmath429 from @xmath422 gives @xmath430 and this agrees ( up to translation ) with figure [ fig : dunfield - intro ] . the following computation shows that the markings are the same : @xmath431 [ [ section ] ] [ conj : sym ] let @xmath432 be a @xmath22presentation for the fundamental group of a 3manifold . let @xmath14 be a vertex of @xmath422 . then there exists a unique vertex @xmath74 of @xmath395 such that @xmath14 and @xmath74 are opposite vertices . furthermore we have @xmath433 the twisted reidemeister torsions of @xcite can be computed in terms of fox derivatives , and the symmetry results for twisted reidemeister torsions proved in @xcite give strong evidence towards this conjecture . to give an explicit example , let us return to dunfield s link . given the group @xmath128 and @xmath434,$ ] write @xmath435 if there exist @xmath436 such that @xmath437 . furthermore , denote @xmath438 the involution of @xmath371 $ ] defined by the inversion map @xmath439 for each @xmath130 . we denote by @xmath432 the presentation from section [ section : dunfield ] . we then note that @xmath440 the relator @xmath33 is conjugate to @xmath441 in particular writing @xmath442 we have the following equality in @xmath59 $ ] : @xmath443 & = & -\ol{(r_x)^{w_2}}.\ea\ ] ] we initially attempted to prove theorem [ mainthm ] just using twisted reidemeister torsions corresponding to finite - dimensional representations , noting that theorem [ mainthm ] follows from proposition [ prop : tapdetecttn ] ( 1 ) together with an affirmative answer to the following question , which is interesting in its own right . i. agol , _ criteria for virtual fibering _ , j. topol . 1 ( 2008 ) 2 , 269284 . i. agol , _ the virtual haken conjecture _ , with an appendix by i. agol , d. groves and j. manning , documenta math . 18 ( 2013 ) , 10451087 . m. aschenbrenner , s. friedl and h. wilton , _ 3manifold groups _ , to be published by the ems series of lectures in mathematics ( 2015 ) r. bieri , w. d. neumann and r. strebel , _ a geometric invariant of discrete groups _ , invent . ( 1987 ) , 451 - 477 . m. boileau and r. weidmann , _ the structure of 3manifolds with two - generated fundamental group _ , topology 44 ( 2005 ) 283320 . m. boileau and h. zieschang , _ heegaard genus of closed orientable seifert 3manifolds _ , invent . 76 ( 1984 ) 455468 . s. buyalo and p. svetlov , _ topological and geometric properties of graph - manifolds _ , st . petersburg math . j. * 16 * ( 2005 ) , no . 2 , 297340 . t. cochran , _ noncommutative knot theory _ , topol . 4 ( 2004 ) , 347398 . t. cochran , k. orr and p. teichner , _ knot concordance , whitney towers and @xmath446-signatures _ , ann . of math . ( 2 ) 157 ( 2003 ) , no . 2 , 433519 . d. cooper and s. tillmann , _ the thurston norm via normal surfaces _ j. math . 239 ( 2009 ) 115 . w. dicks and p. a. linnell , _ @xmath447-betti numbers of one - relator groups _ , math . ( 2007 ) , 855874 . j. dodziuk , p. linnell , v. mathai , t. schick and s. yates , _ approximating @xmath447-invariants , and the atiyah conjecture _ , preprint series sfb 478 mnster , germany . communications on pure and applied mathematics 56 ( 2003 ) , 839873 . n. dunfield , _ alexander and thurston norms of fibered 3manifolds _ , pacific j. math . 200 ( 2001 ) , no . 1 , 4358 . n. dunfield and d. thurston , _ does a random tunnel - number one 3-manifold fiber over the circle ? _ , slides available at ` http://www.math.uiuc.edu/ nmd / preprints / slides / treiste.pdf ` n. dunfield and d. ramakrishnan , _ increasing the number of fibered faces of arithmetic hyperbolic 3-manifolds , _ amer . j. math . 132 ( 2010 ) , no . 1 , 5397 . d. b. a. epstein , _ finite presentations of groups and 3manifolds _ , quart . j. math . oxford ser . ( 2 ) 12 ( 1961 ) , 205212 . r. h. fox , _ free differential calculus i , derivation in the free group ring _ , ann . math . 57 ( 1953 ) , 547560 . s. friedl , _ reidemeister torsion , the thurston norm and harvey s invariants _ , pac . j. math . 230 ( 2007 ) , 271296 . s. friedl and s. harvey , _ non commutative multivariable reidemeister torsion and the thurston norm _ , alg . ( 2007 ) , 755777 . s. friedl and t. kim , _ thurston norm , fibered manifolds and twisted alexander polynomials _ , topology 45 ( 2006 ) , 929953 . s. friedl and t. kim , _ twisted alexander norms give lower bounds on the thurston norm _ , 360 ( 2008 ) , 45974618 . s. friedl , t. kim and t. kitayama , _ poincar duality and degrees of twisted alexander polynomials _ , indiana univ . j. 61 ( 2012 ) , 147192 . s. friedl and t. kitayama , _ the virtual fibering theorem for 3manifolds _ , enseign . math . 60 ( 2014 ) , no . 1 , 79107 . s. friedl and s. tillmann , _ two - generator one - relator groups and marked polytopes _ , preprint ( 2015 ) s. friedl and s. vidussi , _ a survey of twisted alexander polynomials _ , the mathematics of knots : theory and application ( contributions in mathematical and computational sciences ) , editors : markus banagl and denis vogel ( 2010 ) , 4594 . s. friedl and s. vidussi , _ the thurston norm and twisted alexander polynomials _ , preprint ( 2012 ) , to be published by j. reine ang . math . s. harvey , _ higher order polynomial invariants of 3manifolds giving lower bounds for the thurston norm _ , topology 44 ( 2005 ) , 895945 . j. hempel , _ residual finiteness for 3manifolds _ , combinatorial group theory and topology ( alta , utah , 1984 ) , pp . 379396 , ann . of math . , 111 , princeton univ . press , princeton , nj , 1987 . j. hillmann , _ on @xmath447homology and asphericity _ , israel j. math . 99 ( 1997 ) , 271283 . j. hillman , d. silver and s. williams , _ on reciprocality of twisted alexander invariants _ , alg . ( 2010 ) , 20172026 . j. johnson , _ bridge number and the curve complex _ , arxiv : math/0603102v2 . t. kitano , _ twisted alexander polynomials and reidemeister torsion _ , pacific j. math . 174 ( 1996 ) , no . 2 , 431442 . p. h. kropholler , p. a. linnell and j. a. moody , _ applications of a new @xmath234-theoretic theorem to soluble group rings _ , proc . 104 ( 1988 ) , no . 3 , 675684 . t. li , _ rank and genus of 3manifolds _ , j. amer . ( 2013 ) , 777829 . p. linnell and p. schick , _ finite group extensions and the atiyah conjecture _ , j. amer ( 2007 ) , no . 4 , 10031051 . y. liu , _ virtual cubulation of nonpositively curved graph manifolds _ , j. topol . 6 ( 2013 ) , no . 4 , 793822 . r. lyndon and p. schupp , _ combinatorial group theory _ , ergebnisse der mathematik und ihrer grenzgebiete , band 89 . springer - verlag , berlin - new york , 1977 . j. milnor , _ whitehead torsion _ , bull . 72 ( 1966 ) , 358426 . k. morimoto , _ on composite tunnel - number one links , _ topology appl . 59 ( 1994 ) 5971 . k. morimoto and m. sakuma , _ on unknotting tunnels for knots , _ math . ( 1991 ) 143167 . s. friedl and m. nagel , _ twisted reidemeister torsion and the thurston norm : graph manifolds and finite representations _ , preprint ( 2015 ) f. h. norwood , _ every two - generator knot is prime , _ proc . ( 1982 ) 143147 . d. s. passman , _ the algebraic structure of group rings _ , john wiley & sons . xiv ( 1977 ) p. przytycki and d. wise , _ graph manifolds with boundary are virtually special _ , j. topology 7 ( 2014 ) , 419435 . p. przytycki and d. wise , _ mixed 3manifolds are virtually special _ , preprint ( 2012 ) . l. ribes and p. zalesskii , _ profinite groups _ , , ergebnisse der mathematik und ihrer grenzgebiete , 3 . folge , vol . 40 , springer - verlag , berlin , 2010 . j. rosenberg , _ algebraic k -theory and its applications _ , graduate texts in mathematics 147 , springer , new york , 1994 . k. schreve , _ the @xmath447-cohomology of discrete groups _ , ( 2015 ) , ph.d thesis . serre , _ galois cohomology _ , springer - verlag , berlin , 1997 . w. p. thurston , _ a norm for the homology of 3manifolds _ , mem . 59 , no . 339 ( 1986 ) , d. tischler , _ on fibering certain foliated manifolds over @xmath264 _ , topology 9 ( 1970 ) , 153154 . j. l. tollefson and n. wang , _ taut normal surfaces . _ topology 35 ( 1996 ) , 5575 . v. turaev , _ introduction to combinatorial torsions _ , lectures in mathematics , eth zrich ( 2001 ) v. turaev , _ a homological estimate for the thurston norm _ , unpublished note ( 2002 ) , arxiv : math . gt/0207267 m. wada , _ twisted alexander polynomial for finitely presentable groups _ , topology 33 , no . 2 : 241256 ( 1994 ) f. waldhausen , _ some problems on 3-manifolds , _ proc . sympos . pure math . 32 ( 1978 ) , 313322 . c. weinbaum , _ on relators and diagrams for groups with one defining relation _ , illinois j. math . 16 ( 1972 ) , 308322 . d. wise , _ the structure of groups with a quasi - convex hierarchy _ , electronic res . ann . math . ( 2009 ) , 4455 . d. wise , _ the structure of groups with a quasi - convex hierarchy _ , 189 pages , preprint ( 2012 ) , downloaded on october 29 , 2012 from + ` http://www.math.mcgill.ca/wise/papers.html ` d. wise , _ from riches to raags : 3manifolds , right angled artin groups , and cubical geometry _ , cbms regional conference series in mathematics , 2012 .

we study the statistics of the number of connected components and the volume of a random real algebraic hypersurface in defined by a homogeneous polynomial of degree in the _ real fubini - study _ ensemble . we prove that for the expectation of the number of connected components : the asymptotic being in for fixed . + we do not restrict ourselves to the random homogeneous case and we consider more generally random polynomials belonging to a window of eigenspaces of the laplacian on the sphere proving that the same asymptotic holds . as for the volume properties we prove that : both equations ( [ bettiabstract ] ) and ( [ volumeabstract ] ) exhibit expectation of maximal order in light of milnor s bound and the bound

1212.3823

hilbert s sixteenth problem , in its general form , asks for the study of the maximal number and the possible arrangements of the components of a nonsingular real algebraic hypersurface of degree @xmath2 in @xmath0 . it turns out that even the case of curves in @xmath9 is an extremely subtle problem and essentially nothing is known for @xmath10 . the analogous question for a hypersurface in @xmath11 is trivial , as for a _ generic _ choice of the polynomial defining it , its topology is determined ( this is the statement that the cohomology of a complete intersection in the complex projective space is determined by its multidegree and complete intersections are generic ) . thus , if from one hand no technique exists to study the arrangement of the components of a real algebraic hypersurface , on the other hand one still would like to formulate typical " statements : a way for doing that is to replace the word _ generic _ with _ let us start by considering the case of hypersurfaces in @xmath12 , i.e. zeros of ordinary univariate polynomials . in fact here we consider the _ homogenization _ of a univariate polynomial together with its _ projective _ roots rather than the classical polynomial itself , but it is clear that the two objects produce the same statistic . over the complex numbers statements like : @xmath13 make perfect sense . on the other hand the only property we can say to be generic for the real roots of a real polynomial is the one of being distinct . proposing a random viewpoint raises the obvious question : _ what is meant by random ? _ we first look to the seminal work @xcite of kac , who proved that a random univariate polynomial of degree @xmath2 whose coefficients are i.i.n . has asymptotically @xmath14 real roots . even if it might seem natural , the distribution kac considered is not invariant by the action of projective transformations over the extended real line . to produce such an invariant distribution we proceed as follows . we first _ extend _ our space of polynomials by homogenization ; this leads us to consider _ projective _ zeros rather than just on the real line . thus , we consider the vector space @xmath15 of real homogeneous polynomials of degree @xmath2 in two variables . we endow this vector space with a euclidean structure by considering the @xmath16-norm of the restriction of these polynomials to the circle @xmath17 . in an equivalent way we can define for @xmath1 and @xmath18 in @xmath15 their scalar product to be : @xmath19 where @xmath20 is the variable in @xmath21 . notice that since @xmath1 and @xmath18 are homogeneous of the same degree , the previous integral equals the integral of @xmath22 restricted to the circle @xmath17 ( up to a constant factor ) . using this scalar product we can define a _ centered gaussian _ probability distribution on @xmath15 by letting for a measurable subset @xmath23 : @xmath24 where @xmath25 is the constant that makes the previous a probability distribution and @xmath1 is treated as the integration variable that runs over @xmath26 for us a _ random polynomial _ will simply be a polynomial in @xmath15 with the above distribution . having thus replaced _ generic _ with _ random _ , we can make a positive statement in the real setting : @xmath27 we provide now another equivalent definition of a random polynomial . we fix a basis @xmath28 for @xmath15 which is orthonormal for the above scalar product and we consider a random function of the form : @xmath29 what we get in this way is again a random polynomial in the above sense . one way to produce such a basis is by considering _ trigonometric polynomials _ , i.e. functions on the circle satisfying : @xmath30 we denote by @xmath31 the vector space of all the solutions of the previous equation : these must be linear combinations of @xmath32 and @xmath33 . equivalently we can extend such functions by homogeneity to all @xmath21 and we get harmonic polynomials of degree @xmath34 . what is interesting for us is that in this fashion we can write the vector space @xmath15 as : @xmath35 where again @xmath20 is the variable in @xmath21 and the direct sum turns out to be orthogonal with respect to the above scalar product . in particular the above orthonormal basis can be obtained by collecting together the orthonormal bases coming from each @xmath31 , @xmath36 . if we repeat the same construction for the scalar product @xmath37 then the corresponding real random polynomials are called _ kostlan _ distributed ( or _ bombieri - weyl _ ) . the vector space is the same , but in this second case the euclidean structure comes by restricting to the real polynomials the hermitian structure given on the complex ones . the expectation of the number of real roots of a real kostlan polynomial of degree @xmath2 is @xmath38 . already from this comparison we see that the structure of the zeros of a random polynomial is richer than the one of a kostlan , which in turn is richer than for the naive ensemble . the two numbers @xmath39 and @xmath2 are called the _ parameters _ of these distributions . more generally each gaussian distribution on @xmath15 that is invariant by projective transformation has its parameter @xmath40 which is defined to be the square of the expected number of zeros it produces . our interest is in what happens when we increase the number of variables . we let @xmath41 be the vector space of homogeneous polynomials of degree @xmath2 in @xmath42 variables , and consider their zero sets as hypersurfaces in @xmath0 . if in the univariate case as a measure " of our set we could use the number of its points , here we have at least two possible approaches to generalize this idea . the first one is by considering the number of points as a zero - dimensional volume . in this case an upper bound for the size " of our set is provided by means that there exist two constants @xmath43 such that for @xmath2 large enough @xmath44 : @xmath45 the previous formula comes from an application of the integral geometry formula and wirtinger s inequality ( see @xcite ) . in fact if @xmath1 is regular the previous volume is bounded by @xmath46 and since @xmath47 this also gives a nice way to compare the real zero locus to the complex one ( as we did in the univariate case using the fundamental theorem of algebra ) . the second approach is more subtle and regards the number of points as a topological property ( rather than a metric one ) , namely as the number of connected components . it is interesting to notice that there is a quantitative bound also for this number , usually referred to as milnor s bound ( see @xcite ) : @xmath48 the original paper @xcite of milnor does not contain the bound in this form , since a @xmath49 term appears . we will give a proof of ( [ milnor ] ) in the appendix for the smooth case ( which is the case we are interested in , since the set of nonregular polynomials has _ measure zero _ in @xmath41 ) , though the same bound holds true even for singular hypersurfaces . as we already pointed out , for a generic @xmath1 in @xmath41 both the volume and the betti numbers are constant for its complex zero locus . this is certainly not the case for the real zero locus , as for example it might even be empty . to study the expectation of these quantities over the reals we make @xmath41 into a gaussian space . we first define the scalar product of two elements @xmath1 and @xmath18 in it to be : @xmath50 using this scalar product @xmath41 is turned into a probability space by defining for each of its measurable subsets @xmath23 : @xmath51 where now @xmath52 following @xcite , we call @xmath41 with the resulting distribution of probability the _ real fubini - study ensemble_. the expectation of the volume of the zero locus of a random polynomial of degree @xmath2 from the real fubini - study ensemble was computed by p. brgisser in @xcite : @xmath53 in fact in the paper @xcite this result follows from the more striking computation of the expected curvature polynomial of @xmath54 for any projectively invariant distribution . such a result is formulated in terms of the above mentioned parameter @xmath40 , which in this case is the square of the average number of real zeros of a restriction to an @xmath12 . in particular , if we define the _ kostlan _ distribution on @xmath41 by means of the scalar product @xmath55 , the corresponding parameter is @xmath2 and the result we get for the expectation of the volume is @xmath56 ( the general result for a distribution with parameter @xmath40 is @xmath57 ) . in the case @xmath58 and restricting the distribution of probability to @xmath59 , i.e. considering only _ harmonic _ polynomials rather than homogeneous ones , the statistics of the number of connected components was studied by nazarov and sodin . in the breakthrough paper @xcite the authors prove that there exists a constant @xmath60 such that for a random spherical harmonic @xmath1 of degree @xmath2 : @xmath61 moreover , they showed that @xmath62 concentrates exponentially about its mean . in the visionary handwritten letter @xcite p. sarnak claims that using the same ideas as in @xcite it is possible to extend this result to the case of real homogeneous polynomials in the real fubini - study ensemble . the interesting part of these statements ( besides the concentration ) is the lower bound estimate : an upper bound of the same order is given by milnor s one ( or in the spherical harmonic case by courant s nodal domain theorem ) . the statistics of the number of connected components of a kostlan distributed curve was studied by gayet and welschinger : in the paper @xcite the authors prove that such a real curve of degree @xmath2 has on average @xmath63 components . the same authors prove in @xcite that the expectation of _ each _ betti number of a kostlan hypersurface in @xmath0 is bounded by @xmath64 , generalizing in some sense the univariate case . on a different perspective the first author in @xcite , using techniques from random matrix theory , has studied intersection of one or two kostlan distributed quadrics , the asymptotic being this time in the number @xmath4 of variables ; for such random algebraic sets the expectation of the sum of the betti numbers behaves asymptotically as @xmath4 . as we did before , it is possible to give an equivalent definition of the real fubini - study structure on @xmath41 using an @xmath65-orthonormal basis . in fact we can again decompose our vector space into the _ orthogonal _ sum : @xmath66 where now @xmath67 is the vector space of spherical harmonics of degree @xmath34 , i.e. those functions on the sphere @xmath68 satisfying : @xmath69 inspired by the ideas of @xcite and @xcite we study a more general class of random polynomials that contains as special cases both the homogeneous and the harmonic ones . to define such a class , for every @xmath70 $ ] and @xmath71 let us consider the following index set : @xmath72=\{\textrm{natural numbers $ l$ such that $ [ \alpha d]\leq l\leq d$ and $ d - l$ is even}\}.\ ] ] with this definition the @xmath73-_window _ of eigenspaces we are going to consider is the following orthogonal direct sum : @xmath74}}=\bigoplus _ { l\in [ \alpha d , d]}h_{n , l},\quad \alpha\in [ 0,1].\ ] ] we endow @xmath75}$ ] with the probability distribution induced by the real fubini - study ensemble . there is no unified convention for the names of these ensembles . in fact the name _ real _ fubini - study should be distunguished from _ complex _ fubini - study which is another name for what we called kostlan distributed . + in this paper we will only deal with the _ real fubini - study _ ensemble and its restriction to @xmath75}.$ ] for this reason , since no confusion can arise , a _ random polynomial _ for us will simply mean a polynomial from this ensemble . in particular for @xmath76 we get random homogeneous polynomials of degree @xmath2 and for @xmath77 we get random spherical harmonics of degree @xmath2 : @xmath78}= w_{n , d } \quad \textrm{and}\quad h_{[d , d]}=h_{n , d}.\ ] ] in fact we do not need to restrict to the case the @xmath73-window contains only the eigenspaces @xmath67 with @xmath79 even . our choice is motivated by the fact that we want to avoid redundancy in order to get random polynomials for @xmath76 ; one can consider as well the full window . we denote the dimension of @xmath67 by @xmath80 and the one of @xmath81}}$ ] by @xmath82 ; notice that in the case @xmath83 we have @xmath84 . if we consider an orthonormal basis @xmath85 for @xmath67 , we can decompose a random @xmath86 in @xmath81}}$ ] into a sum of random spherical harmonics of different degrees : @xmath87}\sum_{i=1}^{d(n , l)}\xi_{l}^iy_l^i\quad \textrm{with}\quad \xi_l^i\sim n(0 , d_\alpha^{-1}).\ ] ] the introduction of the scaling coefficient @xmath88 is absolutely not necessary , but it has the advantage that @xmath89 which simplifies the notation in the proofs . as for the expectation of metric properties of the zero locus of a random @xmath1 in @xmath81}}$ ] , we prove the following theorem . [ volume]let @xmath70 $ ] and @xmath1 be a random polynomial in @xmath75}$ ] . then : @xmath90 in particular taking expectation on both sides of inequality ( [ introvolume ] ) turns it into an equality ( the volume on the projective space and the one on the sphere being related by a constant that depends only on @xmath4 ) . we notice that the statement for @xmath77 was proved by brard in @xcite and for @xmath76 by brgisser @xcite as stated above in ( [ volumeburg ] ) . in fact it is possible to write the exact value of this expectation using an analog of what we called the parameter of the distribution , defined in this case as @xmath91}l(l+n-1)d(n , l)$ ] ; the expectation of the volume is then @xmath92 the same statement holds for the projective zero locus , as it is double covered by the spherical one . the statistics of the number of connected components is more subtle and requires more work , drawing on the ideas of @xcite . the statement we will prove is the following . [ thm : main ] let @xmath93 $ ] , and @xmath1 be a random polynomial in @xmath81}}$ ] . then : @xmath94 in analogy with the case of the volume , this theorem says that if we take expectation on both sides of milnor s bound ( [ milnor ] ) we turn it into an equality ( the number of connected components on the projective space and those on the sphere being related by a factor of _ two _ ) . notice that the case @xmath58 and @xmath77 gives the spherical harmonic case proved by nazarov and sodin in @xcite . the case @xmath58 and @xmath76 gives sarnak s claim @xcite on curves from the real fubini - study ensemble . after this paper was written , the authors have been informed by f. nazarov and m. sodin that they have an alternative approach to this problem , that works in more general settings ; a summary of their ideas is available at @xcite . the authors have also been informed by p. sarnak that together with i. wigman they are able to provide asymptotics for the expected number of ovals of a random plane curve for the _ naive _ model ( the defining coefficients for the monomials are independent normals with mean zero and variance one ) , the kostlan model and the real fubini - study one . it is interesting to discuss the case @xmath76 and @xmath58 , i.e. random real algebraic curves . the first part of hilbert s sixteenth problem asks to study the configuration of the components of a real algebraic curve of degree @xmath2 in the plane @xmath95 the nullhomotopic components of such curves are called _ ovals _ ; each one of them separates @xmath9 in two components : one of them is homeomorphic to a disk ( and is called the _ interior _ of the oval ) and the other to a moebius band . in the case @xmath2 is even all components are ovals , but if @xmath2 is odd there is a component that is not an oval ( it is homotopic to a projective line and generates the fundamental group of @xmath9 ) . harnack s theorem states that such curves can not have more than @xmath96 components , but the possible arrangements of their ovals is a far more complicated problem ; we refer the reader to the beautiful survey @xcite on the subject . in fact the problem itself of constructing curves with many ovals is a difficult one . gayet and welschinger s theorem @xcite on the exponential rarefaction of maximal curves confirms such difficulty : if we write the equation of a curve putting some coefficients in front of the standard monomials , it is exponentially difficult to get one with many ovals . this approach " of building curves depends on the choice of the basis we want to use for @xmath97 ; if instead of the monomial basis ( which is orthogonal for the bombieri scalar product , inducing the kostlan distribution ) we use the spherical harmonics the result improves dramatically . in fact now , because of theorem [ thm : main ] for @xmath76 and @xmath58 the expectation of the number of components of the curve we obtain is a fraction of the maximal one . the price we pay is that we are no longer allowed to use the simple monomial basis and we need an explicit expression for the polynomials representing spherical harmonics , such an expression being quite complicated . a completely analogous discussion holds for real hypersurfaces : now the fact that what we get applying theorem [ thm : main ] is on average a fraction of the maximal number of components is due to milnor s bound . the method employed here ( based on the ideas of nazarov and sodin ) suggests that more generally one could ask for the expected _ arrangement _ of such ovals . in fact essentially the proof relies on the following observation : given any point @xmath20 on the sphere and a small disk around it of radius @xmath98 there is a positive probability ( independent of @xmath2 ) of finding an oval inside such small disk ( and such oval loops around @xmath20 ) . unfortunately this probability , even if constant , is still very little . a numerical computation of the expectation of the number of components for @xmath58 was performed by m. nastasescu in @xcite : as sarnak writes in his letter a random curve is 4% harnack " . thus , the random version of hilbert s sixteenth problem , to study the expected _ arrangement _ of a random curve , is still open and despite its difficulty it seems to be very fascinating . one of the main difficulties in this framework is to define a statistic ( beyond the number of components ) that captures some features of the arrangement . we suggest one that we call the _ energy _ of the curve @xmath99 . first we say that one _ oval _ @xmath100 is contained into another one @xmath101 if it lies in the interior of @xmath101 . using containment we can give a partial ordering on ovals and construct a forest ( a collection of trees ) representing their arrangement . the vertices of the trees represent the ovals , with the empty ovals represented by leaves , and an oval not contained in any other is represented by a root . we define an energy that is additive for disjoint unions of trees , and multiplicative when appending trees . so the energy of a forest is the sum of the energies of its trees . we define the energy of a tree with only one isolated vertex as @xmath102 , and of a larger tree inductively as twice the sum of the energies of all subtrees that result from removing its root . this description determines the energy of any forest ( for example , see figure [ fig : tree ] ) . and the corresponding forest . the energy is @xmath103 . ] by bzout s theorem the length of each tree can not be more than @xmath104 and if such a tree exists then no other trees can appear ; the corresponding curve is said to be maximally nested and its energy is @xmath105 . on a somehow opposite direction one can consider harnack curves , having @xmath106 ( for @xmath2 large enough ) . such curves have the maximal number of components and satisfy some special condition on the arrangement of their ovals ; we refer the reader to @xcite for the exact definition and to @xcite for surprising connections with planar dimers . thus , using this notion , one open problem that we can state on the expected arrangement is to describe the asymptotic behavior as @xmath2 goes to infinity of the energy of a random curve . the pictures we get of a random curve suggest that the energy has the same order as a harnack curve , so we conjecture that : @xmath107 ( the lower bound follows from theorem [ thm : main ] ) . in section 2 , we specialize to the case @xmath108 . this helps to illustrate the main techniques used in the paper and provides a proof for the value of the parameter @xmath40 used section 6.2 . theorem [ volume ] is proved in section3 and theorem [ thm : main ] in section 4 . section 5 is devoted to the proof of the main technical lemmas for theorem [ thm : main ] . in section 6 we discuss some examples ( more on random curves and random surfaces ) . in the appendix we provide the proof of milnor s bound , we review a construction of hyperspherical harmonics , as well as some of their properties . the authors are grateful to s. basu for his constant support and many useful suggestions and a. lancichinetti for stimulating discussions . to illustrate the main ideas , we first consider a random polynomial in the special case @xmath108 . as mentioned in the introduction , in this case both the zero dimensional volume and the number of connected components coincide . for simplicity of notation we will denote their expectation by : @xmath109 in view of the generalization to higher dimensions , we will give two proofs that @xmath110 . the first one is based on techniques coming from _ integral geometry _ and will produce an exact result ; this proof will generalize to any number of variables producing again an exact result . the second is based on the _ barrier method _ introduced by nazarov and sodin ; it will produce only an approximate result but has the advantage that it generalizes to higher dimensions . we notice that in the case @xmath108 , a real random polynomial is just a trigonometric polynomial with either only even or only odd terms ( depending on whether @xmath2 is even or odd ) . the first proof is obtained following the ideas of @xcite . recall that if we have functions @xmath111 defined on the real line we can consider the random function : @xmath112 then if the coefficients are independently distributed , the expected number of zeros of @xmath1 on the interval @xmath113 $ ] is given by the formula : @xmath114$}\}=\frac{1}{\pi}\int_a^b\| \dot{\gamma}(t ) \|dt,\ ] ] where @xmath115 is the projection of the curve @xmath116 on the unit sphere @xmath117 ( we refer the reader to the beautiful paper @xcite for this and related topics ) . we note that ( [ expintgeom ] ) is proved using a special case of the _ kinematic formula _ from integral geometry . in our case @xmath113=[0 , 2\pi]$ ] and if @xmath2 is even , @xmath118 the curve @xmath119 is given by : @xmath120 and its norm in @xmath121 is constant and equals @xmath122 in particular we have : @xmath123 the norm of @xmath124 is @xmath125 which easily simplifies to : @xmath126 a similar calculation in the case @xmath2 is odd gives the exact same result . we are now in the position to use formula ( [ expintgeom ] ) and we obtain that the expected number of zeros of @xmath1 on @xmath127 $ ] is @xmath128 . thus , the number @xmath129 of expected zeros on the _ real projective line _ is : @xmath130 notice that this is an exact result and gives the asymptotic @xmath131 in this section we only prove that @xmath132 for @xmath2 large enough ; the inequality @xmath133 follows from the fundamental theorem of algebra . the proof is centered around the function described in the following claim whose proof is deferred until the end . there exist numerical constants @xmath134 and @xmath135 such that , for each sufficiently large @xmath2 and each @xmath136 there is a homogeneous polynomial @xmath137 of degree @xmath2 and @xmath16-norm one satisfying : @xmath138 once a point @xmath20 in @xmath17 is fixed , we can decompose @xmath139 where @xmath140 is a gaussian random variable with variance @xmath141 , and @xmath142 is in the orthogonal complement to @xmath143 in @xmath15 . we choose a new gaussian random variable @xmath144 independent of @xmath140 and of @xmath142 with variance @xmath141 , and set : @xmath145 by construction the random functions @xmath146 and @xmath147 have the same distribution as @xmath1 . we notice that we can write : @xmath148 next we apply the following claim to @xmath146 and @xmath147 . there exists a constant @xmath149 such that @xmath150 for every @xmath20 in @xmath151 we also defer the proof of this claim in favor of first seeing how it is used . let us recall that markov s inequality for a positive random variable @xmath152 reads @xmath153 . if we apply such inequality to @xmath154 , because of the previous claim we can find a constant @xmath155 such that : @xmath156 where @xmath134 is provided by the _ existence of the barrier _ claim . the event that @xmath157 and @xmath158 are each simultaneously _ at most _ @xmath155 is an intersection of four events each of which is complementary to the one of the above . thus the probability of their intersection satisfies : @xmath159 we note that these four events are not mutually independent , but this is not needed in the line above . now consider the event @xmath160 that @xmath161 and @xmath162 . if @xmath160 occurs then @xmath1 has a zero ; in the current situation we only need to use one of the two boundary points of an interval centered at @xmath20 . ] on @xmath17 between @xmath20 and @xmath163 . the event @xmath160 happens provided that : ( i ) @xmath164 and ( ii ) @xmath165 evaluated at at @xmath20 and @xmath163 are each bounded in absolute value by @xmath155 . note that ( i ) and ( ii ) are independent , and we have just checked above that ( ii ) happens with probability at least @xmath166 . moreover , @xmath167 for some constant @xmath168 , since @xmath169 has variance @xmath170 . it remains to choose @xmath171 disjoint intervals in @xmath17 each of length @xmath172 . each of them contains a zero of @xmath1 with probability at least @xmath173 and hence : @xmath174 as desired . we will choose our barrier in such a way that the barrier @xmath143 centered at some point @xmath136 is obtained just by precomposing @xmath175 with a rotation that moves the point @xmath176 to @xmath20 , so we only need to define @xmath175 which we do by normalizing the function : @xmath177 notice that @xmath178 is a polynomial of degree @xmath2 , since the denominator divides the numerator . ( @xmath178 is a projectivized chebychev polynomial of the second kind ) . + we have : @xmath179 say by de lhopital s rule with @xmath180 , and for all @xmath2 sufficiently large , we also have @xmath181 thus it suffices to show that the norm of @xmath178 is on the order of @xmath38 as @xmath182 , for then we will obtain the desired properties for @xmath183 ( the normalization of @xmath178 ) . it is easy to show that @xmath184 thus the function @xmath185 has the desired properties with @xmath186 and @xmath187 . let us choose a new orthonormal basis @xmath188 so that all basis elements except @xmath189 vanish at @xmath20 . to see that it is possible to do this , first notice that the subspace @xmath190 of homogeneous polynomials of degree @xmath2 vanishing at @xmath20 has codimension one inside @xmath15 . thus , to get the desired orthonormal basis , we may first orthonormalize a basis for @xmath190 , then take its orthogonal complement @xmath191 which consists of just one function . once normalized , this function gives the basis element @xmath189 . writing our random polynomial in terms of this basis , we have @xmath192 using again the standard basis @xmath193 , the fact that @xmath189 is a normalized trigonometric polynomial of degree @xmath2 means that @xmath194 where we have also used the triangle inequality and the discrete cauchy - schwarz inequality . combining this with the above , @xmath195 which implies that @xmath196 is bounded by a constant ( the gaussian random variable @xmath197 has variance of constant order , and therefore so has the expectation of its modulus ) . we generalize the proof given in @xcite for the _ harmonic _ case to our @xmath73-window of eigenspaces . here the main idea is to use a basis of the eigenfunctions of the laplacian to define a kind of _ veronese _ embedding , transforming the geometry of zeros of such functions on @xmath68 to the geometry of the intersection of the image of @xmath68 with hyperplanes . what is nice about using harmonic functions is that such a veronese embedding turns out to be a dilation , i.e. dilates the riemannian metric by a constant factor . in the language of integral geometry such embeddings are called _ moment maps _ ( generalizing the idea of the _ moment curve _ @xmath115 that we used in the univariate case in section 2.1 ) . let us denote by @xmath80 the dimension of @xmath67 as above ( thus in the case @xmath58 we have @xmath198 and in the general case @xmath199 ) . for each @xmath200 let @xmath201 be an orthonormal basis for @xmath67 and consider the map : @xmath202 whose components are @xmath203 such a map has the following properties : ( i ) it is an immersion and its image is a submanifold of the sphere in @xmath204 of radius @xmath205 ; ( ii ) it is a dilation of ratio @xmath206 where @xmath207 it is the eigenvalue for the eigenspace @xmath67 of @xmath208 . in other words the length of the image of a unit vector in @xmath68 under the differential of @xmath209 is @xmath210 . property ( i ) follows from the addition theorem for spherical harmonics : @xmath211 here @xmath212 denotes the ( reproducing kernel ) zonal harmonic of degree @xmath34 , for which @xmath213 . thus , evaluating the addition formula at @xmath214 we get : @xmath215 which shows the image is contained in the sphere . property ( ii ) is proved in @xcite and goes under the name of _ nice imbeddings property_. for a given @xmath216 we consider thus the map : @xmath217}{\mathbb{r}}^{d(n , l)}\ ] ] whose components are the @xmath209 normalized by the factor @xmath218 , i.e. @xmath219 with this choice of the normalization the addition theorem for spherical harmonics implies that the image of @xmath220 is contained into the unit sphere in @xmath221 . we claim now that @xmath220 is a dilation of ratio @xmath40 , as defined in the introduction . to prove this let us consider a unit vector @xmath222 and let us compute the square of the norm of its image under the differential of @xmath220 : @xmath223 } \frac{\textrm{vol}(s^n)}{d_\alpha}\|d\lambda_{n , l } ( v)\|^2=\sum_{l\in [ \alpha d , d]}\frac{\lambda ( n , l ) d(n , l)}{n d_\alpha}=\delta,\ ] ] where in the second equality we have used the fact that @xmath209 is a dilation of ratio @xmath224 . we apply now the integral geometry formula ( see @xcite ) : @xmath225 property ( ii ) above implies that the map @xmath220 is a covering space ( in fact either a double cover , or a diffeomorphism ) ; let us say it is @xmath226-sheeted . we claim now that the left hand side of ( [ intgeom ] ) equals exactly @xmath227 . in fact if @xmath228 , then @xmath229 equals the image under @xmath220 of the nodal line of the function @xmath230 , where @xmath231 is the basis for @xmath75}$ ] obtained by collecting together the elements from the various orthonormal basis of @xmath67 . since @xmath220 is a covering , the volume of such image ( which is @xmath232 dimensional ) equals @xmath233 times the original volume ; moreover since our gaussian distribution is uniform on the unit sphere in @xmath75}$ ] , then the above integral computes expectation and the conclusion follows . as for the right hand side of ( [ intgeom ] ) again the fact that @xmath220 is a @xmath226-sheeted covering and a dilation of ratio @xmath40 , implies @xmath234 . plugging all this into ( [ intgeom ] ) we finally get : @xmath235 our proof is deeply influenced by the ideas and the exposition of @xcite . in fact for @xmath77 our claim is the direct generalization of the _ lower bound estimate _ in @xcite ; this case is a simple extension of the proof in @xcite and we devote a remark to it at the end of the proof . notice that the upper bound for our expectation is provided by milnor s bound ( [ milnor ] ) , discussed in the appendix , and we only have to prove the lower bound . thus we fix , but not on @xmath236 for the rest of the proof an @xmath237 to start with , for @xmath20 a point on the sphere @xmath68 we consider the following event : @xmath238 if @xmath239 occurs , then @xmath240 has a component inside @xmath241 ( and this component _ loops _ around @xmath20 ) . if we prove that for @xmath242 there exists a constant @xmath243 independent of @xmath20 and @xmath2 such that @xmath244 then we can cover the sphere @xmath68 with @xmath245 disjoint such disks and get the statement ( each one of these disks contributing at least @xmath246 to the expectation in the statement ) . specifically , the radius @xmath247 we are going to use from now on is given by : @xmath248 where @xmath249 is the first point of minimum of the bessel function @xmath250 and @xmath251 for @xmath252 is the first point _ to the right of zero _ where the bessel function @xmath253 reaches a minimum ( in this case @xmath253 already has a minimum at zero , see @xcite ) . the following lemma provides the existence of the so called _ barrier _ function ; we defer its proof until the next section . [ lemma1 ] for every @xmath254 there exists a constant @xmath255 such that for every @xmath71 and every @xmath20 in @xmath68 there exists a homogeneous polynomial @xmath256 of degree @xmath2 and @xmath16-norm one satisfying : @xmath257 at the point @xmath20 . ] using @xmath258 we can decompose @xmath81}}=\textrm{span}\{{b_{x}}\}\oplus\textrm{span}\{{b_{x}}\}^\perp$ ] , thus getting the following decomposition for @xmath86 : @xmath259 we let now @xmath260 be a random variable distributed as @xmath140 but independent of it and we define @xmath261 . notice that both @xmath86 and @xmath262 have the same distribution . the introduction of these new random polynomials allows us to write : @xmath263 and to split our problem into the study of the behavior of @xmath258 and @xmath262 separately . in fact the event @xmath239 happens provided that for some constant @xmath264 the two following events both happen : * @xmath265 ; * @xmath266 and @xmath267 to check that @xmath268 implies @xmath239 we simply substitute the inequalities defining @xmath269 and @xmath270 in ( [ splitf ] ) evaluating respectively at @xmath20 and at @xmath271 . now by definition the two events @xmath269 and @xmath270 are independent and thus : @xmath272 because of this it suffices to show that the probability of each one of them is bounded from below by a positive constant that does not depend on @xmath2 and this will provide a bound from below for the probability of @xmath239 which is independent of @xmath2 . the fact that the probability of @xmath269 is bounded from below independently on @xmath2 immediately follows from lemma [ lemma1 ] : since @xmath84 and @xmath273 the random variable @xmath274 is gaussian with mean zero and variance @xmath275 ; similarly for the boundary . thus for @xmath276 small enough the probability of it being bigger than @xmath276 is uniformly bounded from below . it remains to study @xmath270 ; to this extent notice that this event is given by the intersection of the four events @xmath277 and @xmath278 . for these we do not need independence as : @xmath279 and it is enough to prove that the probability of the complement of each one of them is small . to bound the probability of @xmath280 and @xmath281 we argue as follows . first notice that since @xmath165 are distributed as @xmath1 it is enough to prove the corresponding bound for a random @xmath1 . we consider in @xmath75}$ ] the hyperplane @xmath282 of those functions vanishing at @xmath20 together with its orthogonal complement @xmath191 . such orthogonal complement _ must _ be spanned by a norm - one function @xmath283 of the form : @xmath284}y_ly_l\quad \textrm{with}\quad \sum_{l\in [ \alpha d , d]}y_l^2\leq 1,\ ] ] where @xmath285 is the normalized _ zonal harmonic _ centered at @xmath20 ( notice in particular that this normalization produced the r.h.s . inequality in the above line ) . in fact in each space @xmath67 the orthogonal complement of the zonal @xmath286 centered at @xmath20 vanishes at this point , and we immediately get ( [ bestbarrier ] ) . such zonal function is obtained by composing the function in equation ( [ zonal ] ) with a rotation that moves the north - pole of @xmath68 to @xmath20 ; moreover by setting @xmath287 in equation ( [ zonal ] ) we see that it has order @xmath288 at @xmath20 . using the above orthogonal decomposition @xmath75}= v\oplus v^\perp$ ] , we can write the random @xmath1 as @xmath289 with @xmath290 a random function in @xmath291 thus we get : @xmath292}|y_ly_l(x)|\bigg|\leq { \mathbb{e}}\bigg|\xi_0\bigg(\sum_{l\in [ \alpha d , d]}y_l(x)^2\bigg)^{1/2}\bigg|\leq a_3.\ ] ] we have used ( [ bestbarrier ] ) and the discrete cauchy - schwarz inequality in the second inequality ; for the last inequality we have used the fact that @xmath293 , which implies @xmath294}y_l(x)^2\bigg)^{1/2}=\theta(d^{n/2}).\ ] ] since the variance of @xmath295 has order @xmath245 , when we scale it by @xmath296 we make its variance of constant order , which gives ( [ bestbarr ] ) . the statement on the probability of @xmath297 being smaller immediately follows from markov s inequality . we explain now the argument for the remaining two cases . again since @xmath262 is distributed as @xmath86 , the probability of each @xmath298 is exactly the probability of the following event : @xmath299}}$}\}.\ ] ] hence we consider the positive random variable : @xmath300 bounding the expectation of @xmath152 is more delicate and we devote to it a separate lemma , whose proof is postponed to the next section . [ lemma2 ] for every @xmath93 $ ] there exists a constant @xmath301 such that @xmath302 applying markov s inequality to @xmath152 combined with the previous lemma gives a bound for the probability of @xmath303 concluding the proof . this proof does not apply when @xmath77 , but in that case one can use a rather direct generalization to @xmath42 variables of the proof given in @xcite . namely , their claim 3.1 providing a bound on the expected max over a small disk is still true in more variables and can be proved in the same way , and the barrier function can again be chosen to be the normalized zonal harmonic composed with a rotation . we work out the details for @xmath58 first . we will construct @xmath256 for @xmath20 the north pole and define it for the other points by considering the composition with rotations ( such rotations preserve all the properties of the statement ) . set @xmath304 ( the first minimum point of the bessel function @xmath250 ) and let @xmath305 be the normalized _ zonal _ harmonic of degree @xmath34 centered at the north nole . using the _ legendre polynomial _ @xmath306 of degree @xmath34 we can write : @xmath307where @xmath308 are the standard polar coordinates on @xmath309 . we recall from @xcite the following asymptotic for @xmath306 ( hilb s asymptotic ) : @xmath310 where @xmath250 is the bessel function of the first kind and order zero . the error @xmath311 is always less than @xmath312 the function @xmath250 satisfies the following properties : @xmath313 it is decreasing in @xmath314 $ ] where @xmath315 is a local minimum with @xmath316 . we call @xmath317 the first zero of @xmath250 . let @xmath318 be the maximum of the two values @xmath73 and @xmath319 ; we define @xmath256 as : @xmath320}y_l^0}{\sqrt{\textrm{card}[\mu d , d]}}.\ ] ] since each @xmath305 has norm one and since different zonals are pairwise orthogonal , we immediately see that the @xmath16-norm of @xmath256 is _ one _ , as required from the statement . let us now estimate the value of @xmath256 at @xmath20 . using the integral estimate for the numerator , which provides a bound of the form @xmath321 , and the fact that @xmath322}=\theta(d^{1/2})$ ] , for some constant @xmath243 independent of @xmath2 we have : @xmath323}\sqrt{2l+1}}{\sqrt{\textrm{card}[\mu d , d]}}\geq a_1 d.\ ] ] recall that the number @xmath324 was defined such that @xmath325 i.e. we set : @xmath326 we estimate now the value of @xmath256 on the boundary of a disk of radius @xmath324 . notice that it suffices to show the bound for the numerator in the definition of @xmath256 has order @xmath321 , since we already know the denominator has order @xmath327 . because of hilb s asymptotic we have : @xmath328}\sqrt{\frac{r}{\sin r}}j_0\bigg(\frac{2l+1}{2}r\bigg)\sqrt{2l+1}}_{l(r)}+\underbrace { \sum_{l\in [ \mu d , d]}r_l(r)\sqrt{2l+1}}_{r(r)}.\ ] ] we estimate the error @xmath329 first . notice that the summands @xmath330 are all bounded by @xmath331 so that : @xmath332 next we bound @xmath333 as follows . first notice that since @xmath334 , then for every @xmath34 in the sum @xmath335 . we fix now a @xmath336 and notice that since each term is negative and @xmath250 is decreasing in @xmath337 , then : @xmath338}\bigg(\frac{r}{\sin r}\bigg)^{1/2}j_0(\gamma y)\sqrt{2l+1}.\ ] ] moreover using the expansion @xmath339 we get : @xmath340 where we have used the integral estimate for the sum of the square roots . recalling that @xmath341 , this concludes the proof for @xmath58 . for the case @xmath252 we argue as follows . first recall that we set @xmath251 to be the first point after zero where the bessel function @xmath253 reaches a minimum . the function @xmath253 has the following properties : it vanishes at zero and we denote by @xmath342 its first point of maximum ( it is positive at this point ) ; it is decreasing on the interval @xmath343 $ ] and negative at @xmath251 . we let @xmath344 be its only zero in @xmath343 $ ] and @xmath345 be @xmath346 if @xmath285 is the normalized zonal harmonic centered at @xmath20 , we define the barrier as : @xmath347}y_l}{\sqrt{\textrm{card}[\mu_n d , d]}}.\ ] ] since the zonals are pairwise orthogonal , the @xmath16-norm of @xmath256 is one . as already discussed in the proof of theorem [ thm : main ] , the order at @xmath20 of each zonal @xmath286 is @xmath288 as shown in ( [ zonalorder ] ) , and thus the order of their sum at @xmath20 is @xmath348 . moreover since @xmath349}=\theta(d^{1/2})$ ] , then we immediately get that for some positive constant @xmath135 : @xmath350 the proof of the bound for the value of @xmath256 on @xmath271 involves the same steps as in the case @xmath58 , this time using a generalization of hilb s asymptotic involving the bessel function @xmath351 . we need an explicit expression for @xmath286 first : @xmath352 where @xmath353 is a jacobi polynomial and @xmath354 has order @xmath355 , see ( [ c(n , l ) ] ) in the appendix . using the asymptotic provided in ( @xcite , theorem 8.21.12 ) we have : @xmath356 where @xmath357 @xmath358 is given by ( [ h(n , l ) ] ) , and the error term @xmath311 is always less than @xmath312 since by definition @xmath359 we have @xmath360 and plugging all this into the definition of @xmath256 , we have that @xmath361 equals : @xmath362}\theta(l^{1/2})\bigg\{\bigg(\frac{r}{\sin r}\bigg)^{1/2}j_{\frac{n-2}{2}}(r(2l+n-1)/2)+r_l(r)\bigg\}.\ ] ] for the error term we get a bound of the form @xmath363 as for the principal term , we fix as above a @xmath364 and notice that since each summand is negative and @xmath253 is decreasing in @xmath365 , then : @xmath366}\theta(l^{1/2})j_{\frac{n-2}{2}}(\gamma_n y_n)\leq -c_1d^{n/2}\ ] ] where in the last inequality we have used the integral estimate again . since the distribution is invariant by rotation , it is enough to prove the statement in the case the point @xmath20 equals the north - pole @xmath367 . we need an explicit expression for a real orthonormal basis of @xmath368}$ ] . to this end , we use the inductive construction of an orthonormal basis for @xmath67 , the space of hyperspherical harmonics in @xmath42 variables of degree @xmath34 , given in terms of orthonormal bases @xmath369 for each space @xmath370 with @xmath371 , where @xmath372 is an index set of size @xmath373 . namely , the basis elements as reviewed in the appendix are : @xmath374 in the previous formula @xmath375 is the gegenbauer polynomial ( * ? ? ? * formula ( 4.7.1 ) ) , @xmath376 is the angle from the north - pole , the multi - angle @xmath377 gives the remaining coordinates , and the normalization constant @xmath378 is stated explicitly in ( [ n ] ) ( see also ( * ? ? ? * chapter 3 ) for an exposition of this inductive construction ) . using this basis we can write the random @xmath1 in @xmath75}$ ] as : @xmath379 } \sum_{m=0}^{l } \sum_{j \in i_m}\xi_l^j y_l^m(\theta,\phi ) \\ & = \sum_{l\in [ \alpha d , d ] } \sum_{m=0}^{l } \sum_{j \in i_m}\xi_l^j n_l^m ( \sin\theta)^m p^{\left ( \frac{n-1}{2}+m \right)}_{l - m}(\cos \theta ) y_j(\phi),\end{aligned}\ ] ] with each @xmath380 distributed as @xmath381 , with @xmath382 . in particular , the restriction of @xmath1 to @xmath271 is given by : @xmath383 where the index set @xmath384\ , |\ , l \geq m \}$ ] arises from having reversed the order of summation . notice that the term in the large brackets is a gaussian random variable itself : it is a linear combination of independent random variables , each one of mean zero and variance @xmath88 . we denote such a random variable by @xmath385 ; in other words we define : @xmath386 by the addition formula for independent gaussians we get that @xmath385 is distributed as a gaussian with mean zero and variance @xmath387 where : @xmath388 most importantly , for each @xmath389 the gaussians @xmath385 are _ identically distributed _ across all @xmath390 within the index set @xmath391 . this follows from the fact that in the sum ( [ eq : newxi ] ) the coefficient of @xmath380 depends on @xmath389 but not @xmath390 . using this new notation we can rewrite : @xmath392 using the triangle inequality , we bound the expectation of @xmath393 as : @xmath394 by the previous remark the function : @xmath395 is a random spherical harmonic on @xmath396 with independent identically distributed gaussian coefficients . this allows us to apply the following claim , whose proof is provided in the appendix . @xmath397 where recall that @xmath398 is the dimension of the space of spherical harmonics of degree @xmath389 in @xmath4 variables . returning to @xmath399 , we have : @xmath400 it remains to estimate @xmath401 . we note that the gegenbauer polynomials satisfy ( * ? ? ? * formula ( 7.33.1 ) ) : @xmath402 } \bigg|p^{\left ( \frac{n-1}{2}+m \right)}_{l - m}(x)\bigg|= { l + m + n - 2 \choose l - m}.\ ] ] using this to estimate ( [ eq : sigma ] ) we get : @xmath403 inspecting the constant @xmath404 ( see formula ( [ n ] ) ) , we notice that we can bound part of it by a constant : @xmath405 ( in fact this number goes to zero as @xmath389 goes to infinity ) ; in particular we bound each term in the above sum by a constant times @xmath406 which simplifies to @xmath407 this finally gives : @xmath408 where @xmath409 is a constant that does not depend on @xmath2 . in particular @xmath410 . recalling that @xmath411 ( we have set @xmath412 for simplicity of notation ) and using @xmath413 , we have in ( [ max ] ) : @xmath414 we estimate @xmath415 , so that using the definition of the exponential function we obtain : @xmath416 this concludes the proof . besides the number of ovals of a curve in the projective plane , there are other interesting invariants associated to it that are worth studying ( e.g. the _ energy _ we defined in the introduction ) . one classical invariant is the number of _ empty _ ovals ; let us denote it by @xmath417 . if the curve @xmath99 has even degree @xmath418 , then arnold s inequalities ( see @xcite section 6 ) imply : @xmath419 in particular if a curve has the maximal number of ovals , then @xmath420 the numerical results obtained by m. nastasescu in @xcite have shown that the constant in the bound for @xmath421 is very small : @xmath422 and combining this with arnold s bound would only give @xmath423 on the other hand , as stated in the introduction , the barrier method can be used to give information also on the expected arrangement of the components of the random hypersurface . in fact the event @xmath424 as defined above ( [ omega ] ) implies the existence of an _ empty _ oval inside the disk @xmath425 . thus , placing on the sphere @xmath426 disjoint such disks gives the following : @xmath427 we notice that every @xmath428 defines a function on the real projective space @xmath0 and for a generic @xmath1 the gradient of this function does not vanish on its zero locus . in particular if @xmath99 is a component of @xmath54 , then @xmath429 has two components ; if @xmath430 is the quotient map , we define the _ interior _ of @xmath99 to be the component of @xmath429 whose preimage under @xmath431 is not connected . as we did for curves , we say that a component @xmath100 of @xmath54 contains another @xmath101 if @xmath101 is contained in the interior of @xmath100 ; an empty component is one whose interior is connected . using this notation the above corollary can be stated for hypersurfaces as well and provides the existence of @xmath245 empty components . similarly we can define the _ energy _ of an hypersurface as we did for curves : additive on disjoint unions and multiplicative on nestings . instead of always taking the value @xmath102 , the `` seed '' we take for the energy of an empty component is the sum of its betti numbers ( in the case of an empty oval , this equals @xmath102 ) . as for higher moments of @xmath62 , nazarov ad sodin have proved that for @xmath77 ( random spherical harmonics ) it exponentially concentrates about its mean and sarnak has informed the authors that the same result holds for @xmath76 ( homogeneous polynomials ) . thus the _ typical _ curve has @xmath426 components . the authors expect the same results to hold for any @xmath4 as well . on a different direction gayet and welschinger have studied random curves in the kostlan ensemble ( the _ complex _ fubini - study one ) . in @xcite they proved that maximal curves become exponentially rare in their degree . specifically for any sequence @xmath432 one can consider @xmath433 , the probability that a kostlan curve has more than @xmath434 components , and there are positive constants @xmath99 and @xmath435 such that : @xmath436 for this random model they have also proved that the expectation of the number of components is less than @xmath63 , conjecturing that this was the right asymptotic ( as confirmed by sarnak ) . finally , for _ curves _ there is another random model introduced by the first author in @xcite that has not been discussed so far and has relation with random matrix theory . recall that once we fix three symmetric matrices @xmath437 of order @xmath2 , we can consider the following homogeneous polynomial : @xmath438 if we let the symmetric matrices to be independent random matrices in the _ gaussian orthogonal ensemble _ , @xmath1 becomes a random polynomial of degree @xmath2 ; its distribution is very different from the fubini - study ones . despite this , _ every _ real algebraic curve arises as the zero locus of one such polynomial and we can define random curves in this way ; we call them _ random determinantal curves_. for example , using this description , maximally nested curves correspond to matrices for which there exists a positive definite linear combination @xmath439 and since the probability of the positive definite cone becomes exponentially small in @xmath2 , they are very unlikely ( notice that using the above definition they also have very high energy ) . each symmetric matrix @xmath23 of order @xmath2 also defines a quadratic form @xmath440 by the following identity : @xmath441 it is interesting that under this correspondence @xmath23 is in the gaussian orthogonal ensemble if and only if @xmath442 is kostlan distributed . using this property the first author has computed the expectation of the sum of the betti numbers of the intersection of one or two kostlan distributed quadrics in @xmath443 , showing that it behaves asymptotically as @xmath2 ( see @xcite ) . moreover there is a kind of _ duality _ between the curve above defined and the common zero locus of @xmath444 , as discussed in @xcite . this duality allows to interchange the topology of the random intersection of three quadrics and the one of corresponding random determinantal curve , i.e. to study one using the other . the authors plan to discuss this ensemble and related properties in a forthcoming paper . consider a random polynomial @xmath445 and its zero locus @xmath446 on the sphere @xmath447 . such @xmath448 is a random surface and using the above results , combined with the computation of the expectation of its euler characteristic , we can provide an asymptotic for _ each _ of its betti numbers . applying theorem 1 from @xcite with @xmath449 and @xmath450 we get : @xmath451 where @xmath40 is the parameter of the distribution , which we computed to be @xmath452 . since @xmath448 double covers @xmath453 , its euler characteristic is twice the one of @xmath453 . thus , plugging the value of the parameter in the above formula we get : @xmath454 we notice now that since @xmath448 is the zero locus of a globally defined function on @xmath447 , then it is orientable ( a nowhere vanishing normal is given by @xmath455 ) . we can thus apply poincar duality and get the equality @xmath456 using the definition of the euler characteristic we get : @xmath457 and @xmath458 we notice that in the case @xmath1 has _ even _ degree then it defines a function on @xmath459 as well , thus @xmath453 is orientable and the same asymptotic holds ; in the case @xmath2 is odd , the zero locus on the sphere is still orientable , but the one on the projective space might have nonorientable components ; in particular poincar duality holds only for @xmath460 coefficients . in this case we get similar asymptotic using betti numbers with @xmath460 coefficients . as we said in the introduction , we only need to prove the bound for a regular @xmath1 ( the corresponding zero locus is smooth ) ; the set of nonregular @xmath1 is a proper algebraic subset of @xmath41 , hence with measure ( and probability ) zero . in fact this proof can be modified to obtain a bound of the same order also for nonregular polynomials , using techniques from semialgebraic geometry , but this is far beyond our scope . to start with , notice that it suffices to give a bound for the sum of the betti numbers , since this will bound in particular @xmath461 moreover the universal coefficients theorem implies that the sum of the betti numbers with _ coefficient is bounded by the sum of the betti numbers with @xmath460 coefficients ( see @xcite ) : @xmath462 now we use the so called _ smith s inequality _ , as presented in @xcite . this is the statement that for a topological space @xmath463 with an involution @xmath464 , the inequality @xmath465 holds . in particular , since @xmath1 is a _ real _ polynomial , the complex conjugation is an involution on @xmath466 and smith s inequality reads : @xmath467 we use now the fact that @xmath466 is a _ nonsingular hypersurface _ in @xmath11 of degree @xmath2 . this property has two consequences : first , its cohomology has no torsion and @xmath468 ( lemma 3.1 of @xcite ) ; second , such cohomology is completely determined by @xmath2 and @xmath4 , and in particular : @xmath469 the explicit formula in the above line is a consequence of the _ adjunction _ formula , which provides the euler characteristic of @xmath466 ( see exercise ( 3.7 ) in @xcite ) ; the formula for the sum of the betti numbers immediately follows from poincar duality . finally , since @xmath470 double covers @xmath54 , the same bound holds for : @xmath471 it is interesting to notice that in the case @xmath1 is _ harmonic _ , this bound can be obtained using courant s nodal domain theorem . in fact this theorem implies that the number of nodal domains of @xmath1 on @xmath68 , i.e. @xmath472 , is bounded by @xmath245 and the long exact sequence for the pair @xmath473 provides the bound ( the reader is referred to @xcite for these effective techniques from homological algebra ) . this proof using courant s theorem applies even when the zero set of @xmath1 is not regular , but it does nt work for a _ linear combination _ of harmonic functions . as mentioned in section [ sec : lemma2 ] , an orthonormal basis for @xmath67 can be built up inductively . namely , start with an orthonormal basis for @xmath479}h_{n-1,m}.\ ] ] for each such basis element , we obtain a basis element for @xmath67 . this entails that the dimension of the space @xmath67 is the same as the dimension of the space @xmath480}h_{n-1,m}$ ] . it is instructive to check this using the known formulas for each , that is , to verify directly that @xmath481 which follows from cancellation in the sum along with the recurrence relation for binomial coefficients . we follow ( * ? ? ? * chapter 3 ) in order to state the details resulting from the inductive construction of an orthonormal basis for @xmath67 ( restricted to the sphere @xmath68 ) of degree @xmath34 . suppose an orthonormal basis @xmath369 for @xmath370 has already been constructed for each @xmath371 , where @xmath372 is an index set of size @xmath373 . we can then obtain a spherical harmonic in @xmath42 variables ( restricted to the sphere @xmath68 ) of degree @xmath34 as follows @xmath482 in the previous formula @xmath375 is the gegenbauer polynomial , @xmath376 is the angle from the north - pole , the multi - angle @xmath377 gives the remaining coordinates , and the normalization constant @xmath378 is given by : @xmath483 writing these functions for all @xmath484 and for all @xmath485 generates an orthonormal basis for @xmath67 . in order to produce the normalized zonal harmonic @xmath486 that was heavily utilized in section [ sec : lemma1 ] , take @xmath487 in the formula ( [ spherical ] ) . so the polynomial @xmath488 is constant and : @xmath489 where @xmath490 is the @xmath491-dimensional volume of @xmath396 and @xmath492 the gegenbauer polynomials satisfy ( * ? ? ? * formula ( 4.7.3 ) ) : @xmath493 together with ( [ zonalgeg ] ) this implies that @xmath494 theorem 8.21.12 in @xcite states an asymptotic for the jacobi polynomials . in particular this applies to the polynomial @xmath498 , and taking @xmath499 , we obtain the asymptotic that was used in section [ sec : lemma1 ] . @xmath500 where @xmath501 @xmath502 and the error term @xmath311 is always less than @xmath503 ( see the remark in @xcite immediately following theorem 8.21.12 ) . the proof uses a standard reproducing kernel argument . let @xmath505 be the zonal harmonic in @xmath4 variables with pole at the point @xmath506.this is not the normalized zonal , it is the reproducing kernel as defined in @xcite . namely , @xmath505 acts as a reproducing kernel for the whole space of homogeneous harmonics of degree @xmath389 . in particular , @xmath507 the cauchy - schwarz inequality yields @xmath508 where we have used @xmath509 ( * ? ? ? * , proposition 5.27 ) . now , taking the maximum over the sphere , then taking expectations over both sides of the inequality ( [ eq : cs ] ) , we have : @xmath510 the inequality in the claim now follows from the general inequality @xmath511 for a random variable @xmath448 ( it is an application of cauchy - schwarz again ) . m. nastasescu : _ the number of ovals of a real plane curve _ , senior thesis , princeton 2011 . f. nazarov , m. sodin : _ on the number of nodal domains of random spherical harmonics _ , american journal of mathematics , 131 ( 2009 ) , 1337 - 1357 .

i present several new relations between mutual information ( mi ) and statistical estimation error for a system that can be regarded simultaneously as a communication channel and as an estimator of an input parameter . i first derive a second - order result between mi and fisher information ( fi ) that is valid for sufficiently narrow priors , but arbitrary channels . a second relation furnishes a lower bound on the mi in terms of the minimum mean - squared error ( mmse ) on the bayesian estimation of the input parameter from the channel output , one that is valid for arbitrary channels and priors . the existence of such a lower bound , while extending previous work relating the mi to the fi that is valid only in the asymptotic and high - snr limits , elucidates further the fundamental connection between information and estimation theoretic measures of fidelity . the remaining relations i present are inequalities and correspondences among mi , fi , and mmse in the presence of nuisance parameters . this work has been submitted to the ieee for possible publication . copyright may be transferred without notice , after which this version may no longer be accessible . mutual information , mmse , bayesian estimation , fisher information , nuisance parameters

1010.1508

statistical information theory @xcite constitutes an essential tool for modern signal processing , computation , coding , and communication systems . its core philosophy hinges on the notions of information potential and the ability of systems to encode , transmit and decode information about one or more input parameters . based in statistical estimation theory , fisher information ( fi ) @xcite on the other hand represents the sensitivity of statistical data to one or more input parameters . its inverse , the so - called cramr - rao bound ( crb ) , yields a useful lower bound on the variance of any statistical data - based estimation of those parameters . in spite of the different essential motivations for the two families of information measures , mutual information ( mi ) and fi are closely related at least asympotically in the limit of a large number of conditionally independent measurements @xcite . a recent paper explores the validity of this asymptotic relationship when the number of measurements is not particularly large @xcite . the relation of mi to fi is essentially a local one that is valid only in the limit of either a narrow channel pdf , as in ref.@xcite , or a narrow input pdf , as we shall see in this paper . in the more general case , the mi , as i shall also show , is related more naturally to the minimum mean squared error ( mmse ) of bayesian estimation . unlike previous work @xcite , on this topic , the new relation , a lower bound on the mi , is general and applicable to arbitrary channel and input statistics . it may be regarded as a global generalization of the more restrictive local relations between mi and fi . a number of additional correspondences between the mi and mmse are derived that apply when either more measurements , or channels , are added or multiple input parameters must be estimated at once . in the latter case if the input parameters are statistically independent , then each parameter serves as a nuisance for the other parameters that must , in general , reduce both the mi and the fidelity of estimation for each parameter . these local and global considerations on the fundamental relationship between mi and bayesian estimation error are the subject of this paper . let @xmath0 be an input parameter that is statistically distributed according to the probability density function ( pdf ) @xmath1 @xcite with mean @xmath2 and variance @xmath3 . let @xmath4 be an output variable , _ e.g. , _ a measurement variable , that carries information about @xmath0 , and is distributed according to the pdf @xmath5 . for notational definiteness , let us take these variables to be continuous over appropriate ranges of values , but the analysis of this section applies equally well to discrete random variables too , provided all integrals over such variables are regarded as discrete sums over the corresponding sample spaces . the communication channel , or the measurement system as the case may be , is described by means of the conditional pdf , @xmath6 . in spite of the notation , there is no restriction placed on the number of output variables represented by the symbol . in other words , @xmath4 is in general a multi - dimensional output vector . although i shall for clarity assume initially that the input is one - dimensional , the generalization to multiple input parameters , as we shall see subsequently , is straightforward . the three pdfs are related according to the bayes rule , @xmath7 the mi is defined in terms of the various pdfs by three different entirely equivalent expressions , @xmath8 where for each pdf @xmath9 denotes the corresponding differential entropy defined by averaging the negative logarithm of the pdf over the joint pdf , @xmath10 , @xmath11 and so on . i shall always use the natural logarithm for the definition of entropies in this paper , as it yields the simplest form of the final results . all entropy and information measures are thus expressed in natural units , or nats . by using definitions of form ( [ e3 ] ) in the second of the expressions ( [ e2 ] ) and using the bayes relation ( [ e1 ] ) , we may express the mi as the average @xmath12.\ ] ] by expanding @xmath13 in a taylor series of powers of the deviation @xmath14 , we may transform the logarithmic term in eq . ( [ e4 ] ) , @xmath15,\ ] ] where the @xmath16 dependent moments " of the @xmath0-pdf are defined as @xmath17 by subtracting @xmath2 , the mean value of @xmath0 , from both @xmath18 and @xmath16 inside the integrand in eq . ( [ e6 ] ) and noting that linear deviations from the mean average to 0 , we may easily evaluate the first two @xmath16-dependent moments as @xmath19 we may now expand the logarithm ( [ e5 ] ) to second order in the deviations and note that @xmath20 for all @xmath21 . in view of this result , the only contributing term to the second order is @xmath22 ^ 2 ( \partial \ln \ , p(y|x)/\partial x)^2 $ ] . substituting this term into eq . ( [ e4 ] ) yields to the second order the following expression for the mi , @xmath23 : @xmath24 ^ 2 \ , \int dy\ , p(y|x ) \nonumber\\ & \qquad\times \left[{\partial \ln \ , p(y|x ) \over \partial x}\right]^2\nonumber\\ & = { 1\over 2}\int dx\ , p(x)\ , ( x-\bar x)^2 \ , j(y|x),\end{aligned}\ ] ] where @xmath25 is the fi defined locally at each value of @xmath0 as @xmath26 ^ 2.\ ] ] this is the first important result of the paper . its validity is guaranteed for sufficiently narrow priors for which the higher - order deviations about the input mean are negligible . note the non - local character of this second - order equality ( [ e9 ] ) : the mi is a squared - deviation - weighted average of the fi , the latter evaluated locally over the full sample space of @xmath0 . for multiple - input , multiple - output ( mimo ) channels , the following multi - parameter analog of the second - order result ( [ e9 ] ) is easily derived as well : @xmath27 where @xmath28 denotes the deviation of the @xmath29th component of the input vector @xmath30 from its mean value . it is also possible to extend relations ( [ e9 ] ) and ( [ e11 ] ) to the case of _ discrete _ random input parameters by replacing all integrals over @xmath16 to discrete sums over values in the sample space of @xmath0 , writing instead of eq . ( [ e5 ] ) @xmath31\right\},\ ] ] expanding the logarithm in a power series , and then noting that up to the second order it may be expressed as @xmath32\right\}^2\nonumber\\ & \leq { 1\over 2}\be_x\left\{\left [ { p(y|x)-p(y|x)\over p(y|x)}\right]^2\right\},\end{aligned}\ ] ] where the inequality follows from a simple application of the cauchy - schwarz inequality . the subscript @xmath0 to the expectation - value symbol indicates that the expectation is taken relative to @xmath0 , keeping other variables fixed . an expectation of the rhs above , first over @xmath33 , given @xmath16 , and finally over @xmath16 yields the following upper bound on mi to the second order : @xmath34,\ ] ] where @xmath35 defined by @xmath36 ^ 2\bigg| x , x^\prime\right\}\ ] ] is the chapman - robbins information ( cri ) @xcite . for a fixed value of @xmath0 , the cri when minimized over all possible values that @xmath37 can take yields , via its reciprocal , the tightest lower bound on the error in estimating the discrete variable in the single - test - point optimization subspace . note that the upper bound ( [ e11c ] ) applies to mimo channels as well . the results of this section have a simple interpretation : for a narrow input pdf , the mi , like the fi and cri , is a local sensitivity based measure of information . the more sensitive the channel pdf and thus the data to the input , the larger all these information measures . the gaussian linear channel illustrates this point well . consider the gaussian linear channel in which @xmath0 and @xmath4 are related through a linear gain parameter @xmath38 , a linear bias @xmath39 , and an additive noise @xmath40 distributed according to a zero - mean gaussian pdf of variance @xmath41 : [ e12 ] y = ax+b+n , n~n(0,_n^2 ) . in this case , the fi of @xmath4 , given @xmath42 , is easily computed to be [ e12p ] j(y|x)=a^2_n^2 , independent of @xmath16 . in view of this result , the second - order equality ( [ e9 ] ) becomes @xmath43 times the power snr , which is the ratio of @xmath44 times the @xmath0-variance and the noise variance , [ e13 ] i(x;y)= 12 a^2_x^2_n^2= 12 snr . note that the gaussian - channel result ( [ e13 ] ) is independent of the statistics of @xmath0 . it is also in agreement with the well known expression for the mi of a gaussian channel with a gaussian input pdf , @xmath45 when the latter is expanded to the lowest order in @xmath3 . for input pdfs that have arbitrary width , a different relation between the estimation error and mi can be obtained . the precise relation in this case involves the minimum mean - squared error of bayesian estimation and provides a lower bound on the mi . i next derive this lower bound . a good bayesian estimation is one that reduces the mean squared error ( mse ) to a value below the variance of the input pdf , the so - called prior . the variance of the prior represents the maximum mse incurred by electing to use the mean of the prior as the trivial estimator when no information from data is availaible as , e.g. , in the limit of a vanishing snr . the mse of a bayesian estimator , @xmath46 , of @xmath0 is defined as [ e14 ] _ x=\ { [ x(y)-x]^2 } , where the statistical average is taken over the joint distribution of @xmath0 and @xmath4 . the estimator @xmath47 that minimizes the mse is called the minimum - mse estimator ( mmsee ) @xcite . it is easily shown to be the mean of @xmath0 , given @xmath4 , _ i.e. , _ its _ posterior _ mean , [ e15 ] x_m(y)=(x|y ) = x p(x|y ) dx . its mean value is the mean of the prior , @xmath2 . the mse corresponding to the mmsee is the minimum mse ( mmse ) that provides the tightest possible lower bound for the mse of _ any _ bayesian estimator of @xmath0 . since @xmath48 ^ 2 = \hat x^2(y)-2x\hat x+x^2 $ ] , we may express the mse ( [ e14 ] ) for the mmsee , _ i.e. , _ the mmse as @xmath49+\be [ \hat x_{m}^2(y)]\nonumber\\ & = \be(x^2)-\be[\hat x^2_{m } ] = \sigma_x^2-\sigma_{m}^2,\end{aligned}\ ] ] where the last two equalities are obtained by recognizing that @xmath50 is the mmsee , @xmath51 , and that @xmath0 and @xmath52 both have the same expectation . since variance is always non - negative , the last equality proves that the mmse can never exceed the prior variance . the conditional differential entropy , @xmath53 , sometimes called equivocation , may be expressed as a statistical average over the output , @xmath4 , [ e17 ] h(x|y ) = -where the argument of the @xmath4-average is the conditional entropy , given a fixed value of @xmath4 . but for a given variance , @xmath54 , of the pdf @xmath55 , its entropy is bounded above by the entropy of a gaussian pdf with the same variance @xcite , namely @xmath56 . as a result , the conditional differential entropy ( [ e18 ] ) is bounded above as follows : @xmath57\nonumber\\ & \leq { 1\over 2}\ln(2\pi e)+{1\over 2}\ln \left[\int dy \ , p(y)\,\sigma_{x|y}^2\right],\end{aligned}\ ] ] where the second inequality results from the convexity of the logarithm . to see that the integral on the rhs of the second of the relations ( [ e18 ] ) evaluates to the mmse , we may note that in view of relation ( [ e15 ] ) @xmath58\right\}^2 p(x|y)\ , dx\nonumber\\ & = \int \left[{\hat x}(y)-x\right]^2 p(x|y ) \ , dx,\end{aligned}\ ] ] whose @xmath4-average is simply the mse for the mmsee estimator , namely the mmse . ( to simplify notation here and in the rest of the paper , i have omitted the subscript @xmath59 from the mmse estimator . ) putting results ( [ e18 ] ) and ( [ e19 ] ) together , we arrive at the following upper bound on equivocation : [ e20 ] h(x|y ) ( 2e ) and the corresponding lower bound on the mi ( [ e2 ] ) : [ e21 ] i(x;y ) h(x)-12 ( 2e ) . result ( [ e21 ] ) is the second major contribution of this paper . it demonstrates the precise inverse relationship between the _ minimum _ bayesian estimation error and the _ minimum _ statistical information that can be transmitted by the measurement channel . for an additive , linear gaussian channel with a gaussian input , both inequalities in eq . ( [ e18 ] ) become equalities , the first because in this case @xmath60 is gaussian and the second because @xmath54 is independent of @xmath4 . consequently , for such channel and input , the inequality ( [ e21 ] ) is obeyed as an equality . indeed , since the mmse for this case is simply @xmath61 , while @xmath62 is @xmath63 , we have the well known result , @xmath64 , for mi , where @xmath65 is the power snr and @xmath38 is the linear gain factor of the gaussian channel . the derivative equality obtained in @xcite , [ e22 ] dd i(x;y ) = 12 ^ 2_x , is a simple , immediate consequence of this result specific to gaussian channels . for a non - gaussian channel , the lower bound ( [ e21 ] ) on the mi , @xmath23 , is in general not attainable . i now analyze the poisson channel with a negative - exponential prior to illustrate this fact . consider the linear poisson channel with linear gain ( or , scaling ) factor @xmath38 and linear bias @xmath39 , so the conditional mean of output @xmath4 , given input @xmath0 , is @xmath66 . the conditional poisson probability distribution ( pd ) over the discrete samples of @xmath4 , given @xmath0 , has the form [ e23 ] p(y|x ) = ( ax+b)^yy ! , y=0,1,2, . if we take the prior pdf to be negative exponential with mean @xmath2 , [ e24 ] p(x ) = \ { ll 1|x(-x/|x ) & for x0 + 0 & otherwise , . then by bayes theorem the unconditional @xmath4-pdf takes the form @xmath67\exp(-x/\bar x),\nonumber\\ & y=0,1,2,\ldots .\end{aligned}\ ] ] by a suitable scaling and shift of the integration variable , this integral may be expressed in terms of the incomplete gamma function , [ e26 ] ( y+1,u)=_u^dx ( -x ) x^y , as [ e26 ] p(y)=1y!(a|x)^y(a|x+1)^y+1(b / a|x)(y+1,b(a|x+1)/a|x ) . the following expression for the mean squared mmsee , @xmath68 $ ] , is a simple consequence of the definition ( [ e15 ] ) and the bayes theorem : [ e27 ] = _ y=0^ , where @xmath69 denotes the expression [ e28 ] k(y)=dx x p(x ) p(y|x ) . for the poisson channel and negative - exponential prior , @xmath69 may be expressed in terms of @xmath70 , since the latter has a similar expression as ( [ e28 ] ) with the only difference that the factor @xmath16 is missing from the integrand . to see this , we first write @xmath71 in expression ( [ e28 ] ) and then recognize that for the poisson channel pd given by eq . ( [ e23 ] ) @xmath72 equals @xmath73 times @xmath74 . this yields the following useful form for @xmath69 : [ e29 ] k(y)= ( y+1)a p(y+1 ) - ba p(y ) . substituting this expression into eq . ( [ e27 ] ) and noting that [ e30 ] _ y=0^(y+1 ) p(y+1 ) = y= a|x + b ; _ y=0^p(y)= 1 ; and @xmath75 for the ne prior ( [ e24 ] ) , we obtain the following expression for the mmse ( [ e16 ] ) : [ e31 ] = 2 ^ 2 + 2ba + b^2a^2 - 1a^2_y=0^. we can now numerically evaluate the mmse expression ( [ e31])in the general case of arbitrary @xmath38 and @xmath39 , but for the case of zero bias , @xmath76 , a simple analytical expression can be derived as we now show . expression ( [ e26 ] ) for @xmath70 now greatly simplifies since the incomplete gamma function in that expression becomes complete , taking the value @xmath77 , and the sum in expression ( [ e31 ] ) may now be easily performed analytically , since @xmath78 is related to the sum @xmath79 by two successive applications of the differential operator , @xmath80 . this yields the following simple expression for the mmse when @xmath76 : [ e33 ] = ^21+a . this expression has the desired property of reducing to the prior variance , @xmath81 , in the limit of vanishing snr , @xmath82 , and of vanishing in the opposite limit , @xmath83 . the mi may also be evaluated for the poisson channel and negative exponential prior , most simply via the second of the expressions ( [ e2 ] ) . since @xmath84 , the conditional mean of @xmath85 , given @xmath16 , is simply [ e34 ] -= -(ax+b)[(ax+b)-1 ] + _ y|x ( y ! ) . a subsequent average over the prior @xmath1 then yields the conditional ( discrete ) entropy @xmath86 , which when subtracted from the unconditional output entropy @xmath87 $ ] produces the following exact expression for the mi : @xmath88 \nonumber\\ & -\sum_{y=0}^\infty p(y)\ , \ln [ p(y)\ , y!].\end{aligned}\ ] ] this too can be evaluated numerically . the differential entropy of the negative exponential prior takes a simple analytical form , since @xmath89 whose mean , the differential entropy of @xmath0 , is simply @xmath90 : [ e36 ] h(x)= 1 + . use of this expression and the mmse ( [ e31 ] ) yields the lower bound ( [ e21 ] ) on the mi . i now compare this lower bound numerically with the exact value given by the expression ( [ e35 ] ) . mi vs. the normalized linear gain parameter , @xmath91 , for three different values of @xmath39 , as indicated . the solid curves refer to the exact result ( [ e35 ] ) , while the corresponding dashed curves refer to the lower bound ( [ e21 ] ) . , width=288 ] in fig . 1 i display , as a function of the normalized linear gain parameter @xmath91 , the exact expression ( [ e35 ] ) ( solid curves ) along with the corresponding lower bound ( [ e21 ] ) ( dashed curves ) for three different values of @xmath39 , namely 0 , 50 , and 100 . the lower bound becomes tighter as the gain parameter increases in value , but typically it fails to provide a useful , nontrivial lower bound below a certain threshold value of the gain . indeed , as the exact expression for the lower bound in the case @xmath76 obtained from eqs . ( [ e36 ] ) , ( [ e33 ] ) , and ( [ e21 ] ) , namely @xmath92\end{aligned}\ ] ] shows , the lower bound drops below the trivial lower bound of 0 for @xmath91 below @xmath93 . a similar but higher threshold below which the lower bound ( [ e21 ] ) ceases to be nontrivial is obtained when @xmath39 is non - zero . however , as @xmath39 increases this lower bound becomes increasingly tighter and thus more useful at sufficiently large values of the normalized gain , @xmath91 . mi vs. the linear bias parameter , @xmath39 , for three different values of gain @xmath91 , as indicated . the solid curves refer to the exact result ( [ e35 ] ) , while the corresponding dashed curves refer to the lower bound ( [ e21 ] ) . , width=288 ] in fig . 2 , i plot the exact values and the corresponding lower - bound values for the mi as a function of the linear bias parameter , @xmath39 , for three different values of the gain parameter , @xmath91 . as @xmath39 increases , the mi decreases as expected since the sensitivity of data on the input variable @xmath0 is reduced . raising the linear gain raises the mi , as expected , for each @xmath39 value , as the previous figure shows . again , it is clear that the lower bound ( [ e21 ] ) is useful one for sufficiently large values of @xmath91 and @xmath39 . when mimo channels are involved , we may organize the input and output variables into two different column vectors , say @xmath94 and @xmath95 , where @xmath96 denotes a matrix transpose . the @xmath40-parameter analog of the upper bound ( [ e18 ] ) is simply @xmath97\right\},\ ] ] where @xmath98 denotes the determinant of the positive semi - definite covariance matrix of @xmath99 , given @xmath100 . the determinant of such a matrix is a product of its @xmath40 non - negative eigenvalues , or simply the @xmath40th power of their geometric mean . since the latter can not exceed the arithmetic mean of these eigenvalues , which is @xmath101 times the trace of the matrix , and since the logarithm is a convex function , we have the following inequalities for @xmath53 : @xmath102 \nonumber\\ & \leq { n\over 2 } \ln(2\pi e / n)\ , + \,{n\over 2 } \ln\ , \be\left(\tr\ , \bc_{x|y}\right ) \nonumber\\ & = { n\over 2 } \ln(2\pi e \mmse),\end{aligned}\ ] ] where @xmath103 here is the average minimum mse of a component - wise estimation of @xmath99 , @xmath104^t[\bfx-\hat \bfx(\bfy)]\big|y\right\}\nonumber\\ & = { 1\over n}\be_y [ \tr\,\bc_{x|y } ] , \end{aligned}\ ] ] involving the mmsee @xmath105 for the mimo problem , @xmath106 correspondingly , the mi is lower bounded by [ m27 ] i ( ; ) h()-n2 ( 2e ) . note that in the mimo case each component of the mmsee minimizes the mse of the corresponding input parameter , the one it estimates . as such , the mmsee vector ( [ m26 ] ) as a whole also minimizes the average mse per component of the input vector . i now establish two additional important properties of the mmse not previously reported in the literature but which help strengthen the correspondences with information i have already discussed via relations ( [ e10 ] ) , ( [ e11 ] ) , ( [ e21 ] ) , and ( [ m27 ] ) . the first of these concerns the behavior of the mmse as additional measurements are made . it is well known @xcite that both mi and fi exhibit an additive property , namely @xmath107 which represents the fact that in general an additional measurement only increases information . the conditional information , either @xmath108 or @xmath109 , is a direct measure of the capacity of the measurement @xmath110 to improve information about @xmath0 , given that the measurement @xmath4 has already been made . since the estimation variance is lower bounded by the inverse of the fi , the two relations ( [ e38 ] ) represent a useful inverse relationship between mi and estimation error . but this fundamental relationship is at best a local one since , as i have argued before , the fi and its inverse , the cramr - rao lower bound on estimator variance , are local measures of information and estimation fidelity . i now show that mmse exhibits a similar behavior , which will serve to accord a general global character to this local inverse relationship between information and error . let us consider two measurements , @xmath4 and @xmath110 , of the input parameter @xmath0 . the joint mmse estimator , @xmath111 , has the following mean squared value : @xmath112&=\be\left[\iint dx\ , dx^\prime x\ , x^\prime \ , p(x|y , z)\ , p(x^\prime|y , z)\right]\nonumber\\ = \iint & dx\,dx^\prime x\ , x^\prime \iint dy \ , dz { p(x , y , z ) p(x^\prime , y , z ) \over p(y , z)},\end{aligned}\ ] ] where the bayes theorem was used to replace the _ posterior _ probabilities in terms of the joint pdfs . in terms of the integral , [ e40 ] k(z , y ) dx x p(x , y , z ) , we may write the mean squared value of the joint mmse estimator ( [ e39 ] ) as @xmath113 ^ 2\}.\end{aligned}\ ] ] the first equality follows from substituting the bayes relation , @xmath114 , and definition ( [ e40 ] ) into expression ( [ e39 ] ) and from the unit normalization of the pdf @xmath115 ; the second line follows from the cauchy - schwarz inequality ; the third line from a substitution of the definition ( [ e40 ] ) and the identity , @xmath116 ; and the fourth line from an interchange of the order of the integrals . since the last expression in inequality ( [ e41 ] ) is simply the mean squared value of the mmse estimator , @xmath46 , relative to the measurement @xmath4 alone , we have arrived at the desired result , @xmath117\nonumber\\ & \leq\be(x^2)-\be[\hat x^2(y ) ] = \mmse(y),\end{aligned}\ ] ] where we have used the fact that @xmath118 = \be[\hat x(y , z)]=e(x)$ ] to express the mse , @xmath119 ^ 2]\}$ ] , as the difference of mean squared values of the prior and the estimator . note that for the inequality ( [ e42 ] ) to hold , the two measurements are not required to be conditionally independent , given the input @xmath0 . the second property of mmse relates to the case of multiple input parameters and how the error in the estimation of any one parameter is affected by the presence of the others . but to fully appreciate this property , we must place it in the context of statistical information processing to which i now turn . it is well known that the fidelity of estimation of a parameter , defined here as the smallness of the lower bound on the statistical variance of the estimator , decreases when other parameters are added to the problem . these added parameters , when not of interest , are known as nuisance parameters , and serve to reduce the fidelity , _ i.e. , _ increase the variance , of estimation of the parameter of interest . the essence of this phenomenon is captured well by the fi matrix and its inverse whose diagonal elements provide the cramr - rao lower bounds on the variances of an unbiased estimation of the parameters @xcite . a similar result must hold in the context of statistical information theory as well . it must be possible to show that when the output variables @xmath4 depend on two input parameters , @xmath0 and @xmath120 , that are distributed independently , then the mi between @xmath0 and @xmath4 can not be larger than the mi obtained by computing the mi between @xmath0 and @xmath4 for a fixed value of @xmath120 first and then averaging it over the statistical distribution of the possible values of @xmath120 . the latter , averaged mi represents the information about @xmath0 successfully transmitted through the information channel when @xmath120 is held fixed in each instance , so the statistical dispersion of @xmath120 does not corrupt the data relative to their capacity to carry information about @xmath0 . i now prove this result . let us define @xmath121 as the mi in the case @xmath120 serves as a nuisance parameter , namely as @xmath122 where @xmath123 are defined as before . this expression for mi may also be written as the following average over all three variables : @xmath124\right\ } \nonumber\\ & = -\int p(x , y , u ) \ln\left[{p(x , y)\over p(x ) p(y)}\right ] dx\ , dy\ , du,\end{aligned}\ ] ] as the integral over @xmath125 only affects the joint density @xmath126 , reducing it to the marginal , @xmath10 . in the absence of nuisance , which is indicated by a @xmath127 superscript , the mi is the following @xmath128averaged conditional mi : @xmath129 dx\ , dy\ , du.\ ] ] note that @xmath130 is the same as the more familiar conditional mi , @xmath131 , so the difference between @xmath121 and @xmath130 is equivalently that between @xmath23 and its conditional version , @xmath131 , which , as is well known @xcite , can be of either sign . in view of jensen s inequality applied to the logarithm , the difference between the two mis , ( [ n16 ] ) and ( [ n17 ] ) , has a lower bound , @xmath132 dx\ , dy\ , du\nonumber\\ & \geq-\ln\ \int p(x , y , u ) \left[{p(x , y)\ , p(x|u)\ , p(y|u)\over p(x , y|u)\ , p(x)\ , p(y)}\right ] dx\ , dy\ , du\nonumber\\ & = -\ln\ \int \left[{p(u)\ , p(x , y)\ , p(x|u)\ , p(y|u)\over p(x)\ , p(y)}\right ] dx\ , dy\ , du\nonumber\\ & = -\ln\ \int \left[{p(x|u)\ , p(y , u)\ , p(x , y)\over p(x)\ , p(y)}\right ] dx\ , dy\ , du.\end{aligned}\ ] ] in obtaining the last two equalities above , i have used the bayes rule twice , first via the identity @xmath133 , and then via the identity @xmath134 . when the variables @xmath0 and @xmath120 are statistically independent , @xmath135 , the above inequality simplifies greatly to the form @xmath136 dx\ , dy\ , du\nonumber\\ & = -\ln\,\iint p(x , y ) dx\ , dy\ = -\ln\ 1 = 0,\end{aligned}\ ] ] where i used the fact that @xmath137 and the normalization of the joint pdf @xmath10 . this proves our assertion . note that since i have made no explicit use of the dimensionality of the input and output spaces in this proof , the result is valid for an arbitrary mimo chennel . a correspondence may be drawn with analogous results from statistical estimation theory using fi . one can consider two different estimation problems involving nuisance , one in which the nuisance is also estimated and another in which it is not , which must be treated separately . [ [ estimation - of - both - input - and - nuisance - parameters ] ] estimation of both input and nuisance parameters + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the fi matrix relative to @xmath0 and @xmath120 , when both are unknown , namely @xmath138 , may be expressed in terms of the fi matrix relative to @xmath0 , when @xmath120 is known , namely @xmath139 , in the following block form : @xmath140,\ ] ] where the matrix block @xmath141 refers to the fi matrix relative to @xmath120 alone , when @xmath0 is known , and the off - diagonal blocks @xmath142 and @xmath143 , which are transposes of each other , refer to the cross - sensitivity of the data likelihood relative to @xmath0 and @xmath120 . the presence of the cross - sensitivity matrices , @xmath142 and @xmath143 , tends to increase the crbs since , as one may easily show @xcite that , e.g. , the @xmath144 block of @xmath145 has the form @xmath146 the matrix inequality following from the fact that @xmath147 is a positive matrix . for the bayesian case of priors on @xmath0 and @xmath120 , assumed for the moment to be uncorrelated , the fi matrices relative to these priors on @xmath0 and @xmath120 must be added to the blocks @xmath139 and @xmath141 , respectively , in expression ( [ f1 ] ) . adding these prior - information - based fi submatrices has , as expected , the opposite effect : it decreases the crbs on @xmath0 and @xmath120 , thus improving the fidelity of estimation . [ [ estimation - of - input - without - estimating - nuisance ] ] estimation of input without estimating nuisance + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + in this case , we must integrate over the statistical distribution of nuisance to obtain the needed pdfs from their nuisance - free counterparts . we have , in particular , @xmath148 where the input , output , and nuisance parameters have been organized into three respective column vectors , @xmath99 , @xmath100 , and @xmath149 . if we take the nuisance and input variables to be statistically uncorrelated , @xmath150 , then we may take the gradient of eq . ( [ fn1 ] ) with respect to @xmath151 simply , @xmath152 the inner product of this gradient vector with an arbitrary vector , @xmath153 , of the same length generates a scalar quantity @xmath154 upon writing the integrand in eq . ( [ fn3 ] ) as the bilinear product @xmath155 \nonumber\\ & \times [ p^{1/2}(\bfu ) p^{1/2}(\bfy|\bfx,\bfu)],\end{aligned}\ ] ] squaring both sides of that equation , and then using the cauchy - schwarz inequality , we arrive at the inequality @xmath156 ^ 2 & \leq \int d\bfu\ , p(\bfu ) { 1\over p(\bfy|\bfx,\bfu)}[\blambda^t \bnabla_x p(\bfy|\bfx,\bfu)]^2\nonumber\\ & \qquad\times \int p(\bfu)\ , p(\bfy|\bfx,\bfu)\ , d\bfu . \end{aligned}\ ] ] since the last @xmath157-integral above evaluates simply to @xmath158 according to bayes rule , by dividing both sides by @xmath158 , integrating over @xmath159 , and finally averaging over @xmath99 , we obtain the desired inequality , @xmath160 where the fi matrices in the presence and absence of nuisance are defined as @xmath161 ^ 2 \nonumber\\ & \qquad \times\bnabla_x p(\bfy|\bfx)\ , \bnabla_x^t p(\bfy|\bfx ) \nonumber\\ \bj^{(-)}(\bfy|\bfx)&\defeq \int d\bfu\ , p(\bfu ) \iint d\bfx\ , d\bfy \ , p(\bfx ) \ , p(\bfy|\bfx,\bfu ) \nonumber\\ & \times \left[{1\over p(\bfy|\bfx,\bfu)}\right]^2 \bnabla_x p(\bfy|\bfx,\bfu)\ , \bnabla_x^t p(\bfy|\bfx,\bfu ) . \end{aligned}\ ] ] note that for statistically mutually independent input and nuisance parameters , the prior - based fi for the input paramaters is the same whether the nuisance parameters are present or absent . it then follows from the the non - negative - definiteness of the difference of the data - based fis , @xmath162 , implied by relation ( [ fn6 ] ) , that the corresponding difference between the sums of data - based and prior - based fis is also non - negative - definite . this result embodies the fact that in general nuisance parameters even when they are not estimated , if statistically independent of the input parameters of interest , degrade the fidelity with which the latter can be estimated @xcite . for the more general case when @xmath0 and @xmath120 are statistically correlated , @xmath120 may indeed carry information about @xmath0 through their correlation , in which case the rhs of the inequality ( [ n18 ] ) may be negative allowing for @xmath163 to exceed @xmath164 . as i noted before , in this general case the inequality ( [ n18 ] ) can be of either sign . the corresponding result from statistical estimation theory is based on the fact that any information that @xmath120 has about @xmath0 through its correlations with it yields an additional fi submatrix , @xmath165 , to be added to the @xmath139 block in eq . ( [ f1 ] ) . this submatrix represents information that @xmath120 carries about @xmath0 through the first - order sensitivity of @xmath166 on @xmath16 . unlike the coupling of @xmath120 to the data alone , such additional prior information can reduce the crbs on estimating @xmath0 . when the nuisance parameters are _ not _ estimated , but are correlated with the input parameters of interest , the basic relation ( [ fn2 ] ) used to obtain the desired inequality ( [ fn6 ] ) is itself not valid . also , the prior - based fis are not necessarily the same with and without nuisance . as a result , the data - based fi or prior - based fi or both may contain more information about the parameters of interest in the presence of nuisance than in its absence . as a first example , let us consider a gaussian additive channel with additive noise @xmath40 , @xmath167 where @xmath0 , @xmath120 , and @xmath40 are all independently normally distributed as follows : @xmath168 thus the marginal pdf for @xmath4 as well as its various conditional pdfs are all gaussian too , @xmath169 in view of the variances given in relations ( [ n22 ] ) , we may easily write down the mis of interest here using the well known expression for the differential entropy for a gaussian additive channel @xcite , @xmath170 since @xmath171 , it follows that @xmath172 . for uncorrelated @xmath0 and @xmath173 , the latter serves as a nuisance relative to information about the former in the sense that the terms @xmath174 and @xmath40 in the model ( [ n21a ] ) simply combine to yield an increased noise variance , @xmath175 , on the determination of @xmath0 from @xmath4 . but when the nuisance is removed by holding @xmath173 fixed in each measurement , the noise variance is lower at @xmath41 , leading to increased information about @xmath0 . as my second example , let us modify the gaussian channel represented by eqs . ( [ n21a ] ) and ( [ n21b ] ) simply to include a correlation between @xmath0 and @xmath173 , so only the first of the relations in eq . ( [ n21b ] ) is changed to the conditional relation @xmath176 while the remaining relations are unchanged . in the limit that @xmath177 , the variables @xmath0 and @xmath120 become uncorrelated as in the previous example . thus the largeness of @xmath178 in relation to @xmath179 may be regarded as the strength of the correlation between @xmath0 and @xmath173 . in view of the relation ( [ n24 ] ) and the fact that @xmath180 , the marginal pdf @xmath1 is also gaussian . the conditional pdf , @xmath181 , for @xmath4 , given @xmath173 , may be computed by integrating @xmath182 over @xmath16 . the marginal pdf @xmath5 is then obtained by integrating @xmath183 over @xmath125 . using standard analysis involving gaussian integrals , we may derive the following marginal and conditional pdfs : @xmath184 where the @xmath4-variance may be expressed as @xmath185 having evaluated @xmath1 , we may now evaluate @xmath166 , via bayes rule , as the ratio @xmath186 , @xmath187\nonumber\\ & = { 1\over \sqrt{2\pi\sigma_{u|x}^2}}\exp\left[-{\left(u-\bar{u}_{|x } \right)^2\over 2\sigma_{u|x}^2}\right],\end{aligned}\ ] ] where the conditional mean , @xmath188 , and variance , @xmath189 , are given by the expressions @xmath190 by multiplying @xmath166 with @xmath191 and integrating over @xmath125 , we may obtain the last of the needed conditional pdfs , namely @xmath6 , @xmath192,\ ] ] where the conditional mean and variance may be expressed as @xmath193 we are now in a position to write down both @xmath121 and @xmath194 by use of the gaussian - channel entropy formula in terms of the variances of @xmath5 , @xmath6 , @xmath181 , and @xmath195 , @xmath196 ; \nonumber\\ i^{(-)}(x;y ) & = i(x;y|u)\nonumber\\ & = { 1\over 2 } \ln \left(1+{a^2\sigma_{x|{u}}^2\over \sigma_n^2}\right ) . \end{aligned}\ ] ] note that in the limit of @xmath197 , the two variables , @xmath0 and @xmath173 , are infinitely tightly coupled . in effect , @xmath198 , and the data @xmath4 carry no information about @xmath0 when @xmath173 is held fixed . this is seen in the relation ( [ n30 ] ) . by contrast , @xmath121 is finite in this limit , and thus trivially exceeds @xmath130 . the other limit , @xmath177 , returns us to the case of uncorrelated @xmath0 and @xmath120 variables for which the results ( [ n23 ] ) are recouped and @xmath130 exceeds @xmath121 . the more general cases in which neither of these limits is a good approximation are illustrated in fig . i plot here @xmath121 ( solid curves ) and @xmath130 ( dashed curves ) for the case of normalized @xmath173 variance , @xmath199 , equal to 5 . each variance is normalized the same way by multiplying it with @xmath200 . six different values of @xmath201 , which determines the largness of the conditional mean of @xmath0 , given @xmath173 , were used to generate the various @xmath163 curves . by contrast , @xmath164 is independent of @xmath201 , as seen from the single dashed curve on each plot . a number of observations can be made from these plots . first , for @xmath202 , the variables @xmath0 and @xmath173 are uncorrelated , so in this case the plot of mi in the presence of the nuisance variable , @xmath173 , lies below that for mi when uninfluenced by the nuisance variable . second , as @xmath201 increases , the coupling of @xmath0 and @xmath173 becomes increasingly less sensitive to the noise in @xmath173 . this means that when @xmath4 is measured , its value reveals more information about @xmath0 than when @xmath201 is smaller . when the nuisance is removed , _ @xmath173 is held fixed , then a change of @xmath201 merely changes the mean value of @xmath0 , leaving its variance unchanged , which is the reason why @xmath164 depends neither on @xmath201 nor on @xmath203 . third , as the relative strength , @xmath204 , of the nuisance increases while @xmath201 is held fixed , @xmath163 increases initially since the data @xmath4 possess an increasing amount of information about @xmath0 through the latter s coupling to @xmath173 . however , increasing the strength of the @xmath174 term in eq . ( [ n21a ] ) to large values leads to the data becoming more corrupted than helped by the nuisance , which leads to an eventual decrease of the mi . these two competing tendencies lead to a maximum for each curve ( left top and bottom ) , with the location of the maxima shifting to larger nuisance - parameter strength values with increasing @xmath201 . fourth , comparing the plots for the smaller vs. larger values of @xmath205 ( left panels ) , we see that the tighter the @xmath0-@xmath120 coupling the softer the degradation of mi@xmath206 with increasing strength of the nuisance parameter . finally , as seen from the right - hand panels of the figure , an infinitely tight coupling between @xmath0 and @xmath173 ( for @xmath207 ) yields , through the sensitivity of data to @xmath173 , information about @xmath0 as well . this information about @xmath0 degrades when @xmath208 increases to finite values , the more so the smaller the parameter @xmath201 . when multiple input parameters must all be estimated from the same measurement(s ) , one expects the mmse , like the crb , for estimating any of the parameters to be higher than if the others were not present . i prove this result next . let @xmath209 be two input parameters to be estimated from data @xmath4 . let @xmath210 be the joint prior on the inputs . the mmse estimator for @xmath0 in the absence of the nuisance @xmath120 can be defined in terms of the conditional mmse estimator , @xmath211 given @xmath212 . it is the mmse estimator of @xmath0 for a given value of @xmath120 . its mean squared value has an expression analogous to that found in ( [ e41 ] ) , @xmath213 where @xmath214 stands for the function @xmath215 we also used the bayes - rule identity , @xmath216 , to arrive at the last line of eq . ( [ n32 ] ) . a use of the cauchy - schwarz inequality in eq . ( [ n32 ] ) shows that the mean squared value of the mmse estimator in the absence of nuisance has the lower bound @xmath217,\end{aligned}\ ] ] where a simple substitution of @xmath214 from eq . ( [ n33 ] ) was used to obtain the second relation , bayes rule to obtain the third relation , and the definition of the mmse estimator @xmath46 as the posterior mean of @xmath0 , namely @xmath218 , to arrive at the final relation . since the mmse , as we have noted earlier , may be expressed simply as the mean squared value of @xmath0 minus the mean squared value of the mmse estimator , the desired inequality between the mmse without and with nuisance follows immediately , @xmath219 we deduce from this important result that the presence of the nuisance parameter can _ never _ lower the mmse below that obtained in its absence , _ i.e. , _ when the nuisance has a known value , regardless of whether the priors on @xmath0 and the nuisance @xmath120 is statistically correlated or not . this seems to exclude the possibility that @xmath120 if suitably correlated with @xmath0 may serve , as we observed in sec . [ mi+- ] in the context of mi , as a source of additional information for @xmath0 . the answer to this apparent paradox may be found in the way mmse is defined . since given a value of the nuisance @xmath125 , the mmse estimator minimizes the mse relative to the corresponding conditional prior , @xmath220 , on @xmath0 and measurement pdf @xmath195 , the nuisance - averaged mmse is not characterizable as the mse for a single , nuisance - averaged mmse estimator . the mmse metric thus may not possess the same degree of specificity as the mi or fi metrics when the effect of nuisance must be quantified . we now illustrate the effect of nuisance on the mmse with our previous example of a gaussian channel for which some of the relevant pdfs are given in eqs . ( [ n24])-([n28 ] ) . what we need are the mmse@xmath221 estimators , namely @xmath222 given by expression ( [ n31 ] ) , and @xmath46 by ( [ e15 ] ) . as is well known from the theory of mmse @xcite for gaussian priors and gaussian channel pdfs , each mmse estimator may be expressed as the inverse - variance - weighted sum of its prior and measurement based estimates , @xmath223,\end{aligned}\ ] ] where the multipiers @xmath224 are given by @xmath225 and the unconditional @xmath0-variance , @xmath3 , may be expressed as @xmath226 the various data and prior based estimates and variances used in arriving at the expressions ( [ n36 ] ) have been inferred from the mean values and variances of the pdfs given in eqs . ( [ n24])-([n29 ] ) . [ [ mmse - in - the - absence - of - nuisance ] ] mmse in the absence of nuisance + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + to compute the mean squared values of these estimators , we first subtract and add the appropriate mean values of @xmath4 from it in the expressions ( [ n36 ] ) and then use the fact that @xmath227=\be[(\delta y)^2]+q^2 $ ] , where @xmath228 is the deviation of @xmath4 from its mean and @xmath229 is any quantity independent of @xmath4 . for the mmse@xmath230 estimator , the mean we subtract and add is the conditional mean of @xmath4 , given @xmath125 , namely @xmath231 , so the following conditional squared mean value for it , given @xmath125 , results : @xmath232&=f^{(-)2}\left[{a^2\sigma_{x|u}^2+\sigma_n^2\over a^2\sigma_n^4 } + \alpha^2u^2/f^{(-)2}\right]\nonumber\\ & = { \sigma_{x|u}^2/\sigma_n^2\over { 1\over\sigma_n^2}+{1\over a^2\sigma_{x|u}^2 } } + \alpha^2u^2.\end{aligned}\ ] ] an averaging of this expression over @xmath125 with the help of the result @xmath233 then yields the required mean squared value of the mmse@xmath230 estimator . subtracting this squared mean value from @xmath234 , the latter being simply @xmath235 , generates , according to eq . ( [ e16 ] ) , the mmse@xmath230 , @xmath236 where use was made of relation ( [ n38 ] ) in the second line . [ [ mmse - in - the - presence - of - nuisance ] ] mmse in the presence of nuisance + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + subtracting and adding the mean value of @xmath4 , namely @xmath237 , from @xmath4 inside the expression ( [ n36 ] ) for the estimator @xmath46 and the squaring and averaging over @xmath4 generates the following mean squared value of the mmse estimator in the presence of nuisance : @xmath238&=f^{(+)2}{\sigma_y^2\over \sigma_{y|x}^4 } \left[{a\sigma_x^2+\alpha b\sigma_u^2\over\sigma_x^2}\right]^2\nonumber\\ & + f^{(+)2}{\alpha^2\bar u^2\over \sigma_x^4 } \left[{(a\sigma_x^2+\alpha b\sigma_u^2)^2\over\sigma_{y|x}^2\sigma_x^2}+1\right]^2\nonumber\\ & = \sigma_y^2 \left[{(a\sigma_x^2+\alpha b\sigma_u^2)^2\over\sigma_{y|x}^2 + ( a\sigma_x^2+\alpha b\sigma_u^2)^2/\sigma_x^2}\right]^2\nonumber\\ & + \alpha^2\bar u^2,\end{aligned}\ ] ] where use was made of the definition ( [ n37 ] ) of @xmath239 to simplify both terms on the rhs . in view of relations ( [ n29 ] ) for the conditional mean and variance of @xmath4 , given @xmath42 , and the fact that all pdfs are gaussian , we may express the unconditional variance of @xmath4 , namely @xmath240 , as the sum of conditional variance , @xmath241 , given @xmath0 , and @xmath242 times @xmath3 . this observation greatly simplifies the preceding expression , @xmath243=\alpha^2\bar u^2 + { ( a\sigma_x^2+\alpha b\sigma_u^2)^2\over\sigma_y^2}.\ ] ] the mmse now follows from subtracting expression ( [ n42 ] ) from @xmath244 , a result that can be simplified further in view of the relation between @xmath240 and @xmath241 that we just noted in the previous paragraph , @xmath245 by using relations ( [ n29 ] ) , ( [ n27 ] ) , the alternate form of @xmath240 given by relation ( [ n25 ] ) , and @xmath246 , we may express the mmse in the presence of nuisance in the more explicit form @xmath247 in this form , we may easily compare mmse@xmath206 to the corresponding result ( [ n40 ] ) for mmse in the absence of nuisance . a sequence of steps involving simple algebraic manipulations , followed by a use of the inequality , @xmath248 , easily confirms the general result proved earlier that the presence of nuisance parameters can never reduce the mmse for the estimation of the parameter of interest , @xmath249 this result is illustrated in fig . 4 where we plot both mmse@xmath250 , in units of @xmath251 , as functions of the variable @xmath252 for different values of the nuisance coupling parameter , @xmath253 . the reciprocal of @xmath254 is a measure of the strength of the statistical correlation between @xmath0 and nuisance @xmath120 , while @xmath255 represents the ability of nuisance to carry information about @xmath0 through its statistical correlations with @xmath0 . as @xmath254 becomes larger , the prior on @xmath0 becomes broader and the measurement becomes increasingly more dominant in controlling the mmse whether the nuisance is absent or present . but , as expected , mmse@xmath230 does not depend on the coupling parameter @xmath255 or the nuisance - parameter snr defined as snr@xmath256 . on the other hand , mmse@xmath206 decreases with increasing @xmath255 since the nuisance becomes increasingly more effective - and the data @xmath4 increasingly less so - in controlling the mse . with increasing snr@xmath257 , from 10 to 100 between the two panels of the figure , the nuisance causes an increased error in estimating @xmath0 , as its increased variance leads to an increased variance of the prior on @xmath0 . but in no event does the mmse in the presence of nuisance fall below the mmse without nuisance . the optimal condition under which the presence of nuisance does not degrade the mmse , _ mmse@xmath206 = mmse@xmath230 , is achieved when @xmath258 , as seen from the figures and can also be easily shown analytically from the expressions ( [ n40 ] ) and ( [ n44 ] ) . in this paper i have derived a number of previously unknown relationships between mutual information and the minimum error of estimating a parameter from its measurements . a seoond order linear relation between mi and a prior - averaged , squared - deviation - weighted form of the fi accords added significance to the phrase information " when describing the latter even though its chief claim to this phrase has been in the sense of being the reciprocal of estimation error . a second , more important relation between information and estimation error has been obtained in the fully bayesian context of minimum mean squared error . i have shown , in particular , that the shannon equivocation , @xmath53 , in the differential sense can not exceed @xmath259 , and hence the mi is bounded below by @xmath260 . both these results were generalized to the case of mimo channels . however , the mmse - based lower bound on mi is not easily extendable to the discrete case . ( i exclude here the trivial construct of associating with the pdf @xmath1 of a continuous random parameter @xmath0 a discrete pd involving probabilities @xmath261 computed for finite bins , centered at regularly spaced points @xmath262 that are separated by an interval @xmath263 small compared to the scale over which @xmath1 varies significantly . ) if additional input variables other than those of interest to the estimation problem are present , in general they serve to compromise the fidelity with which the variables of interest may be estimated . the impact of such nuisance variables on estimator performance was elucidated here with formulations based separately on mi , fi , and mmse , and a number of important inequalities were derived that provide valuable insight into information and error - based metrics of performance . the mmse based description of the nuisance is particularly intriguing since it seems to predict a nearly counter - intuitive result that the presence of nuisance , can never improve performance , even when it is strongly coupled to the input and has vanishing variance , _ i.e. , _ independent of its statistical correlations with the input . this may be a peculiarity of how mmse is defined , but surely deserves additional consideration . the author is pleased to acknowledge helpful contributions from s. narravula . funding support from the air force office of scientific research under grants fa9550 - 08 - 1 - 0151 and fa9550 - 09 - 1 - 0495 is gratefully acknowledged . 99 c. shannon , a mathematical theory of communication , " bell syst . j. , * 27 * , pp . 379 - 423 and 623 - 656 ( 1948 ) . t. cover and j. thomas , _ elements of information theory _ , wiley ( new york , 1991 ) . h. van trees , _ detection , estimation , and modulation theory _ , wiley ( new york , 1968 ) . b. clarke and a. barron , information - theoretic asymptotics of bayes methods , " ieee trans . th . , * 36 * , pp . 453 - 471 ( 1990 ) . j. rissanen , fisher information and stochastic complexity , " ieee trans . th . , * 42 * , pp . 40 - 47 ( 1996 ) . n. brunel and j .- p . nadal , mutual information , fisher information , and population coding , " neural computation , * 10 * , pp . 1731 - 1757 ( 1998 ) . k. kang and h. sompolinsky , mutual information of population codes and distance measures in probability space , " phys . lett . , * 86 * , pp . 4958 - 4961 ( 2001 ) . e. challis , s. yarrow , and p. seris , fisher vs shannon information in populations of neurons , " preprint ( 2008 ) . t. duncan , on the calculation of mutual information , " siam j. appl . math . * 19 * , pp . 215 - 220 ( 1970 ) . d. guo , s. shamai , and s. verdu , mutual information and minimum mean - square error in gaussian channels , " ieee trans . th . * 51 * , pp . 1261 - 1282 ( 2005 ) . e. mayer - wolf and m. zakai , some relations between mutual information and estmation error in wiener space , " annals appl . prob . * 17 * , pp . 1102 - 1116 ( 2007 ) . whenever possible , a random variable is denoted by an upper - case roman letter and the values it may take by the corresponding lower - case letter . also , for notational simplicity the same function label @xmath264 is used for the pdfs of different variables , even though the pdfs have , in general , different functional dependences on their arguments . d. chapman and h. robbins , minimum variance estimation without regularity assumptions , " ann . * 22 * , pp . 581 - 586 ( 1951 ) . s. kay , _ fundamentals of statistical signal processing : estimation theory _ , prentice hall ( new jersey,1993 ) , chapters 10 and 11 . c. rao , _ linear statistical inference and its applications _ , wiley ( new york , 1973 ) . a single - parameter , non - bayesian version of this result was proved in a. dandrea , u. mengali , and r. reggiannini , the modified cramr - rao bound and its application to synchronization problems , " ieee trans . * 42 * , pp . 1391 - 1399 ( 1994 ) .

we analyze the spatial orientation of a homogenous sample of 440 elongated planetary nebulae ( pne ) in order to determine the orientation of their apparent major axis respect to the milky way plane . we present some important geometrical and statistical considerations that have been overlooked by the previous works on the subject . the global distribution of galactic position angles ( gpa ) of pne is quantitatively not very different from a random distribution of orientations in the galaxy . nevertheless we find that there is at least one region on the sky , toward the galactic center , where a weak correlation may exist between the orientation of the major axis of some pne and the galactic equator , with an excess of axes with gpa . therefore , we confirm that `` extrinsic '' phenomena ( i.e. , global galactic magnetic fields , shell compression from motion relative to the interstellar medium ) do not determine the morphology of pne on most of the sky , with a possible exception towards the galactic center .

astro-ph0505092

since charles messier registered the first planetary nebulae ( dumbell nebula in vulpecula ) on july 12 , 1764 until the present time , over one thousand of these beautiful objects were found in our galaxy and much more in nearby galaxies . it is well known that pne have different shapes but , in most cases , the projection of a nebula onto the sky has a defined extension axis . besides , the study of the orientation of many kind of phenomena such as supernova remnants ( gaensler 1998 ) , orbits of binary systems ( brazhnikova et al . 1975 ) , rotation of individual stars ( hensberge et al . 1979 ) , has been of broad astrophysical interest . this sort of study provides information about formation , evolution and death of the stars . however , the orientation of the projected major axis of the pne was rarely studied and the results that were found are still contradictory . for instance , shain ( 1956 ) and gurzadyan ( 1958 ) made the pioneering works about the orientation of pne . they worked with very small samples from the curtis catalogue ( 1917 ) and their conclusions were contradictory . shain found that the angle between the semi - major axis of pne and the galactic equator was small for objects with low latitudes . on the other hand gurzadyan did not found any correlation and concluded that the magnetic field could not be influencing the shape of pne . at the present time both papers only have a historical character . the first works to determine spatial orientations of pne with a significant number of objects were those of grinin & zvereva ( 1968 ) , and melnick & harwitt ( 1974 ) . these works were more complete than early ones and both papers concluded that pne are aligned with the plane . the hypothesis explaining these non - random orientations were based in the effect of the ambient magnetic field and interactions with the interstellar medium . the works of phillips ( 1997 ) and corradi , aznar & mampaso ( 1998 ; hereafter cam98 ) are the most recent and contradictory papers about global pne orientations . in both cases , the images examined had high quality and the determination of pa have small errors , but they did not study the orientation of pne over a specific region of the sky . the sample of phillips was culled from a variety of published images complemented with broad band survey plates , whereas the sample of corradi and coworkers comes mainly from three different narrow band surveys . in this paper we visit this topic and make a careful study with new geometrical considerations . the characterization of any found correlation in a certain region of the sky would help to disentangle the role of the galactic magnetic field in determining the morphologies of these sources . alternatively , it could be found that the motion of pne through the interstellar medium leads to some compression of their shells , and result in significant correlations between the apparent outflow axes . the unrestricted total sample was extracted from 3 different sources : digitized sky survey version 2 , dss2 ( red broad band ) ; the macquarie@xmath1strasbourg @xmath2 planetary nebulae catalog , ( mash ; @xmath2 narrow band ) ; and hst archival imagery ( filters f502n and f656n ) . after the starting criteria described in section 2 , the initial number of pne studied in this work resulted in 868 , significantly larger than in previous studies . the analysis is performed through sections 3 to 5 , the results are presented and discussed in sections 6 and 7 . the sample was extracted from the catalogue of galactic planetary nebulae updated version 2000 ( kohoutek 2001 ; hereafter cgpn2000 ) that includes 1510 true pne and the mash survey ( parker et al . 2006 ) with 903 objects , 578 of them are classified as true pn . there is virtually no overlap between the cgpn2000 and the mash catalogs , therefore this last one provided one third of the final sample of truly elongated pne . we included a special subsample of the cgpn2000 which is formed by those objects imaged by the hubble space telescope ( hst ) . the selection criteria employed to obtain our final sample were : 1 . objects classified as true pn . angular size larger than 10@xmath3 . the experience shows that , in general , this limit yields spatial resolution sufficient to determine the orientation of the major axis . exception in the angular size criterium was made for those objects imaged by the hst . 3 . @xmath4 , to avoid information degradation by projection effects over high latitudes objects . 4 . in case of the mash survey we only included pne which at least have one confirmation spectrum . 5 . objects clearly visible in the dss2 plates ( surface brightness @xmath5 17 mag / arcmin@xmath6 ) , this restriction criterium was applied for all the surveys contributing to the sample . elongated shape ( pne with morphology type r were rejected ) . by elongated shape pne we mean bipolar nebulae or those which show two lobes defining unambiguously the direction of their polar outflows , as well as e nebulae with an appreciable ellipticity . at the cases where the object is near to the dss resolution we estimate an empirical limit of major to minor axis length ratio 1.2:1 , as determined from those objects which were in the visual limit of acceptance . in order to extend our sample we included pne with angular size between 4@xmath3 and 10@xmath3 inclusively , that had been observed by hst . such pne , 42 in total number and 27 truly elongated , belong to some of our 4 regions . due to the variety of filters the chosen criterium here was to use the images taken through the f656n or f502n filters whenever possible . to complete the sample we also included those pne measured by cam98 that we could not measure from the dss2 plates . after a careful examination of more than 868 pne in the initial list ( obtained after criteria 1 , 2 and 3 ) , we arrived to the final sample that contains 444 truly elongated pne distributed in four regions , 174 belonging to the mash sample . the remaining objects were planetaries with apparent circular shape , low surface brightness or very peculiar morphology . we measured the position angles ( pa ) of the projected major axis over the sky , for all the pne of the final sample . pa are measured following the usual convention : from the north towards the east . these angles are measured within the equatorial coordinates system so we call them epa . the directions of the elongated axis were estimated visually by fixing the line that better represents the long symmetry axis of the pn . the angle that this line forms with the north was the epa . for bipolar structures the epa was measured with respect to the direction of the outflow of the objects . the uncertainties of epa are estimated following the criterium of cam98 : all measurements were repeated independently by both authors . the largest difference between both measurements was 8@xmath7 . the measuring criteria are exemplified in fig . [ fotito1 ] . in spite of the fact that all previous works determined the epa visually , we tried to obtain those angles through an automated way , but the problem was that this sort of automated algorithms are still not powerful enough to deal with low s / n and overcrowded fields with high galactic background . pne are usually located in highly populated milky way fields and in many cases do not show structure complete at the same surface brightness . this patchy structure and the star images make the software approach still difficult . an example of a more refined criterion of epa measurement can be derived from an attentive comparison between the fig . [ fotito1]c and [ fotito1]d . in spite of the fact that both objects present the characteristic belt of bipolar planetary nebulae , in the case of ngc5189 the structures that appear in both sides of the apparent belt do not have hemispherical appearance as the case shown in fig . [ fotito1]d . the orientation marked in the figure takes into account that the brilliant knot at the e of the main body s external part , is of totally nebular appearance when the original image is inspected . considering that the criterion for determination is based mainly on the external morphology , the major axis of the pn is well determined . this complex object represents a limit case where the measured epa could show the largest scatter when employing other criteria . but as we show in section 7 , our epa values do not show significant deviations from the values obtained by cam98 for objects present in both samples . the epas are measured relative to the equatorial coordinates system . in order to transform this position angle to the galactic system we need to compute the angle @xmath8 , which is subtended at the position of each object by the direction of the equatorial north and the galactic north . @xmath9 where @xmath10 and @xmath11 are the equatorial coordinates of the galactic north pole , @xmath12 and @xmath13 are the equatorial right ascension and galactic latitude coordinates respectively ( both coordinates were extracted from cgpn2000 ) . then the galactic position angle ( gpa ) is defined as the position angle of the major axis of the apparent nebular elongation , measured from the direction of the galactic north towards the east ( table 4 ) . @xmath14 we used the same convention that phillips ( 1997 ) used for measuring the epas , in the interval @xmath15 $ ] . the values are reported and used here in the same range , but some authors ( e.g. cam98 ) report them to the interval @xmath16 $ ] assuming a priori a symmetric problem , in the sense that a preferred orientation of the pne elongations very nearly to the galactic equator will be easier to detect . the problem with this treatment of the epa is that it would blur any other preferred orientation that is not very near to the galactic equator ( fig . [ fotito2 ] ) . the orientations for individual planetary nebulae is presented in table 4 , which is also available on request to the authors . one difference with previous studies is that , in order to avoid information degradation by projection effects , we did not include in our study objects with @xmath17 . the number of objects far from the galactic plane is not statistically significant , although the projection effect with respect to the galactic equatorial plane should be considered in any future larger sample to be analyzed . to study the orientation of the sample , we selected four angular sectors : those defined by the galactic center and anti - center , and the perpendicular directions ( table 1 ) . the regions were defined in angular range in such a way that there was enough number of objects to do a statistical analysis in at least the region towards the galactic center ( 262 objects ) . but they should not be so large in angular extent that any absolute orientation respect to the galactic system would be deleted by the superposition of different projections to the observer . it must be taken into account , that the projection effects at the position of the observer would mask the effect of some preferred orientations if the data are compiled and analyzed as a whole set without any regard to the position respect to the observer and the galactic center . all the previous works have treated the data as a whole set , and we show that in this way important information could have been or might be overlooked in the future . the region that shows a distribution of gpa with some possible non - random characteristics is the region with center towards the galactic center . in the other regions , the low number of objects precludes detection of any clear trend , although we can rule out any strong correlation that involves the majority of the pne in the sample . the distribution of gpa of the regions are presented in fig . [ fotito3 ] . if the distribution of the * n * objects of the sample in each sector were random , with the same probability of finding an elongated object with its axis towards any pa , when * j * angular boxes are defined in the range @xmath15 $ ] , it could be expected an average population of * n / j * objects per bin . the frequency of all the possible values would be an angular positions distribution , with a dispersion @xmath18 . in table 2 we show the expected average per angular bin and the dispersion for each region , and the number of peaks observed above @xmath19 and @xmath20 . the barycenters calculated for each peak , determined within a window of @xmath21 , are summarized in table 2 . a thorough test can be performed by applying a simple binomial test , as proposed by siegel ( 1956 ) and applied by hutsemkers ( 1998 ) . it gives the probability @xmath22 that a random distribution has @xmath23 angles ( of a set of @xmath24 objects ) in the interval @xmath25 $ ] . in our case @xmath26 is the peak barycenter . such probability is defined as : @xmath27 table 3 shows the probability of randomness for each observed peak using @xmath28 and @xmath29 ( note that when @xmath28 the probability that the right peak of region i is random , is only @xmath30 ) . in region i , the distribution has two peaks with a separation of @xmath31 ( or @xmath0 in the opposite sense ) in the directions defined by the peaks . _ it is important to remark that the separate subsamples of pne from dss2 , cam98 and mash each one shows at the region i , an apparent peak in @xmath32 . _ following the clue provided by some similarity of the results for regions iii and iv , we added the distributions found for regions i and ii ( fig . [ fotito4]a ) and for regions iii and iv ( fig . [ fotito4]b ) . the two distributions in fig . [ fotito4]b show right peak over @xmath33 and left peak over @xmath34 , in both cases with a peak separation of @xmath35 . here the separation is considered as the minimum angular distance between both peaks . moreover as both distributions are in opposite sides of the sky , is natural to measure the separation in opposite sense . it should be noted that the position angles are being measured with the nesw convention , without any regard to the absolute orientation of the objects respect to the observer . then if some absolute orientation would be common to most of the objects and considering that the observer is in the center of the sky band over which the objects appear projected , the sense of measuring of the position angles in opposite regions ( e.g. i and ii ) should be inverted by making * 180 - gpa * in the opposite regions . coincidentally , the transformation of the data in regions iii+iv ( fig . [ fotito4]b ) gives a distribution resembling that of regions i+ii : a narrow main peak and a possible secondary one @xmath35 before . we also carried out the kolmogorov - smirnov ( k - s ) test to have an alternative evaluation of the randomness of the observed distribution . this requires the calculation of @xmath36 : the absolute value of the maximum deviation between the observed ( @xmath37 with @xmath38 points ) and theoretical ( @xmath39 with @xmath40 points ) cumulative distribution function ( press et al . 1992 ) . the significance can be written as the following sum : @xmath41 where @xmath42 and @xmath43 that is the effective number of data points of the samples . this test gave us a probability , that our data derived from a random distribution , of 63% for region i , 53% for region ii , 93% for region iii and 28% for region iv . as an alternative we tried to perform a variation of k - s test : the kuiper ( k - p ) test which is more sensible than k - s test in some kind of circular distributions . to perform this test we have to calculate @xmath44 , which is the sum of the maximum distance of @xmath37 above and below @xmath39 . a good approximation for the significance is : @xmath45 where @xmath46 the result that we obtained with k - p test for our region i ( with the strongest signal ) is @xmath47 , which seems to be not a convincing one . these tests could not be fully applicable to our kind of data : caution should be taken when applying all varieties of k - s test , because they lack the ability to discriminate some kind of distributions . for example , we can consider a probability distribution showing a narrow hole within which the randomness probability falls to zero . the existence of even one data point within the hole would rule out such distribution ( because of its cumulative nature , press et al . 1992 ) , the k - s test would require many data points in the narrow hole before signaling a discrepancy . to probe this kind of behavior , we generated a random distribution of 72 gpa and after that we added 10 gpa values distributed in the first bin ( fig . [ fotito5 ] ) . this distribution has an average of 9.1 gpa per bin and a dispersion of @xmath48 . in this way , the first bin shows a signal over @xmath49 . now if we apply both tests to this artificial sample we obtain the next probabilities , that this data follow a random distribution : @xmath50 and @xmath51 ( fig . [ fotito6 ] ) . we have verified that there is no way to generate 67 cases and not even 17 cases of a @xmath49 peak in 100 fully random samples . therefore we conclude that although k - s and k - p tests are well suited to analyze non random global trends in samples , they are strongly insensitive to the presence of a local non - random feature . we studied the orientation of all pne of our sample ( without separating it in regions ) and we do not see a clear preferential orientation of long axis of pn respect to the galactic plane . this distribution of gpa is similar to that observed by cam98 . we tried to double check our results by applying our analysis to the data published by cam98 , performing a suitable analysis to take over the ambiguity of the position angles ( assigned to the range @xmath52 $ ] ) . more precisely , the sample of cam98 includes 208 objects ( counting the object m2 - 55 in only one list ) ; 69 of them are in both samples ( the object that cam98 put in the list of ellipticals as pc 4 , is in fact pb 4 ) ; 85 are outside our four regions ; 50 were not measured because they are too small ; have strange morphology or circular appearance ; and 4 objects are not true pn ( bl cru , crl 2688 , m 1 - 91 , m 1 - 92 ) . 75@xmath53 of the 69 objects that are in both samples , have their gpa in good agreement ( @xmath54 ) with our measurement . moreover , the distribution of gpa from those objects from cam98 whose coordinates are in region i , shows a clear peak over 3 @xmath55 ( fig . [ fotito7 ] , 48 objects ) , whose barycenter is in @xmath56 , a position similar , within the uncertainties , to the main peak in our fig . [ fotito3 ] . this feature could not be detected by cam98 due to the way they calculated the final angles to plot . the differences could be mainly caused by the fact that they employed deep narrow - band images , as shown in fig . [ fotito8 ] . this figure shows an example of the pn th2-a dss image through r broad - band filter and 1.5 resolution . the epa was measured following the ellipsoid apparent major axis . for comparison , [ fotito8]-right shows the same object imaged with gmos at gemini - s 8 m telescope , through a [ oiii]@xmath575007 filter and with 0.7 seeing ( from daz et al . , in preparation ) . the presence of faint blue emission knots can change the major axis determination from the ellipsoid maximum diameter to an approximately perpendicular angle . therefore it can be expected that the measured geometrical properties of the objects change as distinct ionization layers are imaged through different filters , more precisely , the left peak ( e.g. we 1 - 4 ) in our broad band study could be arose in the measuring of the pn equatorial belt apparent axis instead of the faint external envelope . this should not occur in the deep narrow band images of the cam98 sample and could explain the absence of the second peak in that sample . nevertheless , the agreement between the gpa measurements in both samples is very good and the presence of the second peak can just be ascribed to the ability to detect the outer fainter details in the deep narrow band images , which are usually perpendicular to the bright equatorial belt structures more easily detected in the dss imagery . an interesting test to perform is to check if the distribution of gpa in the four regions has a distance modulation and consider objects statistically far and close . as a first order approach we separated objects in each region by its angular size : an angular diameter of 35@xmath3 divided the sample in two halves , objects with large angular size were considered closer than the ones with small angular size . assuming an average radius of 0.1pc the separating distance is 1.2kpc . the results that we obtained in the four regions show that the distribution of gpa for larger pne has the same shape than the smaller pne . so there is not evidence about a distance effect . following the same open minded search we tried to relate the pa of the long axis of pne with respect to the gould belt , the plane defined by nearby stars of young population , mainly o and b stars ( cameron et al . 1994 ) . the result found shows a noisy distribution of pa . besides , several studies have related bipolar pne with binary progenitors ( bond & livio 1990 ) but the antecedents about the orientations of the orbits of binary stars do not provide any comparable result . notwithstanding it must be remarked that a thorough analysis should be made , keeping in mind the new geometric considerations carried out in this work . we could also consider the case in which only the disk population of the pne towards the galactic center would show a preferred orientation . first , we can test the robustness of the result against the possibility of significant contamination of the pne sample toward the galactic center caused by a random orientation of bulge pne . if the bulge pne contamination were as high as 50% , then more than half of the disk pne would need to have a preferred orientation near to the galactic plane , to be detected and identified in the epa distribution . moreover , it would be necessary to disentangle the bulge objects by their radial velocities in order to verify if the preferred orientation for the disk pne is actually much larger than the one reported here . regarding the physical causes , melnick & harwitt ( 1974 ) mentioned the compression of the pne shells resulting from motion through the ism is seldom observed , remarking that the time scale of the pn expansion ( @xmath58years ) would be large enough to show systematic off - centering of the progenitor stars , which heretofore has been detected in a few lopsided objects ( borkowski , sarazin & soker 1990 ) and could be a dominant factor at the very faintest outer envelopes where the nebular density falls below a critical limit of n@xmath59@xmath60 . undoubtedly the most considered hypothesis has been that the pne could eventually expand more in the direction of the ambient or galactic magnetic field field force lines , which are approximately deployed along the milky way plane ( e.g. phillips 1997 ) . the typical energy density of the interstellar magnetic field is lower than @xmath61erg@xmath60 , whereas the energy density ( thermal , excitation and ionization ) is usually larger than @xmath62erg@xmath60 , consequently a correlation with any observable quantity related with galactic magnetic fields ( diaz & weidmann 2008 , in preparation ) would imply that this are at least one or two orders of magnitude higher in some regions towards the galactic center . the data on pa of planetary nebulae show that global preferred orientations are not dominant , even considering a new approach that takes into account the three - dimensional problem . besides the conclusions hold by the results , we presented here some important geometrical and analytical considerations that have been overlooked by the previous works on the subject . it is worth to mention that there could be a preferred orientation of some pne in the zones near to the galactic center , and the corresponding distribution of galactic position angles could have a peak not exactly aligned with the milky way plane : the pne towards the galactic center have an orientation distribution with a possible narrow peak near to the galactic plane ( @xmath32 ) , with a randomness probability as small as @xmath63 . we remark that we did not exclude any object within the selection criteria and galactic coordinates in the range @xmath64 and @xmath65 , and no special attention was given to the possibility that some objects may belong to the galactic bulge population and the consequences over the detected trend are unknown beyond the qualitative discussion of the previous section . furthermore a larger and deeper galactic center sample , optimally observed at nir wavelengths , should be analyzed in order to thoroughly assess the reality of this preferred orientation . the possible implications are of broad astrophysical interest and we hope that they stimulate more detailed studies , in particular , larger samples of galactic pne towards the galactic center should be studied as soon as they become available . this work was partially supported by gustavo carranza through the agencia cordoba ciencia and conicet of argentina . we acknowledge the referee for his useful suggestions , in particular the hint to use the mash survey . rd acknowledges fruitful discussions about the original manuscript with romano corradi in 2005 , and thanks percy gomez for a critical reading of the manuscript . this research has made use of aladin and the multimission archive at the space telescope science institute ( mast ) . stsci is operated by the association of universities for research in astronomy , inc . , under nasa contract nas5 - 26555 . support for mast for non - hst data is provided by the nasa office of space science via grant nag5 - 7584 and by other grants and contracts . the gemini observatory is operated by the association of universities for research in astronomy , inc . , under a cooperative agreement with the nsf on behalf of the gemini partnership : nsf ( usa ) , pparc ( united kingdom ) , nrc ( canada ) , arc ( australia ) , conicet ( argentina ) , cnpq ( brazil ) and conicyt ( chile ) . lccccc g000.0 - 01.8 & 0 & -1.8 & 163 & 42 & mash + m 2 - 19 & 0.2 & -1.9 & 110 & 170 & cam98 + g000.2 - 03.4 & 0.2 & -3.4 & 100 & 160 & mash + ic 4634 & 0.3 & 12.2 & 153 & 27 & dss2 + g000.3 + 07.3 & 0.3 & 7.3 & 90 & 145 & mash + g000.3 + 04.5 & 0.3 & 4.5 & 160 & 36 & mash + k 1 - 4 & 1 & 1.9 & 155 & 33 & dss2 + g001.2 - 05.6 & 1.2 & -5.6 & 138 & 19 & mash + he 2 - 262 & 1.3 & 2.2 & 21 & 78 & hst + g001.5 - 02.4 & 1.5 & -2.4 & 60 & 120 & mash + swst 1 & 1.6 & -6.7 & 126 & 8 & hst + h 1 - 55 & 1.7 & -4.5 & 90 & 151 & hst + g001.8 - 05.0 & 1.8 & -5 & 55 & 116 & mash + g001.9 + 02.1 & 1.9 & 2.1 & 80 & 138 & mash + ic 4776 & 2 & -13.4 & 34 & 100 & cam98 + g002.0 + 06.6 & 2 & 6.6 & 60 & 116 & mash + g002.0 + 01.5 & 2 & 1.5 & 115 & 173 & mash + g002.0 - 03.2 & 2 & -3.2 & 95 & 155 & mash + g002.1 - 02.4 & 2.1 & -2.4 & 130 & 10 & mash + g002.1 - 02.8 & 2.1 & -2.8 & 15 & 75 & mash + h 1 - 54 & 2.1 & -4.2 & 137 & 18 & hst + g002.2 + 05.8 & 2.2 & 5.8 & 5 & 61 & mash + g002.2 - 01.2 & 2.2 & -1.2 & 40 & 99 & mash + g002.3 + 01.7 & 2.3 & 1.7 & 35 & 93 & mash + cn 1 - 5 & 2.3 & -9.5 & 162 & 45 & hst + ngc 6369 & 2.4 & 5.9 & 135 & 44 & dss2 + g002.4 + 03.5 & 2.4 & 3.5 & 46 & 103 & mash + g002.4 + 01.1 & 2.4 & 1.1 & 62 & 120 & mash + g002.4 - 05.0 & 2.4 & -5 & 15 & 76 & mash + h 2 - 37 & 2.4 & -3.4 & 71 & 131 & hst + g002.5 + 04.8 & 2.5 & 4.8 & 16 & 73 & mash + m 1 - 42 & 2.7 & -4.8 & 21 & 83 & cam98 + g002.8 - 04.1 & 2.8 & -4.1 & 65 & 126 & mash + te 1567 & 2.8 & 1.8 & 0 & 58 & hst + g002.9 - 03.0 & 2.9 & -3 & 15 & 75 & mash + g003.0 - 01.7 & 3 & -1.7 & 140 & 20 & mash + g003.1 + 05.2 & 3.1 & 5.2 & 135 & 12 & mash + g003.1 - 01.6 & 3.1 & -1.6 & 100 & 160 & mash + hb 4 & 3.2 & 2.9 & 140 & 18 & hst + g003.3 - 01.6 & 3.3 & -1.6 & 30 & 90 & mash + ic 4673 & 3.5 & -2.4 & 126 & 7 & dss2 + g003.5 + 04.5 & 3.5 & 4.5 & 35 & 92 & mash + g003.5 + 02.6 & 3.5 & 2.6 & 45 & 103 & mash + g003.6 - 03.0 & 3.6 & -3 & 145 & 25 & mash + h 2 - 15 & 3.8 & 5.3 & r & ... & hst + h 1 - 59 & 3.9 & -4.4 & 71 & 132 & hst + g004.0 - 02.6 & 4 & -2.6 & 115 & 175 & mash + g004.0 - 02.7 & 4 & -2.7 & 92 & 152 & mash + g004.1 - 03.3 & 4.1 & -3.3 & 85 & 145 & mash + g004.2 - 02.5 & 4.2 & -2.5 & 140 & 20 & mash + g004.5 + 06.0 & 4.5 & 6 & 60 & 117 & mash + h 2 - 12 & 4.5 & 6.8 & ... & ... & hst + g004.8 - 01.1 & 4.8 & -1.1 & 32 & 92 & mash + h 2 - 25 & 4.9 & 2.1 & 41 & 99 & hst + m 1 - 25 & 4.9 & 4.9 & 40 & 97 & hst + g005.0 + 02.2 & 5 & 2.2 & 166 & 44 & mash + g005.4 - 03.4 & 5.4 & -3.4 & 168 & 49 & mash + g005.9 - 09.8 & 5.9 & -9.8 & 140 & 24 & mash + m 1 - 28 & 6 & 3.1 & 14 & 73 & dss2 + g006.1 + 03.8 & 6.1 & 3.8 & 0 & 58 & mash + g006.1 + 01.5 & 6.1 & 1.5 & 65 & 124 & mash + m 1 - 20 & 6.2 & 8.4 & 102 & 159 & hst + g006.3 + 01.7 & 6.3 & 1.7 & 170 & 49 & mash + g006.4 - 03.4 & 6.4 & -3.4 & 147 & 28 & mash + g006.4 - 05.5 & 6.4 & -5.5 & 140 & 22 & mash + g006.5 + 08.7 & 6.5 & 8.7 & 54 & 111 & mash + g006.5 - 03.9 & 6.5 & -3.9 & 20 & 81 & mash + m 3 - 15 & 6.8 & 4.2 & 122 & 0 & hst + g007.1 + 07.3 & 7.1 & 7.3 & 136 & 13 & mash + g007.1 + 04.9 & 7.1 & 4.9 & 118 & 176 & mash + g007.1 - 05.0 & 7.1 & -5 & 65 & 127 & mash + g007.3 + 01.7 & 7.3 & 1.7 & 30 & 89 & mash + g007.4 + 01.7 & 7.4 & 1.7 & 13 & 72 & mash + m 2 - 34 & 7.8 & -3.7 & 177 & 59 & dss2 + g007.8 + 04.3 & 7.8 & 4.3 & 80 & 138 & mash + h 1 - 65 & 7.9 & -4.4 & r & ... & hst + ngc 6445 & 8 & 3.9 & 149 & 110 & dss2 + m 1 - 40 & 8.3 & -1.1 & 44 & 105 & cam98 + g008.3 + 09.6 & 8.3 & 9.6 & 120 & 177 & mash + he 2 - 260 & 8.3 & 6.9 & 85 & 143 & hst + g008.4 - 02.8 & 8.4 & -2.8 & 85 & 146 & mash + g008.7 - 04.2 & 8.7 & -4.2 & 135 & 17 & mash + g009.0 - 02.2 & 9 & -2.2 & 15 & 76 & mash + g009.0 - 02.4 & 9 & -2.4 & 18 & 79 & mash + g009.4 - 01.2 & 9.4 & -1.2 & 53 & 114 & mash + ngc 6309 & 9.6 & 14.8 & 56 & 114 & dss2 + a 41 & 9.6 & 10.5 & 143 & 21 & dss2 + g009.8 - 01.1 & 9.8 & -1.1 & 10 & 71 & mash + g009.9 + 04.5 & 9.9 & 4.5 & 60 & 119 & mash + g010.0 - 01.5 & 10 & -1.5 & 65 & 126 & mash + ngc 6537 & 10.1 & 0.7 & 38 & 99 & dss2 + g010.2 + 00.3 & 10.2 & 0.3 & 50 & 110 & mash + m 2 - 9 & 10.8 & 18.1 & 179 & 57 & dss2 + ic 4732 & 10.8 & -6.5 & r & ... & hst + ngc 6578 & 10.8 & -1.8 & 144 & 25 & hst + g011.0 + 01.4 & 11 & 1.4 & 60 & 120 & mash + m 2 - 13 & 11.1 & 11.5 & 40 & 98 & cam98 + deht 10 & 11.4 & 17.9 & 175 & 53 & dss2 + ngc 6567 & 11.8 & -0.7 & 107 & 168 & hst + g011.9 + 07.3 & 11.9 & 7.3 & 165 & 44 & mash + g012.1 + 02.8 & 12.1 & 2.8 & 150 & 30 & mash + pm 1 - 188 & 12.2 & 4.9 & r & ... & hst + g012.5 + 04.3 & 12.5 & 4.3 & 113 & 173 & mash + g013.1 + 05.0 & 13.1 & 5 & 85 & 145 & mash + m 1 - 33 & 13.1 & 4.2 & 30 & 90 & hst + g013.6 - 04.6 & 13.6 & -4.6 & 90 & 152 & mash + we 4 - 5 & 13.7 & -15.3 & 141 & 27 & dss2 + sawe 3 & 13.8 & -2.8 & 133 & 15 & dss2 + g014.6 + 02.3 & 14.6 & 2.3 & 50 & 110 & mash + g014.6 + 01.0 & 14.6 & 1 & 175 & 56 & mash + a 44 & 15.6 & -3 & 131 & 13 & dss2 + m 1 - 39 & 15.9 & 3.4 & 80 & 140 & hst + m 1 - 54 & 16 & -4.3 & 104 & 167 & dss2 + g016.0 - 07.6 & 16 & -7.6 & 135 & 18 & mash + g016.3 - 02.3 & 16.3 & -2.3 & 168 & 50 & mash + g016.4 - 00.9 & 16.4 & -0.9 & 139 & 20 & mash + g016.6 + 03.1 & 16.6 & 3.1 & 90 & 151 & mash + g018.0 - 02.2 & 18 & -2.2 & 56 & 118 & mash + g018.5 - 01.6 & 18.5 & -1.6 & 160 & 42 & mash + deht 3 & 19.4 & -13.6 & 13 & 73 & dss2 + cts 1 & 19.8 & 5.6 & 102 & 163 & cam98 + g020.4 + 02.2 & 20.4 & 2.2 & 115 & 176 & mash + m 1 - 51 & 20.9 & -1.1 & 29 & 92 & cam98 + m 3 - 55 & 21.7 & -0.6 & 57 & 120 & cam98 + m 3 - 28 & 21.8 & -0.5 & 2 & 64 & dss2 + m 1 - 57 & 22.1 & -2.4 & 137 & 20 & cam98 + m 1 - 58 & 22.1 & -3.2 & 78 & 140 & hst + mac 1 - 13 & 22.5 & 1 & 26 & 88 & dss2 + g023.4 + 00.7 & 23.4 & 0.7 & 145 & 27 & mash + m 1 - 59 & 23.9 & -2.3 & 116 & 179 & cam98 + m 2 - 40 & 24.1 & 3.8 & 80 & 142 & cam98 + m 4 - 9 & 24.2 & 5.9 & 167 & 50 & dss2 + pe 1 - 17 & 24.3 & -3.3 & 45 & 108 & cam98 + g024.4 - 03.5 & 24.4 & -3.5 & 60 & 122 & mash + a 60 & 25 & -11.7 & 103 & 166 & dss2 + ic 1295 & 25.4 & -4.7 & 69 & 132 & cam98 + ngc 6818 & 25.8 & -17.9 & 13 & 79 & cam98 + pe 1 - 14 & 25.9 & -0.9 & 45 & 108 & cam98 + g026.2 - 03.4 & 26.2 & -3.4 & 7 & 69 & mash + g026.4 + 02.7 & 26.4 & 2.7 & 25 & 87 & mash + g026.9 - 00.7 & 26.9 & -0.7 & 56 & 118 & mash + a 49 & 27.3 & -3.4 & 31 & 94 & cam98 + deht 2 & 27.6 & 16.9 & 44 & 107 & dss2 + g027.6 - 00.8 & 27.6 & -0.8 & 40 & 102 & mash + g027.8 + 02.7 & 27.8 & 2.7 & 50 & 112 & mash + wesb 3 & 28 & 10.3 & 145 & 31 & dss2 + pe 1 - 20 & 28.2 & -4 & 12 & 75 & cam98 + k 3 - 2 & 28.6 & 5.2 & ... & ... & hst + g028.7 - 03.2 & 28.7 & -3.2 & 40 & 102 & mash + a 48 & 29 & 0.5 & 37 & 99 & dss2 + ngc 6751 & 29.2 & -5.9 & 81 & 144 & dss2 + g029.8 + 00.5 & 29.8 & 0.5 & 20 & 82 & mash + a 68 & 60 & -4.3 & 13 & 72 & cam98 + ngc 6886 & 60.1 & -7.7 & 54 & 111 & cam98 + k 3 - 45 & 60.5 & -0.3 & 18 & 78 & cam98 + ngc 6853 & 60.8 & -3.7 & 129 & 6 & dss2 + he 2 - 437 & 61.3 & 3.6 & 77 & 138 & dss2 + m 1 - 91 & 61.4 & 3.6 & 77 & 138 & cam98 + ngc 6905 & 61.4 & -9.6 & 161 & 36 & dss2 + m 2 - 48 & 62.4 & -0.3 & 67 & 125 & dss2 + ngc 6720 & 63.1 & 14 & 57 & 123 & dss2 + m 1 - 92 & 64.1 & 4.3 & 130 & 11 & cam98 + bd+30 3639 & 64.8 & 5 & r & ... & hst + we 1 - 9 & 65.1 & -3.5 & 70 & 126 & dss2 + he 1 - 6 & 65.2 & -5.7 & 122 & 177 & dss2 + he 2 - 459 & 68.4 & -2.7 & 57 & 113 & hst + m 1 - 75 & 68.8 & 0 & 152 & 28 & dss2 + k 3 - 46 & 69.2 & 3.8 & 110 & 168 & dss2 + ngc 6894 & 69.4 & -2.6 & 145 & 20 & dss2 + k 3 - 58 & 69.6 & -3.9 & 83 & 137 & dss2 + m 3 - 35 & 71.6 & -2.3 & 48 & 103 & cam98 + k 3 - 57 & 72.1 & 0.1 & 33 & 90 & cam98 + a 74 & 72.7 & -17.1 & 52 & 99 & dss2 + k 3 - 76 & 73 & -2.4 & 134 & 9 & cam98 + gm 1 - 11 & 73 & -2.2 & 65 & 119 & dss2 + ngc 6881 & 74.5 & 2.1 & 139 & 16 & cam98 + anon & 75.6 & 4.3 & 109 & 166 & dss2 + a 69 & 76.3 & 1.2 & 54 & 108 & dss2 + dd 1 & 78.6 & 5.2 & 81 & 137 & dss2 + m 4 - 17 & 79.6 & 5.8 & 118 & 174 & dss2 + crl 2688 & 80.1 & -6.5 & 14 & 63 & cam98 + a 78 & 81.2 & -14.9 & 146 & 9 & dss2 + ngc 6884 & 82.1 & 7.1 & 174 & 51 & hst + k 4 - 55 & 84.2 & 1 & 85 & 135 & dss2 + a 71 & 84.9 & 4.4 & 15 & 68 & dss2 + hu 1 - 2 & 86.5 & -8.8 & 129 & 172 & cam98 + we 2 - 245 & 87.4 & -3.8 & 137 & 1 & dss2 + ngc 7048 & 88.7 & -1.6 & 15 & 60 & dss2 + ngc 7026 & 89 & 0.3 & 4 & 51 & dss2 + sh 1 - 89 & 89.8 & -0.6 & 50 & 95 & dss2 + k 3 - 84 & 91.6 & -4.8 & 3 & 45 & cam98 + we 1 - 11 & 91.6 & 1.8 & 49 & 95 & dss2 + k 3 - 79 & 92.1 & 5.8 & 132 & 1 & dss2 + m 1 - 79 & 93.3 & -2.4 & 88 & 129 & dss2 + ngc 7008 & 93.4 & 5.4 & 38 & 86 & dss2 + k 3 - 83 & 94.5 & -0.8 & 100 & 142 & cam98 + a 73 & 95.2 & 7.8 & 38 & 87 & dss2 + k 3 - 61 & 96.3 & 2.3 & 164 & 27 & cam98 + m 2 - 50 & 97.6 & -2.4 & 52 & 90 & cam98 + me 2 - 2 & 100 & -8.8 & r & ... & hst + ic 5217 & 100.6 & -5.4 & 90 & 122 & cam98 + a 75 & 101.8 & 8.7 & 94 & 137 & dss2 + a 80 & 102.8 & -5 & 19 & 48 & dss2 + a 79 & 102.9 & -2.3 & 20 & 51 & dss2 + m 2 - 51 & 103.2 & 0.6 & 151 & 4 & dss2 + m 2 - 52 & 103.7 & 0.4 & 116 & 148 & dss2 + klw 8 & 104.1 & -1.4 & 40 & 70 & dss2 + ngc 7139 & 104.1 & 7.9 & 47 & 86 & dss2 + m 2 - 53 & 104.4 & -1.6 & 95 & 124 & dss2 + ngc 7354 & 107.8 & 2.3 & 17 & 45 & dss2 + iras22568 + 6141 & 110.1 & 1.9 & 121 & 146 & cam98 + k 1 - 20 & 110.6 & -12.9 & 154 & 169 & dss2 + kjpn 6 & 111.2 & 7 & 4 & 32 & cam98 + kjpn 8 & 112.5 & -0.1 & 72 & 91 & cam98 + k 3 - 88 & 112.5 & 3.7 & 45 & 66 & dss2 + a 84 & 112.9 & -10.2 & 172 & 6 & dss2 + a 83 & 113.6 & -6.9 & 34 & 47 & dss2 + a 82 & 114 & -4.6 & 46 & 59 & dss2 + we 2 - 262 & 116 & 0.1 & 172 & 4 & dss2 + m 2 - 55 & 116.2 & 8.5 & 38 & 55 & dss2 + m 2 - 56 & 118 & 8.4 & 87 & 100 & cam98 + a 86 & 118.7 & 8.2 & 115 & 125 & dss2 + bv 5 - 1 & 119.3 & 0.3 & 53 & 59 & dss2 + hu 1 - 1 & 119.6 & -6.1 & 9 & 14 & cam98 + ngc 40 & 120 & 9.8 & 15 & 24 & cam98 + k 3 - 64 & 151.4 & 0.5 & 120 & 77 & cam98 + ic 351 & 159 & -15.1 & 3 & 143 & cam98 + ic 2149 & 166.1 & 10.4 & 73 & 11 & dss2 + crl 618 & 166.4 & -6.5 & 99 & 49 & dss2 + pu 2 & 173.5 & 3.2 & 47 & 169 & dss2 + h 3 - 29 & 174.2 & -14.6 & 158 & 107 & dss2 + pu 1 & 181.5 & 0.9 & 50 & 170 & dss2 + wesb 2 & 183.8 & 5.5 & 33 & 152 & dss2 + ngc 2371 & 189.1 & 19.8 & 129 & 61 & dss2 + m 1 - 7 & 189.8 & 7.8 & 153 & 90 & dss2 + j 320 & 190.3 & -17.8 & 174 & 117 & dss2 + hawe 8 & 192.5 & 7.2 & 134 & 71 & dss2 + j 900 & 194.2 & 2.6 & 83 & 21 & dss2 + ncg 2022 & 196.6 & -11 & 32 & 152 & dss2 + a 14 & 197.8 & -3.4 & 175 & 114 & dss2 + a 12 & 198.6 & -6.3 & 15 & 134 & dss2 + a 19 & 200.7 & 8.4 & 65 & 2 & dss2 + we 1 - 4 & 201.9 & -4.7 & 104 & 42 & dss2 + a 13 & 204 & -8.5 & 47 & 165 & dss2 + k 3 - 72 & 204.8 & -3.6 & 139 & 77 & dss2 + a 21 & 205.1 & 14.2 & 136 & 72 & dss2 + m 3 - 2 & 240.3 & -7.6 & 42 & 160 & dss2 + m 3 - 4 & 241 & 2.3 & 141 & 83 & dss2 + m 3 - 1 & 242.6 & -11.6 & 143 & 78 & dss2 + g242.6 - 04.4 & 242.6 & -4.4 & 98 & 37 & mash + ngc 2452 & 243.3 & -1.1 & 116 & 57 & dss2 + a 29 & 244.5 & 12.5 & 133 & 80 & dss2 + g244.6 - 00.3 & 244.6 & -0.3 & 0 & 121 & mash + g249.8 + 07.1 & 249.8 & 7.1 & 55 & 2 & mash + a 26 & 250.3 & 0.1 & 141 & 85 & dss2 + g250.3 - 05.4 & 250.3 & -5.4 & 148 & 88 & mash + g250.4 - 01.3 & 250.4 & -1.3 & 105 & 48 & mash + k 1 - 21 & 251.1 & -1.6 & 160 & 103 & dss2 + k 1 - 1 & 252.6 & 4.4 & 36 & 163 & dss2 + k 1 - 2 & 253.5 & 10.7 & 101 & 52 & dss2 + vbrc 1 & 257.5 & 0.6 & 118 & 65 & dss2 + he 2 - 9 & 258.1 & -0.4 & 31 & 157 & hst + he 2 - 11 & 259.1 & 0.9 & 137 & 86 & dss2 + he 2 - 15 & 261.6 & 3 & 33 & 164 & dss2 + ngc 2818 & 261.9 & 8.5 & 90 & 45 & dss2 + he 2 - 7 & 264.1 & -8.1 & 161 & 105 & dss2 + wray 17 - 20 & 264.4 & -3.6 & 22 & 150 & dss2 + ngc 2792 & 265.7 & 4.1 & 146 & 100 & dss2 + g266.8 - 04.2 & 266.8 & -4.2 & 125 & 73 & mash + g267.5 + 04.6 & 267.5 & 4.6 & 142 & 97 & mash + g268.6 + 05.0 & 268.6 & 5 & 50 & 6 & mash + g269.3 - 00.8 & 269.3 & -0.8 & 80 & 32 & mash + ngc 3132 & 272.1 & 12.3 & 167 & 131 & dss2 + he 2 - 18 & 273.2 & -3.7 & 75 & 29 & dss2 + g273.2 - 00.3 & 273.2 & -0.3 & 123 & 79 & mash + lo 4 & 274.3 & 9.1 & 46 & 11 & dss2 + he 2 - 37 & 274.6 & 3.5 & 119 & 80 & dss2 + g274.8 - 05.7 & 274.8 & -5.7 & 57 & 9 & mash + pb 4 & 275 & -4.1 & 35 & 170 & dss2 + g275.0 + 01.5 & 275 & 1.5 & 13 & 152 & mash + g275.1 - 03.5 & 275.1 & -3.5 & 137 & 92 & mash + he 2 - 25 & 275.2 & -3.7 & 14 & 148 & cam98 + g275.6 - 00.5 & 275.6 & -0.5 & 128 & 86 & mash + he 2 - 29 & 275.8 & -2.9 & 55 & 12 & dss2 + g276.1 - 03.3 & 276.1 & -3.3 & 130 & 86 & mash + ngc 2899 & 277.1 & -3.8 & 117 & 73 & cam98 + g277.4 - 00.7 & 277.4 & -0.7 & 138 & 98 & mash + wray 17 - 31 & 277.7 & -3.5 & 90 & 48 & dss2 + g278.3 - 04.3 & 278.3 & -4.3 & 30 & 167 & mash + he 2 - 32 & 278.5 & -4.5 & 159 & 117 & dss2 + vbrc 3 & 279 & -3.2 & 56 & 16 & dss2 + g279.1 - 03.1 & 279.1 & -3.1 & 57 & 16 & mash + he 2 - 36 & 279.6 & -3.1 & 144 & 105 & dss2 + g279.6 + 02.2 & 279.6 & 2.2 & 141 & 105 & mash + g281.3 - 07.1 & 281.3 & -7.1 & 143 & 100 & mash + g282.5 - 02.1 & 282.5 & -2.1 & 147 & 111 & mash + he 2 - 48 & 282.9 & 3.8 & 129 & 99 & dss2 + hf 4 & 283.9 & -1.8 & 13 & 160 & dss2 + eso 215 - 04 & 283.9 & 9.7 & 143 & 118 & dss2 + g284.5 + 03.8 & 284.5 & 3.8 & 127 & 98 & mash + g284.6 - 04.5 & 284.6 & -4.5 & 10 & 154 & mash + ic 2553 & 285.4 & -5.3 & 7 & 151 & cam98 + he 2 - 52 & 285.5 & 1.5 & 105 & 76 & cam98 + g285.5 - 03.3 & 285.5 & -3.3 & 83 & 50 & mash + ic 2448 & 285.7 & -14.9 & 137 & 88 & dss2 + he 2 - 47 & 285.7 & -2.7 & ... & ... & hst + wray 17 - 40 & 286.2 & -6.9 & 102 & 67 & dss2 + g286.3 - 03.1 & 286.3 & -3.1 & 145 & 113 & mash + g286.3 - 00.7 & 286.3 & -0.7 & 130 & 100 & mash + hf 38 & 288.4 & 0.3 & 94 & 69 & dss2 + g288.4 - 01.8 & 288.4 & -1.8 & 107 & 80 & mash + hf 39 & 288.9 & -0.8 & 43 & 18 & dss2 + he 2 - 57 & 289.6 & -1.6 & 119 & 95 & dss2 + hf 48 & 290.1 & -0.4 & 136 & 113 & dss2 + g291.3 + 03.7 & 291.3 & 3.7 & 55 & 36 & mash + g291.3 + 08.4 & 291.3 & 8.4 & 175 & 158 & mash + he 2 - 64 & 291.7 & 3.7 & 81 & 62 & cam98 + g291.9 - 04.0 & 291.9 & -4 & 44 & 21 & mash + g292.4 + 00.8 & 292.4 & 0.8 & 153 & 134 & mash + wray 16 - 93 & 292.7 & 1.9 & 134 & 117 & dss2 + he 2 - 67 & 292.8 & 1.1 & 120 & 102 & cam98 + g292.8 + 00.6 & 292.8 & 0.6 & 168 & 150 & mash + he 2 - 70 & 293.6 & 1.2 & 137 & 121 & dss2 + lo 6 & 294.1 & 14.4 & 150 & 140 & dss2 + ngc 3918 & 294.6 & 4.7 & 138 & 126 & dss2 + he 2 - 72 & 294.9 & -0.6 & 101 & 87 & dss2 + he 2 - 71 & 296.5 & -6.9 & r & ... & hst + ngc 3195 & 296.6 & -20 & 13 & 155 & cam98 + g297.6 - 01.6 & 297.6 & -1.6 & 90 & 79 & mash + he 2 - 76 & 298.2 & -1.7 & 83 & 75 & dss2 + ngc 4071 & 298.3 & -4.8 & 43 & 34 & dss2 + g298.6 - 01.5 & 298.6 & -1.5 & 50 & 41 & mash + g298.7 - 07.5 & 298.7 & -7.5 & 158 & 147 & mash + k 1 - 23 & 299 & 18.4 & 3 & 179 & dss2 + g299.0 + 03.5 & 299 & 3.5 & 140 & 133 & mash + hatr 1 & 299.4 & -4.1 & 131 & 124 & dss2 + he 2 - 82 & 299.5 & 2.4 & 162 & 157 & dss2 + bl cru & 299.7 & 0.1 & 31 & 25 & cam98 + lo 9 & 330.2 & 5.9 & 82 & 120 & dss2 + he 2 - 153 & 330.6 & -2.1 & 41 & 86 & dss2 + he 2 - 159 & 330.6 & -3.6 & 173 & 39 & dss2 + g330.7 - 02.0 & 330.7 & -2 & 95 & 139 & mash + wray 16 - 189 & 330.9 & 4.3 & 72 & 112 & dss2 + pc 11 & 331.1 & -5.8 & r & ... & hst + g331.3 + 01.6 & 331.3 & 1.6 & 145 & 6 & mash + he 2 - 145 & 331.4 & 0.5 & 113 & 156 & dss2 + he 2 - 165 & 331.5 & -3.9 & 36 & 83 & dss2 + he 2 - 161 & 331.5 & -2.7 & 48 & 94 & dss2 + mz 3 & 331.7 & -1 & 10 & 55 & dss2 + he 2 - 164 & 332 & -3.3 & 82 & 129 & dss2 + g332.0 - 04.3 & 332 & -4.3 & 35 & 82 & mash + g332.5 - 02.2 & 332.5 & -2.2 & 117 & 163 & mash + hatr 6 & 332.8 & -16.4 & 55 & 119 & dss2 + he 2 - 152 & 333.4 & 1.1 & 150 & 14 & dss2 + hatr 3 & 333.4 & -4 & 157 & 26 & dss2 + mewe 1 - 6 & 334.3 & -1.4 & 125 & 172 & dss2 + ic 4642 & 334.3 & -9.3 & 143 & 18 & dss2 + g334.3 - 13.4 & 334.3 & -13.4 & 43 & 102 & mash + hatr 4 & 335.2 & -3.6 & 94 & 144 & dss2 + he 2 - 169 & 335.4 & -1.1 & 3 & 51 & dss2 + eso 330 - 02 & 335.4 & 9.2 & 4 & 45 & dss2 + ds 2 & 335.5 & 12.4 & 115 & 154 & dss2 + pc14 & 336.2 & -6.9 & 106 & 159 & cam98 + k 2 - 17 & 336.8 & -7.2 & 139 & 13 & dss2 + g337.0 + 08.4 & 337 & 8.4 & 0 & 41 & mash + g337.3 + 00.6 & 337.3 & 0.6 & 124 & 171 & mash + g337.8 - 04.1 & 337.8 & -4.1 & 50 & 101 & mash + g338.0 + 02.4 & 338 & 2.4 & 130 & 176 & mash + ngc 6326 & 338.1 & -8.3 & 54 & 110 & dss2 + he 2 - 155 & 338.8 & 5.6 & 61 & 106 & dss2 + g338.9 + 04.6 & 338.9 & 4.6 & 30 & 75 & mash + g339.1 + 00.9 & 339.1 & 0.9 & 164 & 32 & mash + g339.4 - 06.5 & 339.4 & -6.5 & 13 & 67 & mash + g340.0 + 02.9 & 340 & 2.9 & 0 & 47 & mash + sa 1 - 6 & 340.4 & -14.1 & 42 & 106 & dss2 + lo 11 & 340.8 & 12.3 & 61 & 104 & dss2 + lo 12 & 340.8 & 10.8 & 86 & 130 & dss2 + g340.9 + 03.7 & 340.9 & 3.7 & 141 & 8 & mash + ngc 6026 & 341.6 & 13.7 & 63 & 106 & dss2 + g341.7 + 02.6 & 341.7 & 2.6 & 68 & 116 & mash + g342.0 - 01.7 & 342 & -1.7 & 90 & 142 & mash + ngc 6072 & 342.1 & 10.8 & 71 & 116 & dss2 + sp 3 & 342.5 & -14.3 & 144 & 28 & dss2 + h 1 - 3 & 342.7 & 0.7 & 146 & 17 & dss2 + he 2 - 207 & 342.9 & -4.9 & 56 & 112 & dss2 + pe 1 - 8 & 342.9 & -2 & 61 & 114 & dss2 + suwt 3 & 343.6 & 3.7 & 77 & 127 & dss2 + g343.9 - 01.6 & 343.9 & -1.6 & 116 & 169 & mash + h 1 - 6 & 344.2 & -1.2 & 145 & 18 & dss2 + h 1 - 7 & 345.2 & -1.2 & 66 & 120 & cam98 + tc 1 & 345.2 & -8.8 & r & ... & hst + mewe 1 - 11 & 345.3 & -10.2 & 41 & 102 & dss2 + ic 4637 & 345.4 & 0.1 & 0 & 53 & dss2 + he 2 - 175 & 345.6 & 6.7 & 94 & 143 & dss2 + g345.8 + 02.7 & 345.8 & 2.7 & 58 & 109 & mash + vd 1 - 6 & 345.9 & 3 & 110 & 162 & dss2 + ic 4663 & 346.2 & -8.2 & 86 & 146 & dss2 + a 38 & 346.9 & 12.4 & 83 & 130 & dss2 + g347.0 + 00.3 & 347 & 0.3 & 58 & 111 & mash + g347.2 - 00.8 & 347.2 & -0.8 & 97 & 151 & mash + g347.4 + 01.8 & 347.4 & 1.8 & 161 & 33 & mash + ic 4699 & 348 & -13 & 28 & 92 & cam98 + g349.1 - 01.7 & 349.1 & -1.7 & 114 & 169 & mash + ngc 6337 & 349.3 & -1.1 & 139 & 15 & dss2 + ngc 6302 & 349.5 & 1.1 & 83 & 137 & dss2 + h 1 - 26 & 350.1 & -3.9 & 88 & 146 & dss2 + g350.9 - 02.9 & 350.9 & -2.9 & 53 & 110 & mash + h 2 - 1 & 350.9 & 4.4 & 7 & 59 & hst + g351.1 - 03.9 & 351.1 & -3.9 & 170 & 47 & mash + m 1 - 19 & 351.2 & 4.8 & 0 & 52 & hst + g352.2 + 02.4 & 352.2 & 2.4 & 96 & 150 & mash + k 2 - 16 & 352.9 & 11.4 & 141 & 12 & dss2 + g353.3 - 02.9 & 353.3 & -2.9 & 55 & 113 & mash + wray 16 - 411 & 353.7 & -12.8 & 87 & 152 & dss2 + g353.8 - 03.7 & 353.8 & -3.7 & 177 & 55 & mash + g353.9 - 05.8 & 353.9 & -5.8 & 36 & 96 & mash + g354.0 + 04.7 & 354 & 4.7 & 132 & 6 & mash + g354.0 - 00.8 & 354 & -0.8 & 108 & 165 & mash + g354.5 + 04.8 & 354.5 & 4.8 & 112 & 166 & mash + g354.5 - 03.9 & 354.5 & -3.9 & 31 & 90 & mash + g354.9 - 04.4 & 354.9 & -4.4 & 41 & 100 & mash + g355.0 + 05.8 & 355 & 5.8 & 25 & 79 & mash + g355.1 + 03.7 & 355.1 & 3.7 & 10 & 65 & mash + g355.3 + 05.2 & 355.3 & 5.2 & 10 & 64 & mash + g355.3 - 04.1 & 355.3 & -4.1 & 130 & 9 & mash + hf 2 - 1 & 355.4 & -4 & 40 & 99 & cam98 + m 3 - 14 & 355.4 & -2.4 & 164 & 43 & dss2 + g355.4 + 03.6 & 355.4 & 3.6 & 167 & 42 & mash + g355.6 + 04.1 & 355.6 & 4.1 & 24 & 79 & mash + g355.6 - 02.3 & 355.6 & -2.3 & 50 & 108 & mash + g355.9 + 04.1 & 355.9 & 4.1 & 45 & 100 & mash + g355.9 - 04.4 & 355.9 & -4.4 & 131 & 10 & mash + m 1 - 30 & 355.9 & -4.3 & 98 & 157 & hst + h 1 - 9 & 356 & 3.6 & ... & ... & hst + h 2 - 26 & 356.1 & -3.4 & 132 & 11 & hst + th 3 - 3 & 356.5 & 5.1 & 102 & 157 & dss2 + g356.5 + 02.2 & 356.5 & 2.2 & 65 & 121 & mash + g356.6 + 02.3 & 356.6 & 2.3 & 33 & 89 & mash + g356.6 - 01.9 & 356.6 & -1.9 & 153 & 31 & mash + h 1 - 39 & 356.6 & -4 & 31 & 90 & hst + m 2 - 24 & 356.9 & -5.8 & 62 & 123 & dss2 + m 3 - 7 & 357.1 & 3.6 & ... & ... & hst + g357.2 + 07.2 & 357.2 & 7.2 & 55 & 109 & mash + g357.3 + 01.3 & 357.3 & 1.3 & 50 & 107 & mash + trbr 4 & 357.6 & 1 & 26 & 84 & dss2 + g357.6 - 03.0 & 357.6 & -3 & 134 & 13 & mash + g357.6 - 06.5 & 357.6 & -6.5 & 37 & 98 & mash + g357.8 + 01.6 & 357.8 & 1.6 & 155 & 32 & mash + bl d & 358.2 & -1.1 & 149 & 28 & dss2 + m 3 - 39 & 358.5 & 5.4 & 35 & 91 & dss2 + ngc 6563 & 358.5 & -7.3 & 60 & 121 & dss2 + g358.8 - 07.6 & 358.8 & -7.6 & 34 & 96 & mash + m 3 - 9 & 359 & 5.1 & 116 & 173 & dss2 + m 1 - 26 & 359 & -0.7 & ... & ... & hst + a 40 & 359.1 & 15.1 & 72 & 125 & dss2 + 19w32 & 359.2 & 1.2 & 50 & 108 & cam98 + hb 5 & 359.3 & -0.9 & 76 & 135 & dss2 + g359.3 - 02.3 & 359.3 & -2.3 & 4 & 63 & mash + g359.7 - 04.4a & 359.7 & -4.4 & 0 & 60 & mash +

we present the discovery of a brown - dwarf companion to the 130400 myr - old g8 v star hd 203030 . separated by 11 ( 487 au in projection ) from its host star , hd 203030b has an estimated mass of . the-band spectral type of l7.5.5 places hd 203030b near the critical l / t transition in brown dwarfs , which is characterized by the rapid disappearance of dust in sub - stellar photospheres . from a comparative analysis with well - characterized field l / t transition dwarfs , we find that , despite its young age , hd 203030b has a bolometric luminosity similar to the gyr - old field dwarfs . adopting a radius from current models of sub - stellar evolution , we hence obtain that the effective temperature of hd 203030b is only k , markedly lower than the k effective temperatures of field l / t transition dwarfs . the temperature discrepancy can be resolved if either : ( 1 ) the ages of field brown dwarfs have been over - estimated by a factor of.5 , leading to under - estimated radii , or ( 2 ) the lower effective temperature of hd 203030b is related to its young age , implying that the effective temperature at the l / t transition is gravity - dependent .

astro-ph0607514

after more than a decade of study , the fundamental parameters of brown dwarfs remain poorly constrained . unlike hydrogen - burning stars , whose effective temperatures and luminosities are approximately age - independent ( at a fixed mass and composition ) on the main sequence , brown dwarfs cool and dim continuously as they age . only one of the fundamental sub - stellar parameters , luminosity , can be estimated independently of the others , through the use of trigonometric parallax and empirically determined bolometric corrections . any attempts to resolve the remaining degeneracies among temperatures , ages , masses , and radii rely on the fortuitous discovery of brown dwarfs in association with stars or clusters of known age . in such cases , brown - dwarf ages can be inferred by demonstrating the physical association of the brown dwarfs , usually as common proper motion secondaries or as members of stellar associations , with stars that have known ages . brown - dwarf masses can be measured from orbital astrometry and radial - velocity monitoring of close ( @xmath7510 au ) binary systems , in which at least one component is sub - stellar . finally , brown - dwarf radii , and hence , effective temperatures and surface gravities , can be measured in sub - stellar eclipsing binaries . the union of these three fortuitous cases , eclipsing brown - dwarf binaries associated with stars , offers the best opportunity for empirical determination of sub - stellar parameters . however , such unusual systems are extremely rare the first one , 2massj 053521840546085 , has only recently been reported @xcite . in all other cases involving brown dwarfs , the estimation of their properties relies to various degrees on the use of theoretical models of sub - stellar evolution ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? consequently , sub - stellar evolution models remain , for the most part , empirically unconfirmed . partial tests have been carried out for systems other than the ideal eclipsing double - line systems . @xcite and @xcite used astrometric measurements of the multiple system ab dor a / b / c to test the mass age luminosity relation at the stellar / sub - stellar boundary . the two analyses offer examples of how sub - stellar evolution models can be tested without considering all degenerate sub - stellar parameters in this case , excluding effective temperature . we note , however , that the conclusions of the two teams differ ( see * ? ? ? * ; * ? ? ? * for further discussion ) , in part because the stellar parameters themselves ( in particular , the stellar ages ) , although used as a reference in either study , may often not be known to the desired or believed level of accuracy . in a separate example , @xcite and @xcite tackle the full set of degenerate sub - stellar parameters using high - resolution spectroscopic observations of brown dwarfs in the upper scorpius and taurus associations . their studies rely on model spectra of brown - dwarf photospheres , constructed independently of sub - stellar evolutionary models , to resolve the mass effective temperature luminosity degeneracy . even though model - dependent , they allow a self - consistent comparison between sub - stellar cooling and photospheric models in the context of empirical data . in the present paper we take a similar approach and discuss the degeneracies among sub - stellar age , luminosity , and effective temperature ( i.e. , excluding mass ) in the context of models of sub - stellar evolution . we address these fundamental properties with the help of a new brown dwarf that we discovered as a common proper motion companion to the 130400 myr - old main - sequence g8 v star hd 203030 ( hip 105232 ; [ sec_observations ] ) . the companion , hd 203030b , has a spectral type of l7.5 , and thus lies very near the transition between l- and t - type sub - stellar photospheres , characterized by the settling of dust and the appearance of methane absorption in the near - ir spectra of brown dwarfs . because dust - settling occurs over a very narrow range of effective temperatures , 15001300 k ( spectral types l6t4 ) , a fact inferred both theoretically @xcite and semi - empirically @xcite , the effective temperature of hd 203030b is expected to be constrained very well . given the known age and luminosity of hd 203030b , we can find its radius and thus offer a constraint on the theory ( [ sec_properties ] ) . our approach is not fully empirical , because it relies on the semi - empirical results of @xcite and @xcite ; both studies rely on evolution models for the sub - stellar age mass radius relations . however , by adopting the same theoretical evolution models @xcite as in these two studies , our analysis provides a test of the self - consistency of the theoretical models ( [ sec_lt_teff ] ) . hd 203030b was discovered in the course of a direct imaging ao survey for sub - stellar companions conducted at the palomar 5 m and keck ii 10 m telescopes . the observation strategy for the survey is described in @xcite and @xcite . here we briefly describe aspects relevant to hd 203030 . hd 203030b was first detected at the palomar 5 m telescope , using the facility ao system palao @xcite and the palomar high - angular resolution observer ( pharo ; * ? ? ? * ) camera . palao allows diffraction - limited ( 0@xmath009 at @xmath8-band ) imaging in the near - ir with @xmath8-band strehl ratios routinely above 50% for natural guide stars brighter than @xmath9 mag . pharo employs a 1024@xmath101024 pix hgcdte hawaii detector , allowing a 25@xmath07@xmath1025@xmath07 field of view ( fov ) at a pixel scale of 25.09 mas pix . we used a 0@xmath097-diameter opaque coronagraphic spot in pharo to occult the primary in deep ( 60 sec ) exposures . hd 203030 was imaged at palomar on three occasions : 28 august 2002 , 16 july 2003 , and 26 june 2004 ( table [ tab_observations ] ) . during the first ( discovery ) epoch , we obtained a total of 24 individual 60-sec exposures of the primary with the coronagraph . subsequent epoch imaging was shorter ( 612 min ) , as dictated by the necessity to obtain only astrometric follow - up of the already discovered candidate companions . in addition to the deep coronagraphic exposures , at each of the three epochs we obtained shallow ( 210 sec ) non - coronagraphic @xmath8-band exposures of hd 203030 in a five - point dither pattern , to use as references for photometry and astrometry . we also took five two - second @xmath11- and @xmath12-band unocculted exposures at first epoch , to measure the colors of the brightest candidate companions . a 1% neutral density ( nd ) filter was placed in the optical path in pharo during the second and the third imaging epochs to prevent the primary from saturating in the unocculted @xmath8-band exposures . the short @xmath13 exposures from the first epoch , however , were unattenuated and the primary was saturated . a final ( fourth ) set of @xmath14 imaging observations was obtained with the keck ii ao system @xcite on 12 july 2005 . on keck we used the near - ir camera nirc2 ( k. matthews et al . , in preparation ) in its 20 mas pix scale and with a 1@xmath00-diameter partially transmissive coronagraph . both ao systems were employed in natural guide star mode , using hd 203030 ( @xmath15 mag ) itself as the guide star , which allowed @xmath8-band strehl ratios around 50% for all observations . the reduction of the imaging data followed the standard steps of sky - subtraction , flat - fielding , and bad - pixel correction , as described in detail in @xcite . the individual processed images were median - combined to produce a final high signal to noise image . the final first - epoch image of hd 203030 is shown in figure [ fig_hd203030_image ] . a total of seven candidate companions are visible in the pharo fov , only one of which , hd 203030b ( indicated with an arrow near the eastern edge of the image ) , was confirmed as a bona - fide proper - motion companion . the brightest field object ( `` object 1 '' in fig . [ fig_hd203030_image ] ) was a background star that was visible also in the short exposures and was used to bootstrap the relative photometry and astrometry of hd 203030b . measurements of the relative position of hd 203030b with respect to that of hd 203030a were obtained at all four imaging epochs . whenever possible , relative astrometry was obtained from images showing both the primary and the companion : the unocculted palomar images from the first epoch , in which the primary was saturated , and the occulted keck images from the last epoch , during which we used the partially transmissive nirc2 coronagraph . by fitting gaussians to the psfs of hd 203030a ( or to its wings , if saturated ) and of hd 203030b , we could measure each of their positions with 25 mas ( 0.040.10 pix ) precision . this was not possible for the images taken during the second and third palomar epochs . because we used the pharo nd filter for the short exposures at these epochs , the companion was not visible in the exposures in which the primary was unocculted . and because the pharo coronagraph is opaque , the primary was obscured in the deep coronagraphic exposures , in which the secondary was visible . in the latter cases we estimated the position of the occulted primary by either : ( 1 ) bootstrapping its position in the coronagraphic images from the position of `` object 1 '' ( fig . [ fig_hd203030_image ] ) visible in the shallow exposures taken at the same epoch , or ( 2 ) inferring its position from the peak of the two - dimensional cross - correlation of the central @xmath16 pix section of the coronagraphic image with a copy of itself inverted around its central pixel . the latter technique takes advantage of the fact that the extended psf halo of the primary is the dominant source of signal in the image , and that the halo is approximately radially symmetric on angular scales larger than the radius of the coronagraph . using either of these two approaches we constrained the position of the primary behind the coronagraph to within 13 mas ( @xmath60.5 pix ) . at this precision level of astrometry , the dominant uncertainty is no longer measurement error , but systematic errors from the astrometric calibration . in particular , accurate knowledge of the focal plane distortion on the detector is required . the distortions of the pharo 25 mas pix and nirc2 20 mas pix cameras were mapped to high precision in @xcite and @xcite . we applied the respective distortion corrections to the measured pixel coordinates , and calibrated the absolute pixel scales and orientations of the two detectors at each epoch from observations of wds 18055 + 0230a binary star with a well - known astrometric orbit from the sixth catalog of orbits of visual binary stars @xcite . the four epochs of astrometry spanning nearly three years confirmed hd 203030b as a common proper motion companion of hd 203030a . at first epoch , hd 203030b was separated by @xmath17 ( @xmath18 au ) from the primary , at a position angle @xmath19 . the changes in the position angle ( @xmath20 ) and separation ( @xmath21 ) between the pair at subsequent epochs are shown with the data points in fig . [ fig_hd203030_astrometry ] . the astrometric errors of the first - epoch measurement ( from 28 august , 2002 ) are added in quadrature to the errors of the measurements from the subsequent epochs . the solid line in each of the panels in figure [ fig_hd203030_astrometry ] represents the `` association '' line , synonymous with common proper motion . the dashed lines in the panels trace the expected change in the relative position of hd 203030b with respect to hd 203030a , had the companion been a background star with a negligible apparent proper motion . the dotted lines on either side of the dashed lines represent the @xmath22 uncertainties in the expected motion based on the _ hipparcos _ proper motion errors of the primary . hd 203030b is inconsistent with being an unrelated `` stationary '' background star at the 13@xmath23 level . because hd 203030 is not a member of a known open cluster , which could create the possibility of hd 203030b being just another kinematic member of the same cluster seen in projection , we assume that hd 203030a and b form a physical pair . near - ir @xmath12 and @xmath8 magnitudes of hd 203030b were obtained from the individual unocculted 1.8 sec exposures taken at palomar during the first epoch of observations ( 28 august 2002 ) . as the companion was not visible in the short @xmath11-band exposure , the @xmath11-band magnitude of hd 202030b was obtained from a deeper ( 50 sec ) @xmath11-band coronagraphic image of hd 203030 taken with keck ao on 14 july 2005 , in which the companion was detected . the primary was saturated in the short palomar exposures and was occulted by the coronagraph in the keck @xmath11-band exposures , thus preventing a direct measurement of the magnitude difference between hd 203030 a and b. however , the night of 28 august 2002 was photometric , and we were able to calibrate the near - ir photometry with respect to another program star hd 13531unocculted @xmath13 images of which were obtained through the nd 1% filter on the same night and at the same airmass . hd 13531 is not known to be variable ( its _ hipparcos _ photometry is constant to within 0.015 mag ; * ? ? ? * ) , and is therefore an adequate photometric standard . the near - ir transmissivity of the nd 1% filter was established in @xcite . the @xmath12- and @xmath8-band fluxes of hd 203030b and hd 13531 were measured in 0@xmath065-diameter apertures ( encompassing 2 airy rings at @xmath8 ) . such wide apertures ( @xmath68 palao psfs across ) were chosen to mitigate as much as possible the anisoplanatic distortion ( 510% in fwhm ) of the palao psf toward the edge of the 25@xmath07@xmath1025@xmath07 pharo field , and showed a high degree of self - consistency ( [email protected] mag ) . experiments with smaller apertures , 0@xmath0180@xmath036 in diameter , with sizes proportional to the ( variable ) fwhm of the psf , reproduced our wide - aperture photometry to within 0.10 mag ( though with a larger intrinsic scatter ) without showing any systematic trends with aperture size . estimates of the local background were obtained from the mean of the pixel counts in 0@xmath025-wide annuli with inner radii between 0@xmath050 and 1@xmath025 ( chosen to be at least 0@xmath025 larger than the radius at which the stellar radial profile blended into the background ) . we used the 2mass @xmath12 and @xmath8 magnitudes of hd 13531 to calibrate the transform from instrumental @xmath12 and @xmath8 magnitudes to apparent magnitudes . the @xmath11-band magnitude of hd 203030b was bootstrapped from the deeper nirc2 images with respect to that of `` object 1 '' ( fig . [ fig_hd203030_image ] ) . we used the same 0@xmath065-diameter apertures , as for the @xmath12 and @xmath8 bands above . at @xmath11-band the poorer ao performance ( strehl ratio @xmath2510% ) and the much more severe psf elongation due to anisoplanetism limited the accuracy of our measurements all near - ir photometry of hd 203030b is listed in table [ tab_photometry ] . finally , we note that hd 13531 itself was found to have a faint ( @xmath26 mag ; @xmath27 mag ) stellar proper - motion companion at an angular separation of @xmath28 @xcite . the presence of the secondary ( hd 13531b ) affects the 2mass magnitudes of the primary by 0.01 mag at @xmath11 and by 0.02 mag at @xmath12 and @xmath8 , which we take into account . at its relatively large angular separation from hd 13531a ( 89 times the fwhm of the palao psf ) , hd 13531b has a negligible effect on our photometric measurements . spectroscopic observations of hd 203030b were obtained on 14 jul 2005 with keck ao after the association of the companion with the primary had been established . with nirc2 we used the 20 mas pix camera in conjunction with the 80 mas - wide slit and the lowres grism for r@xmath292000 @xmath2-band spectroscopy . we obtained three 10 min @xmath2-band spectra at an airmass of 1.031.06 with dithers along the slit between exposures . conditions were photometric and the fwhm of the psf as measured from the acquisition images was 50 mas . telluric and instrumental features in the spectra were identified and corrected from observations of a nearby b9.5v @xcite star , hd 201859 , obtained immediately afterwards . images of internal flat field and arc lamps were taken at the beginning of the night to calibrate the detector response and the wavelength dispersion of the system . the individual frames were first flat - fielded and then pair - wise subtracted . the spectra of hd 203030b were then extracted using the iraf task apall . we adopted an aperture width of 14 pix ( @xmath30 ) centered on the spectrum of hd 203030b , which was traced by a high - order ( 610 ) polynomial . we estimated the background in 40 pix ( @xmath31 ) wide bands centered 50100 pix ( @xmath32@xmath33 ) away from the object . the large widths of the slit and of the extraction aperture were chosen to minimize chromatic effects arising from the dependence of the fwhm of the ao psf on wavelength ( e.g. , * ? ? ? * ) . the success of this extraction approach was judged by the fact that three individual companion spectra extracted from wide apertures were consistent among each other , while variations in the ao psf on time - scales of minutes are expected to induce chromatic differences among spectra extracted from narrower apertures . because hd 203030b is separated by almost 12@xmath34 from hd 203030a , stray light from the primary was not an issue in the extraction of the companion spectrum . the spectra of the telluric standard were extracted in a similar fashion as those of hd 203030b , using the same aperture width , though with the background bands farther from the star to avoid contamination from the brighter psf halo . we calibrated the wavelength dispersion of the extracted spectra with the arc lamp spectra and , after interpolating over the intrinsic br@xmath35 absorption in the telluric standard , we divided the extracted spectra of hd 203030b by those of the standard . we then median - combined the resulting three companion spectra and multiplied them by a 10500 k blackbody . the final hd 203030b @xmath2-band spectrum was smoothed to match the instrumental resolution ( @xmath36 , as measured from the fwhm of the arc lines ) and is shown compared to spectra of known l6t0 dwarfs in figure [ fig_hd203030b_spectrum ] . the @xmath2-band spectrum of hd 203030b ( fig . [ fig_hd203030b_spectrum ] ) exhibits molecular absorption features due to h@xmath37o in the blue end and co in the red end , as is typical of late - m to t0 dwarfs . no other molecular or atomic features are seen . in particular , the spectrum does not show either doublet absorption at 2.21 @xmath38 , which is seen in the spectra of @xmath24l2 dwarfs at similar ( @xmath39 ) resolution , or ch@xmath40 absorption at 2.20 @xmath38 , characteristic of @xmath41t0 dwarfs @xcite . we thus infer that the spectral type of hd 203030b is intermediate between l2 and t0 . for a more precise spectroscopic classification of hd 203030b we rely on @xmath2-band spectroscopic indices from the nirspec brown dwarf spectroscopic survey ( bdss ; * ) , designed to follow the l dwarf optical classification scheme of @xcite . following previous work on ir spectroscopic classification , ( * ? ? ? * and references therein ) define three molecular indices in the @xmath2 band , measuring the onset of h@xmath37o , ch@xmath40 , and co absorptions at 2.05 @xmath38 , 2.20 @xmath38 , and 2.30 @xmath38 , respectively . the indices are defined as the ratios of the median flux values in 40 - wide regions on either side of the molecular bandhead . they are therefore applicable to data with somewhat lower than the original @xmath42 resolution of the bdss data , and are well - suited for our @xmath36 nirc2 spectrum . we consider only the @xmath2-band ch@xmath40 and co indices of @xcite , for which we measure values of @xmath43 and @xmath44 , respectively , where the index means and their errors are obtained from measurements on the three independent hd 203030b spectra . we do not use their @xmath2-band h@xmath37o index because it requires measurements blue - ward of 2.0 @xmath38 , where our spectrum of hd 203030b is cut off by the transmissivity of the nirc2 lowres grism . comparing to data from @xcite and @xcite , we find that the 1@xmath23 spectral type ranges based on the ch@xmath40b and co indices are l7t0 and m7l8 , respectively . as already noted in @xcite and @xcite , the co index is of little utility for spectral typing of ultra - cool dwarfs because it displays very little sensitivity to their effective temperatures . hence , we use only the result from the ch@xmath40b index . having excluded the possibility that hd 203030b is a t dwarf , based on the lack of methane absorption at 2.20 @xmath38 , we adopt a final spectral type of [email protected] for it on the l - dwarf classification scheme of ( * ? ? ? * extending to l8 in the l dwarfs ) . if , instead , we adopt the l - dwarf spectroscopic classification system of ( * ? ? * based on lower - resolution spectra , but extending to l9 ) with the appropriately defined 2.20 @xmath38 methane index , the spectral type of hd 203030b would be [email protected] ( methane index value of [email protected] ) . both classifications are consistent with : ( 1 ) a visual comparison of the @xmath2-band spectrum of hd 203030b to the spectra of l6t0 dwarfs ( fig . [ fig_hd203030b_spectrum ] ) , ( 2 ) the near - ir colors of hd 203030b ( @xmath45 ; cf . table [ tab_photometry ] ) , and ( 3 ) the empirical near - ir absolute magnitude vs. spectral type relations for l and t dwarfs from @xcite and @xcite . hence , the spectral type of hd 203030b is later than those of the coolest hydrogen - burning stars ( @xmath29l4 ) , and it is a brown dwarf . given that hd 203030b is younger than typical field brown dwarfs , a lower - than - dwarf surface gravity may make its spectral classification with respect to field dwarfs inadequate . our spectrum of hd 203030b does not allow us to assess its surface gravity empirically because it lacks features with well expressed gravity sensitivity , the only one at @xmath2 band being the 2.21 @xmath38 doublet which disappears after l2 in the l dwarfs @xcite . models predict that at the age of the primary ( 130400 myr ; see [ sec_age ] below ) the surface gravity of hd 203030b is @xmath60.5 dex lower than the gravities of @xmath293 gyr - aged field l dwarfs . this may affect the strength of the observed co and h@xmath37o absorption , which are weakly sensitive to gravity . enhanced absorption by water vapor that produces peaked @xmath12-band continua is indeed observed in younger , @xmath71050 myr old , l dwarfs with low surface gravities@xcite . our spectrum of hd 203030b does indeed exhibit a slightly peaked continuum shape near 2.1@xmath38an effect has not been previously observed in l dwarfs . however , the effect of low surface gravity on co and h@xmath37o absorption at @xmath2 band is likely too weak by the age of hd 203030b , as it is not distinctive even in the younger ( 150 myr ) early - l dwarf 2mass j014158234633574 @xcite . moreover , our estimate of the spectral type of hd 203030b does not depend on co and h@xmath37o absorption indices , but on the ch@xmath40 index , which , in the lack of methane absorption , measures the slope of the continuum at 2.2 @xmath38far from any co and h@xmath37o bandheads . the peaked appearance of our @xmath2-band spectrum near 2.1@xmath38 could , on the other hand , can be traced to a possible chromatic effect in one of the telluric spectra at wavelengths @xmath72.1@xmath38 . therefore , we believe that our estimate of the spectral type of hd 203030b is not affected by its potentially lower surface gravity . fundamental sub - stellar parameters are degenerate with age . ages , masses , effective temperatures , and radii are empirically unconstrained in isolated brown dwarfs . only sub - stellar bolometric luminosities can be estimated from measurements of trigonometric parallaxes . the brown dwarf hd 203030b offers an opportunity to further resolve some of these degeneracies by the additional constraint on its age because of its association with hd 203030 . below , we first discuss the factors that determine the age of the primary and hence , of the secondary , by assumption of coevality . we then use our knowledge of the age and the heliocentric distance of hd 203030 to obtain model - dependent estimates of the mass and effective temperature of its sub - stellar companion . hd 203030 is a g8 v star @xcite and its age can be constrained from a number of photometric and spectroscopic indicators established from g stars in well - studied open clusters of known ages . a probabilistic estimate of the age of hd 203030 can also be made based on comparison of its galactic space motion to the space motions of known stellar populations and moving groups . the relevant stellar age indicators and the inferred ages are discussed below . [ [ position - on - the - color - magnitude - diagram . ] ] position on the color - magnitude diagram . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + hd 203030 has an apparent @xmath46-band magnitude of 8.48 mag @xcite and a _ hipparcos _ parallax of @xmath47 mas @xcite . thus , given an absolute @xmath46-band magnitude of 5.42 mag and a @xmath48 color of 0.75 mag @xcite , hd 203030 sits on the main sequence , and is between 0.1 gyr and 10 gyr old . [ [ stellar - rotation . ] ] stellar rotation . + + + + + + + + + + + + + + + + + stellar rotation periods increase with age as a result of angular momentum transfer outward from the core to the envelope , and then via loss to stellar winds or to a circumstellar disk ( e.g. , * ? ? ? * ; * ? ? ? * ; * ? ? ? @xcite find a photometric period of 4.1 days for hd 203030 in the _ hipparcos _ data . this period lies in between the rotation periods of late - g stars in the @xmath49 per ( 0.22.4 days ; * ? ? ? * ) and in the coma / hyades clusters ( 612 days ; * ? ? ? * ) , and is within the range of g - star rotation periods observed in the pleiades ( 0.46 days ; @xcite ; @xcite ) . assuming that the photometric periodicity of hd 203030 is due to star spots and stellar rotation , the age of hd 203030 is between that of @xmath49 per ( 90 myr ; * ? ? ? * ) and that of the hyades ( 625 myr ; * ? ? ? * ) , and possibly comparable to the age of the pleiades ( 125 myr ; * ? ? ? [ [ chromospheric - h - k - emission . ] ] chromospheric h & k emission . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + the strength of emission at 3968 and 3933 in the cores of the h & k chromospheric absorption bands is a well - known measure of chromospheric activity , and a proxy for the evolution of the stellar dynamo . a widely adopted indicator is the index , which measures the relative strength of emission with respect to the stellar continuum @xcite . the value of the index for hd 203030 has been obtained from high - resolution optical spectra of the star by four separate groups , the individual measurements being : 4.20 @xcite , which we convert to using the @xmath48 color of hd 203030 and an expression from @xcite . ] , 4.36 ( d. soderblom , 2000 , private communication ) , 4.37 @xcite , and -4.47 ( white , gabor , & hillenbrand , submitted ) . the mean value of the index from the four sources is @xmath50 ( standard error ) . using an empirical age- relation from @xcite , @xmath51 where @xmath52 is the stellar age in years and @xmath53 , we calculate that the age of hd 203030 is @xmath54 myr . [ [ coronal - x - ray - emission . ] ] coronal x - ray emission . + + + + + + + + + + + + + + + + + + + + + + + similarly to chromospheric activity , coronal x - ray luminosity is another proxy for the strength of the stellar dynamo that can be applied to determine stellar ages @xcite . in a study of age diagnostics for young sun - like ( g and early k ) stars , hillenbrand et al . ( in preparation ) find , based on an analysis of open cluster data , that the age of sun - like stars is related to their x - ray luminosity following @xmath55 where @xmath56 is the stellar age in myr and @xmath57 is the x - ray luminosity in erg s@xmath58 . the x - ray luminosity of hd 203030 from the _ rosat _ all - sky survey faint source catalog @xcite is @xmath59 , where we have converted count rates to flux using the conversion factor in @xcite . from this , we find that the age of hd 203030 is @xmath60 myr , where the intrinsic scatter of relation ( [ eqn_lx ] ) , 0.16 dex ( hillenbrand et al . , submitted ) , has been added in quadrature to the error in @xmath57 . [ [ absorption . ] ] absorption . + + + + + + + + + + + + lithium is destroyed in the convective envelopes of late - type stars , and so a high photospheric lithium abundance is a signature of stellar youth @xcite . @xcite measure a @xmath61 equivalent width of 56 m for hd 203030 in @xmath62 spectra . this is a factor of @xmath292 larger than the measured equivalent widths in @xmath63 spectra of stars of similar colors in the 625 myr - old hyades @xcite , a factor of @xmath293 smaller than the equivalent widths in @xmath64 spectra of late - g stars in the 125 myr - old pleiades @xcite , and is within the range of equivalent widths in @xmath65 spectra of analogous stars in the m 34 ( ngc 1039 ) open cluster ( 250 myr ; * ? ? ? hd 203030 is therefore younger than 625 myr , probably older than 125 myr , and likely of order 250 myr old . [ [ uvw - galactic - space - motion . ] ] @xmath66 galactic space motion . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + young stars inherit the kinematics of their birth environment for tens to hundreds of myr before their space motion is disrupted by dynamical interactions with the galactic environment . the galactic velocity vector of a star can therefore be used to obtain a probabilistic estimate of the star s age , when compared to the space motions of comoving groups of stars with known ages . based on its galactic space motion ( @xmath67 km s@xmath58 , @xmath68 km s@xmath58 , @xmath69 km s@xmath58 ) , @xcite classify hd 203030 as a probable kinematic member of the ic 2391 super - cluster , with an estimated age of 3550 myr @xcite . to summarize , based on its chromospheric and coronal activity , hd 203030 is @xmath29250 myr old , with a possible age range of 130400 myr . this age is fully consistent with its stellar rotation rate and photospheric lithium absorption , and is also supported by its main - sequence position in a color - magnitude diagram . however , this age does not agree with the young kinematic age of hd 203030 quoted in @xcite . nevertheless , such discrepancies are not unusual in kinematically selected samples of young stars . indeed , we note that the @xmath66 velocity vector of hd 203030 inhabits an intermediate position between the loci of the 20150 myr local association , the 100300 myr castor moving group , and the 600 myr hyades super - cluster ( fig . 1 in * ? * and references therein ) . if hd 203030 were a member of the local association or the castor moving group , instead of the ic 2391 moving group , its kinematic age would be consistent with the ages derived from the other criteria . we determine the absolute @xmath8-band magnitude of hd 203030b ( table [ tab_photometry ] ) , by assuming that it is at the same heliocentric distance as the primary ( _ hipparcos _ parallax @xmath47 mas ; * ? ? ? to convert the absolute magnitude to bolometric luminosity , we use a calibration of the @xmath2-band bolometric correction bc@xmath70 vs. l t spectral type from @xcite . we take into account that the calibration of @xcite is for the spectral classification system of @xcite , on which the spectral type of hd 203030b is [email protected] ( [ sec_sptype ] ) . we also note that the bc@xmath70 vs. spectral type relation from @xcite is for the mko near - ir filter set , while our data are on the cit filter set . however , the synthetic transformations of @xcite show that the mko @xmath2 and cit @xmath8 filters produce nearly identical results with an offset of 0.000.02 mag for l6l9 dwarfs . this small offset has been added to our error budget . finally , we note that the bolometric corrections in @xcite are compiled from data for @xmath41 gyr - old , high - surface gravity field dwarfs , and may need to be corrected for the expected @xmath60.5 dex lower surface gravity of the @xmath29250 myr - old hd 203030b . the sense of this correction is unknown empirically , as the body of data on young l dwarfs is extremely limited . from the models of @xcite and @xcite we infer that the magnitude of the correction over a 0.5 dex change in surface gravity is @xmath240.05 mag , and that it may be either positive or negative . therefore , we treat it as an added error term . averaging the bolometric corrections for the six l6l9 dwarfs in table 6 of @xcite , we find bc@xmath71 mag for hd 203030b . the bolometric magnitude of hd 203030b is then @xmath72 mag and , assuming @xmath73 mag for the sun @xcite , its luminosity is @xmath74 . the bolometric luminosity of hd 203030b is thus comparable to those of single field l6t5 dwarfs ( mean @xmath75 with a standard deviation of 0.15 dex ; cf . table 6 in @xcite ) . that is , at the 130400 myr age of hd 203030b , transition l / t dwarfs are not more luminous than their @xmath41 gyr - aged field counterparts . this is an unusual but significant result , given that at its @xmath60.5 dex lower anticipated surface gravity the luminosity of hd 203030b is expected to be @xmath60.5 dex higher than that of older l / t dwarfs in the field ( provided that the effective temperature at the l / t transition is independent of age ) . as we discuss in the following sections , this translates in an unusually low effective temperature for hd 203030b ( [ sec_teff ] ) , and indicates either a necessity to re - evaluate the effective temperatures and ages of l / t dwarfs in the field or a gravity dependence in the spectral type effective temperature relation near the l / t transition ( [ sec_discussion_teff ] ) . the effective temperature of hd 203030b can be inferred from its age and bolometric luminosity by adopting an age - dependent theoretical estimate of its radius . using the age and luminosity values obtained above , the models of @xcite predict a mean radius of 1.06 ( at the the mean age and luminosity of hd 203030b ) with a range of 1.001.22 ( corresponding to the full age range and 1@xmath23 luminosity range of hd 20303b ) , and a mean effective temperature of 1206 k with a range of 10901280 k. these values and their ranges are also consistent with those inferred from the models of ( * ? ? ? * dusty ) and ( * ? ? ? * cond ) . alternatively , the effective temperature of hd 203030b can be inferred from its spectral type , by comparison to the compilation of spectral type and effective temperature data for brown dwarfs with known parallaxes in @xcite and @xcite . this seemingly empirical approach is also model - dependent because , despite the use of measured bolometric luminosities , @xcite and @xcite also adopt brown - dwarf radii based on theoretical evolutionary models ( from @xcite ) . nevertheless , because these authors use only ensemble mean values for the radii ( @xmath76 ; * ? ? ? * ) or the ages ( 3 gyr ; * ? ? ? * ) of field brown dwarfs , their results do not hinge on the details of the theoretical models . from 12 distinct l dwarfs listed in table 7 of @xcite and table 6 of @xcite , we find that the effective temperatures of l dwarfs of spectral types l7 or later ( i.e. , consistent with that of hd 203030b ) fall in the range 13301650 k , with a mean of @xmath77 k ( where we have excluded the temperature estimate for sdss j042348.57041403.5 , now resolved into a l6/t2 binary ; * ? ? ? the individual estimates have errors of approximately 100200 k , arising from a combination of uncertainties in the photometry , parallaxes , and adopted radii or ages of the brown dwarfs . we note that the two ranges obtained for the effective temperature of hd 203030b , 10901280 k , based on its luminosity and the assumption of a model - dependent radius , and 13301650 k , from its spectral type and a comparison with the semi - empirical calibrations of @xcite and @xcite , are only marginally consistent with each other , with hd 203030b appearing on average 240 k cooler . the lower inferred effective temperature of hd 203030b is a direct consequence of the assumption of a @xmath615% larger radius ( because of its young age ) and similar bolometric luminosity ( [ sec_l ] ) compared to late - l dwarfs in the field . we estimate the mass of hd 203030b using the sub - stellar evolution models from the lyon @xcite and arizona @xcite groups . figure [ fig_hd203030b_mass ] shows the allowed locus of values ( thick rectangle ) around the adopted age and luminosity ( solid dot ) of hd 203030b overlaid on the evolutionary tracks . we have only plotted the cond suite of models @xcite from the lyon group , as the evolution of sub - stellar luminosity predicted by the dusty @xcite and cond models is nearly identical @xcite . based on these predictions , the mass of hd 203030b is between 0.012 and 0.031 , with the adopted mean luminosity and age values corresponding to 0.023 . if , instead , we estimate the mass of hd 203030b from the effective temperature ( @xmath77 k ) corresponding to its spectral type from the calibrations of @xcite and @xcite , then its mass would be @xmath78 . however , the predicted bolometric luminosity of hd 203030b in this case would be a factor of 1.32.3 greater than the inferred one . at a mass ratio of only @xmath790.02 , the hd 203030 a / b system is among the lowest mass ratio binaries discovered through direct imaging to date , along with hr 7328 a / b @xcite , ab pic a / b @xcite , and gq lup a / b @xcite . all of these very low mass ratio ( @xmath80 ) binaries are young ( @xmath24400 myr ) and have projected separations @xmath4100 au , with estimated orbital periods near the @xmath29@xmath81-day peak of the g - star binary period distribution of @xcite . wide very low mass ratio systems may thus not be highly unusual at young ages and will be important for understanding the architecture of planetary systems in conjunction with their closer - in counterparts found from radial - velocity surveys . the uncharacteristically low effective temperature of the young hd 203030b sets it apart from older l7l8 field dwarfs with known parallaxes . because the age of hd 203030b is known , unlike the ages of field brown dwarfs , this result has the potential of providing an important constraint for understanding the physics of the l / t transition . however , prior to delving into the implications , we critically consider all factors underlying our estimate of the effective temperature of hd 203030b . our estimate of the effective temperature of hd 203030b hinges on the following assumptions and factors : 1 . physical association and coevality of hd 203030a and b ; 2 . accuracy of the photometry of hd 203030a , b , and of the reference hd 13531a , or variability in any of the above ; 3 . accuracy of the spectral classification ; 4 . age of hd 203030a ; 5 . heliocentric distance of hd 203030a ; 6 . unknown multiplicity of hd 203030b ; 7 . bolometric correction for hd 203030b ; 8 . accuracy of the theoretical evolutionary models . we critically address each of these issues below . 1 . if hd 203030b was older than the primary , the theoretical estimate of its radius could be sufficiently small ( @xmath250.9 ) to drive its mean effective temperature ( corresponding to its mean age and luminosity ; [ sec_age ] , [ sec_l ] ) above 1330 k and to make it consistent with the effective temperatures of l / t dwarfs in the field . however , the physical association of hd 203030b with its primary was already demonstrated beyond reasonable doubt in [ sec_astrometry ] . while the possibility remains that hd 203030b is an old field brown dwarf captured by a younger hd 203030a , it is highly unlikely . an overestimate of the apparent brightness of hd 203030b by @xmath410.3 mag would be sufficient to resolve the effective temperature discrepancy . this remains a possibility , especially given that precise ( @xmath240.1 mag ) ao photometry is notoriously difficult to obtain because of variations in the quality of the ao psf on time - scales of seconds . our observations on 28 august , 2002 , benefitted from photometric conditions and relatively stable ao performance , with @xmath8-band strehl ratios ranging between 40% and 60% . furthermore , our choice of wide ( @xmath68 fwhm ) photometric apertures alleviates the effect of variations in the total enclosed power that might be caused by changes in the ao correction . both of these factors contributed to the high degree of self - consistency of our @xmath8-band ao photometry : @xmath240.05 mag among individual exposures ( [ sec_phot ] ) . we therefore believe that a systematic error in our photometry is not the culprit for the inferred discrepancy in the effective temperature . + even though the photometry may be accurate , any of hd 203030a , hd 203030b , or hd 13531a ( the photometric reference ) , could be variable by @xmath410.3 mag at @xmath8 , and thus possibly account for the fact that hd 203030b appears cooler than expected ( because of being under - luminous ) . indeed , hd 203030a and hd 13531a are both young ( @xmath240.5 gyr ) stars and therefore expected to be variable to some extent . furthermore , variability amplitudes as high as 0.5 mag have been observed in l / t transition dwarfs on time - scales of 0.33 hours @xcite . however , as already pointed out ( [ sec_phot ] ) , the _ hipparcos _ photometry of hd 13531 shows no variability , hence hd 13531 appears to be an adequate photometric standard . in addition , our differential photometry between hd 203030b and `` object 1 '' from the individual 60 sec @xmath8-band coronagraphic exposures spanning 1.3 hours on aug 28 2002 does not exhibit deviations larger than 0.1 mag from the mean ( r.m.s . scatter = 0.06 mag ) . barring the highly unlikely scenario in which hd 203030b and `` object 1 '' follow photometric variations of similar magnitudes , periods , and phases , we infer that hd 203030b is not strongly variable , and therefore its low effective temperature can not be explained by a period of low intrinsic luminosity . spectral classification of l dwarfs in the near - ir is still not as well established as in the visible . optically - typed l dwarfs do not follow a unique and monotonic spectral sequence in the near - ir , indicating that the two wavelength regions probe different photospheric physics @xcite . it is therefore possible that the spectral type of hd 203030b inferred from its @xmath2-band spectrum ( fig . [ fig_hd203030b_spectrum ] ) does not correspond to the same spectral type in the visible . however , there is very little room for error in the present spectral classification . on one hand , the spectral type hd 203030b is not later than l8 ( on the l dwarf classification scheme of * ? ? ? * ) , i.e. , it is not a t dwarf , because its @xmath2-band spectrum lacks methane absorption at 2.20 @xmath38 ( [ sec_sptype ] ) . on the other hand , if the spectral type of hd 203030b was earlier than l7 , its low effective temperature would be even more discrepant with those of similarly - type field l dwarfs . we are therefore confident that our @xmath2-band spectral classification of hd 203030b is accurate to the claimed precision . 4 . an older age for hd 203030 would imply a longer contraction time for the sub - stellar companion and hence , a smaller radius . using the models of @xcite , we find that , given the measured bolometric luminosity , an age @xmath41 gyr would be required at to make the brown dwarf radius sufficiently small ( @xmath250.9 ) to raise its effective temperature above 1330 k. however , as discussed in [ sec_age ] , all age - dating criteria indicate that hd 203030 is younger than the hyades ( 625 myr ) . if hd 203030 was at a larger heliocentric distance , hd 203030b would be intrinsically more luminous , and hence , hotter . a 13% error in the _ hipparcos _ parallax could account for the observed discrepancy in effective temperatures . we note that this would be of the same relative magnitude and sense as the error in the _ hipparcos _ distance to the pleiades @xcite . it is possible that such systematic zonal errors of similar magnitude exist also in other parts of the sky ( e.g. , the coma cluster ; * ? ? ? however , the inferred _ hipparcos _ systematic error leading to the pleiades distance problem is of order 1 mas . this would result in only a @xmath64% error in the parallax ( 24.48 mas ) of hd 203030 : too small to explain the effective temperature discrepancy . the existence of an unresolved binary companion to hd 203030b , a hypothetical hd 203030``c '' at @xmath254 au from the secondary , would change the inferred bolometric luminosity of the secondary . indeed , previously unknown binarity at the l / t transition has been recently invoked as a possibility in explaining the observed brightening in @xmath11 absolute magnitude from late - l to early - t spectral types @xcite . however , a companion to hd 203030b would have the opposite effect : it would decrease the bolometric luminosities of the individual components ( hd 203030b and `` c '' ) , without much affecting their spectral types , and would thus exacerbate the problem . 7 . in estimating the absolute @xmath8-band magnitude of hd 203030b we used bolometric corrections from @xcite . the accuracy of the bolometric corrections is limited by the presently small amount of data on l and t dwarfs and by the even smaller body of knowledge of the non - raleigh - jeans ir spectral energy distributions of brown dwarfs at @xmath43 @xmath38 . only recently has _ spitzer_/irs spectroscopy over 5.538@xmath38 confirmed that the bolometric corrections of @xcite are accurate to within 0.10 mag @xcite , albeit for a small number of l and t dwarfs . regardless , systematic uncertainties in the @xmath2-band bolometric corrections of l and t dwarfs do not affect our comparison of effective temperatures in relative terms , because both studies that calibrate l and t spectral types vs. effective temperature @xcite use the same scale of bolometric corrections as adopted by us . finally , our calculation of the effective temperature for hd 203030b hinges on the accuracy of the sub - stellar evolutionary models of @xcite . without further observational constraints ( e.g. , dynamical mass , surface gravity ) it is not possible to empirically test the accuracy of these theoretical predictions . however , the radii of the brown dwarfs in the studies of @xcite and @xcite are also estimated based on these models . our approach in estimating the effective temperature of hd 203030b from its bolometric luminosity and age is therefore fully consistent with the methods employed in these two studies . having considered and rejected all possible factors that may have lead us to underestimate the effective temperature of hd 203030b ( @xmath5 k ) in comparison to the temperature expected from its [email protected] spectral type ( @xmath77 k ) , we conclude that hd 203030b is indeed cooler than late - l dwarfs in the field . we pointed out in [ sec_discussion_reliability ] that the discrepantly low effective temperature of hd 203030b arises mostly from the larger sub - stellar radius adopted for hd 203030b , because of its younger age but similar luminosity compared to older brown dwarfs of the same spectral type . having established the robustness of our age and luminosity determinations of hd 203030b ( [ sec_discussion_reliability ] ) , we conclude that either : ( 1 ) the effective temperatures of brown dwarfs across the l / t transition in @xcite and @xcite have been over - estimated on average by 240 k because of unresolved binarity , ( 2 ) the ages of l / t transition dwarfs in the field have been over - estimated , or equivalently , their radii under - estimated ( and hence , their effective temperatures over - predicted ) by theoretical models , or ( 3 ) the range of effective temperatures over which the l / t transition occurs extends to cooler temperatures than found in the empirical studies of @xcite and @xcite , e.g. , due to variations in surface gravity . we address each of these possibilities in turn below . the effective temperatures of field brown dwarfs at the l / t transition may have been over - estimated if the majority of these brown dwarfs are unresolved close binaries , i.e. , if they appear over - luminous . indeed , @xcite find that @xmath2950% of late - l to early - t dwarfs turn out to be binaries when imaged at high angular resolution ( e.g. , with ao , or with _ hst _ ) , including one of the l / t brown dwarfs considered in the studies of @xcite and @xcite , sdss j042348.57041403.5 . we test if unresolved binarity is a factor by focusing only on these l / t transition dwarfs that have been imaged at high angular resolution . seven of the l7l8 dwarfs in @xcite have a priori known multiplicity from _ hst_/wfpc2 @xcite , i.e. , they have either been resolved into binaries , or have remained unresolved . the mean effective temperature of these is @xmath82 k. it is similar to , if not somewhat higher than , the mean effective temperature , @xmath83 k ( both sets of errors denote standard deviations ) , of the remaining four l7l8 dwarfs that have not been observed at high angular resolution . therefore , it is unlikely that unresolved binarity has lead to a significant over - estimate of the effective temperatures across the l / t transition . the second hypothesis ( overestimated ages leading to underestimated radii ) is not directly testable with the existing observational data , as the radii of l and t dwarfs have not been measured empirically . nevertheless , following our discussion in [ sec_teff ] , we construct an indirect test , by comparing the effective temperatures of all known l7l8 dwarf companions to stars ( with known ages and distances ) to the effective temperatures of field late - l dwarfs with known trigonometric parallaxes from the studies of @xcite and @xcite given a self - consistent application of the theoretical models @xcite , the two sets of late - l dwarfs , companions vs. free - floating , should have identical effective temperatures . the inferred model - dependent effective temperatures of all l7l8 companions to stars ( gj 584c , @xcite ; gj 337c , @xcite ; hd 203030b , this paper ) are all @xmath241345 k , which is @xmath41100 k below the mean for field l7l8 dwarfs with known trigonometric parallaxes ( [ sec_teff ] ; * ? ? ? * ) . while the corresponding deviation in each case is only 12@xmath23 , the statistical significance of the overall effect is @xmath499.5% ( i.e. , @xmath843@xmath23 ) . if the effect is real , it indicates that the predicted radii of field late - l dwarfs are 1015% too small ( making their effective temperatures too high ) compared to those of the late - l companions . equivalently , this translates into a factor of 1.52 over - estimate of the adopted ages for late l dwarfs in the field , according to the models . indeed , the mean ages of the three l7l8 companions ( ranging between 0.3 and 2.0 gyr ) are a factor of @xmath841.5 smaller than the 2.9 gyr mean statistical age of late - l dwarfs in the field ( @xcite ; also based on models from @xcite ) . while the youngest of the late - l companions , hd 203030b , was the product of a targeted search for brown dwarfs around @xmath240.5 gyr old stars @xcite , and so should be excluded from this age comparison , gj 584c and gj 337c were not . thus , barring a systematic under - estimate of the mean ages of these two stellar primaries by a factor of @xmath291.5 ( not at all unlikely , given that their maximum ages are 2.5 and 3.4 gyr , respectively ; * ? ? ? * ; * ? ? ? * ) , it is possible that the ages of field late - l dwarfs have been overestimated by a factor of @xmath291.5 . a similar age overestimate is inferred when using the models of @xcite and @xcite . that is , the result does not hinge on peculiarities in either set of models . thus , sub - stellar cooling rates may have been under - predicted , or effective temperatures in the initial conditions may have been over - estimated , in both the arizona and the lyon models . however , given the strong reliance of this result on the ages of two isolated stars , which are not determined as accurately as the ages of stars in stellar associations , we can not claim to have discovered any inaccuracy in the theoretical models with certainty . the hypothesis needs to be re - visited once accurate bolometric luminosities are determined for a larger number of sub - stellar objects with well - known ages , e.g. , in open clusters . the last hypothesis ( that the l / t transition extends to cooler temperatures ) is the default one , as it reconciles previous determinations of the effective temperature at the l / t transition with the present data , without seeking constraints on the underlying physics . a physical interpretation may nonetheless focus on the role of metallicity and surface gravity in the l / t transition ( e.g. , * ? ? ? the metallicity of hd 203030b may be assumed to be the same as that of the primary , for which @xcite measure @xmath85}=-0.01 $ ] , i.e. , solar , and likely not very different from that of most field l dwarfs . however , as already discussed ( [ sec_sptype ] ) , the gravity of the @xmath29250 myr - old hd 203030b is expected to be @xmath60.5 dex lower than the gravities of @xmath41 gyr - old field brown dwarfs , as corresponding to its 1015% larger adopted radius . while we determined that gravity has not affected our estimates of the effective temperature and luminosity of hd 203030b in a systematic way , its effect on photospheric chemistry in brown dwarfs is unknown . we note that , among stars , late - type giants have 200600 k lower effective temperatures than dwarfs of the same spectral type ( c.f . * ; * ? ? ? hence , a similar trend among sub - stellar objects is not at all unlikely . a dependence on surface gravity in the effective temperature at which the l dwarf sequence merges onto the t dwarf sequence is indeed predicted by the `` unified cloudy model '' of @xcite . these authors propose such a dependence to explain the observed brightening of early - t dwarfs at @xmath11 band @xcite , arguing that the brighter early - t dwarfs have lower surface gravities because they are younger . we note that @xcite s theory does not agree with the fact that the young hd 203030b is not brighter than its older late - l counterparts in the field ( [ sec_l ] ) . furthermore , @xcite and @xcite have recently demonstrated that the @xmath11-band brightening across the l / t transition is not an age - spread effect , but a real feature of brown dwarf evolution , as might be explained by disruption of condensate clouds at a fixed effective temperature @xcite . still , with hd 203030b we have now shown that the effective temperatures of different l / t transition objects may not be identical . it therefore remains an open question whether surface gravity may have a role in setting the effective temperature at the l / t transition . planned optical and @xmath11-band spectroscopy of hd 203030b , combined with theoretical modeling of gravity - sensitive alkali , metal - hydride , and vo features will attempt to address this issue . unfortunately , the magnitude of the effect that we will be trying to measure is of order of the accuracy with which of modern models of sub - stellar photospheres can match gravity - sensitive features in brown - dwarf spectra . the role of gravity on the effective temperatures of l dwarfs at a fixed spectral type may thus be better addressed with younger ( albeit earlier - type ) objects , such as the recently discovered 2massw j1207334393254b @xcite and 2mass j014158234633574 @xcite . we have discovered a proper motion sub - stellar companion to the 130400 myr - old sun - like star hd 203030 . the companion , hd 203030b , is 11@xmath09 away from the primary , at a projected orbital separation of 487 au . assuming coevality with the primary star , sub - stellar evolution models place the companion mass at @xmath1 . from @xmath2-band ao spectroscopy , we determine a spectral type of [email protected] for hd 203030b . the brown dwarf is thus near the spectroscopic l / t transition , characterized by diminishing amounts of dust in sub - stellar photospheres . because of its association with a main - sequence star of a known age and heliocentric distance , hd 203030b offers a rare chance to probe the l / t transition in a setting in which most of the degeneracies characteristic of sub - stellar models are resolved . we find that the bolometric luminosity of hd 203030b is comparable to that of @xmath41 gyr - old field dwarfs of similar spectral types , despite the @xmath60.5 dex lower gravity anticipated at the young age of hd 203030b . as a result , theoretical models of sub - stellar evolution predict a @xmath6240 k cooler effective temperature for hd 203030b compared to the effective temperatures of its older counterparts in the field . we consider three hypotheses for the discrepancy : ( 1 ) that the bolometric luminosities and effective temperatures of late - l dwarfs in the field have been over - estimated because of unresolved binarity ; ( 2 ) that the effective temperatures of late - l dwarfs in the field may have been over - estimated ( equivalently , their radii under - estimated ) because theoretical models may have over - predicted their ages ; and ( 3 ) that the spectral type effective temperature for late - l dwarfs is not age- and gravity - independent . based on multiplicity studies of field brown dwarfs from high - angular resolution observations with _ hst _ and ao , we do not find a significant discrepancy between the effective temperatures ( or luminosities ) of brown dwarfs with known and unknown multiplicities , and rule out the first possibility . we test the second hypothesis by comparing the effective temperatures of late - l secondaries of main - sequence stars ( with known ages and distances ) to late - l dwarfs in the field . we find evidence that the late - l companions are @xmath41100 k cooler and @xmath291.5 times younger than their isolated counterparts . however , because ages of individual stars can not be claimed with such accuracy , the significance of this result is only marginal . the last hypothesis is a compromise between the discrepant effective temperatures of hd 203030b and late - l dwarfs in the field , and draws a parallel with the dependence of spectral type and effective temperature on surface gravity in stars @xcite . both remaining hypotheses await testing from larger samples of brown dwarfs with known distances and ages , either as members of known stellar associations or as companions to stars . the authors acknowledge many insightful discussions with travis barman , j. davy kirkpatrick , and ian mclean during the preparation of the manuscript . we are grateful to our support astronomers and engineers randy campbell , rick burress , and jeff hickey for their guidance in using keck and palomar ao , to keith matthews and dave thompson for advice with nirc2 spectroscopy , and to our telescope operators at the palomar and keck ii telescopes , jean mueller , karl dunscombe , and cynthia wilburn . we also wish to extend special thanks to those of hawaiian ancestry , on whose sacred mountain of mauna kea we are privileged to be guests . without their generous hospitality , the spectroscopic observations presented herein would not have been possible . s. a. m. acknowledges support from a _ spitzer space telescope _ postdoctoral fellowship . this work was supported by nasa grant nng05gj37 g and nasa / jpl contract 1224566 . ccccc 2002 aug 28 & yes & @xmath86 imaging & palomar + 2003 jul 16 & no & @xmath8 imaging & palomar + 2004 jun 26 & no & @xmath8 imaging & palomar + 2005 jul 14 & yes & @xmath87 imaging , @xmath2 spectroscopy & keck +

starting from a model hamiltonian for the normal state of the topological insulator bise , we construct a pseudospin basis for the single - particle wavefunctions . considering weak superconducting pairing near the fermi surface , we express the recently proposed superconducting order parameters for cu doped bise in this basis . for the odd parity states , the-vectors specifying the order parameter can have unusual momentum dependence for certain parameter regimes . some peculiar results in the literature for surface states are discussed in light of the forms of these s . properties of the even parity states are also illuminated using this pseudospin basis . results from this single - band description are compared with those from the full two - band model .

1301.0863

the recent prediction @xcite of the existence of three - dimensional topological insulators ( ti ) and their experimental confirmation in bi@xmath0se@xmath1 and related compounds @xcite have generated a lot of excitement , as these ti form a new class of material which is distinct from ordinary band insulators , metals etc , possessing peculiar properties such as topologically protected surface states and usual electrodynamics @xcite . these properties arises from spin - orbit coupling , leading to band - inversions at some regions of the brillouin zone . interestingly , bi@xmath0se@xmath1 , when doped with copper , is found to become superconducting @xcite.@xcite some unusual properties , such as existence of zero bias conductance peak in tunneling experiments @xcite and absence of pauli limiting in upper critical field @xcite , seem to suggest unconventional character of the cooper pairing , though the situation is not without controversy @xcite . there is a lot of attention to the theoretical aspects of superconductivity in this compound , in particular possible odd parity pairing states . early on , fu and berg @xcite , starting from an effective hamiltonian for the normal state of bi@xmath0se@xmath1 near the zero momentum @xmath5 point with two orbitals per unit cell , considered various models of superconducting states with momentum - independent pairs when expressed in terms of this basis . properties of these models have subsequently analyzed by many , mostly focusing on the surface states @xcite . on the other hand , superconductivity in systems with strong spin - orbit coupling have been much studied in the past , in particular in the context of heavy fermions @xcite . there , the usual language used is that the normal quasiparticles are described by a pseudospin basis , obeying certain symmetry properties , and then the superconducting state order parameter and pair wavefunctions are expressed in terms of this basis . the superconducting states then can be classified by crystal symmetries into group representations , and the pairing states are expressed in terms the sets of basis functions appropriate to the relevant group representations . in this formulation , for crystals with inversion symmetry , even parity superconductors always have pairing wavefunctions even in momentum @xmath3 and are singlets in pseudospins , whereas odd parity superconductors have pairing wavefunctions that matrices in pseudopsin space with each opponent odd in @xmath3 . this matrix structure is usually expressed in terms of a `` d - vector '' which specifies the corresponding pseudospin structure via familiar pauli @xmath6 matrices . properties of the superconductors can then be directly obtained by examining these order parameters , including the possibility of surface states . many questions then arise . if the chemical potential @xmath7 , of the doped bi@xmath0se@xmath1 is sufficiently large compared with the pairing potential @xmath8 , ( which is likely to be the case since @xmath9 according to @xcite , whereas the transition temperature is @xmath10k . the measured gap in tunneling is indeed @xmath11 , though the precise value is controversial @xcite ) , pairing should effectively only take place within one normal state band . how well then can we understand the superconducting properties of cu : bi@xmath0be@xmath1 within a single band picture ? what is the pairing order parameter in the pseudospin basis ? can we understand the andreev bound states in this way ? in particular , what is the origin of the very peculiar dispersions found for the odd parity states found in @xcite ? in this paper , we report such an attempt . in sec [ sec : back ] , we review the model of @xcite . pseudospin wavefunctions are constructed in sec [ sec : wf ] . they would be applied to the superconducting phases , first for the bulk in [ sec : bulk ] , then the surface states in [ sec : surface ] . we shall show that many of the results in the literature can be understood in this way . we conclude in [ sec : concl ] . the appendix gives further discussions on the surface states and topology of some of the superconducting phases . in this section , we review the model of @xcite . we begin with the normal state . the effective hamiltonian which captures the physics near @xmath12 is given by h_n ( k ) = m _ x + v_z k_z _ y + v _ z ( k_x s_y - k_y s_x ) [ hn ] here @xmath3 , @xmath13 represent the wavevector and its components , @xmath14 , @xmath15 are velocities , @xmath16 represents the two ( mainly ) @xmath17 orbitals in the quintuple layer of bi@xmath0se@xmath1 , and @xmath18 are the pauli matrices for the spin . equation ( [ hn ] ) is basically a dirac hamiltonian . the energies for the quasiparticles are @xmath19 where @xmath20 with @xmath21 , thus consist of a conduction and a valence band . bi@xmath0se@xmath1 possesses @xmath22 ( @xmath23 ) symmetry , which includes parity . eq ( [ hn ] ) is indeed invariant under the parity when this operator is taken as @xmath24 @xcite : eq ( [ hn ] ) is left unmodified if we substitute @xmath25 , @xmath26 , with all other variables unaltered . equation ( [ hn ] ) is actually invariant under the higher symmetry group @xmath27 as it is obviously unchanged under any continuous rotation about the @xmath28-axis instead of only @xmath29 for @xmath30 . one can check easily that ( [ hn ] ) is invariant under reflection about any vertical reflection planes . since it is rotationally symmetric about @xmath28 , it is sufficient to check any one single vertical reflection plane . for example , under reflection in the x - z plane , @xmath31 , @xmath32 , ( [ hn ] ) remains indeed unchanged . @xcite to discuss the surface states associated with the ti , boundary conditions for the wavefunction are needed . unfortunately , this point seems to be somewhat controversial @xcite . for definiteness , we follow @xcite here . the boundary condition for the wavefunction @xmath33 at the @xmath34 plane for a crystal occupying @xmath35 is taken to be @xmath36 , and hence it has no projection in the @xmath37 orbital . topological surface states in the form of a dirac cone exist when @xmath38.@xcite for spin along @xmath39 , the bound state energy is @xmath40 . hence , the positive energy branch has spin along @xmath41 . to account for the situation of be@xmath0si@xmath1 , @xcite we need to take @xmath42 , though we shall consider arbitrary relative signs of @xmath15 , @xmath43 and @xmath14 below for comparison purposes . since the bulk energies are given by @xmath44 , and so within this model , the surface states are always separated from the continuum for any given @xmath45 . now we consider the superconducting states , first for the bulk . for time - reversal and inversion symmetric systems , there is a pair of degenerate states at any given momentum @xmath3 , forming a pseudospin @xmath46 . cooper pairing occurs between opposite momenta @xmath3 and @xmath47 , and can be classified into even parity , pseudospin singlet and odd parity , pseudospin triplet states.@xcite these superconducting states can further be classified by their different symmetries under the crystal symmetries into different representations in group theory.@xcite these representations depend only on the point group ( but not the space group ) . possible forms of the corresponding momentum and pseudospin dependence in each group representation expressed in the form of basis functions : the general form of the order parameter can be a linear combination of the independent basis functions of the same symmetry , each term possibly multiplied by a momentum dependent function which is invariant under the particular group under consideration . they have in particular been listed for the cubic @xmath48 , tetragonal @xmath49 , and hexagonal @xmath50 groups @xcite . for @xmath50 we have the group representations @xmath51 , @xmath52 , @xmath53 , @xmath54 , @xmath55 and @xmath56 , with each of the above either even ( @xmath57 ) or odd ( @xmath58 ) parity . the corresponding table for @xmath30 appropriate for bi@xmath0se@xmath1 was not listed in these references , but can be trivially obtained from those for @xmath50 since @xmath50 would reduce to @xmath30 if we discard rotations about @xmath59 of odd multiples of @xmath60 and three of the horizontal rotational axes . in this case , @xmath51 is no longer distinguishable from @xmath52 , and similarly for @xmath53 and @xmath54 , and @xmath55 and @xmath56 . the resulting group representations and their basis function are listed in the first two columns of table [ table ] , following ref @xcite . for simplicity , we do not list all the possible independent basis functions , but mainly those which would appear again in the later part of this paper . for the complete basis function set , we refer the readers to the literature @xcite . in @xcite , various types of momentum independent ( local ) pairing in the orbital and spin basis of eq ( [ hn ] ) were considered . since this formulation involves states of eq ( [ hn ] ) without _ a priori _ distinguishing the conduction and valence band , we shall refer to this as the `` full two - band description '' @xcite the symmetry of these states were already discussed in @xcite , and we list each these states with their corresponding symmetries in column ( iii ) of table [ table ] . in this table , we have followed the notation of @xcite and use @xmath61 to label the two orbitals instead of @xmath62 of eq ( [ hn ] ) . for convenience of comparison with other works in the literature @xcite , we also list in columns ( iv ) and ( v ) these order parameters in matrix form , which we shall denote @xmath63 and @xmath64 and referred to as `` nambu i '' and `` nambu ii '' . @xmath65 is the matrix order parameter in the ordinary nambu notation after generalization to two orbitals , that is , if we use the operators as @xmath66 , where @xmath67 and @xmath68 s are annihilation and creation operators , and @xmath69 the two orbitals . if we use instead @xmath70 , as done in @xcite , ( or @xmath71 , as in @xcite ) the order parameter matrix is @xmath64 . the two notations are related simply by @xmath72 ( @xmath73 ) . the factorizing out of @xmath74 to the right has the effect of what has been done in the @xmath75he literature @xcite , where the order parameter matrix is written as @xmath76 , so that @xmath77 transforms as a vector under spin - rotations . in this way , it is clear from column ( v ) that the two entries listed under @xmath78 are related by a @xmath79 rotations about the @xmath28-axis . in making this table , we have made use of the gauge symmetry of superconductivity to simplify the matrices ( by removing factors like @xmath80 or @xmath81 ) . however , we have kept the correct relative phase between the two partners within @xmath82 , so that they have the correct relative transformation properties . @llllll@ ( i ) & & ( ii ) & ( iii ) & ( iv ) & ( v ) + even parity & & & & & + @xmath83 & & 1 & @xmath84 & @xmath85 & 1 + & & & @xmath86 & @xmath87 & @xmath24 + @xmath88 & & i m @xmath89 & & & + @xmath90 & & re @xmath91 & & & + & & i m & & & + odd parity & & & & & + @xmath92 & & @xmath93 ; @xmath94 & @xmath95 & @xmath96 & @xmath97 + @xmath98 & & @xmath99 & @xmath100 & @xmath101 & @xmath102 + @xmath78 & & re @xmath103 ; @xmath104 ; @xmath105 & @xmath106 & @xmath107 & @xmath96 + & & i m & @xmath108 & @xmath109 & @xmath110 + table [ table ] shows the correspondence between each local pair and the possible basis functions . however , it does not tell us directly what exactly the momentum dependences are , since multiplication of any basis function by a function invariant under the crystal symmetry is also an as good basis function . it also does not tell us , in particular for the case of odd - parity pairing , what linear combinations between the different inequivalent basis functions ( e.g. @xmath94 and @xmath93 for @xmath92 ) that we should take . we shall see later that this information is in fact useful for understanding the surface state spectrum . to do this , we have to first construct the single particle pseudospin wavefunctions obeying the correct crystal symmetries , as we shall do in section [ sec : wf ] . we shall then take up the task of showing how the pairings listed in column ( iii ) correspond to the basis functions listed in column ( ii ) in sec [ sec : bulk ] . now we construct the pseudospin wavefunctions . for each @xmath3 , we shall denote the two degenerate states by @xmath111 and @xmath112 . we shall demand that the corresponding pairs at @xmath47 are related to those at @xmath3 by the relations | - k , > & = & p @xmath113 the time - reversal . @xmath114 was already taken to be @xmath24 , and we shall take @xmath113 as @xmath115 times the complex conjugate . correspondingly , we have | k , > = p t |k , > [ pt ] and | - k , > = p |k , > [ p2 ] note that @xmath116 , as @xmath117 . @xmath118 and @xmath119 are also required to satisfy certain rotational symmetry properties , to be treated in details below . these requirements basically enable us to roughly think of @xmath120 and @xmath121 as spin - up " and spin - down " respectively.@xcite we shall first deal with eq ( [ p ] ) and ( [ t ] ) , ignoring the rotational properties for the moment . we shall thus first find an intermediate basis @xmath122 and @xmath123 obeying equations ( [ p ] ) and ( [ t ] ) without worrying about the rotational properties . for this purpose , let us first introduce spin wavefunctions which diagonalize the spin part of eq ( [ hn ] ) , using |s = z k > = ( 1 + i e^i _ k ) |s = -z k > = ( i e^-i _ k + 1 ) for spins along @xmath124 . here @xmath125 is the azimuthal angle of @xmath3 in the @xmath126 plane . these spin wavefunctions satisfy @xmath127 . it is then straight - forward to diagonalize eq ( [ hn ] ) . we shall define @xmath122 for @xmath128 ( the northern hemisphere " ) to be the one associated with spin along @xmath129 : |k , e^ i k r ( e_k + v k _ + m + i v_z k_z ) ( 1 + i e^i _ k ) [ ka1 ] where the first column matrix denotes the part in orbital space and the second part denotes the spin space . here @xmath130 is the energy of the particle ( which can be @xmath131 ) , and @xmath132^{1/2}$ ] is a renormalization factor . the other state @xmath123 for @xmath128 , as well as states for @xmath133 , are obtained by the symmetry requirements ( [ p]),([t ] ) and ( [ pt ] ) @xcite . we thus have , for @xmath128 , |k , i k r ( m - i v_z k_z + e_k + v k _ ) ( i e^ - i _ k + 1 ) [ kb1 ] and |- k , > e^ - i k r ( m + i v_z k_z + e_k + v k _ ) ( 1 + - i e^i _ - k ) [ -ka1 ] |- k , > e^ - i k r ( e_k + v k _ + m - i v_z k_z ) ( - i e^- i _ -k + 1 ) [ -kb1 ] we have written the last two wavefunctions for wavevectors in the southern hemisphere " using the labels @xmath134 with @xmath128 . this is for convenience later since we shall always be consider cooper pairs between @xmath3 and @xmath47 , and it is sufficient to write these pairs with @xmath128 for the pair due to the fermionic antisymmetry of wavefunctions . note that @xmath135 . we now proceed to find the wavefunctions @xmath136 and @xmath119 with the desired @xmath114 , @xmath113 , and rotational properties . one can of course directly study the wavefunction themselves . however , a more convenient way to proceed is to evaluate some physical quantity with known transformation properties ( _ c.f . _ @xcite ) . for this , we consider the spin operators projected onto our two - dimensional hilbert space @xmath136 and @xmath119 for each @xmath3 point . these operators are thus then also @xmath6 matrices . this spin operator is related to the effective magnetic moment of our quasiparticles , if the orbital contributions can be ignored ( which can indeed be the case if the relevant orbitals are just @xmath17 and the mixing to @xmath137 can be ignored ) . we shall therefore denote them as @xmath138 . anyway , the operators for this effective spin moment @xmath139 , @xmath140 , are simply _ j^ eff ( k ) = ( < k , | s_j | k , > & < k , | s_j | k , > + < k , | s_j | k , | s_j | k , > ) [ def - seff ] where the @xmath141 inside the matrices are the spin pauli matrices as in eq ( [ hn ] ) . the prime @xmath142 is to remind us that we are using the @xmath143 and @xmath144 basis at this moment . viewed as operators , we thus have _ j^ eff ( k ) = _ , = , | k , > < k , | s_j | k , > < k , | straight - forward calculations using eqns ( [ ka1 ] ) and ( [ kb1 ] ) give , for @xmath128 , ( those with @xmath133 can be found later by using eq ( [ p ] ) and ( [ t ] ) , this guarantees the correct properties under parity and time - reversal)@xcite _ x^ eff ( k ) = ( - sin _ k & |a_k| cos _ k e ^ - i ( _ k + _ k ) + _ y^ eff ( k ) = ( _ k & |a_k| sin _ k e ^ - i ( _ k + _ k ) + - i e ^ i ( _ k + _ k ) & 0 ) [ sz1 ] where a_k |a_k| e^ - i _ k ( e_k + v k _ ) ( m - i v_z k_z ) is a factor generated by the overlap of the orbital wavefunctions in eq ( [ ka1 ] ) and ( [ kb1 ] ) , and so |a_k| = [ aabs ] and e^ - i _ k = ( sgn e_k ) [ alpha ] the set of @xmath6 matrices @xmath145 , @xcite with @xmath146 etc do not yet have the desired transformation properties under rotation . we now construct a new basis @xmath147 , @xmath119 so that the corresponding pauli matrices @xmath148 for the pseudospin do transform like an axial vector . since the system has complete rotational symmetries about @xmath59 , we must require @xmath149 . to do this , we simply have to find a basis so that @xmath150 is diagonalized . this can be done by choosing |k , > = ( |k , > - i e^ i ( _ k + _ k ) |k , > ) [ transf1a ] and |k , > = ( |k , > - i e^ -i ( _ k + _ k ) |k , > ) [ transf1b ] where the phase factor @xmath151 is at this time arbitrary . note that we have demanded that @xmath119 be related to @xmath147 by eq ( [ pt ] ) . in this new basis , we find _ z^eff ( k ) = | a_k| _ z [ mzf ] and _ x^eff = ( 0 & ( |a_k| cos _ k + i sin _ k ) e^ -i ( _ k + _ k + 2 _ k ) + ( |a_k| cos _ k - i sin _ k ) e^ i ( _ k + _ k + 2 _ k ) & 0 ) _ y^eff = ( 0 & ( |a_k| sin _ k - i cos _ k ) e^ -i ( _ k + _ k + 2 _ k ) + ( |a_k| sin _ k + i cos _ k ) e^ i ( _ k + _ k + 2 _ k ) & 0 ) to proceed further , it is simplest to examine the radial and azimuthal components of @xmath152 and @xmath153 , _ i.e. _ @xmath154 and @xmath155 , and similarly for @xmath156 and @xmath157 , _ i.e. _ _ r ( 0 & e^-i _ k + e^i _ k & 0 ) and _ ( 0 & -i e^-i _ k + i e^i _ k & 0 ) evidently due to the existence of vertical reflection planes at arbitrary angles with respect to the x - axis , @xmath158 ( @xmath159 ) must simply be proportional to @xmath156 ( @xmath157 ) but would not involve the other component . one sees that we can choose @xmath151 to satisfy _ k + 2 _ k = 0 [ theta ] or @xmath160 . we shall adopt the first choice . in this case we get _ r^eff = | a_k| _ r [ mrf ] and _ ^eff = _ [ mpf ] for this choice , a pseudospin along the positive azimuthal direction would correspond also to an effective magnetic moment and hence spin along the same direction . ( the alternate choice would give @xmath161 and @xmath162 instead . ) back to the cartesian form , we have _ x^eff ( k ) = _ x - ( 1 - |a_k| ) k_x ( k_x _ x + k_y _ y ) _ y^eff ( k ) = _ y - ( 1 - |a_k| ) k_y ( k_x _ x + k_y _ y ) which explicitly shows that @xmath163 has the same transformation properties as @xmath164 respectively . note that the procedure above also gives us the effective @xmath57-factor for the effective moments . for magnetic moment along @xmath28 and the radial component @xmath165 , eq ( [ mzf ] ) and ( [ mrf ] ) show that they are reduced by the factor @xmath166 given in eq ( [ aabs ] ) , but there is no reduction for the @xmath167 component . for @xmath3 on the fermi surface , @xmath168 , @xmath169 for @xmath3 parallel or antiparallel to @xmath59 . it decreases for increasing @xmath170 , and for @xmath3 in the x - y plane , @xmath171 , which can be substantially less than unity , as in the case relevant to the experiments @xcite . this effective moment would be relevant when considering questions such as pauli limiting of upper critical field @xcite , or spin susceptibilities measured by knight shifts . returning to the pseudospin basis , eq ( [ transf1a ] ) and ( [ transf1b ] ) become |k , > = ( |k , i ( _ k + _ k ) states at @xmath172 , @xmath128 , can be obtained by using eq ( [ p ] ) : |-k , > = ( |-k , > - i e^ i ( _ k + _ k ) |-k , > ) [ -kaf ] and |-k , > = ( |-k , > - i e^ -i ( _ k + _ k ) |-k , > ) [ -kbf ] with @xmath173 and @xmath174 available in eq ( [ ka1 ] ) ( [ kb1 ] ) , ( [ -ka1 ] ) ( [ -kb1 ] ) , this completes our construction of the pseudospin basis . we shall express the cooper pair wavefunctions in terms of it in the next section . before we proceed , since we would also be interested in surface bound states in the superconducting states in sec [ sec : surface ] , we consider reflection of quasiparticles at a surface in the normal state before we end this section . we consider a crystal occupying @xmath35 , with a surface at @xmath34 . consider incident wavevector @xmath175 , @xmath128 . due to our ways of writing wavefunctions for wavevectors in the southern hemisphere , it is convenient to write the reflected wavevector as @xmath176 where @xmath177 and use eq ( [ -kaf ] ) and ( [ -kbf ] ) , and note that @xmath178 . straight - forward algebra shows that @xmath147 is reflected only into @xmath179 , and similarly for @xmath180 . indeed , the wavefunctions @xmath181 @xmath182 or @xmath183 , with the reflection coefficient @xmath184 @xmath185 satisfy the boundary condition @xmath186 , since one can easily verify that @xmath187 and similarly with @xmath180 . hence the reflection of quasiparticles at @xmath34 in the normal state does not alter the pseudospin species , nor is the phase shift dependent on the incident species . hence , in the model of eq ( [ hn ] ) , the @xmath34 surface is _ pseudospin - inactive _ , a result that we would use in sec [ sec : surface ] . our single band description here is unable to capture the surface states of a ti . these states are superposition of states from both the conduction and valence bands . the implication of this for the surface states of the superconducting phases would be discussed in sec [ sec : surface ] . it is straight - forward to obtain the order parameters for the superconducting states in our pseudospin basis . we discuss each of the phases listed in table [ table ] in turn . we confine ourselves to momentum independent pairing within the @xmath188 , @xmath189 basis of eq ( [ hn ] ) . generalization to additional momentum dependence is straight - forward . so far for cu : bi@xmath0se@xmath1 , superconductivity has been found only for @xmath190,@xcite but we shall also consider general sign of @xmath7 in the following . @xmath191 : both the intra - orbital opposite spin pairing " @xmath192 and the inter - orbital singlet pairing " @xmath193 has @xmath83 symmetry . in general , they are expected to be mixed . however , since they have also been discussed separately in , e.g. , @xcite , we shall also first do the same likewise , and consider a general linear combination later . for ease of referral , we shall refer these two states as @xmath194 and @xmath195 respectively . @xmath194 : the pair wavefunction is @xmath192 . the corresponding form in the pseudospin language is just @xmath196 $ ] here the prime over the sum means that @xmath3 is restricted to the `` upper hemisphere '' ( as those in the other hemisphere is already included by antisymmetry ) , and @xmath197 can be evaluated using ( [ kaf])-([-kbf ] ) and eq ( [ ka1])-([-kb1 ] ) . we find that this simply reduces to @xmath198 $ ] , with no additional momentum dependent factors . if the pairing term in the superconducting hamiltonian is taken as @xmath199 where @xmath200 are the creation operators of orbit @xmath188 and spin @xmath189 and @xmath201 indicates the hermitian conjugate , then the corresponding term in the pseudopin basis simply reads @xmath202 . this has just the familiar form for momentum independent conventional @xmath189-wave pairing . the quasiparticle energies in the superconducting state , measured with respect to the chemical potential @xmath7 , are @xmath203 with @xmath204 in the familiar form , for @xmath205 e^2_s = ( _ k - ) ^2 + ( _1)^2 [ esa11 ] where we have taken the gauge where @xmath206 is real . the corresponding formula for @xmath207 is e^2_s = ( _ k + ) ^2 + ( _1)^2 . [ esa11n ] note that since we started with a normal metal and then introduce the superconducting pairing , we necessarily have @xmath208 implicitly , and weak - superconducting pairing actually requires further that @xmath209 . for the full two - band description , the hamiltonian in the nambu - ii notation is just @xmath210 ( see table [ table ] ) , where @xmath211 are the pauli matrices in particle - hole space . now there are instead two pairs of allowed @xmath204 due to the presence of two bands , given by @xmath212 , thus including both eq ( [ esa11 ] ) and ( [ esa11n ] ) irrespective of the sign of @xmath7 . @xmath213 : the pair wavefunction is @xmath214 . following the same procedure described above gives the result @xmath215 in the pseudospin basis . if the pairing term in the hamiltonian is written as @xmath216 with @xmath217 real , the corresponding quasiparticle energies are @xmath203 with e_s^2 = ( _ k ) ^2 + ( ^_1 ) ^2 [ ea11 ] with the upper and lower signs for @xmath190 and @xmath207 respectively . again the energy gap is isotropic in @xmath3 space and is given here simply by @xmath218 . note that the system becomes gapless if @xmath219 even for finite @xmath217 . on the other hand , in the full two - band description , with hamiltonian @xmath220 , we obtain again two pairs of energies , given by e_s^2 = _ k^2 + ^2 + ( ^_1)^2 2 ^1/2 considering the lower energy branch ( where @xmath221 for @xmath222 ) and taking the weak - pairing @xmath223 approximation , we recover eq ( [ ea11 ] ) , as expected . generally , the system is gapped whenever @xmath224 and @xmath225 . if @xmath226 , we recover the normal state , and states at momentum @xmath3 such that @xmath227 have @xmath228 . if @xmath229 , gaplessness occurs if @xmath230 can be satisfied , and at positions slightly different from the one - band result @xmath227 with correction due to finite @xmath217 . general @xmath83 : the most general @xmath83 pairing wavefunction is a linear combination of that of @xmath194 and @xmath195 . the general pairing hamiltonian is @xmath231 . the corresponding expression in pseudospin language is again just the linear combination of those given above for @xmath194 and @xmath195 . we shall , for simplicity , restrict ourselves only to the case where time - reversal symmetry is preserved , thus @xmath232 and @xmath217 can be chosen to be real simultaneously , though either can be positive or negative . the energy spectrum is , in the single band description e_s^2 = ( _ k ) ^2 + ( _1 + ^_1 ) ^2 [ es1 g ] the system is gapped if @xmath233 . the corresponding result in the full two - band description is e_s^2 = _ k^2 + ^2 + ( _1)^2 + ( ^_1)^2 2 ^1/2 [ es1f ] the lower energy branch reduces to eq ( [ es1 g ] ) in the weak - pairing limit . if the full expression ( [ es1f ] ) is used , one can check that the system is gapped whenever @xmath234 . ( thus recovering the one band result since there @xmath7 must be finite ) . if @xmath235 , gaplessness still requires the condition @xmath236 be satisfied . hence we have the following : ( 1 ) : if @xmath237 , we need @xmath238 and @xmath239 for gaplessness . note that for @xmath240 , the latter happens if and only if @xmath241 . we reproduce the weak superconducting pairing results when @xmath242 and @xmath217 . \(2 ) for @xmath243 , the system is gapped whenever @xmath224 and @xmath225 . ( 2a ) if @xmath244 , gaplessness occurs only if @xmath245 . ( 2b ) if @xmath226 , the system reduces to @xmath194 , and the system is fully gapped unless @xmath232 also vanishes . we shall use these results when we discuss the surface states and topology in sec [ sec : surface ] and the appendix . @xmath246 : the pairing wavefunction , @xmath247 becomes _k if the pairing term in the hamiltonian is written as @xmath248 , with @xmath249 real and positive , the corresponding form in pseudospin is @xmath250 with , when quasiparticles are taken at the fermi energy @xmath7 , d_x , y = _ 1u [ da1u ] d_z = _ 1u ( sgn ) [ dza1u ] as compared with the balian and werthamer ( bw ) state @xcite where @xmath251 ( or @xmath252 anti - parrallel to @xmath3 , after a gauge transformation ) , the @xmath4 here has a very peculiar form . the ratio between the in - plane and @xmath28 component is = [ dratio ] besides the anisotropy factors from the velocities , an extra factor @xmath253 arises , which suppresses the @xmath254 component relative to @xmath255 if @xmath256 . this , in retrospect , is actually not surprising since the pairing is between opposite spins in the original @xmath188 and @xmath189 basis . a pure opposite pseudospin pairing would have @xmath77 parallel or antiparallel to @xmath59 . the @xmath257 components of @xmath77 are actually generated by spin - orbit coupling . moreover , we note that the relative signs between @xmath258 and @xmath255 depends on the signs of the various parameters of the system , in particular @xmath259 . we shall come back to this when we discuss the surface bound states . we note here also that this peculiar relative sign and magnitudes between the components of @xmath77 is allowed here due to the inequivalence between @xmath28 and @xmath260-@xmath261 under the relevant @xmath30 symmetry . for a cubic system such as yptbi , @xcite @xmath94 and @xmath93 necessarily comes in the combination @xmath262 . despite the peculiar form for @xmath77 , the energy gap turns out to be isotropic in the weak - coupling limit . the square of this gap is given by @xmath263 , which is which works out to be simply @xmath264 when we restrict ourselves to particles near the fermi surface , where @xmath265 . for the energy gap , the anisotropies due to eq ( [ dratio ] ) and the overall factors @xmath266 in eq ( [ da1u ] ) and ( [ dza1u ] ) cancel each other . the quasiparticle energies are just @xmath267 . this phase is fully gapped ( provided @xmath208 , which , as mentioned , is necessarily the case for a single - band weak - pairing superconductivity picture to be meaningful ) . in the full two - band description , the quasiparticle energies are already worked out in @xcite : e_s^2 = _ k^2 + ^2 + _ 1u^2 2 ^1/2 [ esa1u ] which reduces to what has been just given in the weak - pairing limit . in this two - band description , the system can still be gapless when @xmath268 provided @xmath269 ( with gapless point at @xmath270 ) . we shall use this result later in the appendix . @xmath271 : the pair wavefunction @xmath272 becomes @xmath273 $ ] . if the pairing term is written as @xmath274 , the corresponding @xmath252 is , in the weak pairing limit , d ( k ) = _ 2u [ da2u ] the magnitude of the gap is just @xmath275 . this is the usual planar phase in the @xmath75he literature , and is regaining attention due to its analogy with topological insulators in two - dimensions ( e.g. @xcite ) . in three - dimension however , this state has point nodes in the gap at the north and south poles of the fermi surface , where @xmath276 , @xmath277 both vanish . the expression for the quasiparticle energies in the two - band description is e_s^2 = _ k^2 + ^2 + _ 2u^2 2 ^1/2 [ esa2u ] which reduces to the above results in the weak - pairing limit . the state is gapped at all @xmath278 . gaplessness can occur if the condition @xmath279 can be satisfied . @xmath280 : this is a two - dimensional representation , as @xmath281 is the time - reversed of @xmath282 . generally , the superconducting state can be a superposition of the two . let us first consider the state @xmath283 ( we have inserted an @xmath284 factor for later convenience . ) if the pairing term in the hamiltonian is given by @xmath285 , then we have d ( k ) & = & _ u \ { ( sgn ) ( + ) v_z k_z r_+ - z . + ( sgn ) ( - ) v_z k_z ( ( 2 _ k ) x + i ( 2 _ k ) ) r_- } [ deu ] this @xmath252 is complex ( @xmath286 ) reflecting the fact that the state @xmath282 has broken time - reversal symmetry . the first two terms in eq ( [ deu ] ) are proportional to @xmath287 and @xmath288 listed under @xmath82 in table [ table ] . the last term has a more complicated momentum dependence , but since for small @xmath45 , it is simply proportional to @xmath289 , the third independent basis function listed in table [ table ] , in this limit . the spectrum for this state is complicated since it is `` non - unitary '' , that is , the energy of the two pseudospin - species at the same @xmath3 point are typically unequal , due to the lack of time - reversal symmetry . we shall not investigate this phase in detail , but turn to the time - reversal symmetric states within this two - dimensional manifold . let us consider then @xmath290 . this state is just the linear combination of the one discussed above and its time - reversal conjugate . the @xmath252 vector for this state is therefore simply twice the real part of eq ( [ deu ] ) , and so d ( k ) = _ u \ { ( sgn ) ( + ) v_z k_z x - k_x z . ( sgn ) ( - ) v_z k_z ( ( 2 _ k ) x + ( 2 _ k ) y ) } [ deu1 ] corresponding to the basis functions listed in the first line under @xmath82 in table [ table ] . despite its complicated form , the square of the gap , obtained from @xmath263 , is given simply by _ u^2 this state has two point nodes , as it is gapless for @xmath3 parallel to @xmath291 . the result is in accordance with the full two - band result , which is e_s^2 = _ k^2 + ^2 + _ u^2 2 ^1/2 and is just eq ( [ esa2u ] ) with @xmath292 . the gap - squared in the weak - coupling limit is _ u^2 for momenta on the fermi surface . the point node for this phase has also been noted in @xcite . @xmath252 for the state @xmath293 is evidently twice the imaginary part of eq ( [ deu ] ) . it is just eq ( [ deu1 ] ) rotated by @xmath79 about @xmath59 . now we consider the surface states in the superconducting phases . we shall focus on the odd parity states since they have received more attention in the literature . ( see however the appendix ) for weak superconductors ( pairing potential much smaller than fermi energy ) , surface bound states are most conveniently discussed quasiclassically . bound states can be formed at the surface since the quasiparticle with incident wavevector @xmath294 sees a different order parameter from when it is reflected into @xmath295 . since the surface @xmath34 is pseudospin inactive @xcite , the problem maps to the evaluation of the quasiparticle bound states at a one - dimensional junction , where the order parameter for @xmath296 is different from that for @xmath297 . here , the order parameter for @xmath296 can be identified with that of @xmath294 , and @xmath297 with @xmath295 . in all situations relevant to us , the magnitude of the order parameter @xmath298 for @xmath294 and @xmath295 are identical , only the phase @xmath299 s are different . the effective phase difference for the junction is @xmath300 . the absolute value of the bound state energy is just @xmath301 if assume that the order parameter is constant up to the surface . since we shall be interested also in the sign of @xmath302 , we give a short discussion of it @xcite . for normal state with particle - like dispersion , where the energy is increasing with the magnitude of the wavevector , the bound state energy for positive @xmath303 momentum is @xmath304 for @xmath305 , and @xmath306 for @xmath307 , and vice versa for negative @xmath303 momentum . the above signs should be reversed for hole - like normal state dispersions @xcite . for our normal state , the spectrum is particle - like for @xmath205 , and hole - like if @xmath207 . for time - reversal symmetric odd parity superconductors , @xmath252 can be chosen real . it is best to work with the quantization axes @xmath308 ( no relation to @xmath303 above ) which is perpendicular to both @xmath309 and @xmath310 . we shall choose @xmath308 to be parallel to @xmath311 . then the order parameter with @xmath308 as the quantization axis is given by ( - d_x_s + i d_y_s & 0 + 0 & d_x_s + i d_y_s ) hence diagonal in pseudospin space . consider first the `` down '' component . we can write @xmath312 where @xmath313 is just the angle for @xmath314 in the @xmath315 - @xmath316 plane , measured counterclockwise from the @xmath315 axis . hence the phase difference @xmath317 for the junction `` is just @xmath318 , the angle between @xmath319 and @xmath320 , with @xmath321 , and with @xmath322 for particle - like normal state spectrum . for pseudospin ' ' up " along @xmath308 , we can rewrite @xmath323 , and the effective phase difference is @xmath324 . the bound state energy is @xmath325 for a particle - like normal state spectrum . summarizing , the bound state energy is positive ( negative ) if the pseudospin is parallel ( antiparallel ) to @xmath326 . we are particularly interested in comparing this sign with the bound state in the normal phase of our ti . we recall that , in our model , the energy is positive if the spin is parallel @xmath327 , where @xmath328 is the surface normal ( pointing outward from sample ) . hence , if we focus on the relative sign between the superconducting and the normal state , we can state that : the bound state dispersion for the normal and superconducting phases has the same sign if @xmath326 is parallel to @xmath327 . + we shall call this situation as `` regular relative to normal '' ( rn ) . conversely , we shall call it `` anomalous relative to normal '' ( an ) . actually , another meaningful comparison would be to the surface bound state for the bw phase where @xmath77 is parallel to @xmath3 . in that case , then the dispersion has the same ( opposite ) sign as the bw phase if @xmath311 is parallel ( antiparallel ) to @xmath329 . since @xmath294 and @xmath295 differ only by the component along the surface normal , we see that @xmath329 is simply parallel to @xmath330 . we shall mainly be focusing on the first comparison , though the comparison with the bw phase can be directly read - off from the expressions below . before we proceed further , we remark here that the above single band argument only takes into account bound states formed by superposition of particle and holes of the same band , with energies close to the fermi level . for the ti in its normal phase however , there are bound states at @xmath331 formed by superposition of the conduction and valence band . hence , our single band approximation for superconducting bound states is applicable only when the states at @xmath332 are sufficiently far away from @xmath7 so that we can ignore the hydridization of our states with these due to the superconducting pairing @xmath8 , _ i.e. _ , we need @xmath333 . for @xmath334 , we thus we expect that we can capture the superconducting bound states only when @xmath335 , the fermi momentum . more discussion on this will be given below . now we apply our above results to the odd - parity phases . @xmath92 : @xmath252 is available in eq ( [ da1u ] ) and ( [ dza1u ] ) . we write @xmath336 and @xmath337 , @xmath128 . since the system is rotationally symmetric about @xmath59 , let us consider @xmath338 . we get ( sgn ) d_in d_out = 2 | _ 1u|^2 y with @xmath339 . we see that the dispersion is rn if @xmath340 , but an if @xmath341 . the magnitude of the group velocity of the bound state , @xmath342 , can also be obtained easily . for small @xmath276 , @xmath343 , since @xmath299 is close to @xmath344 , so @xmath345 . using @xmath346 , we get | d e_b / d k_x | = |_1u| ( 1 - ( ) ^2 ) ^-1/2 for @xmath347 , this group velocity is reduced compared with the bw phase by a factor @xmath348 ( see also @xcite ) . in our single band description , the bound state energy approaches the bulk gap when @xmath45 approaches the fermi momentum @xmath349 , here given by @xmath350 , as the effective phase difference @xmath317 vanishes when @xmath294 and @xmath295 becomes parallel . the bound state spectrum is thus of the form in fig [ fig1](a ) or fig [ fig1](b ) according to whether @xmath351 ( when @xmath15 taken as @xmath352 according to sec [ sec : back ] ) . in contrast , the bound state spectrum in the full two - band calculation for @xmath341 is schematically shown in fig [ fig1 ] ( c ) @xcite . though our one - band model captures correctly the sign of the group velocity for @xmath353 , it does not capture the behavior at larger @xmath45 . this is in retrospect not surprising . for large @xmath45 , the cooper pairing plays no role , and the sign for the dispersion must be the same as the corresponding normal phase . in the later case , when the system is a ti , @xmath341 and the positive energy branch has spin along @xmath41 . hence , for @xmath341 , the sign of the dispersion at small and large @xmath170 must be opposite , as in fig [ fig1](c ) . this sign change has been found earlier by other authors @xcite . an alternative view of this sign change has been given by @xcite . ( see also appendix [ sec : app ] for more discussions . ) @xmath98 : @xmath4 is given in eqn ( [ da2u ] ) , which is independent of the sign of @xmath43 and @xmath14 . since @xmath252 is independent of @xmath354 , @xmath355 . there are no surface bound states ( in our approximation where the order parameter is constant up to the surface ) . this conclusion is in agreement with fig 5 a , c of @xcite . note that the gap magnitude vanishes for normal incidence , where @xmath356 . we consider the state @xmath357 , the first line under @xmath78 in table [ table ] . ( the result for the second line is the same except for a @xmath79 rotation about @xmath28 ) . @xmath252 is available in eq ( [ deu1 ] ) . for near normal incidence , it can be rewritten as d ( k ) = _ u \ { - z + ( sgn ) x + ( sgn ) y } the factor @xmath358 only gives a small correction to the results below and will be ignored for simplicity . for the @xmath276-@xmath354 plane , @xmath359 , ( sgn ) d_in d_out = 2 _ u^2 y whereas @xmath360 . thus we have rn if @xmath340 and an if vice versa . for the later case , as argued in the last paragraph for @xmath92 , the dispersion should be the same as the normal phase for large @xmath170 , hence we again expect a sign change for the group velocity at some @xmath170 . this is in accordance with fig 6a of @xcite ( recall @xcite ) and of @xcite . before we depart from this section , we would like to make a remark on the more general case where the momentum dependence of the term @xmath43 in eq ( [ hn ] ) is included . all the above single - band calculations can be simply generalized to this case . for weak superconducting pairing , it is the value of @xmath361 at the fermi surface that is physically relevant . for a ti , @xmath362 where @xmath363 and @xmath364 has opposite signs ( we shall ignore possible anisotropies in the @xmath364 term as they do not affect the arguments below ) . thus it is possible that the sign of @xmath361 for @xmath3 at the fermi surface be different from @xmath365 . hence , @xmath366 can be positive for a ti , even though @xmath367 is required , and the spectra can change from an to rn with increasing @xmath368 or @xmath369 ( for @xmath92 and @xmath78 ) , at the point where @xmath370 in the @xmath371 limit . this is a simple explanation of the finding of @xcite . in this respect , thus rigorously speaking , rn or an is not an indication of topological character of the underlying normal phase , but rather the @xmath252 configuration of the superconducting phase . in this paper , we have constructed a pseudospin basis to describe the normal state of bi@xmath0se@xmath1 . superconductivity is then expressed in this basis . using this approach , many of our previous knowledge in unconventional superconductivity can then be directly applied , especially for the bulk . we have also shown that many features of the surface bound states can also be understood in this way . although we have concentrated on the surfaces parallel to the bi@xmath0se@xmath1 quintuple layers in this paper , the same considerations are applicable to other surfaces as well as systems with other symmetries . this picture however misses the topological properties of the normal phase , which in turn some features of the surface bound states in the superconducting phases , but only at momenta comparable with or larger than the fermi momenta in the weak - coupling limit . this research was supported by the national science council of taiwan under grant number nsc101 - 2112-m-001 - 021-my3 . i also thank lei hao , t. k. lee , liang fu for discussions and communications of their results which motivate me to this work , and bor - luen huang for his help in preparing figure 1 . we consider discuss some topological aspects and the surface states for the @xmath92 and @xmath83 phases . let us first consider the @xmath92 phase , and begin with simple continuity arguments . for simplicity , we shall consider non - vanishing @xmath15 and @xmath14 , varying only @xmath43 . for a single - band model with @xmath252 in eq ( [ da1u ] ) and ( [ dza1u ] ) , we note that at @xmath219 , @xmath252 only has @xmath59 component and is odd in @xmath354 . the bulk state then has a line node on the equator , and the surface state is simply a flat band independent of @xmath170 . this is how the surface state spectra evolve between fig [ fig1](a ) and fig [ fig1](b ) when @xmath43 changes sign . time - reversal symmetric superconductors in three - dimension can be characterized by a winding number @xmath372 . @xcite this value can be evaluated by first transforming the hamiltonian into an off - diagonal form . the hamiltonian for the @xmath92 phase in the nambu - ii notation is , in the one - band model , @xmath373 . here @xmath374 is the kinetic energy measured with respect to the chemical potential @xmath7 . for example , for a quadratic band with particle - like dispersion , @xmath375 with @xmath205 , whereas for a hole - like band , @xmath376 with @xmath207 . here @xmath377 is an effective mass ( not to be confused with @xmath43 in eq ( [ hn ] ) ) . the hamiltonian becomes off - diagonal under a rotation in @xmath378 space , such as @xmath379 , @xmath380 . then the hamiltonian becomes h_s = ( 0 & h_s + h_s^ & 0 ) [ hod ] where @xmath381 . the winding number can be evaluated from w = d^3 k _ abc tr [ w ] where @xmath382 , @xmath383 is the fully antisymmetric tensor , and @xmath384 is a unitary matrix . here @xmath385^{1/2}$ ] . for the bw phase with @xmath77 parallel to @xmath3 , we get @xmath386 . @xcite for our state with eq ( [ da1u ] ) and ( [ dza1u ] ) , we get instead w = ( sgn ) sgn ( _ 1u v_z ) independent of the sign for @xmath43 , since both @xmath258 would change sign under a sign change of @xmath43 . thus , the winding number seems insufficient to indicate the possible change in the surface state spectra ( and the associated bulk topology ) between fig [ fig1 ] ( a ) and ( b ) . now we turn to the full two - band model , and again first employ only continuity arguments . the state is gapped if @xmath268 so long as @xmath387 . hence , one can change sign of @xmath43 without going through any gapless phase , provided the above inequality is satisfied . hence when @xmath388 changes sign , it can not affect which spin species is connected to the @xmath389 band at large @xmath170 , even though the sign of the dispersion can change for smaller @xmath170 . hence , continuity argument shows that fig [ fig1](c ) should evolve to fig [ fig1](a ) when @xmath388 changes from @xmath352 to @xmath390 . we can also examine the winding number @xmath372 in the full two - band model . the hamiltonian in the nambu - ii notation @xmath391 becomes off - diagonal by the same rotation in @xmath378 space mentioned before , with @xmath392 . since the state is gapped so long as @xmath387 and the winding number can not change within a gapped phase , one can first turn off @xmath43 and then @xmath7 , provided @xmath268 . the eigenvalues in eq ( [ esa1u ] ) become degenerate and we can use @xmath393 , with @xmath394 unitary , in eq ( [ w ] ) . a direct evaluation gives @xmath395 ( for @xmath396 ) irrespective of the sign of @xmath388 . there is no topological phase transition between @xmath341 and @xmath390 within the @xmath92 phase . we note that the single band model therefore produces a spurious topological change when @xmath43 changes sign . lastly , we consider the even parity @xmath83 phase . in the single - band picture , this is the ordinary @xmath189-wave superconductor . there is no phase difference for the order parameter at @xmath294 and @xmath295 , and so no bound states are expected . the surface states for @xmath194 and @xmath195 phases have been investigated by @xcite . they showed that there are no surface states for @xmath194 , in accordance with above . interestingly , they found that surface states survives for @xmath195 . the surface states seem to be in the form of two dirac cones , related by the particle - hole symmetry of the superconductivity , and crossing each other at the chemical potential . we do not have a simple explanation of this in our single - band picture . hao and lee @xcite noted that the @xmath217 term does not split the crossing at the fermi level . it is unclear whether this reflects any topology of the @xmath195 phase , such as a possible even winding number that is allowed for a time - reversal symmetric even parity superconducting state @xcite . we here simply note that , if a general @xmath83 is considered , the pure @xmath195 state and the pure @xmath194 state are connected in the sense that one can find a path in parameter space which connects them with the state remaining fully gapped ( see sec [ sec : bulk ] ) . the @xmath194 phase is topologically trivial , and no surface states are expected . thus the surface states of @xmath195 are expected to be destroyed in general once a finite @xmath232 is introduced . this is in accordance with the argument of @xcite where they showed that @xmath232 would introduce a finite matrix element coupling the two surface states of @xmath195 . y. xia , d. qian , d. hsieh , l. wray , a. pal , h. lin , a. bansil , d. grauer , y. s. hor , r. j. cava , and m. z. hasan , nature phys . 5 , 398 ( 2009 ) . h. zhang , c .- x . qi , x. dai , z. fang and s .- c . zhang , nature physics * 5 * , 438 ( 2009 ) d. hsieh , d. qian , l. wray , y. xia , y. s. hor , r. j. cava and m. z. hasan , nature ( london ) * 452 * , 970 ( 2008 ) d. hsieh , y. xia , l. wray , d. qian , a. pal , j. h. dil , j. osterwalder , f. meier , g. bihlmayer , c. l. kane , y. s. hor , r. j. cava and m. z. hasan , science * 323 * , 919 ( 2009 ) y. l. chen , j. g. analytis , j .- h . chu , z. k. liu , s .- k . mo , x. l. qi , h. j. zhang , d. h. lu , x. dai , z. fang , s. c. zhang , i. r. fisher , z. hussain , z .- x . shen , science , * 325 * , 178 ( 2009 ) d. hsieh , y. xia , d. qian , l. wray , j. h. dil , f. meier , j. osterwalder , l. patthey , j. g. checkelsky , n. p.ong , a. v. fedorov , h. lin , a. bansil , d. grauer , y. s. hor , r. j. cava and m. z. hasan , nature * 460 * , 1101 ( 2009 ) x .- l . qi , t. l. hughes , s .- c . zhang , phys . 78 , 195424 ( 2008 ) x .- l . qi , r. li , j. zang and s .- c . zhang , science , * 323 * , 1184 ( 2009 ) y. s. hor , a. j. williams , j. g. checkelsky , p. roushan , j. seo , q. xu , h. w. zandbergen , a. yazdani , n. p. ong , and r. j. cava , phys . 104 , 057001 ( 2010 ) bi@xmath0te@xmath1 is also found to be superconducting under pressure @xcite . however , the bulk carriers are holes which are located away from the @xmath12 @xmath5 point , a rather different situation from cu : be@xmath0se@xmath1 . we shall not discuss this compound here any further in this paper . j. l. zhang , s. j. zhang , h. m. weng , w. zhang , l. x. yang , q. q. liu , s. m. feng , x. c. wang , r. c. yu , l. z. cao , l. wang , w. g. yang , h. z. liu , w. y. zhao , s. c. zhang , x. dai , z. fang and c. q. jin , proc . u. s. a. * 108 * , 24 ( 2011 ) p. w. anderson , phys . b 30 , 1549 ( 1984 ) g. e. volovik and l. p. gorkov , pisma zh . 39 , 550 ( 1984 ) [ jetp lett . 39 , 12 ( 1984 ) ] ; zh . 88 , 1412 ( 1985 ) [ sov . phys jetp 61 , 843 ( 1985 ) ] . k. ueda and t. m. rice , phys . rev . b 31 , 7114 ( 1985 ) . e. i. blount , phys . rev . b 32 , 2935 ( 1985 ) . m. sigrist and k. ueda , rev . 63 , 239 ( 1991 ) s. yip and a. garg , phys . rev . b 48 , 3304 ( 1993 ) robert joynt and louis taillefer , rev . 74 , 235 ( 2002 ) different models have also been employed in the literature , e.g. , @xcite , hence some modifications must be taken to compare those results with here . in model i of @xcite , the spin part in eq ( [ hn ] ) was instead written as @xmath397 . hence all the spins in @xcite must be rotated by @xmath79 about z in order to compare with ours here . fortunately , this would not alter anything for pairings between opposite spins , that is , for the @xmath83 , @xmath92 and @xmath98 phases in table [ table ] . for the two - dimensional odd parity state @xmath82 which involves pairing between parallel spins , this rotation brings in the factors @xmath398 for @xmath399 and @xmath400 respectively , thus interchanges the two entries under @xmath82 in table [ table ] . we shall not try to provide the comparison with other models in this paper . we are using this term with a different meaning from the usual `` two - band superconductivity '' in the literature . here , we only have one fermi surface for the corresponding normal phase in the absence of pairing potentials . there are no entries listed under columns ( iii ) , ( iv ) and ( v ) for @xmath88 and @xmath401 , since they do not arise if we insist that there is no momentum dependence when the pairs are expressed in the orbital and spin basis . they can however occur if we allow more general momentum dependence . we shall not worry about how we define states at @xmath402 . we can define these by approprate limiting procedures . this can lead to discontinuities of these wavefunctions as functions of @xmath3 ( see @xcite ) , but these complications would not occur for the pair wavefunctions , our major concern in this paper . we can separate the reflection due to the vaccum - sample surface and the andreev reflection due to the order parameter variations by imagining inserting a thin normal layer near the surface , as done in , _ e.g. _ @xcite . since the reflection coefficient for the hole is given by the complex conjugate of that of the particle , the phase shift @xmath408 from the reflection coefficient @xmath409 of sec [ sec : wf ] cancels , if we ignore the small wavevector difference between the particle and hole .

we explore situations in which certain stochastic and high - dimensional deterministic systems behave effectively as low - dimensional dynamical systems . we define and study moment maps , maps on spaces of low - order moments of evolving distributions , as a means of understanding equations - free multiscale algorithms for these systems . we demonstrate how nonlinearity arises in these maps and how this results in the stabilization of metastable states . examples are shown for a hierarchy of models , ranging from simple stochastic differential equations to molecular dynamics simulations of a particle in contact with a heat bath . equation - free , multiscale , moment map , metastable states .

cond-mat0508551

an equation - free framework has recently been developed as a means of computationally analyzing the dynamical behavior of a large class of complex , multiscale dynamical systems . the systems may be either stochastic or deterministic with many of degrees of freedom and subject to random initial data . the key observation behind the equation - free framework is that in many cases the quantities of interest are averages or low - order moments of evolving distributions which are smooth in space and time and which evolve effectively as closed , low - dimensional systems . in effect the low - order moments evolve as though they are governed by reduced closed equations , even though the reduced equations are not analytically available . algorithms for performing scientific computing tasks such as numerical integration , or bifurcation and stability analysis of these unavailable reduced equations have been developed ( e.g. coarse projective integration , coarse newton - gmres , see @xcite and references therein ) . these algorithms are based on traditional , continuum numerical analysis , wrapped " around direct fine - scale simulation . the purpose of this paper is to establish a mathematical framework for understanding the behavior of coarse dynamics and coarse bifurcation methods on problems which exhibit metastable behavior . the systems we consider are exemplified by the following model . a particle , called the distinguished particle , with position @xmath0 and momentum @xmath1 sits in a potential well @xmath2 . it is coupled via linear springs to a large number @xmath3 of particles comprising a heat bath ; see figure [ fig : sketch ] . the potential well considered here is a slightly asymmetric double well . the full system is an @xmath4 degree of freedom hamiltonian system . a detailed description of the model , including the choice of spring constants , masses , and initial data for the bath particles is given below and in @xcite . the important point is that the dynamics of this simple model ( and the others that we consider in this paper ) is typical of many more complex molecular and stochastic systems in which the state is primarily confined to a few conformations ( here defined by the minima of @xmath5 ) with rare switching events between them . this is illustrated in figure [ fig : intro_qp ] with a typical trajectory and time series for the distinguished particle . other ( bath ) particles ( three are shown as illustration ) . the full system has @xmath4 degrees of freedom . , width=240 ] most of the time the distinguished particle is located in one of the two potential wells , but occasionally it makes a jump between wells.,width=192 ] consider now the dynamics of an ensemble of trajectories for the model system . figure [ fig : intro_ensemble ] illustrates the evolution of an ensemble of trajectories all with the same initial conditions @xmath6 for the distinguished particle , but with different initial data for the bath particles . ( the total initial energy of the bath is approximately the same for all realizations . [ sec : models ] . ) over a time of order @xmath7 the initial density evolves to a nearly gaussian density centered near the bottom of the right well , where it remains roughly constant for some time . while not immediately evident , during the initial @xmath7 time units , the density is never very far from gaussian . however , it is evident from the left - hand plot that a small percentage of the realizations are located in the left well at time @xmath7 . over a much longer time scale , @xmath8 , the density evolves to the bimodal equilibrium distribution and is hence far from gaussian . thus the system exhibits _ metastability _ with near equilibration within one well dominating over medium time scales , before the system ultimately converges to an equilibrium distribution which sees both wells . the time it takes the system to reach equilibrium clearly depends on the potential barrier height ; the time scale of the intermediate time evolution to the well bottom depends only on properties of the particular well . realizations for the model system . on the left density plots of position and momentum of the distinguished particle are shown at four times as labeled . for clarity only the first @xmath9 realizations are plotted . the initial conditions are @xmath10 . the right shows the trajectory and time series for the ensemble expectations . red points indicate the four times shown on the left . there are @xmath11 particles in the heat bath . , width=384 ] our aim is to study the behavior of coarse dynamics and coarse bifurcation methods on problems which exhibit metastable behavior of this type . in 2 we introduce a hierarchy of model problems , all of which exhibit rare transitions between a small number of states , and which we then use throughout as illustrations . in 3 we define the discrete - time _ moment map _ @xmath12 for the first @xmath13 moments of the ensemble of solutions to a time evolving system . specifically , @xmath12 will be a low - dimensional map for only the low - order moments , defined in a general setting which applies to both systems of ordinary differential equations ( odes ) , with randomness from initial data , and stochastic differential equations ( sdes ) with randomness from initial data and brownian driving noise . figure [ fig : intro_moments ] illustrates the _ first - order moment map _ for the heat bath example . the short - term dynamics of the map resemble those of the ensemble , but significantly , the moment map has _ stable fixed points _ corresponding to means of metastable densities centered in each well . the second - order moment map additionally captures the widths of the metastable measures . the algorithms we study are based on these maps . the heart of the paper is 4 . it is devoted to the study of the moment map , when applied to a variety of model systems . we use a combination of exact solutions for gaussian problems , approximate solutions for metastable systems and numerical experiments . of central interest is the observation that the moment maps stabilize the metastable states in the model problems of interest . the moment map is a nonlinear map , defined from the linear flow of probability densities for odes and sdes , with the nonlinearity entering through the process of repeatedly projecting onto the space of moments . this process of _ nonlinearization _ creates interesting fixed points which are associated with metastable behavior , and are amenable to low dimensional bifurcation analyses ; related issues are addressed in @xcite . 5 contains our concluding remarks . and momentum @xmath1 of the distinguished particle . the map has three fixed points ; two stable foci and one unstable saddle shown as blue cross . two trajectories are shown ; one evolving to each of the stable fixed points . the right trajectory has initial condition @xmath14 , corresponding to that in figure [ fig : intro_ensemble ] . there are @xmath11 particles in the heat bath . , width=192 ] we consider three example systems in this paper . two are sdes and one is the ode heat bath model described in the introduction , and illustrated in figure [ fig : sketch ] . a major thrust of this paper is to establish , through computational experimentation , that the moment map stabilizes metastable behavior arising from the slow dynamics between potential wells with large energy barriers ; this gives rise to nonlinear phenomena , such as bifurcations , in the moment map . such phenomena can be illustrated in both sde models , and in the ode heat bath model . furthermore , in various parameter limits , the sdes can be derived as approximations for the heat bath , further justifying their study . this model problem is defined by the hamiltonian @xmath15 where @xmath16 are the position and momentum of a distinguished particle of unit mass in a potential field @xmath17 . the @xmath18 s and @xmath19 s are the coordinates and momenta of @xmath3 particles that are referred to as `` heat bath '' particles . the @xmath20-th heat bath particle has mass @xmath21 and interacts with the distinguished particle via a linear spring with stiffness constant @xmath22 . if the distinguished particle were held fixed it would be the anchor point of @xmath3 independent oscillators with frequencies @xmath23 . the numerical experiments are all conducted with mass @xmath24 initial conditions for the distinguished particle are : @xmath25 , @xmath26 . the initial data for the heat bath particles , @xmath27 and @xmath28 , are randomly drawn from a gibbs distribution with inverse temperature @xmath29 . the gibbs measure is conditioned by the ( non - random ) initial data @xmath30 and @xmath31 . for fixed @xmath16 the hamiltonian is quadratic in @xmath32 , and hence the corresponding measure is gaussian . it is easily verified that @xmath33 where @xmath34 are mutually independent sequences of i.i.d . random variables . this leaves the specification of the parameters @xmath22 and @xmath21 . for our purposes the only important property of these parameters is that the frequencies @xmath23 cover an increasingly large range in an increasingly dense manner as the number of particles @xmath3 increases . hence what we actually specify is the frequencies . these are chosen to be random and uniformly distributed in @xmath35 $ ] , @xmath36.\ ] ] it is important to note that in addition to the initial data , the model specification itself contains this random element . we shall be careful to distinguish between the two types of randomness . from the frequencies the spring constants and masses are given by : @xmath37 with @xmath38 . see @xcite for further details . the parameters @xmath39 and @xmath29 are fixed at @xmath40 the potential considered in this paper is @xmath41 where @xmath42 and @xmath43 are parameters with @xmath43 typically small . for large @xmath3 and @xmath39 the distinguished particle @xmath0 in the heat bath model can be approximated by the sde @xmath44 a theorem justifying this approximation can be proved using the techniques of weak convergence , by taking the limit @xmath45 ( @xcite ) and then @xmath46 ( @xcite ) . in the absence of noise ( the zero temperature limit @xmath47 ) this damped hamiltonian system exhibits decay towards stationary points with zero velocity and positions at the critical points of @xmath48 the presence of noise ( finite @xmath29 ) then induces transitions between the minima of @xmath5 , with time - scales determined by the well - depths relative to the size of noise . the stochastic dynamics between potential wells are also present in simple one dimensional sdes . a particular instance of such a one dimensional sde follows from for @xmath49 . in the limit @xmath50 , the solutions of converge strongly @xcite or weakly @xcite to solutions of the sde @xmath51 we will use this problem to illustrate the moment map , and its properties on systems exhibiting metastable dynamics within potential wells . in most respects the sde systems derived above are simpler to treat , and computationally they are far less expensive to simulate than the full heat bath model . therefore , when we later use these models as examples , we will study the models in the opposite order from what has just been presented . we start with the 1d sde and examine its behavior extensively and then consider more briefly 2d sde and the heat bath system of odes . the central objects of our study are maps on moments . the basic ingredients are an evolution equation ( either an sde or system of odes ) , a space of low - order moments , and a measure determined uniquely by low - order moments . we refer to the latter as _ lifting _ and choice of the lifting operator is an essential ingredient in the method . we start with the sde case , then describe the situation for odes . for both the sde and the odes the flow on probability densities is linear . the nonlinearity inherent in the moment map comes from the relationship between the probability density function and its moments . after describing the moment map for the sde and the ode , the section concludes with some general remarks . let @xmath52 solve the following it sde , driven by brownian motion @xmath53 @xmath54 this includes and as special cases . we will consider ensembles of solutions of this equation , with ensembles taken over multiple driving noises and random initial data . let @xmath55 be the @xmath56 moment of @xmath57 , with expectation taken with respect to both the driving brownian motion @xmath58 and random initial data , the latter being assumed independent of the brownian motion . denote the first @xmath13 moments of @xmath59 by @xmath60 . let @xmath42 be a measure on @xmath61 determined by exactly @xmath13 moments , with density @xmath62 here @xmath63 denotes the dependence of the density on the @xmath13 moments and we require that the @xmath13 moments of @xmath64 are exactly those given by @xmath65 we now define the map @xmath12 on @xmath13 moments by @xmath66 where the initial data @xmath67 is distributed with density @xmath68 and @xmath69 ( note that if @xmath70 then @xmath12 is the identity . ) we refer to this as the _ moment map_. it depends on three choices that need to be made when calculating the moment map : a. the evolution time @xmath71 ; b. the _ number _ of moments we choose to use ; c. the lifting step : the way choose to distribute the initial density based on the moments . we will return to this dependence in more detail below . the two examples of the measures which will be used throughout this paper are the dirac measure and the gaussian measure , uniquely determined by the first moment and the first two moments , respectively . in practice we typically find this map through monte carlo simulation , but it is insightful to describe the definition of the map through the fokker - planck equation for . this linear pde for the probability densities @xmath72 propagated by is @xmath73 where @xmath74 here @xmath75 is the adjoint of the generator for the process @xmath76 we denote the solution by @xmath77 from @xmath78 we can construct @xmath79 by . in general the moment map is _ nonlinear _ because of the nonlinear dependence of @xmath68 on @xmath63 and the nonlinearity of the map from @xmath72 to @xmath80 given by . thus we have constructed a nonlinear map on @xmath61 from the linear flow on @xmath81valued densities . it is helpful to consider figure [ fig : mm ] illustrating the definition of the moment map . one should view the map as a composition of three steps : ( 1 ) _ lifting _ from a space of moments ( subset of @xmath61 ) to a space of probability densities ( captured by our choice of @xmath82 ) ; ( 2 ) time evolution of the density by the underlying process ( a linear flow , given by the map @xmath83 ) ; ( 3 ) projection back to moment space by integrating against the time - evolved measure . in moment space a density @xmath84 is uniquely determined . this initial density is evolved by the system dynamics ( e.g. the fokker - planck equation ) . from the resulting density @xmath85 , @xmath71 time units later , a new point in moment space is determined . this point is defined to be the image of @xmath86 under the moment map . , width=336 ] in the case of a single moment , in this paper , we take @xmath42 to be a dirac measure at @xmath63 and then @xmath87 can be calculated from @xmath88 in the case of two moments we have @xmath89[x(\tau)-{\mathbb e}x(\tau)]^t \}\ ] ] and we take @xmath42 to be a gaussian measure with mean and covariance determined by these moments . it is convenient to express the moment map in terms of the mean @xmath90 and covariance matrix @xmath91 . we obtain the map @xmath92 the functions @xmath93 are defined as follows . let @xmath94 then @xmath95 is given by @xmath96 and @xmath97 is given by @xmath98 there is a connection between particle filters and the moment map , although the former represent the desired density as a sum of several delta functions , or gaussians , not just as one @xcite . the moment map can be defined for deterministic problems of the form @xmath99 here @xmath100 and @xmath101 and the randomness is assumed to come entirely from the initial data . in systems characterized by a separation of time scales , it is sometimes the case that one can write an effective reduced models in terms of a subset of ( typically slow ) variables . under such appropriate conditions , we might , for example , be interested in finding a map in terms of the first @xmath13 moments of @xmath59 alone . thus the measure @xmath42 must be chosen on @xmath102 so that it is uniquely characterized by @xmath63 , the first @xmath13 moments of @xmath100 . it is natural to choose @xmath42 to be an invariant measure for the flow , conditioned by knowledge of the first @xmath13 moments of @xmath59 ; if the flow is hamiltonian then a gibbs measure is often used . we denote the density associated with this measure by @xmath103 this occurs in the heat bath example considered in the introduction where @xmath59 represents coordinate and momentum of the distinguished particle , whilst @xmath104 represents the heat bath coordinates and momenta . rather than the fokker - planck equation we have the liouville equation for propagation of probability densities . this is @xmath105 and we denote the solution by @xmath106 in the case of a dirac mass we take @xmath107 and then the map @xmath108 here @xmath109 is chosen so that @xmath110 is the density of @xmath42 conditional on @xmath111 generalization to gaussian , and higher moment problems , is also possible . * _ notation . _ we use @xmath63 to represent a point in the moment space up to some order @xmath13 , which will be made explicit for each particular example we consider . in practice , the coordinates used to describe the moment space are dictated by the particular problem . for example , in our case we use the mean and standard deviation as coordinates when considering @xmath112 and @xmath113 . for @xmath112 and @xmath114 we use the two means , the two standard deviations , and the cross correlation as coordinates . * _ usage . _ in discussing moment maps we often do not distinguish between a point @xmath63 in moment space and the uniquely determined density @xmath82 based on this point . that is , we sometime speak of the moment map as mapping @xmath115 to @xmath116 and sometime speak of the moment map as taking mapping density @xmath117 to density @xmath118 . * _ relation to optimal prediction_. the map @xmath12 can be used to generate an approximate vector field by defining @xmath119 for the ode case and @xmath42 a dirac at @xmath63 , the limit @xmath120 coincides with the vector field found by the method of optimal prediction @xcite ; this is demonstrated in @xcite . in general the method of optimal prediction leads to errors which grow linearly in time @xmath121 @xcite . the approach we study here attempts to overcome this error growth by closing the system with a larger number of moments . * _ previous work _ for simple problems in chemical kinetics , which are modeled by birth - death processes , the equation for the first moment is a closed ode , in the limit of a large number of independent particles , and the moment map studied here then works well in the dirac mass case @xcite . for more complex problems , such as lattice boltzmann , a closed effective pde may sometimes be found , using first and second moments and , again , the moment map works well in this case @xcite . in this paper we study examples where no closure is proven to exist , and demonstrate the properties of the moment map . in particular we study the relevance of fixed points of this map to the identification of metastable states . although no rigorous analysis is presented , the numerical studies show that the moment map has some merit as a method for elucidating long - term dynamics of large systems , through low dimensional dynamical systems studies . * _ lifting _ initialization of the detailed simulation consistently with coarse - grained observables is the _ lifting _ step in equation - free computation @xcite . this step is obviously not unique , as there exist many ways of initializing a distribution conditioned on a few of its lower order moments . our choice of a particular measure depending only on the lower order moments allows for a systematical initialization of the fine scale dynamics , a concept that goes back to ehrenfest @xcite , and is an important component of our computational approach . in the case of systems with metastability , different effective dynamics will be deduced ( different closures will be obtained ) depending ( a ) on the time scale of the observation ( the time horizon of the simulation @xmath71 with the fine scale solver ) and ( b ) on the nature of the lifting from the moments ( the choice of @xmath122 ) . over very short times , and initializing with a dirac delta function , the simulation will effectively sample the local gradient of the well ; over medium times , and initializing _ within one well _ , one will observe equilibration within this well ; and over very long times ( no matter what the initialization ) one will observe the approach to the equilibrium density . if we want to study the system over medium time scales , it is obviously important to use a time @xmath71 in the construction of the moment map that is short compared to the escape time between wells , but long enough for the noise ( dynamics ) to allow the sampling of the features of the well bottom . on a longer time - scale it is necessary to incorporate the transition time between wells as is done , for example , in the method of conditional averaging @xcite . * _ computational savings_. the moment map can lead to computational savings in two primary ways : a. the map @xmath12 can be used in finite dimensional bifurcation and continuation studies ; b. the estimated vector field @xmath123 can be used to advance the moments over several multiples of the time - step @xmath71 . + in case ( i ) savings can arise from using accelerated methods , such as newton iteration in a continuation environment , to find fixed points . in case ( ii ) savings arise by considering maps of the type @xmath124 to advance the moments through @xmath125 time units , using solution of the full problem or only over @xmath71 time units . ( if @xmath126 this simply reduces to the moment map ) . the above formula constitutes a _ projective forward euler _ explicit coarse integrator . much more sophisticated integrators , including multistep and implicit ones can also be used ; a rigorous analysis of savings in case ( ii ) has only recently been initiated @xcite . these ideas have been applied to a wide range of problems , both deterministic and stochastic , see e.g. @xcite and other references in @xcite . an approach related to case ( ii ) , which can be fully optimized for variance reduction and so forth when explicit time - scale separation occurs between the @xmath59 and @xmath104 dynamics in , is outlined in @xcite ; for a rigorous analysis see @xcite . here we explore a series of examples of moment maps . we start with examples based on the ou process for which explicit representation is possible . in these cases the moment maps are linear . we then proceed to the more interesting nonlinear maps arising from systems with double - well potentials . consider the ou process @xmath127 this is essentially the simplest example of and corresponds to in the case of a quadratic potential . the exact solution of this process is @xmath128 for first - order moment map on @xmath129 , we take initial data with dirac measure with density @xmath130 . the map on the first moment @xmath129 is explicitly @xmath131 from we have @xmath132 the map @xmath12 is linear and has a unique , globally attracting fixed point at @xmath133 . figure [ fig : ou_mm ] shows a phase portrait for this simple map . before discussing this we consider the second - order moment map with mean @xmath129 and standard deviation @xmath134 as co - ordinates . in this case the initial data has density @xmath135 from we then have that the moment map on @xmath136 is given by @xmath137\}^{\frac12}.\end{aligned}\ ] ] while not a linear map on @xmath136 , this is linear on @xmath138 . figure [ fig : ou_mm ] shows a phase portrait for map . the map has the unique fixed point @xmath139 which is globally attracting . the solution to is gaussian if @xmath140 is gaussian . hence the second - order moment map gives exact time @xmath71 samples of the distribution of the sde . this unique fixed point of the map characterizes the unique invariant ( gaussian ) measure of . in contrast , the first - order moment map , with @xmath42 a dirac , can only approximate the solution . although this map does not quantitatively represent the solution , since it contains no information about the width of the measure , it captures the correct dynamics of the first - order moment and shows that probability mass initially far from the origin will be transported inward . , @xmath141 , and @xmath142 . , width=384 ] we now present a detailed study of moment maps for the one - dimensional sde with double - well potential . in this section we focus on the dynamics of these maps using numerical simulations . in the next section we analyze the maps , in particular the nonlinearity of the maps , in more detail . after suitable scaling sde can be rewritten in the notation of as @xmath143 we consider the double - well potential @xmath144 for @xmath145 the potential is symmetric about zero and for @xmath43 small this symmetry is weakly broken . the potential has two local minima for @xmath146 and one minimum otherwise . the first - order ( dirac - based measure ) and second - order ( gaussian - based measure ) moment maps for this equation are _ nonlinear_. in particular these maps have multiple stable fixed points which undergo bifurcations as the potential ( @xmath42 or @xmath43 ) is varied . we are interested in these fixed points and their stable and unstable manifolds as a function of @xmath42 for @xmath43 fixed . we shall consider two cases , the slightly asymmetric case @xmath147 and the symmetric case @xmath145 . we resort solely to numerical studies of the moment maps throughout this section . in brief , we use monte carlo simulations to evolve densities forward over time interval @xmath71 , as in the evolution from @xmath64 to @xmath122 in figure [ fig : mm ] . this numerically determines the moment maps . by employing additional techniques we can compute steady states and bifurcations . the effective fixed point , bifurcation and continuation calculations require estimates of the jacobian of the moment map ( or its action ) . this is achieved by using nearby initializations of the moment map ( see @xcite as well as @xcite and the monograph @xcite for matrix - free implementations of newton - gmres ) . we first consider the moment maps for the slightly asymmetric potential with representative values of @xmath42 and map time @xmath71 . figure [ fig : sde1_pp ] shows phase portraits for both first - order and second - order moment maps . each map has three fixed points . these are shown together with unstable , and for the second - order map also stable , manifolds of the unstable fixed point . figure [ fig : sde1_phi ] shows one iteration of both the first- and second - order maps . consider the first - order map . the initial density @xmath148 is a dirac delta at @xmath149 , here a point slightly to the right of the unstable fixed point near zero . after time @xmath150 the mean of the density has moved to the right and hence the map takes @xmath149 to the right in this case . the density spreads considerably but over @xmath150 it remains nearly symmetric ( the mean is indistinguishable from the maximum ) . the initial density for the next iteration is a dirac delta displaced to the right . for the second - order map the initial density @xmath151 is a gaussian centered at @xmath149 with width @xmath134 . here @xmath152 corresponds to a point on the lower branch of the stable manifold of the saddle fixed point . after @xmath150 time units the density has spread and the mean has moved slightly to the right . the map thus corresponds to substantial increase in @xmath134 and small increase in @xmath153 . the density @xmath85 is slightly non - gaussian as can be seen in comparison with the initial ( gaussian ) density for the next iteration . figure [ fig : sde1_fixed ] shows all fixed points in figure [ fig : sde1_pp ] . in each case we plot the density @xmath82 corresponding to each fixed point in moment space as well as the density @xmath85 . the stable fixed points of the second - order map are the metastable measures centered in each well . specifically , the gaussian measures @xmath154 corresponding to the stable fixed points are indistinguishable from the evolved densities @xmath85 . as we shall show this is independent of @xmath71 over a very large range of @xmath71 . intuitively this is because on any time scale , up to the very long time scale needed to reach the equilibrium distribution , these gaussian measures are approximately invariant . the stable fixed points of the first - order map are not invariant measures . however , starting from initial condition @xmath155 where @xmath149 is indistinguishable from the potential minimum , the density simply fills out ( symmetrically ) , the ( locally quadratic ) well bottom . note that @xmath150 is close to , but not quite , the time necessary to reach the metastable density starting from the dirac measure . the unstable fixed points for the maps are understood as follows . for the first - order map , the unstable fixed point is at the local maximum of the potential . starting from a dirac delta @xmath156 , the density spreads symmetrically , since the maximum is locally quadratic and hence symmetric . hence after time @xmath150 the mean is still at local maximum of the potential . only for times @xmath71 long enough for the density to fill the two wells , and hence have a mean different from the local maximum , would the fixed point be different from the local maximum . for such times the fixed point would approximately be the mean of the equilibrium density . for the second - order map , the saddle fixed point corresponds to a gaussian @xmath157 which is quite broad . the evolved density @xmath85 is far from gaussian ; it simply has the property that its first two moment agree with those of the initial gaussian . the saddle fixed point is quantitatively sensitive to the map time @xmath71 ( see below ) . qualitatively , however , for any value of @xmath71 the saddle fixed points are broad gaussians . , @xmath147 , @xmath150.,width=384 ] . the bold red curve shows @xmath158 , the thin green curve shows @xmath85 , and the dash bold red curve shows @xmath159 . ( see figure [ fig : mm].),width=384 ] , shown with bold blue curves , corresponding to fixed points for first - order ( left ) and second - order ( right ) moment maps shown in figure [ fig : sde1_pp ] . thin green curve show the evolved density @xmath85 for each case . ( for the two stable fixed point of the second - order moment map the evolved density @xmath85 is indistinguishable from @xmath82 . ) the potential @xmath160 is shown as a dashed curve . , width=384 ] the _ stable _ fixed points of both maps correspond to metastable states ( measures ) of the underlying process . the metastable states are very nearly gaussian measures , because the wells are locally quadratic , and they are thus well captured by the low dimensional moment maps : the densities corresponding to the stable fixed points of the second - order map are virtually indistinguishable from the metastable states . the fixed points of the first - order map capture the means of the metastable distributions . the stable fixed points are insensitive to the value of @xmath71 over a large range of @xmath71 . ( see figure [ fig : sde1_bd_tdepend ] below ) . this lack of sensitivity to @xmath71 suggests that these fixed points are meaningful characteristics of the observed dynamics over a range of observation time scales . the unstable ( saddle ) fixed point for the first - order moment map is also insensitive to the value of @xmath71 but the unstable fixed point for the second - order map is sensitive to @xmath71 . this suggests that the fixed points of the first - order moment map provides a useful description of the dynamics ( for relatively short times ) close to both the saddle and the well bottoms ; the second - order map fixed points provide a meaningful description of the effective dynamics close to the bottoms of the two wells , _ but not in the neighborhood of the saddle _ ; this is essentially because a reduced equation in terms of the second - order moments does not appear to successfully close in the neighborhood of the saddle . we now turn to the behavior of the fixed points as a function of well - depth . figure [ fig : sde1_bd ] shows bifurcation diagrams for each moment map as a function of @xmath42 , including @xmath42 for which the potential has a single well ( basically @xmath161 ) . local extrema of the potential are shown for comparison . the right - most end of each bifurcation diagram ( @xmath162 ) corresponds to the phase portraits just considered . the fixed points for the first - order map follow the extrema of the potential closely for all @xmath42 , including near the saddle node bifurcation . the second - order fixed points do not . in figure [ fig : sde1_other_fixed ] we show fixed points in cases where the potential is far from locally quadratic . one case is @xmath163 and the other is @xmath164 , very near the value corresponding to the saddle - node bifurcation of the second - order map . in both cases the first - order fixed point is at the potential minimum while the mean of the second - order fixed point is noticeably different from the minimum as can also be seen in figure [ fig : sde1_bd ] . where the non - quadratic aspect of the potential is apparent . thin green curve show the evolved density @xmath85 for each case . the top case is @xmath163 . the bottom is @xmath164 , very near the saddle - node bifurcation for the second - order map in figure [ fig : sde1_bd ] . the potential @xmath160 is shown as a dashed curve . , width=384 ] we show in figure [ fig : sde1_bd_tdepend ] how steady states for the second - order moment map are affected by the choice of the map time @xmath71 . to understand what the figure shows , it is helpful to first consider the fixed points for @xmath161 . neither @xmath149 nor @xmath134 vary significantly with @xmath71 and @xmath134 is not large . while not as easy to see , the stable fixed points for @xmath165 , behave similarly . the stable fixed points correspond to the uppermost and lowermost branches in the @xmath149-plot and the bottom branch in the @xmath134-plot . both stable fixed points have approximately the same , relatively small , value of @xmath134 . the intermediate , generally unstable , fixed points have larger values of @xmath134 ; these vary quantitatively and qualitatively with @xmath71 . the number of unstable fixed points can change with @xmath71 , e.g. at @xmath162 , @xmath150 there is one saddle fixed point between the two stable fixed points ( figure [ fig : sde1_pp ] ) whereas at @xmath162 , @xmath166 there are three fixed points between the two stable points ( similar to figure [ fig : sde1_pp_sym ] ) . the conclusion is that the stable fixed points of the second - order moment map which correspond to stable or metastable measures of the sde are largely insensitive to the time @xmath71 used to define the map . the other fixed points typically correspond to broad distributions and depend quantitatively on @xmath71 . this dependence of the fixed points of the moment map on @xmath71 suggests that they are not useful features of the effective dynamics which in turn suggests that the effective behavior _ does not usefully close _ at the second moment level in the neighborhood of these fixed points . : @xmath167 blue , @xmath150 red , @xmath168 yellow , @xmath142 green , @xmath166 black . , width=240 ] finally we consider the moment maps for the symmetric double - well potential . figure [ fig : sde1_pp_sym ] shows phase portraits , similar to figure [ fig : sde1_pp ] for the asymmetric case , while figure [ fig : sde1_fixed_sym ] fixed points , similar to figure [ fig : sde1_fixed ] for the asymmetric case . figure [ fig : sde1_bd_sym ] shows a bifurcation diagram as a function of @xmath42 . an important observation from the data presented in the symmetric case is that the moment map produces stable equilibria which are far from metastable states . the gaussian measures corresponding to these fixed points are very broad . . same conventions as in figure [ fig : sde1_fixed ] . top plots show the right stable fixed point . the middle plot shows the right saddle for the second - order map . the bottom plot shows the middle fixed point ( saddle for the first - order map and stable fixed point for the second - order map ) . the other points figure [ fig : sde1_pp_sym ] are obtained by symmetry . , width=384 ] we now present some analysis of the nonlinear map . consider the sde . the adjoint of the generator for this process is @xmath169 the equation has a unique invariant density @xmath170 , in the null - space of @xmath75 , and given explicitly by @xmath171 the operator @xmath172 is self - adjoint in the space @xmath173 ( see proposition 2.2 , @xcite ) . we let @xmath174 denote the weighted @xmath173 inner product @xmath175 and we write the eigenvalue problem @xmath176 with eigenvalues ordered so that @xmath177 we may choose the normalization @xmath178 since @xmath179 we have @xmath180 now , the solution @xmath72 of the fokker - planck equation can be expanded in the eigenfunctions @xmath181 as @xmath182 where @xmath183 notice that @xmath184 in all cases , because @xmath185 is a probability density function . assume @xmath160 in is a double well potential with deep wells relative to the noise . then the analysis and numerical evidence in @xcite suggests that it is reasonable to assume that @xmath186 thus there exists a spectral gap and , noting that @xmath187 , this suggests that after times @xmath71 of order @xmath188 the density @xmath72 can be well - approximated by only the first two eigenfunctions @xmath189 and @xmath190 . assuming that @xmath191 at the two well bottoms , then @xmath192 is well - approximated as the weighted sum of two gaussians @xmath193 ; this may be verified from . we may assume that the gaussians are normalized to be probability densities , and then define their means and standard deviations by @xmath194 ^ 2 g_{\pm}(x)dx={\bar \sigma}_{\pm}^2.\ ] ] we then write @xmath195 where @xmath39 determines the relative weight of the two gaussian contributions . furthermore , by orthonormality , it may be shown that @xmath196.\ ] ] we can now use the approximations , to understand the nonlinearity in the moment map for the case of the double well potential . we focus on the second - order map . in this case @xmath197 will be gaussian . we choose @xmath198 for some @xmath121 of order @xmath199 and set @xmath200 then , due to the spectral gap in the eigenvalues , we have for @xmath201 , @xmath202 for @xmath203 the key to the dynamics of the moment map is manifest in the formula . given any gaussian with mean and standard deviation @xmath204 we project the density @xmath205 onto the basis @xmath206 then the evolution shows that @xmath78 is given by and @xmath207 since @xmath208 depends nonlinearly on @xmath204 , through @xmath209 , we have a nonlinear map @xmath210 in principle this map can be computed explicitly , though this requires knowing the eigenfunctions @xmath190 and @xmath189 accurately . this analysis can be applied to our computational studies of the double well potential to gain further insight into the moment map . we employ the symmetric double well potential considered at the end of [ sec : double ] . in this case the eigenfunctions @xmath190 and @xmath189 are accurately approximated by and with @xmath211 . we use @xmath212 in the computations which follow . ( we shall see that @xmath213 so that @xmath214 . ) figure [ fig : an_enlarge ] shows the moment map in the coordinates suggested by the preceding analysis . the corresponding evolution of the density is shown for one iterate of the map . we wish to view the evolution of the density @xmath72 in terms of the amplitudes @xmath215 of projections onto the eigenfunctions @xmath216 . we know that @xmath217 so there is no need to show this amplitude . the essential amplitude is @xmath218 . the effect of all the higher amplitudes @xmath219 , @xmath220 can be summarized by a single scalar @xmath221 defined as @xmath222 with @xmath223 consider point 1 in figure [ fig : an_enlarge ] . to the right is shown the gaussian density @xmath224 determined by a point @xmath225 in moment space . this density has a significant projection onto the higher modes @xmath216 , @xmath220 and @xmath221 is significantly greater than zero . from the analysis we expect the density to evolve such that @xmath221 decays to zero on a time scale faster than the dynamics of @xmath218 . the thick black curves in [ fig : an_enlarge ] show this evolution . there is little change in @xmath218 as the system evolves toward @xmath226 . the decay to @xmath226 is ultimately exponential as is shown in figure [ fig : expplot ] . from this we estimate that @xmath227 . after time @xmath212 the system is at point 2 with corresponding density shown at the right . the resulting density is no longer gaussian as expected from . the second - order moment map is obtained by determining @xmath228 , the mean and standard deviation of density 2 . for the next iteration of the map one constructs a gaussian density with this mean and standard deviation , point 3 in figure [ fig : an_enlarge ] , and repeats . the dashed line connecting points 2 and 3 illustrates this . thus the moment map takes gaussian @xmath229 determined by @xmath225 ( point 1 ) to @xmath230 determined by @xmath228 ( point 3 ) where the map can be again be iterated . note that the evolution from point 1 to point 2 is due to linear flow of the fokker - planck equation . nonlinearity results from the projection of density at point 2 back to @xmath228 , i.e. going from point 2 to point 3 introduces nonlinearity into the map . further note that the mean has moved slightly to the right after one iteration of the map . hence , on the next iteration less of the mass will move into the left well , ( as in point 2 ) . in this way the map stabilizes the meta - stable states corresponding to localization of density into a single well . now consider exactly the same evolution seen in the moment - map coordinates @xmath136 in figure [ fig : xs_enlarge ] . the green curve is that generated by @xmath231 , @xmath232 , @xmath233 , @xmath220 . this curve is also shown as green in figure [ fig : an_enlarge ] . this is the slow manifold for the system . starting from point @xmath234 , point 1 , the evolution of @xmath136 as @xmath72 evolves is shown in bold . this is the decay of the modes @xmath216 , @xmath220 . after time @xmath71 the system reaches @xmath235 , point 2 . even though the transient dynamics is such as to move initially away from the slow manifold , the evolution brings the system back , as it must . the projection back to gaussian density does not change @xmath129 and @xmath134 so points 2 and 3 appear the same in figure [ fig : xs_enlarge ] . we know that as the system evolves from point 1 to point 2 the density became non - gaussian . we show this in figure [ fig : xz_enlarge ] by showing the dynamics out of the @xmath136 plane . here @xmath236 is given by @xmath237 with @xmath238 and is similar in spirit to @xmath221 . @xmath236 is a measure of how far the density is from gaussian whereas @xmath221 measure how far the density is from the slow manifold . however , these have opposite roles and behavior . gaussian densities correspond to @xmath239 , but since such densities are not in general on the slow manifold they correspond to @xmath240 . starting from point 1 in figure [ fig : xz_enlarge ] , @xmath236 is necessarily zero since the density ( figure [ fig : an_enlarge ] point 1 ) is gaussian . as the system evolves toward the slow manifold and becomes non - gaussian , @xmath236 increases ( point 2 ) . projection back to gaussian moves the system vertically downward to @xmath239 ( point 3 ) . now consider the case @xmath241 , near the local minima of the potential wells ( we consider only the right well , the left is the same ) . one sees in figures [ fig : an_enlarge ] and [ fig : xz_enlarge ] that points on the slow manifold correspond to nearly gaussian densities . this can be most clearly seen in [ fig : xz_enlarge ] where the slow manifold falls to near @xmath239 . the end point of the green curve is @xmath242 . in figure [ fig : an_enlarge ] we seen that the red points ( the gaussian densities in @xmath243 coordinates ) fall to very nearly @xmath226 meaning that these densities are almost exactly represented by a sum of @xmath189 and @xmath190 . the small gap between the slow manifold and the gaussian density ( seen in both figures [ fig : an_enlarge ] and [ fig : xz_enlarge ] ) reflects that fact that the metastable density is not exactly gaussian . in figure [ fig : xs ] we show the dynamics in @xmath244 coordinates showing all the fixed point of the moment map . this is similar to figure [ fig : sde1_pp_sym](right ) except that here @xmath212 . the figure shows trajectories starting from four initial conditions . note that the slow manifold accurately captures the unstable manifold of the saddle fixed points . , @xmath221 coordinates . red dots show points with gaussian densities @xmath245 . the gaussian densities corresponding to points labeled 1 and 3 are show to the right . the evolution of the density over time @xmath71 is shown by bold curves ( actually fine series of bold points ) with green crosses indicating the the final time . the density corresponding to point 2 is shown to the right . dashed lines indicate the projections back to gaussian density which preserve the mean and standard deviation of the distribution . , width=384 ] . , width=192 ] coordinates . the points correspond to exactly the same points as in figure [ fig : an_enlarge ] . the bold curves ( actually fine series of bold points ) show the evolution of @xmath244 as the density evolves between iterates of the map . the green curve is the slow manifold ( @xmath226 ) also shown in [ fig : an_enlarge ] . , width=240 ] coordinates . the points correspond to exactly the same points as in figures [ fig : an_enlarge ] and [ fig : xs_enlarge ] . the bold curves ( actually fine series of bold points ) show the evolution of @xmath246 as the density evolves between iterates of the map . the green curve is the slow manifold.,width=240 ] coordinates . same as figure [ fig : xs_enlarge ] except over a larger range of coordinates and same as [ fig : sde1_pp_sym ] except here @xmath212 . , width=240 ] in this section we consider moment maps for the 2d sde ( [ eq : sde2 ] ) presented in [ ssec:2dsde ] . the dynamics are potentially richer than for the 1d sde considered up to now . nevertheless , the stabilization of metastable states is the same as for the 1d sde . as before we examine maps for both first - order and second - order moments . the first - order map can be written as @xmath247 where @xmath248 and @xmath249 are means of @xmath0 and @xmath1 respectively . the second - order map can be written as @xmath250 where @xmath251 and @xmath252 are the standard deviations of @xmath0 and @xmath1 respectively and @xmath253 is the cross correlation . other coordinates could be used for the five - dimensional phase space . we again consider the slightly asymmetric double - well potential ( [ eq : pot ] ) for parameters similar to those use for the 1d sde . all results have been obtained numerically from simulations of equations ( [ eq : sde2 ] ) for the case @xmath254 , @xmath255 , @xmath256 . , @xmath147 , @xmath150 . , width=384 ] figure [ fig : sde2_pp ] shows typical phase portraits for the first - order and second - order moment maps . the first - order map exhibits the dynamics typical of bistable damped oscillators . similar to the 1d case , the fixed points are located at , or very close to , @xmath257 where @xmath258 is a local extremum of the potential . the second - order map has five fixed points , similar to those seen in figure [ fig : sde1_pp_sym ] , including a stable fixed point with a very broad gaussian measure . typically trajectories of interest approach either one or the other of the stable fixed points corresponding to the potential minima . the moment maps for the 2d sde stabilize metastable states in a very similar way to that found for the 1d sde . . only a few representative ellipses are shown . the curve shows the path of the center of the ellipses . the other trajectory starts at @xmath259 . in this case every points on the trajectory is shown . two stable fixed points of the map are shown as bold blue ellipses . parameters are @xmath162 , @xmath147 , @xmath150 . , width=288 ] figure [ fig : sde2_ellipse ] provides a better view of the dynamics of the second - order moment map and shows how trajectories approach the stable fixed points . a gaussian measure in two variables can be visualized as an ellipse in the plane specified by five numbers : the center , the semi - major and semi - minor axes , and the orientation . so for each point in the five - dimensional phase space of the moment map we plot an ellipse in the plane . one can think of each ellipse as corresponding to the level set of a gaussian density . two trajectories are shown , one with the ellipse at every iteration of the map plotted , and the other with only a few representative ellipses plotted . one can see how densities evolve under the map toward stable equilibria . trajectories from almost all initial condition distributions that are not too broad , ( i.e. , initialized within one of the two metastable wells , as discussed above ) evolve to one or the other of the two fixed points similarly to what is shown in figure [ fig : sde2_ellipse ] . for completeness we show in figure [ fig : sde2_bd ] bifurcation diagrams for the moment maps for the 2d sde . these bifurcation diagrams reveal much the same features as for the 1d sde . fixed points of the first - order moment map track the potential minima . for the second - order map , stable fixed points corresponding to metastable states exist except in regions near where the potential bifurcates from single- to double - well . clearly , close to such parameter values , the separation of time scales between equilibration in one well and transition to the other is no longer present , and the fixed points we find depend on the map reporting horizon @xmath71 . fixed points , both stable and unstable , corresponding to broad distributions , separate the stable fixed points . while not shown , we find that the fixed points corresponding to the stabilized metastable states are essentially independent of the time @xmath71 over which the map is defined ; this suggests that the map is a good effective description in their neighborhood , but not a useful one close to the @xmath71-dependent fixed points . the ability to test the sensitivity of the map dynamics and fixed points to the parameter @xmath260 as well as the ability to use maps of different orders , allow potential tools with which to test online " the validity of a given map as an effective model of the detailed dynamics . , @xmath150 . , width=384 ] finally we consider the dynamics of a particle in a heat bath as described in [ ssec : heat ] . moment maps for this example are of the type defined for ode systems in [ ssec : ode ] . nevertheless we expect the moment map for the ode system to behave very much like that of the 2d sde since the sde is known to capture the dynamics of the heat bath model in the limit of a large bath . from a computational viewpoint in fact there is not much distinction between the stochastic ( sde ) and the deterministic ( ode ) cases since in both case we numerically compute the moment maps using monte carlo simulations to evolve densities over time interval @xmath71 . there are two related new issues , however . the first is that @xmath3 , the number of particles in the heat bath , is now a parameter which could potentially affect the dynamics of the system . the other related issue is that there is a minimum time interval over which we should define the moment map . for any given bath size @xmath3 there is a maximum frequency @xmath261 of the oscillators in the bath , where @xmath262 . we therefore should take the map time @xmath71 to be at least of the same order as the @xmath263 . we always use @xmath264 . other than the preceding two points the only significant difference between the heat bath and the 2d sde is that the heat bath requires significantly more computation to evolve densities forward in time . hence the moment map is much more expensive to compute for the heat bath than for the sde . for this reason , we have limited our studies to a fixed potential : equation with @xmath265 , @xmath147 . this is the potential used for the simulations shown in the introduction . figure [ fig : bath_pp ] shows phase portraits of the moment map for the heat bath model . the first - order map shows the expected three fixed points at the local extrema of the potential and exhibits the dynamics of bistable damped oscillators . the map has indeed stabilized the metastable states with a well defined boundary ( the stable manifold of the saddle ) separating the basins of attraction . the location of the fixed points does not depend in any significant way on the number of particles in the bath or on the map time @xmath71 . the stable and unstable manifolds do vary somewhat , primarily with the map time @xmath71 , but are always qualitatively as seen in figure [ fig : bath_pp ] . for the second - order map we show phase portraits in the style of figure [ fig : sde2_ellipse ] for the 2d sde system . we show two trajectories for each of three cases . the middle case is for @xmath11 , the value used for simulations shown in the introduction . the maximum oscillator frequency in the bath is @xmath266 and so we use a map time of @xmath267 . we show a map starting at @xmath268 . this initial condition corresponds exactly to the initial density used figure [ fig : intro_ensemble ] . one sees the similarity between the trajectory for the moment map and the evolution of the density in figure [ fig : intro_ensemble ] . however , the `` leaking '' of mass into the left potential well observed in figure [ fig : intro_ensemble ] is prevented by the moment map via the mechanism illustrated in figure [ fig : an_enlarge ] . the moment map is nonlinear and has the two stable fixed points where the density is only metastable . most trajectories evolve to one or the other of these fixed points . for the same potential we have investigated the effect of @xmath3 , the number of particles in the heat bath , on the moment map . we find that the map is very insensitive to @xmath3 at least for @xmath269 . for large @xmath3 one can use a smaller map time @xmath71 and the dynamics do depend weakly on @xmath71 . specifically , the fixed points for the three cases in figure [ fig : bath_pp ] are very slightly different , though this can not be seen on the scale of the figure . this difference is not due directly to @xmath3 , but to the value of @xmath71 . particles is used . for the second - order map the following are used : ( top ) @xmath270 , @xmath150 ; ( middle ) @xmath11 , @xmath271 ; ( bottom ) @xmath272 , @xmath273 . , width=384 ] in this paper we have introduced a mathematical framework intended to outline and clarify some aspects of the coarse - grained approach to analyzing stochastic and deterministic systems . in particular we have given a precise definition of the moment map . these are maps on the ( low - dimensional ) space of low - order moments of probability measures . we have considered these maps both for stochastic systems and for deterministic systems with random initial data . while the underlying evolution of densities in both systems is linear , the moment maps are typically nonlinear . our main focus has been understanding the origins of this nonlinearity . in this paper we sought to establish a connection between the dynamics of coarse - grained observables ( such as moments of evolving realization ensembles ) and the nonlinear dynamics one expects at the deterministic limit . contemporary estimation techniques would allow us to recover both the deterministic and the stochastic component of an effective _ stochastic _ model ( e.g. @xcite ) . then , instead of using integral changes of coarse - grained observables , we could directly seek the extrema of an underlying effective potential , or even the extrema of the corresponding equilibrium density ( see , e.g. @xcite ) . we have presented results for a number of model systems . we have first considered the simple ou process for which a full analysis is possible . then using a combination of numerical studies and analysis , we have considered in most detail a one - dimensional sde with a double - well potential . this system provides the simplest example of a nonlinear moment map . in particular this example shows how the moment map can stabilize metastable densities of the underlying linear flow . we have additionally presented numerical results for moment maps computed for two other systems with double - well potentials - a two - dimensional sde and a deterministic ode system with many degrees of freedom . maps for both of these systems show the basic features found for the one - dimensional sde , namely nonlinearity and the stabilization of metastable densities . one of the important issue that arises naturally in this computational framework is the importance of _ the observer_. how long does a physical observer have to wait before declaring that an observed quantity is at steady state ? this is clearly related to our testing the dependence of the map fixed points to the map reporting horizon . we also saw that the initialization of computational experiments ( whether within or not within a well ) can be vital in the existence of an effective reduced model ( this is related , as we mentioned , to conditional averaging techniques ) . k. theodoropoulos y .- h . qian and i. g. kevrekidis _ coarse " stability and bifurcation analysis using timesteppers : a reaction diffusion example _ proc . 97*(18 ) ( 2000 ) , pp.9840 - 9843 . i. g. kevrekidis , c. w. gear , j. m. hyman , p. g. kevrekidis , o. runborg and k. theodoropoulos , _ equation - free coarse - grained multiscale computation : enabling microscopic simulators to perform system - level tasks _ , comm . sciences * 1*(4 ) ( 2003 ) pp.715 - 762 ; original version : physics/0209043 at arxiv.org . a. g. makeev , d. maroudas , a. z. panagiotopoulos and i.g.kevrekidis , _ coarse bifurcation analysis of kinetic monte carlo simulations : a lattice gas model with lateral interactions _ , j. chem * 117*(18 ) ( 2002 ) pp.8229 - 8240 . a. g. makeev and i.g . kevrekidis , _ equation - free multiscale computations for a lattice - gas model : coarse - grained bifurcation analysis of the no+co reaction on pt(100 ) _ , chem . , * 59*(8 - 9 ) ( 2004 ) pp.1733 - 1743 . k. theodoropoulos , k. sankaranarayanan , s. sundaresan and i.g.kevrekidis , _ coarse bifurcation studies of bubble flow lattice boltzmann simulations _ chem.eng.sci . , * 59 * ( 2004 ) pp.2357 - 2362 ; also nlin.ps/0111040 at arxiv.org . r. erban , i. g. kevrekidis and h. g. othmer , _ an equation - free computational approach for extracting population - level behavior from individual - based models of biological dispersal _ , submitted to _ physica d _ , nov . 2004 ; also physics/0505179 at arxiv.org . c. siettos , m. d. graham and i. g. kevrekidis , _ coarse brownian dynamics for nematic liquid crystals : bifurcation , projective integration and control via stochastic simulation _ j. chem * 118*(22 ) ( 2003 ) pp.10149 - 10157 ( 2003 ) ; also cond - mat/0211455 at arxiv.org . r. rico - martinez , c. w. gear and i.g.kevrekidis , _ coarse projective kmc integration : forward / reverse initial and boundary value problems _ , j. comp . * 196*(2 ) ( 2004 ) pp.474 - 489 ; nlin.cg/0307016 at arxiv.org . g. hummer and i. g. kevrekidis _ coarse molecular dynamics of a peptide fragment : free energy , kinetics and long time dynamics computations _ j. chem . phys . * 118*(23 ) ( 2003 ) pp . 10762 - 10773 ; also physics/0212108 . j. cisternas , c. w. gear , s. levin and i.g.kevrekidis , _ equation - free modeling of evolving diseases : coarse - grained computations with individual - based models _ , london , * 460*(2004 ) pp.27621 - 2779 ; also nlin.ao/0310011 . d. i. kopelevich , a. z. panagiotopoulos and i. g. kevrekidis , _ coarse - grained kinetic computations of rare events : application to micelle formation _ , j. chem . phys . * 122 * 044908 ( 2005 ) ; cond - mat/0407220 at arxiv.org .

given a factor code from a one - dimensional shift of finite type onto an irreducible sofic shift , if is finite - to - one there is an invariant called the degree of which is defined the number of preimages of a typical point in . we generalize the notion of the degree to the class degree which is defined for any factor code on a one - dimensional shift of finite type . given an ergodic measure on , we find an invariant upper bound on the number of ergodic measures on which project to and have maximal entropy among all measures in the fibre . we show that this bound and the class degree of the code agree when is ergodic and fully supported . one of the main ingredients of the proof is a uniform distribution property for ergodic measures of relative maximal entropy .

1001.5323

it is a well - known result that a 1-dimensional irreducible shift of finite type on a finite alphabet has a unique measure of maximal entropy , the measure developed by shannon and parry @xcite . in contrast , we consider the relative case in which one is given a factor code @xmath5 from a shift of finite type @xmath1 to a sofic shift @xmath2 , and a measure @xmath3 on @xmath2 . in this case ergodic measures on @xmath1 in the fibre @xmath4 having maximal entropy in the fibre , so - called measures of relative maximal entropy , are not well understood . measures of relative maximal entropy appear frequently in different topics in symbolic dynamics , and have been applied to the problem of computing the hausdorff dimension of certain sets by gatzouras and peres @xcite . there are connections of measures of relative maximal entropy with functions of markov chains @xcite , measures that maximize a weighted entropy functional @xcite , the theory of pressure and equilibrium states @xcite , relative pressure and relative equilibrium states @xcite , and compensation functions @xcite . other uses of such measures arise from their application in the mathematics of information transfer @xcite and information - compressing channels @xcite . unlike in the case considered by shannon and parry where the measure of maximal entropy is unique , it is known that measures of relative maximal entropy may not be unique @xcite . it is natural to ask how many of such measures there may be for a particular factor code @xmath6 and an ergodic invariant measure on y. a previous result on this problem is the result of petersen , quas , and shin @xcite which established an explicit finite upper bound on the number of measures of relative maximal entropy . they considered the case ( always achievable by recoding ) that @xmath0 is a 1-block code from a 1-step shift of finite type . in this case , the main result of @xcite states that the number of measures of relative maximal entropy is bounded above by the number of preimages of any symbol @xmath7 of @xmath2 which has positive probability with respect to the measure on @xmath2 . the bound in @xcite suffers from not being invariant under conjugacy and becomes arbitrarily large simply by recoding to a higher block presentation . to avoid this issue one possibility is to take the minimum of this bound over all shifts of finite type @xmath8 which are conjugate to @xmath1 . however , no algorithm for computing this minimum has been found . on the other hand , in this paper we find a simpler conjugacy - invariant upper bound which can be strictly better and is never worse than the minimum - over - conjugates bound mentioned above . given a finite - to - one factor code @xmath0 from a shift of finite type @xmath1 to a sofic shift @xmath2 , the is defined to be the minimal cardinality of the set of preimages of a point in @xmath2 . this is widely - studied and known to be invariant under recoding @xcite . one can show that the number of preimages of every transitive point in @xmath2 is exactly the degree of @xmath0 . previous efforts to find an analogue of this when @xmath0 is infinite - to - one have proved unsuccessful @xcite . given @xmath9 in @xmath2 , we define an equivalence relation on @xmath1 which is respected by recoding . the relation is motivated by communicating classes in markov chains . loosely , two points @xmath10 and @xmath11 in @xmath12 lie in the same equivalence class `` transition class '' if one can find an element of @xmath12 matching either of @xmath10 and @xmath11 from @xmath13 to an arbitrarily large positive coordinate @xmath14 and matching the other at @xmath15 . the minimal number of transition classes ( always finite ) in @xmath12 when @xmath9 runs over the points in @xmath2 is called the of the code . we show that the number of transition classes over every transitive point of @xmath2 is exactly the class degree . if @xmath0 is finite - to - one the degree and the class degree of @xmath0 agree . we prove an analogue of theorem 1 in @xcite : given an ergodic measure @xmath3 on @xmath2 , for distinct ergodic measures @xmath16 and @xmath17 of relative maximal entropy , the relative independent joining of @xmath16 and @xmath17 assigns measure zero to the set of points @xmath18 such that @xmath19,@xmath20 are in a common transition class . using a combinatorial characterization of class degree and further arguments , we show that the number of transition classes above a typical point of @xmath3 is an upper bound on the number of ergodic measures of maximal entropy in the fibre @xmath21 . this bound is equal to the class degree of the code when @xmath3 is ergodic and fully supported . one key ingredient to this work is a relative version of the uniform conditional distribution property for measures of maximal entropy . we follow techniques developed by burton and steif @xcite ( which were related to an earlier work of lanford and ruelle @xcite ) to show a uniform conditional distribution property for measures of relative maximal entropy . throughout the paper , @xmath1 will denote a one - dimensional shift of finite type on a finite alphabet ( sft ) and the shift map will be denoted by @xmath22 , or simply by @xmath23 when there is no ambiguity . we will deal with a factor code @xmath0 from @xmath1 to a sofic shift @xmath2 . a detailed description of these notions is given in @xcite . a triple @xmath24 is called a when @xmath25 is a factor code from a sft @xmath1 to a sofic shift @xmath2 . the alphabet of a shift space @xmath1 is denoted by @xmath26 and the @xmath27-algebra on @xmath1 generated by cylinder sets of @xmath1 is denoted by @xmath28 . the set of all @xmath14-blocks that occur in points of @xmath1 is denoted by @xmath29 , and the language of @xmath1 , @xmath30 , is the collection @xmath31 . let @xmath10 be in @xmath1 and @xmath32 . the configuration which occurs in @xmath10 on @xmath33 is denoted by @xmath34 . if @xmath35 $ ] is a @xmath36-interval we sometimes denote @xmath34 by @xmath37}$ ] . we say two factor triples @xmath24 and @xmath38 are , and denote it by @xmath39 , if @xmath1 is conjugate to @xmath8 under a conjugacy @xmath40 , @xmath2 is conjugate to @xmath41 under a conjugacy @xmath42 , and @xmath43 . let @xmath24 be a factor triple where @xmath0 is a 1-block factor code induced by the map @xmath44 ( b stands for block ) . the map @xmath45 naturally extends to elements of @xmath30 . above every @xmath2-block @xmath46 of length @xmath14 there is a set of @xmath1-blocks @xmath47 of length @xmath14 which are sent to @xmath46 by @xmath45 ; i.e. , @xmath48 . given @xmath49 , define @xmath50 and let @xmath51 a is a block @xmath46 such that @xmath52 for some @xmath53 . such an index @xmath54 is called a of @xmath46 . a factor code @xmath0 has a if there is a magic block of @xmath0 of length @xmath55 . * theorem 9.1.7)[thm : magicsymbol ] let @xmath24 be a factor triple . there is a factor triple @xmath56 conjugate to @xmath24 such that @xmath57 is a 1-block code with a magic symbol . by a well - known result of parry @xcite , generalizing an earlier result of shannon @xcite , in one dimension every irreducible shift of finite type on a finite alphabet has a unique measure of maximal entropy . burton and steif @xcite give a counterexample to this statement in higher dimensions . however , they show that such measures all have the uniform conditional distribution property stated in theorem [ bust ] ( which is related to an earlier work of lanford and ruelle @xcite ) . given a finite set @xmath58 , the boundary of the complement of @xmath33 is @xmath59 . * proposition 1.19)[bust ] let @xmath1 be a 1-step d - dimensional sft and @xmath60 be a measure of maximal entropy on @xmath1 . then the conditional distribution of @xmath60 on any finite set @xmath61 given the configuration on @xmath62 is @xmath60-a.s . uniform over all configurations on @xmath33 which extend the configuration on @xmath63 . given a factor triple @xmath24 and an ergodic measure @xmath3 on @xmath2 , there can exist more than one ergodic measure of relative maximal entropy over @xmath3 ; i.e. , there can be more than one ergodic measure on @xmath1 which projects to @xmath3 and has maximal entropy among all measures in the fibre @xmath21 , see ( * ? ? * example 3.3 ) . we use lemma [ lem : limit ] and follow techniques developed by burton and steif in the proof of theorem [ bust ] to show the uniform conditional distribution property for measures of relative maximal entropy in theorem [ cuniformdist ] . [ lem : limit ] ( * ? ? ? * theorem 4.7 ) let @xmath64 be a probability space . let @xmath65 be a finite sub - algebra of @xmath66 and let @xmath67 be an increasing sequence of sub-@xmath27-algebras of @xmath66 with @xmath68 . then @xmath69 . [ cuniformdist ] let @xmath6 be a 1-block factor code from a 1-step , 1-dimensional sft @xmath1 to a sofic shift @xmath2 . let @xmath3 be an invariant measure on @xmath2 , and @xmath60 be an invariant measure of relative maximal entropy over @xmath3 . given a finite set @xmath32 and @xmath9 in @xmath2 the conditional distribution of @xmath60 on @xmath12 restricted to @xmath33 given the configuration on @xmath62 is @xmath60-a.s . uniform over all configurations on @xmath33 which extend the configuration on @xmath63 and map to the same configuration in @xmath2 under the factor code @xmath0 . let @xmath33 be a finite subset of @xmath36 . let @xmath70 be a configuration of @xmath2 on @xmath33 . pick a configuration @xmath71 of @xmath1 on @xmath63 such that @xmath72 . starting from @xmath60 , we define a measure @xmath73 on @xmath1 by uniformizing over pre - images of @xmath70 that have @xmath71 on the boundary . we then show that if @xmath60 does not have the required uniform conditional distribution property then @xmath73 has greater entropy than @xmath60 , but still is an element of the fibre @xmath21 . this is a contradiction and will therefore establish the required uniform conditional distribution property of @xmath60 . let @xmath74 be the set of all configurations on @xmath33 which extend @xmath71 and map to @xmath70 under the factor code @xmath0 . let the @xmath36-interval @xmath75 $ ] be large enough so that @xmath76 and let @xmath77 be the partition of @xmath36 by translates of @xmath78 ( @xmath79 ) , i.e. @xmath80 $ ] . let @xmath81 and @xmath82 be the corresponding translates of @xmath33 and @xmath83 in @xmath84 . given @xmath85 let @xmath86 be the partition of @xmath1 generated by the configurations of @xmath1 on @xmath87 , and @xmath88 be the @xmath27-algebra generated by @xmath86 . when @xmath89 $ ] is a @xmath36-interval we sometimes denote @xmath90 by @xmath91 . considering @xmath78 above , @xmath92 is the finite @xmath27-algebra generated by the partition @xmath93\right._{m-1 } : x_{-m}x_{-m+1}\dots x_{m-1}\in\mathscr l(x)\},\ ] ] and @xmath94 . we obtain the measure @xmath95 from the measures @xmath60 and @xmath71 as follows . define the map @xmath96 by @xmath97 and @xmath98 for each integer @xmath14 . since @xmath99 is in @xmath100 and each element of @xmath100 extends @xmath71 , the assumption that @xmath1 is a 1-step sft implies that @xmath101 lies in @xmath1 . for each @xmath102 in @xmath103 we have @xmath104 since @xmath101 and @xmath10 are the same except having alternative @xmath105 s in the same positions . let @xmath106 be a cylinder set of @xmath1 . define @xmath107 where @xmath108 is the bernoulli @xmath109 measure on @xmath103 . the measure @xmath95 is not necessarily invariant under @xmath23 ; however , for each cylinder set @xmath106 we have @xmath110 . so the new measure @xmath73 on @xmath1 defined by @xmath111 is @xmath23-invariant . since for each cylinder set @xmath112 of @xmath2 we have @xmath113 we deduce that both measures @xmath95 and @xmath73 are in the fibre @xmath114 define an equivalence relation on @xmath1 as follows . given @xmath115 in @xmath1 , say @xmath116 if either @xmath117 or else @xmath118 , @xmath119 , and @xmath120 in @xmath100 . denote the equivalence class containing @xmath10 by @xmath121 . such equivalence classes form a sub - partition of @xmath122 . let @xmath65 be the @xmath27-algebra generated by these equivalence classes . we show @xmath123 as follows , using lemma [ lem : parts ] which appears below . @xmath124 [ lem : parts ] reusing previous notations , we have a. @xmath125 . b. @xmath126 . equality occurs if and only if @xmath127 for each @xmath128 in @xmath100 and @xmath129 in @xmath130 such that for each @xmath10 in @xmath131 we have @xmath132 and @xmath34 in @xmath100 . c. @xmath133 . by definition , for a measure @xmath134 in @xmath135 we have @xmath136 where @xmath137 is in @xmath138 and @xmath139 is in @xmath130 ( if @xmath140 is zero we define @xmath141 ) . let @xmath10 be in @xmath139 . if @xmath142 or @xmath143 then for every @xmath137 in @xmath138 we have either @xmath144 or @xmath145 which both imply @xmath146 . let @xmath147 be the set consisting of elements in the partition @xmath130 so that for any point @xmath10 in @xmath148 we have @xmath149 and @xmath34 in @xmath100 . then @xmath150 there are exactly @xmath151 pairwise disjoint sets @xmath137 in @xmath138 defined by blocks which agree everywhere except on @xmath33 , and form a partition of @xmath152 . let these sets be denoted by @xmath153 where @xmath154 for each @xmath155 . it follows that @xmath156 let @xmath157 . by definition of @xmath95 , for each configuration @xmath128 in @xmath100 , each cylinder set @xmath158 in @xmath159 , and each @xmath160 we have @xmath161 it follows that @xmath162 ( use the same argument we had before equation to get the first equality above ) . then ( a ) follows from lemma [ lem : limit ] and the fact that @xmath163 is an increasing sequence of @xmath27-algebras with @xmath164 . \(b ) is a standard property of entropy . to show ( c ) let @xmath165 denote the @xmath27-algebra @xmath166 ( the choice of notation is intended to remind the reader that @xmath167 is a sub-@xmath27-algebra of @xmath168 ) . any set @xmath169 in @xmath165 splits up into @xmath170 elements of @xmath171 , @xmath172 of equal measure ( where @xmath173 is the number of occurrences in @xmath169 of blocks that are randomized under @xmath95 ) . each of these sets has the property that @xmath174 . accordingly we see that @xmath175 . taking a limit we obtain @xmath176 we then notice that @xmath177 since the measures @xmath60 and @xmath95 agree on elements of @xmath178 . finally it is clear that @xmath179 . combining these three equalities and inequalities we obtain @xmath180 as required . _ proof of theorem [ cuniformdist ] ( continued ) . _ since @xmath60 is a measure of relative maximal entropy it follows that all of the inequalities in equations are forced to be equalities . in particular , we have @xmath181 . then lemma [ lem : parts](b ) implies that @xmath182 for each configuration @xmath128 in @xmath100 and @xmath129 in @xmath130 such that for each @xmath10 in @xmath129 we have @xmath132 and @xmath34 in @xmath100 . note that given a finite set @xmath32 , both configurations @xmath183 on @xmath33 and @xmath184 on @xmath63 with @xmath72 are chosen arbitrarily . by choosing different configurations and noting that @xmath185 , equation implies that for any configuration @xmath128 of @xmath1 occurring at @xmath33 we have @xmath186 where @xmath187 is the number of configurations of @xmath1 occurring at @xmath33 which extend @xmath188 and project to @xmath189 . now for a positive integer @xmath190 let @xmath191 $ ] , and let @xmath192 be the partition of @xmath36 by translates of @xmath193 where @xmath194 , i.e. @xmath195 $ ] . let @xmath196 . since @xmath197 and @xmath198 are increasing sequences with @xmath199 and @xmath200 it follows from lemma [ lem : limit ] that @xmath201 when @xmath0 is a finite - to - one factor code from a sft @xmath1 to a sofic shift @xmath2 , there is a uniform upper bound on the number of pre - images of points in @xmath2 @xcite . the minimal number of @xmath0-pre - images of points in @xmath2 is called the of the code and denoted by @xmath202 . [ thm : samenumber ] @xcite let @xmath6 be a finite - to - one factor code from a sft @xmath1 to an irreducible sofic shift @xmath2 . then every doubly transitive point of @xmath2 has exactly @xmath203 preimages . theorem 9.1.11 in @xcite shows this result when @xmath1 is irreducible . however , the proof only depends on the irreducibility of @xmath2 . [ cor : conjdeg ] let @xmath24 and @xmath38 be conjugate factor triples . then we have @xmath204 . a family @xmath205 of sequences in a sft @xmath1 is if , for each integer @xmath54 , @xmath206 are all distinct . [ prop : degreemagic ] @xcite let @xmath25 be a finite - to - one 1-block factor code from a sft to an irreducible sofic shift @xmath2 . the preimages of a transitive point are mutually separated . if @xmath0 has a magic symbol @xmath7 then @xmath207 . for a proof , the reader is referred to proposition 9.1.9 and exercise 9.1.3 of @xcite . we will find a quantity analogous to the degree when @xmath0 is an infinite - to - one factor code . this will be done by developing the following equivalence relation on @xmath1 . figure [ fig : transition ] illustrates definition [ defn : equiv ] . [ defn : equiv ] suppose @xmath24 is a factor triple and @xmath208 . we say there is a from @xmath10 to @xmath11 and denote it by @xmath209 if for each integer @xmath14 , there exists a point @xmath20 in @xmath1 so that 1 . @xmath210 , and 2 . @xmath211 for some @xmath212 . write @xmath213 if the above conditions do not hold . we write @xmath214 , and say @xmath10 and @xmath11 are in the same if @xmath215 and @xmath216 . the relation @xmath217 is an equivalence relation . denote the set of transition classes in @xmath1 over @xmath218 by @xmath219 . sometimes we denote @xmath220 simply by @xmath221 when there is no ambiguity in understanding @xmath0 . say @xmath222{\rightarrow}[x']$ ] if @xmath215 ( well - defined ) . use the notation @xmath222\nrightarrow [ x']$ ] otherwise . the following fact is derived from definition [ defn : equiv ] immediately . [ fact : class ] let @xmath25 be a 1-block factor code from a 1-step sft @xmath1 to a sofic shift @xmath2 . let @xmath115 project to the same point under @xmath0 and @xmath223 where @xmath224 is a strictly increasing sequence in @xmath36 . then we have @xmath214 . note that fact [ fact : class ] gives an obvious case when two points lie in the same transition class . simple examples allow one to find equivalent points without having a common symbol at the same time . let @xmath24 be a factor triple . the minimal number of transition classes over points of @xmath2 is called the of @xmath0 and denoted by @xmath225 . the following proposition shows that conjugate factor triples have the same class degree . [ prop : conjinv ] let @xmath24 and @xmath38 be conjugate factor triples under conjugacies @xmath226 and @xmath227 . then a. for each @xmath9 in @xmath2 , @xmath228 b. @xmath229 \(a ) is clear by noting that given @xmath230 in @xmath1 , @xmath231 if and only if @xmath232 , and ( b ) is a direct corollary of ( a ) . the following two theorems give an upper bound on the number of transition classes over points of @xmath2 . later in theorem [ thm : classdegree ] we give a finitary characterization of the number of such classes . [ thm : symboloccurs ] let @xmath25 be a @xmath55-block factor code from a 1-step sft @xmath1 to a sofic shift @xmath2 . let @xmath9 in @xmath2 and let @xmath7 be a symbol in @xmath2 occurring infinitely often in positive coordinates of @xmath9 . then @xmath233 . let @xmath234 be a strictly increasing sequence of integers such that @xmath235 for each @xmath236 . let @xmath237 and suppose @xmath238 contains @xmath239 distinct transition classes @xmath240 . form a set @xmath241 where @xmath242 is an arbitrary point of @xmath243 . since @xmath244 contains exactly @xmath245 symbols , it follows that for each @xmath236 there are at least two points in @xmath246 with the same @xmath247th coordinate . the pigeonhole principle implies that there is a subsequence @xmath248 of @xmath234 and at least two points @xmath10 and @xmath11 in @xmath246 with @xmath249 for each @xmath250 . fact [ fact : class ] implies that @xmath214 . this contradicts the assumption that @xmath10 and @xmath11 are in different transition classes . a similar proof to the above implies the following theorem . [ thm : classsize ] let @xmath25 be a @xmath55-block factor code from a 1-step sft @xmath1 to a sofic shift @xmath2 . then @xmath251 for each @xmath218 . moreover , @xmath252 is the uniform upper bound on the number transition classes over each right transitive point of @xmath2 . we show when @xmath25 is a factor code from a sft @xmath1 to an irreducible sofic shift @xmath2 , there are exactly @xmath225 transition classes over a transitive point of @xmath2 . in order to show this , we introduce another quantity @xmath253 in definition [ defn : cstar ] , defined concretely in terms of blocks . since the class degree is invariant under conjugacy of factor triples , by recoding we may focus only on 1-step sfts and 1-block factor codes . [ defn : tb ] let @xmath25 be a @xmath55-block factor code from a 1-step sft @xmath1 to a sofic shift @xmath2 . let @xmath46 be a @xmath2-block of length @xmath254 . let @xmath14 be an integer in @xmath255 , and @xmath256 be a subset of @xmath257 . we say @xmath258 @xmath259 @xmath14 if there is a block @xmath260 with @xmath261 , and @xmath262 . a triple @xmath263 is called a of @xmath0 if every block in @xmath264 is routable through a symbol of @xmath256 at time @xmath14 . the cardinality of the set @xmath256 is called the of the transition block @xmath263 . figure [ fig : depth ] illustrates definition [ defn : tb ] . is a transition block with @xmath265 . the blocks @xmath266 are routable through members of @xmath256 at time @xmath14 via blocks @xmath267.,height=192 ] [ defn : cstar ] let @xmath268 a of @xmath0 is a transition block of depth @xmath253 . [ ex : mtb ] in figure [ fig : mtb ] , we display an example of a labeled graph which defines an infinite - to - one 1-block factor code @xmath0 . we see that @xmath269 is a minimal transition block of @xmath0 of depth 1 . for example , observe that block @xmath270 is routable through @xmath271 at time 1 by considering @xmath272 . , height=115 ] we need to develop some definitions and lemmas before proving theorem [ thm : classdegree ] below . given a factor triple @xmath24 , a point @xmath9 in @xmath2 and a block @xmath273 occurring in @xmath9 , it may be the case that some preimages of @xmath273 can not be extended to a point in @xmath1 which projects to @xmath9 under @xmath0 . here we show that in order to recognize those non - extendable blocks we do not need to deal with a bi - infinite sequence . a finite extension of a block is sufficient to determine if the block can be extended to a point above @xmath9 . [ defn : exten ] let @xmath24 be a factor triple and @xmath9 be in @xmath2 . let @xmath273 be a block of @xmath2 occurring in @xmath9 at a @xmath36-interval @xmath274 $ ] . a block @xmath46 of @xmath2 is called a of @xmath273 if @xmath46 occurs in @xmath9 at @xmath275 $ ] for some integer @xmath276 and @xmath277}=u\ } \\ & = \{u\in{\pi^{-1}}_\text{b}(v)\colon \exists\ , b\in{\pi^{-1}}_\text{b}(w),\,b_{[l , l+n - m]}=u\}. \end{split}\ ] ] [ lem : exten ] let @xmath24 be a factor triple and @xmath9 in @xmath2 . let @xmath273 be a block occurring in @xmath9 at a @xmath36-interval @xmath274 $ ] . there is a @xmath9-synchronizing extension of @xmath273 . let @xmath278}=u\}.\ ] ] for @xmath276 let @xmath279}),\ , b^l_{[l , l+n - m]}=u\}.\ ] ] we show for some @xmath280 we have @xmath281 . then it will imply that @xmath282}$ ] is a @xmath9-synchronizing extension of @xmath273 . clearly for all @xmath276 we have @xmath283 . we show @xmath284 for some @xmath280 as follows . first note that since @xmath0 is a factor code , @xmath87 is not empty . it follows that the sequence @xmath285 is a non - increasing sequence of non - empty finite sets . thus there is @xmath280 such that for each @xmath286 we have @xmath287 . we show @xmath284 . let @xmath288 be in @xmath289 . then @xmath288 is in @xmath290 for each @xmath286 . it follows that for each @xmath286 there is @xmath291 in @xmath292})$ ] with @xmath293}=u$ ] . for each @xmath286 consider the cylinder set @xmath294 . the sequence @xmath295 is a non - increasing sequence of non - empty compact subsets of @xmath1 . it follows that @xmath296 is not empty . let @xmath297 . then @xmath298}=u$ ] and @xmath299 which implies that @xmath288 is in @xmath87 . here we introduce a relation between symbols and transition classes . [ defn : belong ] let @xmath25 be a @xmath55-block factor code from a 1-step sft @xmath1 to a sofic shift @xmath2 . let @xmath9 be in @xmath2 . let @xmath14 be in @xmath36 , @xmath54 be a symbol in @xmath300 , and @xmath106 be a transition class in @xmath221 . say @xmath54 @xmath106 @xmath14 , and denote it by @xmath301 , if @xmath302 in other words , the set of transition classes in @xmath221 containing a point with @xmath14th coordinate being @xmath54 , is the same as the set of transition classes to which there is a transition from the class @xmath106 ( including @xmath106 itself ) . say @xmath54 is if there is no transition class @xmath106 in @xmath221 for which @xmath303 . example [ ex : transient ] illustrates the definition above . [ ex : transient ] consider the directed labeled graph in figure [ fig : transient ] which presents a 1-block factor code @xmath6 where @xmath304 and @xmath305 . let @xmath9 be the point @xmath306 in @xmath2 . by definition [ defn : equiv ] , there are 3 distinct transition classes in @xmath221 as follows : , height=96 ] 1 . transition class @xmath307=\{x_1\}$ ] where @xmath308 . 2 . transition class @xmath309 $ ] where @xmath310 is a point in @xmath12 which does not contain symbols @xmath311 or @xmath312 but contains only @xmath245 s and @xmath313 s from some time onwards , for example : @xmath314 . 3 . transition class @xmath315 $ ] where @xmath316 is a point in @xmath12 which does not contain symbols @xmath245 or @xmath313 but contains only @xmath311 s and @xmath312 s from some time onwards , for example : @xmath317 . clearly @xmath318 and @xmath319 , but not vice versa . @xmath320 , @xmath321 , and symbol @xmath322 is transient at any time . note that the class degree of the factor code @xmath0 is 2 ( consider a point of @xmath2 in which the block @xmath323 occurs at some time ) . [ nontransient ] let @xmath25 be a @xmath55-block factor code from a 1-step sft @xmath1 to a sofic shift @xmath2 , and @xmath10 be a point in @xmath1 . there is @xmath324 such that for each @xmath325 the symbol @xmath326 is not transient . in fact , @xmath327 can be found in such a way that for each @xmath325 , @xmath326 belongs to the class @xmath222 $ ] at time @xmath14 ; i.e. , @xmath328 $ ] . let @xmath299 . suppose that the transition class @xmath106 in @xmath221 satisfies @xmath222\nrightarrow c$ ] . then there exists @xmath329 such that for any @xmath330 in @xmath12 with @xmath331 we have @xmath332 . denote the smallest such @xmath54 by @xmath333 . note that @xmath333 may be @xmath13 ( but not @xmath15 ) . let @xmath334\nrightarrow c\}$ ] . let @xmath325 . we show @xmath335{\rightarrow}c\}=\{c\in{\mathscr{c}}(y)\colon \textrm{there is } z\in c \textrm { with } z_n = x_n\},\ ] ] which will imply the result @xmath328.$ ] let @xmath106 be a transition class in @xmath221 with @xmath222\to c$ ] . it follows that @xmath336 for some @xmath330 in @xmath106 . therefore , there is a point @xmath19 in @xmath12 as follows . @xmath337 for some @xmath338 . since @xmath339 from time @xmath280 onwards , it follows that @xmath19 is in the class @xmath106 . moreover , we have @xmath340 by the construction . it follows that @xmath106 belongs to the set in the right hand side of equation . now let @xmath106 be an equivalence class in @xmath221 with @xmath222\nrightarrow c$ ] . suppose , for a contradiction , @xmath106 contains a point @xmath330 with @xmath341 . form the point @xmath19 as follows . @xmath342 note that since @xmath1 is a 1-step sft and @xmath341 , the point @xmath19 is in @xmath1 . moreover @xmath19 maps to @xmath9 by the construction . noticing that @xmath339 from the time @xmath14 onwards , shows that @xmath19 lies in the class @xmath106 . recall that @xmath343 . by the argument in the first paragraph of the proof , any point in @xmath12 with @xmath333th coordinate equal to @xmath344 does not belong to @xmath106 . this contradicts that @xmath19 is a point in @xmath106 . [ lem : s(c ) ] let @xmath25 be a @xmath55-block factor code from a 1-step sft @xmath1 to a sofic shift @xmath2 . let @xmath218 . there is @xmath324 such that for each @xmath325 and @xmath106 in @xmath221 there is a symbol @xmath54 in @xmath345 which belongs to the class @xmath106 at time @xmath14 ; i.e. , @xmath346 . let @xmath347 for some finite @xmath245 . let @xmath348 be a set containing an arbitrary point @xmath349 of @xmath243 for each @xmath243 . by lemma [ nontransient ] there is a @xmath350 such that for each @xmath351 the symbol @xmath352 belongs to @xmath243 at time @xmath14 . let @xmath353 . note that @xmath324 . moreover , for each @xmath325 and @xmath243 the symbol @xmath354 belongs to the class @xmath243 at time @xmath14 . [ prop : ypi ] let @xmath25 be a @xmath55-block factor code from a 1-step sft @xmath1 to a sofic shift @xmath2 . let @xmath3 be an invariant measure on @xmath2 . then @xmath355 let @xmath9 in @xmath2 and let @xmath356 be the infimum of the set of @xmath327 s with the properties given in the statement of lemma [ lem : s(c ) ] . we show @xmath357 which will imply the result immediately . note that @xmath358 if and only if @xmath359 . this shows @xmath360 . for @xmath361 let @xmath362 we have @xmath363 . since @xmath3 is @xmath23-invariant it follows that @xmath364 . therefore @xmath365 for @xmath3-a.e . @xmath9 in @xmath2 . let @xmath1 be a shift space . a point @xmath10 in @xmath1 is if for each block @xmath169 which occurs in @xmath10 , @xmath169 occurs infinitely often in the positive coordinates of @xmath10 . note that given any invariant measure @xmath60 on @xmath1 , @xmath60-almost every point of @xmath1 is recurrent . [ thm : classdegree ] let @xmath0 be a 1-block factor code from a 1-step sft @xmath1 to a sofic shift @xmath2 . let @xmath9 be a point of @xmath2 and @xmath245 be the minimal depth of a transition block occurring in @xmath9 . then @xmath366 . moreover , if @xmath9 is recurrent then @xmath367 . we show @xmath366 by finding a transition block of depth @xmath368 occurring in @xmath9 . we construct such a transition block in the following 4 stages . let @xmath369 and suppose @xmath370 . choose an integer @xmath371 satisfying the properties given in the statement of lemma [ lem : s(c ) ] . * stage @xmath372 * we claim there is @xmath373 such that for each @xmath374 , there is @xmath375 so that @xmath376 is not transient . suppose there is no such @xmath377 . it follows that for each @xmath378 there is a point @xmath379 in @xmath12 such that @xmath380 is transient for all @xmath381 . consider the sequence @xmath382 , and let @xmath10 be the limit of a convergent subsequence of it . clearly @xmath383 . however , @xmath384 is transient for each @xmath385 , contradicting lemma [ nontransient ] . * stage @xmath386 * we claim there is @xmath387 and a set of symbols @xmath388 with @xmath389 such that for each @xmath54 in @xmath390 there is a block @xmath391 which begins with @xmath54 and ends with @xmath392 . see figure [ fig : n2 ] . let @xmath393 be a transition class in @xmath221 and @xmath394 be a point in @xmath393 . for each symbol @xmath54 in @xmath390 ( non - empty by the choice of @xmath371 ) , there is a bi - infinite sequence @xmath395 in the class @xmath393 with @xmath396 such that @xmath395 matches @xmath394 from some time @xmath397 onwards . let @xmath398 and @xmath399 note that @xmath400 . rename @xmath401 for each @xmath402 as @xmath392 , and let @xmath403 . * stage @xmath404 * we claim there is @xmath405 such that for each point @xmath10 in @xmath238 there is a point @xmath11 in @xmath238 so that @xmath406 for every @xmath407\cup [ n_4,\infty)$ ] , and @xmath408 . see figure [ fig : n3 ] . let @xmath10 be a point in @xmath12 . by stage 1 , there is @xmath409 such that @xmath376 is not transient ; i.e. , @xmath376 belongs to some transition class @xmath393 at time @xmath190 . it follows that there is a point @xmath19 in the class @xmath393 with @xmath410 . the new block @xmath411 is an allowable block of @xmath1 . moreover , this block maps to the block @xmath412 note that the symbol @xmath413 belongs to the class @xmath393 at time @xmath377 . so by stage 2 , there is a block starting at @xmath413 ending at @xmath392 which maps to @xmath414 denote this block by @xmath415 connect blocks in equations and at @xmath413 to get the following allowable block of @xmath1 @xmath416 observe that having @xmath376 in @xmath417 implies that @xmath10 must belong to a transition class @xmath418 with @xmath419 . on the other hand having @xmath392 in @xmath420 implies that there is a point @xmath271 in @xmath393 with @xmath421}=\dots x_{n_1}\dots x_tu_{t+1}\dots u_{n_2}v_{n_2 + 1}\dots v_{n_3 - 1}a_e\ ] ] which matches @xmath10 from some time @xmath422 onwards for some @xmath423 . let @xmath424\cup [ j,\infty),\,~x'_{n_3}=a_e\}.\ ] ] we claim that there is @xmath425 such that for each @xmath426 with @xmath427 for some @xmath409 , there exists a point @xmath428 with @xmath406 for each @xmath429\cup [ n^e,\infty)$ ] , and @xmath430 . then letting @xmath431 will complete the proof of stage 3 . suppose , for a contradiction , that there does not exist such an @xmath432 . then there is a sequence @xmath433 with @xmath434 for some @xmath435 , such that @xmath436 . let @xmath437 be the limit of a convergent subsequence of @xmath433 . clearly @xmath437 is in @xmath238 and @xmath438 which contradicts the fact that for each @xmath10 in @xmath12 , @xmath439 . * stage _ 4_. * let @xmath46 be a @xmath9-synchronizing extension of the block @xmath440 occurring in @xmath441 $ ] for some @xmath442 and @xmath443 . we claim @xmath444 is a transition block of depth @xmath368 . let @xmath445 . by the definition of @xmath9-synchronizing extension , there is @xmath10 in @xmath12 with @xmath446}=u$ ] . by stages 1 , 2 , and 3 there is a point @xmath11 in @xmath12 with @xmath406 for each @xmath407\cup[n_4,\infty)$ ] , and @xmath408 . having @xmath447 , @xmath408 , @xmath448 and @xmath449})=v$ ] simply means that @xmath288 is routable through a symbol of @xmath450 at time @xmath451 . since @xmath288 in @xmath452 is arbitrary it follows that @xmath453 is a transition block . because @xmath450 has @xmath368 elements it follows that @xmath46 is a transition block of depth @xmath368 . now suppose @xmath9 is recurrent . then a transition block of depth @xmath245 which occurs in @xmath9 , it occurs infinitely often in the positive coordinates of @xmath9 . apply the same method which was used in the proof of theorem [ thm : symboloccurs ] to see @xmath454 . since by the first part of the theorem @xmath455 we obtain @xmath367 when @xmath9 is recurrent . [ cor : classdegree ] let @xmath25 be a factor code from a sft @xmath1 to an irreducible sofic shift @xmath2 . then a. @xmath456 . b. there are exactly @xmath457 transition classes over every right transitive point of @xmath2 . c. given an ergodic measure @xmath3 on @xmath2 , @xmath3-almost every point of @xmath9 has the same number of transition classes over it . we call this number @xmath458 . let @xmath263 be a transition block of @xmath0 of depth @xmath253 and let @xmath9 be a right transitive point of @xmath2 . notice that @xmath46 occurs infinitely often in the positive coordinates of @xmath9 . then ( a ) and ( b ) are clear by theorem [ thm : classdegree ] . now let @xmath3 be an ergodic measure on @xmath2 . let @xmath459 be a transition block of @xmath0 with @xmath460>0 $ ] such that @xmath461>0\}.\ ] ] note that @xmath47 occurs infinitely often in the positive coordinates of @xmath3-almost every point of @xmath2 . thus ( c ) is obtained as a direct corollary of theorem [ thm : classdegree ] . we now show that for a finite - to - one factor code , the degree and the class degree of the code are the same . [ thm : finitemagic ] let @xmath25 be a finite - to - one factor code from a sft to an irreducible sofic shift @xmath2 . then @xmath462 . corollary [ cor : conjdeg ] and proposition [ prop : conjinv](b ) imply that the degree of a code and the class degree of a code are both invariant under conjugacy of factor triples . so by recoding , without loss of generality , we may assume @xmath1 is a 1-step sft and @xmath0 is a 1-block factor code . let @xmath9 be a transitive point of @xmath2 . by proposition [ prop : degreemagic ] the preimages of @xmath9 are mutually separated . it follows that each transition class above @xmath9 has only one point . therefore , @xmath463 . in section [ sec : unif ] we mentioned that although every 1-dimensional irreducible sft has a unique ergodic measure ( parry measure ) of maximal entropy , there can be more than one ergodic measure of relative maximal entropy over an ergodic measure ; i.e. , given a factor triple @xmath24 and an ergodic measure @xmath3 on @xmath2 , there can exist more than one ergodic measure on @xmath1 that projects to @xmath3 under @xmath0 and has maximal entropy among measures in the fibre @xmath21 , see ( * ? ? ? * example 3.3 ) . in this section we show that the number of such measures over a fully supported ergodic measure can be no more than the class degree . [ thm : final ] let @xmath24 be a factor triple and @xmath3 be a fully supported ergodic measure on @xmath2 . the number of ergodic measures of relative maximal entropy over @xmath3 is at most @xmath225 . since entropy is a conjugacy invariant the following observation follows immediately . [ obs : numberofmrme ] let @xmath24 and @xmath464 be conjugate factor triples . let @xmath3 be an ergodic measure on @xmath2 and @xmath465 be its corresponding ergodic measure on @xmath41 . the number of ergodic measures of relative maximal entropy over @xmath3 is the same as the number of ergodic measures of relative maximal entropy over @xmath465 . in 2003 , petersen , quas , and shin @xcite found an upper bound on the number of measures of relative maximal entropy . let @xmath466 denote the minimum number of symbols in @xmath244 as @xmath7 runs over the symbols in @xmath467 for which @xmath468>0 $ ] . * corollary 1)[pqs ] let @xmath25 be a 1-block code from a 1-step sft @xmath1 to a sofic shift @xmath2 . let @xmath3 be an ergodic measure on @xmath2 . the number of ergodic measures of maximal entropy over @xmath3 is at most @xmath466 . this bound suffers from not being invariant under conjugacy . for example , the full 2-shift and any higher block presentation of it give different bounds on the number of ergodic measures of maximal entropy which map to the trivial measure on the full 1-shift . one possibility to avoid having a non - invariant upper bound is to take the minimum of the bound in @xcite over all conjugate factor triples with 1-block factor codes . this even improves the original bound . however , there is no known algorithm for computing this minimum . here we show that given an ergodic measure @xmath3 on @xmath2 , the number of transition classes over a @xmath3-generic point of @xmath2 is an upper bound on the number of measures of relative maximal entropy over @xmath3 . proposition [ prop : conjinv](b ) verifies that this bound is invariant under conjugacy of factor triples and proposition [ minbound ] shows that it beats the bound mentioned above obtained by minimizing the bound in @xcite over conjugate factor triples . [ minbound ] let @xmath6 be a 1-block factor code from a sft @xmath1 to a sofic shift @xmath2 . let @xmath3 be an ergodic measure on @xmath2 . then we have @xmath469 where @xmath9 is a @xmath3-generic point of @xmath2 . equality holds if @xmath0 is finite - to - one and @xmath3 is fully supported . let the minimum of @xmath470 over all factor triples @xmath464 which are conjugate to @xmath24 and @xmath471 is 1-block , be attained at @xmath472 where @xmath1 is conjugate to @xmath473 under a conjugacy @xmath474 , @xmath2 is conjugate to @xmath475 under a conjugacy @xmath476 , and @xmath477 . denote the measure @xmath478 by @xmath479 . suppose that the minimum of @xmath480 over all @xmath7 in @xmath481 with @xmath482>0 $ ] is attained at a symbol @xmath483 . let @xmath484 be a @xmath479-generic point of @xmath475 . there is a strictly increasing sequence of integers @xmath224 with @xmath485 . theorem [ thm : classsize ] implies that @xmath486 . then by proposition [ prop : conjinv](a ) we have latexmath:[$|{\mathscr{c}}_\pi(y)|=|{\mathscr{c}}_{\bar\pi}(\bar y)|\leq part of the result . now suppose @xmath0 is finite - to - one and @xmath3 is fully supported . since @xmath3 is fully supported we have @xmath488 therefore the minimum of @xmath470 over all @xmath464 which are conjugate to @xmath24 and @xmath471 is 1-block is attained at a factor triple @xmath472 with a magic symbol @xmath489 . by proposition [ prop : degreemagic ] we have @xmath490 which implies @xmath491 since the degree of a code is invariant under conjugacy of factor triples we deduce that @xmath492 recalling that when @xmath0 is finite - to - one the degree and the class degree of @xmath0 are equal , the proof is complete . note that equality in proposition [ minbound ] does not always hold . for example , consider the trivial factor code @xmath493 . then @xmath494 ; however , if @xmath495 is a factor triple conjugate to @xmath496 then @xmath497 must be strictly greater than 1 and therefore @xmath498 for any ergodic measure @xmath3 on @xmath2 . let @xmath25 be a 1-block factor code from a sft @xmath1 to a sofic shift @xmath2 and @xmath3 be an ergodic measure on @xmath2 . let @xmath499 be invariant measures in the fibre @xmath21 . the @xmath500 of @xmath499 over @xmath3 is defined as follows : if @xmath501 are measurable subsets of @xmath1 then @xmath502 writing @xmath503 for the projection @xmath504 onto the @xmath54th coordinate , it is shown that for @xmath505-almost every @xmath506 , @xmath507 is the same for each @xmath54 @xcite . we will use theorem [ orthogonal ] below which is the main theorem from @xcite to prove a stronger theorem ( theorem [ equivagonal ] ) . [ orthogonal ] ( * ? ? ? * theorem 1 ) let @xmath25 be a 1-block factor code from a 1-step sft @xmath1 to a sofic shift @xmath2 . let @xmath3 be an ergodic measure on @xmath2 , and two distinct ergodic measures @xmath16 and @xmath17 be measures of relative maximal entropy over @xmath3 . then @xmath508 [ equivagonal ] let @xmath24 be a factor triple . let @xmath3 be an ergodic measure on @xmath2 , and two distinct measures @xmath16 and @xmath17 be ergodic measures of relative maximal entropy over @xmath3 . then @xmath509 . first we show that , without loss of generality , we may assume @xmath1 is a 1-step sft and @xmath0 is a 1-block factor code with a magic symbol . suppose @xmath38 is a factor triple conjugate to @xmath24 and @xmath510 is a conjugacy from @xmath1 to @xmath8 . by the proof of proposition [ prop : conjinv](a ) we have @xmath511 moreover , the corresponding measure to @xmath512 under the conjugacy @xmath513 is @xmath514 where @xmath515 and @xmath516 are corresponding measures to @xmath16 and @xmath17 under @xmath40 . it follows that @xmath517 therefore by recoding , without loss of generality , we may assume @xmath1 is a 1-step sft and @xmath0 is a 1-block factor code . let @xmath518 be in @xmath1 , @xmath14 be in @xmath36 , and @xmath519 be a symbol in @xmath26 . write @xmath520 if @xmath232 and there are integers @xmath521 and @xmath522 such that @xmath523 it is worth mentioning that the relation @xmath524 is not in general an equivalence relation on @xmath1 . we have @xmath525 now suppose , for a contradiction , that @xmath526 . there must be an integer @xmath14 and a symbol @xmath519 in @xmath26 such that @xmath527 by applying @xmath528 , without loss of generality , we may assume @xmath529 . since we have @xmath530 it follows that there is an integer @xmath531 so that @xmath532 considering only blocks of length @xmath533 , there must be blocks @xmath534 in @xmath1 such that @xmath535}=a,\,v_{[-k,\,k]}=b \big\}>0.\ ] ] both @xmath246 and @xmath169 are routable through @xmath519 at time @xmath14 ; i.e. , there are blocks @xmath536 in @xmath264 with @xmath537 , @xmath538 , @xmath539 , @xmath540 , and @xmath541 . let @xmath542 be the set @xmath543 . basic properties of conditional expectation imply that @xmath544}|{\pi^{-1}}({\mathscr{b}}_y)\right ) & = { \mathbb{e}}\left({\mathbb{e}}_{\mu_1}\left(\mathbf 1_{\left._{-n}[a]\right.}|{\pi^{-1}}({\mathscr{b}}_y)\vee \sigma(x _ { g^c})\right)|{\pi^{-1}}({\mathscr{b}}_y)\right ) \\ & = { \mathbb{e}}\left({\mathbb{e}}_{\mu_1}\left(\mathbf 1_{\left._{-n}[a']\right.}|{\pi^{-1}}({\mathscr{b}}_y)\vee \sigma(x _ { g^c})\right)|{\pi^{-1}}({\mathscr{b}}_y)\right ) \\ & = { \mathbb{e}}_{\mu_1}\left(\mathbf 1_{[a']}|{\pi^{-1}}({\mathscr{b}}_y)\right ) , \end{split}\ ] ] where the second equality follows from theorem [ cuniformdist ] . similarly we have @xmath545}|{\pi^{-1}}{\mathscr{b}}_y)= { \mathbb{e}}_{\mu_2}(\mathbf 1_{[b']}|{\pi^{-1}}{\mathscr{b}}_y).\ ] ] let @xmath546}=a,\,v_{[k,\,k]}=b \big\}$ ] . since @xmath547\right.\times\left._{-k}[b]\right.$ ] we have @xmath548\times [ b ] ) \\ & = \int_y{\mathbb{e}}_{\mu_1}(\mathbf 1_{[a]}|{\pi^{-1}}{\mathscr{b}}_y){\mathbb{e}}_{\mu_2}(\mathbf 1_{[b]}|{\pi^{-1}}{\mathscr{b}}_y)\circ{\pi^{-1}}\,d\nu \\ & = \int_y{\mathbb{e}}_{\mu_1}(\mathbf 1_{[a']}|{\pi^{-1}}{\mathscr{b}}_y){\mathbb{e}}_{\mu_2}(\mathbf 1_{[b']}|{\pi^{-1}}{\mathscr{b}}_y)\circ{\pi^{-1}}\ , d\nu&\text{using } \eqref{expe}\text { and } \eqref{expec } \\ & = ( \mu_1\otimes\mu_2)([a']\times [ b ' ] ) \\ & = 0\end{aligned}\ ] ] where the last equality follows from theorem [ orthogonal ] since @xmath539 . this contradicts equation . the following theorem is a more general case of theorem [ thm : final ] . [ bound ] let @xmath24 be a factor triple and @xmath3 be an ergodic measure on @xmath2 . the number of ergodic measures of relative maximal entropy over @xmath3 is at most @xmath458 . let @xmath549 be a factor triple conjugate to @xmath24 under conjugacies @xmath550 and @xmath227 . by observation [ obs : numberofmrme ] , the number of ergodic measures of relative maximal entropy over @xmath3 and @xmath478 is the same . moreover , by proposition [ prop : conjinv](a ) , for each @xmath9 in @xmath2 we have @xmath551 . therefore by recoding , without loss of generality , we may assume @xmath1 is a 1-step sft and @xmath0 is a 1-block factor code . suppose , for a contradiction , that there are @xmath552 ergodic measures @xmath499 on @xmath1 of relative maximal entropy over @xmath3 . form the relatively independent joining @xmath505 on @xmath553 of the measures @xmath499 . note that for @xmath505-a.e . @xmath554 , @xmath507 has @xmath458 transition classes over it . the assumption @xmath552 implies that for @xmath505-a.e . @xmath555 there are distinct @xmath556 such that @xmath557 ; i.e. , @xmath558 at least one of the sets @xmath559 must have positive @xmath505-measure . it follows that @xmath560 this contradicts theorem [ equivagonal ] . let @xmath1 be a one - sided topologically mixing shift of finite type . an invariant measure @xmath60 on @xmath1 is a corresponding to @xmath561 if there are constants @xmath562 and @xmath563 such that @xmath564)}{\exp\left(-pn+(s_nf)(x)\right)}\leq c_2\ ] ] for every @xmath565 and @xmath157 , where @xmath566 . let @xmath24 be a factor triple and @xmath3 be a fully supported ergodic measure on @xmath2 . for any function @xmath571 the number of ergodic measures of maximal pressure of @xmath313 in the fibre @xmath4 is at most @xmath225 . d. blackwell . the entropy of functions of finite - state markov chains . in _ transactions of the first prague conference on information theory , statistical decision functions , random processes held at liblice near prague from november 28 to 30 , 1956 _ , pages 1320 . publishing house of the czechoslovak academy of sciences , prague , 1957 . d. gatzouras and y. peres . the variational principle for hausdorff dimension : a survey . in _ ergodic theory of @xmath572 actions ( warwick , 19931994 ) _ , volume 228 of _ london math . lecture note ser . _ , pages 113125 . cambridge univ . press , cambridge , 1996 . b. marcus , k. petersen , and s. williams . transmission rates and factors of markov chains . in _ conference in modern analysis and probability ( new haven , conn . , 1982 ) _ , volume 26 of _ contemp . _ , pages 279293 . amer . math . soc . , providence , ri , 1984 . k. petersen . information compression and retention in dynamical processes . in _ dynamics and randomness ( santiago , 2000 ) _ , volume 7 of _ nonlinear phenom . complex systems _ , pages 147217 . kluwer acad . publ . , dordrecht , 2002 .

we investigate the mass function of cold , dusty clumps in 11 low- and high - mass star - forming regions . using a homogeneous fitting technique , we analyze the shape of each region s clump mass function and examine the commonalities among them . we find that the submillimeter continuum clump mass function in low - mass star - forming regions is typically best fit by a lognormal distribution , while that in high - mass star - forming regions is better fit by a double power law . a single power law clump mass distribution is ruled out in all cases . fitting all of the regions with a double power law , we find the mean power law exponent at the high - mass end of each mass function is , consistent with the salpeter result of . we find no region - to - region trend in with the mass scale of the clumps in a given region , as characterized by their median mass . similarly , non non - parametric tests show that the _ shape _ of the clump mass function does not change much from region to region , despite the obvious changes in the intrinsic mass scale . this result is consistent with the hypothesis that the clump mass distribution is determined by a highly stochastic process , such as turbulent fragmentation . it may also suggest that the data reduction and analysis techniques strongly affect the shape of the derived mass function .

astro-ph0607095

advances in submillimeter continuum imaging techniques in the past decade have dramatically improved our understanding of the structure of the dense , cold interstellar medium ( ism ) in star - forming regions . in every region studied , the cold , dusty ism exhibits a clumpy , often filamentary structure whose density maxima correspond to sites of present or future star formation activity . the primary metric used to describe this inhomogeneous structure is the clump / core pc ) , dense condensations thought to be the immediate precursors of individual stars or small multiple systems . we use the term `` clump '' more generally to denote discrete structures which may form zero , one , or many stars and may therefore be much more massive than cores . in this sense , cores may be considered a subset of clumps . ] mass function . submillimeter continuum observations of the core mass function in nearby regions of low - mass star formation , such as serpens @xcite , @xmath3 oph @xcite , and orion b @xcite , have provided compelling evidence that the mass function of cold , dense cores mirrors the stellar initial mass function ( imf , e.g. @xcite , @xcite ) . it appears , therefore , that these dusty cores are the immediate precursors of stars . two main arguments are typically advanced in favor of this interpretation . first , both the stellar and core mass functions are well fit by salpeter - like power laws @xcite above @xmath41 @xmath5 . second , both the core and stellar mass functions appear to flatten out below @xmath41 @xmath5 and peak at about 0.1 @xmath5 @xcite . in other words , the core mass function and the stellar imf have similar _ shapes _ and intrinsic mass _ scales_. thus , by applying a more or less constant core - to - star mass conversion efficiency , the core mass function can be converted into the stellar imf . in papers i and ii in this series , we showed that , despite the large difference in the median clump mass , the clump mass function in massive star - forming regions ngc 7538 and m17 has a similar shape to that in low - mass star - forming regions ( @xcite ; papers i and ii hereafter ) . in both ngc 7538 and m17 , we found that double power law fits to the clump mass functions produce exponents which are consistent with the salpeter mass function @xcite . the similarities in their shapes suggests that the clump mass functions in low- and high - mass star - forming regions may have similar origins . what if the mass function of massive clumps were found to resemble that of massive stars , as is the case for low - mass stars ? that would suggest that massive clumps could also be converted to massive stars ( or small multiple systems ) on a more or less one - to - one basis . such a finding would lend weight to theories which suggest that massive stars form by the collapse of individual clumps @xcite , rather than those which suggest that massive stars form by coalescence or competitive accretion @xcite . to address these important matters , it is necessary to be able to accurately fit and interpret clump mass functions . in paper i , we argued that variations in analytic technique might account for much of the observed variation in the power - law exponents derived from fits to submillimeter continuum clump mass functions . in particular , we suggested that the systematic differences between the power - law exponents of clump mass functions in low- and high - mass regions probably result from the fact that the former are typically fit with two power law segments , while the latter are usually fit with only one ( compare @xcite and @xcite with @xcite and @xcite ) . in this paper , we extend and develop these arguments using a more thorough study of the available data on the submillimeter continuum clump mass function . in [ sec : howtofit ] , we discuss how best to fit a clump / core mass function , paying special attention to means by which the fits may be misinterpreted . in [ sec : funcform ] , we apply a consistent fitting methodology to determine which of several plausible functions best represents the clump / core mass function in a variety of star - forming regions . finally , in [ sec : compare ] , we use non - parametric tests to determine whether the shape of the clump / core mass function varies significantly from region to region or changes with the mass scale studied . clump mass functions are commonly expressed in two forms : differential and cumulative . the differential mass function ( dmf ) , @xmath6 , has the advantages of permitting a simple poisson uncertainty analysis but suffers the arbitrariness of binning . binning can be a serious impediment to the accurate interpretation of stellar mass functions @xcite . thus , the differential mass function is not suitable for use when the total number of clumps , @xmath7 , is small . the cumulative mass function ( cmf ) , @xmath8 , requires a more complicated uncertainty analysis , but it suffers none of the problems associated with binning and therefore is suitable even in the small @xmath7 regime . the cmf is therefore preferable for studies of the submillimeter clump / core mass function where the total number of objects is typically @xmath9 . however , the cmf is easily misinterpreted , as we now demonstrate . consider a very simple mass function whose dmf is well - fit by a single power law of the form @xmath10 where @xmath11 is the number of objects in a mass bin of width @xmath12 , and @xmath13 and @xmath14 are constants . it is sometimes assumed that the complementary cumulative mass function is also a power law whose exponent differs from @xmath14 by 1 : @xmath15 for @xmath16 . the expectation is therefore that , if the dmf is well represented by a single power law , the cmf will be too . in measured mass functions , this may not be the case . a measured mass function contains a finite sample of clumps with an upper mass limit , @xmath17 . this upper mass limit can be either a real cutoff on the mass distribution or the result of finite sampling . in either case , the dmf will remain the simple power law of equation [ eq : dmfeq ] , but the cmf will deviate from equation [ eq : pl1 ] . integrating the dmf of equation [ eq : pl1 ] to an upper limit of @xmath17 , we obtain @xmath18 the practical significance of the difference between equations . [ eq : pl1 ] and [ eq : pl2 ] depends on the range of clump masses ( i.e. the value of @xmath19 ) , the value of the power - law exponent , @xmath14 , and the number of clumps in the sample , n@xmath20 . in figures [ fig : fake_dmfs ] and [ fig : fake_cmfs ] , we use simulated data sets to show how the shape of both the dmf and cmf depend on each of these parameters . to simulate a clump mass function , we randomly draw n@xmath20 masses from the interval @xmath21 to @xmath17 following the distribution @xmath22 . we do this for representative values of @xmath7 = 50 , 100 , and 1000 . we simulate power law exponents of @xmath14 = -1.5 , -2.0 , and -2.5 . this range encompasses the exponents of the salpeter imf ( @xmath14 = -2.35 ; @xcite ) , the co clump mass function ( @xmath23 ; @xcite ) , and the values typically found in submillimeter continuum clump surveys . we vary the ratio @xmath24 by holding @xmath21 constant at 1 ( the mass units are arbitrary ) and taking @xmath17 values of @xmath25 , and 10@xmath26 . for each combination of n@xmath20 , @xmath14 , and @xmath27 , we produce a representative data set by averaging 1000 randomly - generated sets . figures [ fig : fake_dmfs ] and [ fig : fake_cmfs ] show the resulting dmfs and cmfs , respectively . note that we follow the popular convention of normalizing @xmath8 by n@xmath28 . clumps drawn randomly from a mass function of the form @xmath29 ( _ thick dashed line _ ) , with n@xmath20 and @xmath14 values as shown . to produce each plotted dmf , 1000 simulated data sets were averaged together . the four simulated dmfs plotted in each panel vary in the range of masses from which the simulated masses were drawn : 1100 ( _ solid line _ ) , 11000 ( _ dotted line _ ) , 110@xmath30 ( _ short dashed line _ ) , and 110@xmath26 ( _ dot - dashed line _ ) . the mass units are arbitrary . the number of bins in each mass function is set to @xmath31 . [ fig : fake_dmfs],width=576 ] . as in fig . [ fig : fake_dmfs ] , the line types indicate the clump mass range from which simulated masses ( arbitrary units ) were drawn : 1100 ( _ solid line _ ) , 11000 ( _ dotted line _ ) , 110@xmath30 ( _ short dashed line _ ) , and 110@xmath26 ( _ dot - dashed line _ ) . the thick dashed line represents the power law distribution from which the clump masses were drawn . for sufficiently steep power laws ( @xmath32 ) , the cmf is likely to be well - fit by a single power law , even when the total number of clumps is small . however , for shallower power laws ( @xmath33 ) and cmfs spanning smaller clump mass ranges , the plotted cmf deviates significantly from the single power law form . [ fig : fake_cmfs],width=576 ] by construction , the dmfs shown in figure [ fig : fake_dmfs ] all follow the input mass distributions . however , as shown in figure [ fig : fake_cmfs ] , the cmfs clearly deviate from the power law form suggested by equation [ eq : pl1 ] . as seen in the right column of figure [ fig : fake_cmfs ] , a steep power law ( @xmath34 ) overwhelms the effects of both @xmath7 and @xmath17 , recovering essentially the pure single power law in all cases . as seen in the middle and left columns of figure [ fig : fake_cmfs ] , however , shallower power laws ( @xmath35 ) often cause the cmf to deviate from the pure power law behaviour expected from equation [ eq : pl1 ] . this is especially the case when the ratio of @xmath19 is small ( solid line in fig . [ fig : fake_cmfs ] ) . it is therefore likely that fitting real data sets with single power laws of the form of equation [ eq : pl1 ] would give highly misleading power law exponents . to achieve accurate results , fitted functions must account for the maximum clump mass , as in equation [ eq : pl2 ] . many of the simulated cmfs shown in figure [ fig : fake_cmfs ] strongly resemble the cmfs derived from observations of real star - forming regions . to account for their curvature , these observational cmfs are typically fitted with two power law segments ( e.g. @xcite ) . that intrinsically single power - law cmfs should so closely resemble those typically fit with two power laws implies that great care should be taken in interpreting the shape of the cmf and in choosing functions with which to fit it . in simple cases , such as the single power - law simulated here , a check of the dmf should help constrain the functional forms which might be fitted to the cmf . even in this simple case , which is simpler than is likely to be encountered in real data , care must be taken when fitting the cmf in order to recover the correct power - law exponent : accounting for @xmath17 is clearly essential . in cmfs measured from real observations , the effects of incompleteness , small - number statistics , and binning can all lead to uncertainty about the functional form of the clump mass function . we believe that particular care is warranted where the observed clump / core mass function encompasses the peak in the stellar imf at @xmath40.1 @xmath5 @xcite . in such cases , incompleteness and a real physical turnover in the mass function are likely to be difficult to distinguish . we now turn to the question of the functional form of the clump / core mass function . the form most commonly employed in studies of low - mass star - forming regions is a double power law above @xmath40.1@xmath5 with a break between the power law segments at @xmath40.5@xmath5 @xcite . this is essentially the same form as found for the galactic field single - star imf @xcite . however , a double power law is not the only plausible shape for the clump mass function . in paper ii , we showed that the m17 clump mass function is better fit by a lognormal distribution than by a double power law . moreover , in [ sec : howtofit ] above , we noted that the finite clump mass range in any real clump mass function adds curvature to the cmf . this curvature can make a mass spectrum which is intrinsically a single power law appear to adopt a shape similar to those usually fit with a double power law . additional uncertainties about the shape of the clump / core mass function arise due to the diversity of techniques used by different authors to fit such mass functions . to eliminate the effects of differing fitting techniques , we have gathered all of the published lists of clump / core masses from unbiased millimeter and submillimeter continuum surveys of star - forming regions and fit them in a homogeneous manner . in table [ tab : studies ] , we summarize these 11 different clump / core mass functions measured in 7 different star - forming regions . in cases where the original authors used more than one set of assumptions about the dust temperature when calculating the clump masses , we use the masses calculated assuming a uniform dust temperature , which is the method most widely used . the only exceptions are the studies by @xcite and @xcite , which divide the clumps into several groups , each with a single dust temperature , before calculating their masses . not all of the studies made comparable attempts to distinguish between starred and starless cores / clumps , so we have used the complete list of all objects in each case ( except our own study of m17 from paper ii , where we have two comparable data sets in which clumps coincident with msx sources were removed in a consistent way ) . the clumps in these 7 star - forming regions span more than 5 orders of magnitude in mass ( 0.0516000 @xmath5 ) and more than two orders of magnitude in linear size . the median clump masses range from 0.17 @xmath5 in @xmath3 oph to 470 @xmath5 in rcw 106 . for comparison with the theory of turbulent fragmentation , we also include in our analysis a list of clumps extracted from a computational hydrodynamic simulation of 4.6 jeans masses of turbulent , self - gravitating gas ( run b5 of @xcite , tp04 hereafter ) . there are two caveats to any comparison made between the tp04 and observational mass functions . first , the dimensionless nature of the tp04 simulations means the absolute mass scale is arbitrary . second , the number of jeans masses simulated by tp04 is considerably less than is found in real star - forming regions , which could have consequences for the number and mass distribution of fragments formed . clccccccc 1 & @xmath3 oph & 1300 & 62 & 0.0085 & 0.17 & 3.2 & 24 & 1 + 2 & @xmath3 oph & 850 & 55 & 0.011 & 0.19 & 6.33 & 29 & 2 + 3 & orion b & 850 & 82 & 0.02 & 0.8 & 9.20 & 106 & 3 + 4 & orion b & 850 & 75 & 0.03 & 0.86 & 30.34 & 120 & 4 + 5 & lagoon / m8 & 450 & 850 & 37 & 0.13 & 6.93 & 34.1 & 287 & 5 + 6 & m17 & 450 & 96 & 0.07 & 13 & 160 & 1700 & 6 + 7 & m17 & 850 & 105 & 0.12 & 11 & 120 & 2100 & 6 + 8 & ngc 7538 & 450 & 77 & 0.11 & 24 & 2700 & 8800 & 7 + 9 & ngc 7538 & 850 & 67 & 0.21 & 34 & 3000 & 9300 & 7 + 10 & w43 & 1300 & 48 & 0.29 & 135 & 3600 & @xmath36 & 8 + 11 & rcw 106 & 1200 & 95 & 0.42 & 470 & @xmath37 & @xmath38 & 9 + 12 & simulation & & 381 & 0.0014 & 0.045 & 25 & 57 & 10 + in the literature , differential clump mass functions are usually fit by either a single power law , @xmath39 or a double power law , @xmath40 where @xmath41 and @xmath2 are , respectively , the power - law exponents below and above the break mass , @xmath42 . the curvature seen in observed clump cmfs makes it is clear that the single power law of equation [ eq : dmf1pl ] can not be applied to the cmf . however , many authors have made simple double power - law fits to the cmf , as per @xmath43 equation [ eq : cmf2pl ] is the first function with which we will fit observed cmfs , and we call it 2pl . we can derive two other potential fitting functions by integrating equations [ eq : dmf1pl ] and [ eq : dmf2pl ] to obtain their corresponding cmfs . as per the discussion of [ sec : howtofit ] , we integrate to a finite upper mass limit , @xmath17 . the single power law cmf with an upper mass limit , which we call 1plmm , is given by : @xmath44 similarly the double power law cmf with an upper mass limit , which we call 2plmm , is given by : @xmath45 inspired by similar work on the stellar mass function , we also consider a lognormal dmf : @xmath46~~ , \label{eq : dmflogn}\ ] ] whose complementary cmf , which we call logn , is given by : @xmath47~~. \label{eq : cmflogn}\ ] ] we constructed a cmf for each region using the published masses . to assign weights to each point for use in the calculation and minimization of @xmath48 , we make the same assumptions as in paper ii , namely that the mass uncertainties , @xmath49 , are distributed roughly as @xmath50 , and that the cmf can be approximately represented by a power law . under these assumptions , the uncertainties in @xmath8 , @xmath51 , are approximately given by @xmath52 ( see paper ii ) . this method of assigning weights avoids the use of individual mass uncertainties , which are frequently unavailable . to assess the uncertainty in each fitted parameter , we must again choose a method which does not rely on the mass uncertainties . for each cmf , we construct and fit 10@xmath26 monte carlo realizations of the data using the bootstrap method ( random sampling with replacement , e.g. @xcite ) . using the resulting distributions of fitted parameters , which are typically symmetric about the best - fit values , we construct 95% ( @xmath53 ) confidence limits on each parameter and report these as our uncertainties . using the distribution of @xmath48 values , we compute the probability , @xmath54 , of obtaining by chance a @xmath48 value poorer than that from the best fit to the original cmf . higher @xmath54 values indicate a better quality of fit . figures [ fig:2plcmfs][fig : logncmfs ] show all eleven measured cmfs with the best fit versions of these functions : 2pl , 1plmm , 2plmm , and logn . the cmf of the simulated clumps from tp04 is included for comparison . the @xmath54 values for all of the fits are summarized in table [ tab : fitpvals ] . . each cmf is fit by a pure double power law ( 2pl ) , as described by equation [ eq : cmf2pl ] . the panel labels indicate the data set i d from table [ tab : studies ] , goodness - of - fit probability , @xmath54 , and best - fit parameters , @xmath55 , @xmath2 , and @xmath42 . for visual clarity , the maximum clump mass in each plot has been normalized to unity ; the best fit parameters given are for fits to the unnormalized data.[fig:2plcmfs ] ] . each cmf is fit by a single power law ( 1plmm ) described by equation [ eq : cmf1plmm ] . the panel labels indicate the data set i d from table [ tab : studies ] , goodness - of - fit probability , @xmath54 , and best - fit parameters , @xmath14 and @xmath17 . for visual clarity , the maximum clump mass in each plot has been normalized to unity ; the best fit parameters given are for fits to the unnormalized data.[fig:1plcmfs ] ] . each cmf is fit by the double power law ( 2plmm ) described by equation [ eq : cmf2plmm ] . the panel labels indicate the data set i d from table [ tab : studies ] , goodness - of - fit probability , @xmath54 , and best - fit parameters , @xmath14 and @xmath17 . for visual clarity , the maximum clump mass in each plot has been normalized to unity ; the best fit parameters given are for fits to the unnormalized data.[fig:2plmmcmfs ] ] . each cmf is fit by a lognormal clump mass distribution ( logn ) , as described by equation [ eq : cmflogn ] . the panel labels indicate the data set i d from table [ tab : studies ] , goodness - of - fit probability , @xmath54 , and best - fit parameters , @xmath56 and @xmath57 . for visual clarity , the maximum clump mass in each plot has been normalized to unity ; the best fit parameters given are for fits to the unnormalized data.[fig : logncmfs ] ] lccccc @xmath3 oph ( 1.3 mm ) & 1 & 0.84 & 0.95 & * 0.97 * & 0.87 + @xmath3 oph ( 850 @xmath58 m ) & 2 & 0.84 & 0.52 & 0.92 & * 0.95 * + orion b ( 850 @xmath58 m ) & 3 & 0.79 & 0.45 & 0.77 & * 0.96 * + orion b ( 850 @xmath58 m ) & 4 & 0.83 & 0.22 & 0.68 & * 0.86 * + lagoon & 5 & * 0.76 * & 0.21 & 0.69 & 0.74 + m17 ( 450 @xmath58 m ) & 6 & 0.92 & 0.45 & * 0.96 * & 0.83 + m17 ( 850 @xmath58 m ) & 7 & 0.46 & 0.49 & 0.69 & * 0.78 * + ngc 7538 ( 450 @xmath58 m ) & 8 & 0.80 & 0.56 & * 0.86 * & 0.68 + ngc 7538 ( 850 @xmath58 m ) & 9 & * 0.84 * & 0.59 & 0.81 & 0.70 + w43 & 10 & 0.98 & 0.60 & * 0.99 * & 0.59 + rcw 106 & 11 & 0.98 & 0.52 & * 0.99 * & 0.62 + run b5 of tp04 & 12 & 0.90 & 0.47 & * 0.98 * & 0.75 + figures [ fig:2plcmfs][fig : logncmfs ] and the p values shown in table [ tab : fitpvals ] indicate that none of the observational cmfs is best fit by a single power law , 7 are best fit by some form of the double power law ( either 2pl or 2plmm ) , and 4 are best fit by a lognormal clump mass distribution ( logn ) . before discussing each case in turn , we encourage the reader to keep in mind that small differences in the @xmath54 values between two almost equally good fits are probably not meaningful . in calculating the @xmath54 values , we had to use the bootstrap approximation in place of the frequently - unavailable clump mass uncertainties , and this makes the @xmath54 values themselves uncertain . @xmath59 the @xcite 1.3 mm core cmf ( data set 1 ) is best fit by the 2plmm distribution , although the 1plmm fit is only marginally poorer . similarly , the @xcite 850 @xmath58 m core cmf ( set 2 ) is best fit by the logn distribution , but the 2plmm fit is only marginally poorer . given that the two data sets represent the same clump sample , we expect that the same function should describe them . either the 2plmm or logn functions could serve this purpose , but the 2plmm function gives a slightly higher mean @xmath54 value . interestingly , the logn fit , which has only 2 parameters , nearly matches the quality of the 2plmm function , which has 5 . * orion b * in this region the lognormal mass function provides the best fit to both measurements of the cmf . the 2pl function provides an almost equally good fit to data set 4 , but not to data set 3 . visual inspection of figures [ fig:2plcmfs ] and [ fig : logncmfs ] affirms the superiority of the logn fit . if we exclude the massive outlier clump at the high - mass end of set 4 , the @xmath54 value for the logn fit rises to 0.99 and that for the 2pl fit falls to 0.78 . we suspect that the massive outlier is probably a composite clump which would resolve into multiple less - massive objects if observed at higher angular resolution . * lagoon * although @xcite argued for a single power law fit to this mass function ( set 5 ) , the @xmath54 values indicate that the 2pl function provides a significantly better fit ( @xmath60 versus @xmath61 ) . the lognormal fit , with @xmath62 , is only marginally poorer than the 2pl fit . note that the study from which these data were obtained is unique in having used a by - eye clump extraction method . * m17 * as in paper ii , we find that the 450 @xmath58 m cmf ( set 6 ) is better fit by a double power law ( 2plmm , in this case ) , while the 850 @xmath58 m cmf ( set 7 ) is better fit by a lognormal distribution . both the logn and 2plmm functions might be said to provide credible fits to both data sets , though again the logn function does this using three fewer parameters than does the 2plmm function . * ngc 7538 * in this case , the @xmath54 values show that some form of double power law ( 2plmm for set 8 and 2pl for set 9 ) provides the best fit to both measurements of the cmf . the three most massive clumps in each cmf correspond to the irs 1 - 3 , irs 9 , and irs 11 regions . each of these regions is already known to be forming more than one star and therefore likely contains significant unresolved structure . for example , the irs 13 region contains 3 compact hii regions @xcite . if we omit these outliers from the analysis , the @xmath54 values for the lognormal fits become 0.97 and 0.69 for data sets 8 and 9 , respectively . the corresponding @xmath54 values for the 2plmm fits become 0.99 and 0.96 . therefore , we conclude that , although the logn function remains a contender , the ngc 7538 cmf is probably best fit by a double power law . * w43 and rcw 106 * these two massive star - forming regions provide the clearest cases of cmfs ( sets 10 & 11 ) which are best fit by double power laws . in both cases , the 2plmm fits are favored , with @xmath63 , while the 2pl fits both give @xmath64 . no other function provides a comparably good fit . * run b5 of tp04 * here again , the 2plmm function provides the best fit , with @xmath65 . visual inspection of the plot suggests that the logn function fits equally well , although the @xmath54 values suggest otherwise . to summarize , the general trend in the fits is toward lognormal mass functions among the low - mass star - forming regions and some form of the double power law mass function among the massive star - forming regions . the two major sources of uncertainty in this analysis are the choice of weights and the use of the bootstrap technique . both would be mitigated if the original clump mass uncertainties were available . if we use uniform weights in making the fits , instead of the default weights of @xmath66 , the logn function then provides the best fit in 5 cases , the 2plmm in 4 , and they tie in 3 . with regard to the use of the bootstrap technique , we note that although using the real clump mass uncertainties would be preferable , the @xmath54 values obtained herein using the bootstrap technique show the same trend as those obtained in paper ii using the proper uncertainties . we strongly encourage the reporting of both random and systematic clump mass uncertainties in future studies . a potential fault with the 2plmm fits can be seen by comparing the best fit parameters from the 2pl fits ( fig . [ fig:2plcmfs ] ) with those from the 2plmm fits ( fig . [ fig:2plmmcmfs ] ) . the 2pl @xmath67 values range around the salpeter value of @xmath68 but the mean 2plmm value is @xmath69 . moreover , the mean ratio of the break masses derived from the 2plmm and 2pl fits is 1.6 . the 2plmm fits are achieving the excellent @xmath54 values they do by shifting the break mass up and fitting the top end of the mass function with a very steep power law one which is not obviously related to the power laws which fit the co clump or stellar mass functions , both of which are significantly shallower . it is also important to note that the 2plmm fits achieve their excellent @xmath54 values using 5 adjustable parameters , compared to 4 for the 2pl fits and only 2 for the logn fits . for the remainder of the discussion in this paper , we will refer primarily to the power law exponents derived from the 2pl fits , which are most suitable for direct comparison to those obtained by other authors . previous studies of the core mass function in regions of low - mass star formation have emphasized the agreement between the shapes of the core and stellar mass functions ( e.g. @xcite ) , when both are fit with similar numbers of power laws over similar mass ranges . the core mass functions in @xmath3 oph , orion b , and serpens are all well fit by a salpeter - like power law above about 1 @xmath5 ( @xmath70 ) , flatten out somewhat below about 1 @xmath5 ( @xmath71 ) , and peak at about 0.1 @xmath5 . in table [ tab : fitparams ] , we summarize the best - fit parameters for the 2pl and logn fits to the 12 cmfs studied herein . fitting all of the cmfs in the same way , we find @xmath2 values ranging between -1.8 and -3.1 , comparable to the range of power law exponents found in fits to the @xmath72 stellar imf @xcite . the mean of the @xmath2 values is [email protected] , or [email protected] if we include only the 7 observational cmfs which are best fit by double power laws . both measures of the mean @xmath2 are compatible with the salpeter imf , despite the differences in the clump mass ranges over which the various @xmath2 values were derived . similarly , if the low - mass ends of all 11 observational cmfs are fit with power laws , the resulting exponents , @xmath41 , range between -1.2 and -1.7 , with a mean of [email protected] . this result also accords with the shallowing of the stellar imf at its low - mass end , despite the differences in the clump mass scales over which this shallowing occurs . perhaps most interesting is that , as shown in figure [ fig : alphaplot ] , we find no trend in the @xmath2 values with the median clump mass in each region . this result applies over nearly four orders of magnitude in median clump mass . the power law indices derived from fits to the mass functions of submillimeter continuum clumps / cores offer no clear way to distinguish between low - mass regions , such as @xmath3 oph , and very high - mass regions , such as rcw 106 . the median clump mass in rcw 106 is nearly 500 @xmath5 , which suggests that most of the clumps are probably cluster - forming objects , yet the rcw 106 clump mass function and the @xmath3 oph core mass function may be fit with compatible power laws . lccccccccc @xmath3 oph ( 1.3 mm ) & 1 & [email protected] & [email protected] & 0.03 @xmath73 0.01 & 0.43 @xmath73 0.09 & & 0.2 @xmath73 0.2 & [email protected] & [email protected] + @xmath3 oph ( 850 @xmath58 m ) & 2 & [email protected] & [email protected] & 0.016 @xmath73 0.008 & 0.6 @xmath73 0.2 & & 0.2 @xmath73 0.1 & [email protected] & [email protected] + orion b ( 850 @xmath58 m ) & 3 & [email protected] & [email protected] & 0.31 @xmath73 0.05 & 1.3 @xmath73 0.2 & & 1 @xmath73 1 & [email protected] & [email protected] + orion b ( 850 @xmath58 m ) & 4 & [email protected] & [email protected] & 0.39 @xmath73 0.07 & 1.4 @xmath73 0.2 & & 1.0 @xmath73 0.6 & [email protected] & [email protected] + lagoon & 5 & [email protected] & [email protected] & 3.7 @xmath73 0.8 & 8 @xmath73 1 & & 6 @xmath73 2 & [email protected] & [email protected] + m17 ( 450 @xmath58 m ) & 6 & [email protected] & [email protected] & 6 @xmath73 1 & 18 @xmath73 2 & & 18 @xmath73 8 & [email protected] & [email protected] + m17 ( 850 @xmath58 m ) & 7 & [email protected] & [email protected] & 4.1 @xmath73 0.7 & 21 @xmath73 3 & & 17 @xmath73 10 & [email protected] & [email protected] + ngc 7538 ( 450 @xmath58 m ) & 8 & [email protected] & [email protected] & 6 @xmath73 3 & 90 @xmath73 30 & & 30 @xmath73 30 & [email protected] & [email protected] + ngc 7538 ( 850 @xmath58 m ) & 9 & [email protected] & [email protected] & 1.0 @xmath73 0.7 & 90 @xmath73 40 & & 30 @xmath73 30 & [email protected] & [email protected] + w43 & 10 & [email protected] & [email protected] & 20 @xmath73 10 & 300 @xmath73 90 & & 110 @xmath73 80 & [email protected] & [email protected] + rcw 106 & 11 & [email protected] & [email protected] & 80 @xmath73 30 & 1000 @xmath73 200 & & 400 @xmath73 300 & [email protected] & [email protected] + run b5 of tp04 & 12 & [email protected] & [email protected] & 0.013 @xmath73 0.002 & 0.085 @xmath73 0.007 & & 0.06 @xmath73 0.02 & [email protected] & [email protected] + mean values & & & * [email protected] * & & & & & * [email protected] * & * [email protected] * + , versus median clump mass for the 11 millimeter / submillimeter clump mass functions described in table [ tab : studies ] . the @xmath2 values are derived from the pure double power law fits ( 2pl , eq . [ eq : cmf2pl ] ) . the horizontal dashed line indicates the salpeter exponent , @xmath74 @xcite . the error bars represent the 95% ( @xmath42@xmath75 ) confidence limits from table [ tab : fitparams ] . no trend in @xmath2 with mass is apparent.[fig : alphaplot ] ] we emphasize four key results from this fitting exercise . first , there is the tendency toward lognormal cmfs in less massive star - forming regions and power - law cmfs in massive star - forming regions . second , there is the agreement among the @xmath2 and @xmath41 values from region to region , despite the significant differences in the relevant clump mass scales . third , there is the presence of a break between the steep and shallow sections of the mass function in high - mass star - forming regions . if this break is real , it implies that the clump mass function is _ not _ a single unbroken power law above 1 @xmath5 . this result holds true despite the uncertainties in the absolute mass calibration in each study ( see [ sec : massunc ] ) finally , there is the agreement between the power law exponents of the clump and stellar mass functions , again despite the differences in mass scales . there is already ample theoretical evidence that lognormal mass functions can result when any sufficiently stochastic process , such as turbulent fragmentation , governs the evolution of the clump masses . moreover , lognormal mass functions can be viewed as one extremum of a continuum of shapes whose other extremum is the power - law mass function . numerical simulations have repeatedly shown that turbulent fragmentation can naturally produce both power - law and lognormal mass functions whose shapes are largely independent of the intrinsic mass scale @xcite . several authors have invoked the central limit theorem to demonstrate analytically that lognormal mass functions arise whenever a large number of random parameters act to set the clump masses @xcite . conversely , power - law mass functions can arise when fewer random parameters affect each clump s mass . these random parameters could derive from successive turbulent fragmentations or n - body gravitational encounters among clumps . we might therefore imagine the following scenario to explain the trend toward lognormal mass functions in low - mass regions and power - law mass functions in high - mass regions ( see table [ tab : fitpvals ] ) . a molecular cloud undergoes turbulent fragmentation , producing an initially power - law clump mass function ; as the clumps continue to fragment and interact , their mass function approaches a lognormal distribution ; by the time the `` clumps '' have become `` cores '' , they have achieved a fully lognormal mass distribution . as the objects furthest down the chains of turbulent fragmentation and n - body interactions , the masses of the cores would have been maximally influenced by random processes . in this scenario , spatially unresolved observations of a set of cores whose mass distribution is lognormal could effectively restore the less - evolved double power law mass function by blurring together smaller cores into semblances of their antecedent clumps . this hypothesis would explain the trend toward double power law mass functions seen in the more distant regions , such as ngc 7538 , w43 , and rcw 106 , where the spatial resolution of the observations is nt sufficient to adequately distinguish all of the cores from one another . if the core / clump mass function in both low- and high - mass star - forming regions were found to be consistent with the stellar imf ( uncorrected for unresolved multiples ) , it would relax the need to invoke processes such as coalescence and competitive accretion @xcite to explain the formation of more massive stars . such consistency would also argue in favor of theories which predict that massive stars can form by a scaled - up version of molecular cloud core collapse with disk accretion @xcite . as shown in table [ tab : fitparams ] , the mean power law exponents from the double power law fits to the 11 observational cmfs are @xmath76 and @xmath77 . @xcite cites power law exponents for the mean galactic field single - star imf of @xmath78 for @xmath79 and @xmath80 for @xmath81 . the agreement between the @xmath2 values of the clump and stellar mass functions is intriguing . unlike the @xmath41 values , which may be strongly affected by incompleteness in the clump mass functions ( see next section ) , the @xmath2 values should be quite reliable . thus , the agreement between the stellar and clump @xmath67 values reinforces the ideas that the stellar imf originates in the mass function of clumps in a fragmented molecular cloud and that the stellar mass function can be obtained by an essentially one - to - one conversion of clumps into stars ( or binaries ) . if the high - mass end of the stellar imf can be obtained by converting clumps into stars on a more - or - less one - to - one basis , there may be no need to invoke processes such as coalescence or competitive accretion . @xcite has shown that the variation in the measured power law exponents for the stellar imf is compatible with normal measurement variation and need not represent physical region - to - region variation . we suggest that the same description could apply equally well to the range of power law exponents derived from clump mass functions in different regions ( see fig . [ fig : alphaplot ] ) . more measurements of @xmath2 would be required to confirm this hypothesis in the cases of both the stellar imf and the clump mass function . @xcite has shown that real region - to - region variations in the imf are physically plausible , but may be masked by small - number statistics , which is certainly a concern with the clump / core mass functions taken to date . can we identify a mass scale at which the apparent agreement between the clump and stellar mass functions breaks down ? of the parameters which control the shapes of the fitted functions , namely @xmath57 for the lognormal fits and the @xmath14 values for the power law fits , none shows a trend with the median clump mass of the regions ( see table [ tab : fitparams ] ) . thus , the fits suggest that the shape of the clump mass function is not very sensitive to the clump mass range being fit . there is also no clear break between the cmfs which do mirror the shape of the stellar imf and those which do not , despite the breadth in the clump mass ranges . again , this surprising result suggests that the intrinsic shape of the stellar imf may be apparent in the clump mass function on mass and linear scales substantially larger than those previously considered . both interferometry and the chopping technique used by single - dish submillimeter telescopes filter out emission on some spatial scales . clump - finding algorithms may have a similar effect . therefore , all of the observations used to generate the cmfs discussed in this paper are subject to some form of spatial filtering , and hence to incompleteness and uncertainty . we are therefore motivated to ask whether spatial filtering and related effects introduced during data acquisition and reduction can introduce a sufficient number of random perturbations to the clump masses to ensure that a lognormal mass function is obtained . spatial filtering of flux can systematically alter the derived clump masses . the magnitude of the discrepancy would depend on the size ( and therefore the mass ) of each clump . although this possibility should be explored further , several arguments suggest it does not substantially alter the shape of the clump mass function . first , the mass function of clumps extracted from infrared dust extinction maps , which are essentially free of spatial filtering , is very similar to those derived from submillimeter continuum observations ( j. alves , private communication ) . if the clump extraction algorithms exerted a strong spatial filtering effect , we would need to understand how at least three different methods used by the authors of the papers studied here all produce similar mass functions . moreover , the effects of spatial filtering would affect observations of both low- and high - mass star - forming regions . if it were a significant effect , it seems unlikely that observations of low - mass star - forming regions would still manage to produce a core mass function which appears to agree with the stellar imf . finally , we note that the clump cmf extracted from the turbulent fragmentation simulations of tp04 is not subject to observational spatial filtering and its shape is entirely compatible with those of the observational cmfs ( tp04 used a custom clump extraction technique ) . for these reasons , it is difficult to explain how the shape of the clump mass function , and its similarity from region to region might be accounted for by spatial filtering . whether or not the clump mass functions are strongly affected by spatial filtering , they are certainly all affected by incompleteness . for spatially extended objects , the detection threshold is a limiting surface brightness , meaning that a clump of any mass could go undetected if its flux were spread thinly across the sky . thus , the clump sample may be incomplete in any mass range . at the moment , there is no reliable way to correct millimeter and submillimeter continuum clump / core mass functions for incompleteness . if incompleteness is a significant determinant of the shape of each cmf , its effects are likely to be felt most strongly at their low - mass ends , where it may be partly responsible for the shallower slope of the mass function . consequently , the degree of incompleteness might influence the position of the break , @xmath42 , between the two power laws when a double power law is fitted to a given mass function . in most studies to date , the distribution of clumps in mass - radius space shows a ( sometimes slight ) separation from the detection threshold at its high - mass end @xcite . hence , we do not think it likely that the fitted @xmath2 values are strongly affected by sensitivity - induced incompleteness . the @xmath55 values , however , are likely to be significantly impacted by incompleteness . there are several ways to test for the related effects of incompleteness and spatial filtering . high spatial resolution observations of the more distant star - forming regions in the sample will reveal the extent to which currently unresolvable structure will change the shape of the clump / core mass function . observations of more distant regions should also mitigate the effects of chopping by limiting the loss of emission to spatial scales much larger than those of typical molecular cloud cores . comparison of submillimeter continuum and dust extinction maps of identical regions should provide strong constraints on the effect of spatial filtering in the former type of observations . finally , observations with total power detectors , such as scuba-2 , will eliminate the effects of chop - based spatial filtering . scuba-2 s greater sensitivity will also help reveal those clumps not detectable by the present generation of submillimeter and millimeter detectors . it should be noted that , while there are significant uncertainties in the absolute calibration of clump masses , these uncertainties are not large enough to fundamentally change the conclusions of the preceding analysis . the calibration of the absolute mass scale in each study we reference relies on assumptions about the dust temperature and emissivity in each region . typically , as in papers i and ii , values typical of each region are applied to all of the clumps it contains . the dust emissivity in star - forming regions may be uncertain by up to a factor of 2 ( i.e. it may range between @xmath412 ) and the temperature may be uncertain by perhaps a factor of four ( i.e. the mean clump temperature is likely to lie within 1040 k ) . in the most extreme case , whereby the mean temperature and dust emissivity in a region are assumed to be 10 k and 1 , respectively , but are actually 40 k and 2 , the clump masses would all be overestimated by a factor of @xmath430 . however , referring to table [ tab : fitparams ] , we see that even this large uncertainty is insufficient to account for the observed variation in the peak mass , @xmath82 from the logn fits or the break mass , @xmath42 , from the 2pl fits . both of these parameters range over @xmath83 orders of magnitude , which is a much greater variation than can be accounted for by the largest plausible uncertainties in the dust emissivity and temperature . in addition , the mass calibrations are unlikely to be incorrect by the same amount or in the same direction in all of the studies . finally , we emphasize that uncertainties in the absolute mass calibration do not affect the _ shapes _ of the various clump mass functions . the clump mass function is one of the most important descriptors of the initial conditions for star formation . it tells us how the masses of pre - stellar condensations are distributed , leaving us to determine how such a distribution could give rise to the observed distribution of stellar masses . thus , it is important to understand how the clump mass function varies from region to region , between low- and high - mass star - forming regions , and whether it varies more or less than the stellar imf . this information will help evaluate theories which posit different formation mechanisms for stars of different masses . for example , if low- and high - mass star forming regions had systematically different clump mass functions but compatible stellar mass functions , it would be evidence that the mechanisms of low- and high - mass stars were different . in figure [ fig : allcmfs ] , we plot all 11 of the observational cmfs from table [ tab : studies ] in the same panel . to facilitate comparisions among the mass functions , they have all been renormalized such that they share a common median clump mass . this renormalization preserves the shape of each mass function . like figure [ fig : alphaplot ] , figure [ fig : allcmfs ] shows that there is no clear trend in the shape of the mass function with the median mass in each region . particularly at their high - mass ends , where incompleteness is least significant , there is no clear segregation between the cmfs of low- and high - mass star - forming regions . . the masses of the clumps in each cmf have been rescaled such that all clump sets have a common median mass . cmfs of low - mass star - forming regions ( @xmath3 oph and orion b ) are plotted with solid lines , that of m8 with a thick solid line , and those of the massive star - forming regions ( m17 , ngc 7538 , w43 , and rcw 106 ) with dotted lines . the new mass scale is arbitrary . [ fig : allcmfs],width=576 ] in the previous section , we described how parameters derived from fits to the clump mass function vary from one star - forming region to another . however , to more precisely quantify the extent to which the shape of the clump mass function varies from region to region , it is best to use non - parametric techniques . such techniques do not rely on prior knowledge of the functional form of the clump mass function . one such test is the kolmogorov - smirnov ( k - s ) test , which measures the probability , @xmath54 , that two sets of clump masses represent random samplings from the same parent distribution . we compared each of the mass functions listed in table [ tab : studies ] to the others using the k - s test . to establish a baseline for interpreting the k - s test @xmath54 values , we compare similar sets of mass functions . in figure [ fig : same_mfs ] , we plot pairs of independent measurements of the cmf in each of the four regions where multiple measurements are available ( @xmath3 oph , orion b , ngc 7538 , and m17 ) . the @xmath54 values for these comparisons range from 0.72 for @xmath3 oph to 0.96 for orion b. the lowest @xmath54 value results from comparing studies @xcite of @xmath3 oph which differ in their authors , continuum wavebands , clump extraction techniques , and assumptions about the clump temperatures . the two intermediate @xmath54 values come from our own studies of ngc 7538 ( paper i ) and m17 ( paper ii ) . each pair differs only by the continuum waveband of the observations . the highest @xmath54 value derives from the comparison of the two orion b studies @xcite , and they differ by authors , clump extraction techniques , and assumptions about temperatures , but not in continuum waveband . in principle , each pair of mass functions should be the nearly identical , as they represent different measures of essentially the same thing . thus , the magnitude of the difference between the two should primarily reflect differences introduced during data reduction and analysis . hence , we set our threshold for statistically significant similarity between two cmfs at @xmath84 , the lowest measured probability for comparisons of similar data sets . note that these results agree well with those from the previous section : we find the highest k - s @xmath54 value when comparing the two orion b studies , which were both found to be well - fit by lognormal distributions in [ sec : funcform ] , and the lowest k - s @xmath54 value when comparing the @xmath3 oph mass functions , which were found to be best - fit by different functions . value for the comparison of the two cmfs . [ fig : same_mfs],width=480 ] table [ tab : allps ] shows the @xmath54 values for pair - wise k - s test comparisons of all of the mass functions described in table [ tab : studies ] . clearly , a direct k - s comparison of regions such as @xmath3 oph and rcw 106 would yield a @xmath54 value of 0 because their clump mass ranges do not overlap . to enable a comparison of the _ shapes _ of any two mass functions regardless of their intrinsic mass scales , we renormalize each pair of mass functions so they share a common median clump mass , as we did for all of the cmfs in figure [ fig : allcmfs ] . in figure [ fig : ksmulti ] we plot the pair - wise comparisons of a representative set of 6 clump mass functions along with the @xmath54 value for each comparison . the @xmath54 values for the 15 unique comparisons range from 0.73 to 1.00 , implying that , in every case , when the mass functions are set on a common mass scale , there is a significant likelihood that they represent random samples from the same clump mass distribution . examining the @xmath54 values for all 55 unique inter - comparisons between the 11 observational cmfs ( table [ tab : allps ] ) , we find that only 11 fall below the threshold of statistical significance ( @xmath85 ) , though none is lower than @xmath86 . of these 11 , 4 are associated with the @xcite study of orion b , which differs from the rest in having large numbers of duplicate clump masses . long runs of identical masses open gaps between the @xcite cmf and the others . the k - s test interprets these gaps as statistically significant discrepancies when , in fact , they may only reflect the number of significant figures in the reported clump masses . this effect may account for the lower average @xmath54 values for comparisons involving the @xcite orion b cmf , compared to those involving the very similar study of orion b by @xcite , whose @xmath54 values indicate statistically significant similarity in every case . lcccccccccccc @xmath3 oph ( 1.3 mm ) & 1 & 0.72 & 0.87 & 0.95 & 0.82 & * 0.63 * & 0.80 & 0.83 & * 0.60 * & 0.73 & 0.81 & * 0.68 * + @xmath3 oph ( 850 @xmath58 m ) & 2 & & 0.77 & 0.95 & * 0.65 * & 0.81 & 1.00 & 0.98 & 0.77 & 0.90 & 0.96 & 0.96 + orion b ( 850 @xmath58 m ) & 3 & & & 0.96 & * 0.67 * & * 0.68 * & * 0.64 * & 0.83 & * 0.60 * & 0.76 & 0.77 & * 0.69 * + orion b ( 850 @xmath58 m ) & 4 & & & & 0.92 & 0.91 & 0.94 & 0.87 & 0.79 & 1.00 & 0.99 & 0.98 + lagoon & 5 & & & & & 0.90 & * 0.70 * & * 0.56 * & 0.72 & 0.91 & * 0.64 * & 0.70 + m17 ( 450 @xmath58 m ) & 6 & & & & & & 0.77 & 0.88 & 0.82 & 0.98 & 0.99 & 0.81 + m17 ( 850 @xmath58 m ) & 7 & & & & & & & 0.91 & 0.78 & 0.96 & 0.99 & 0.99 + ngc 7538 ( 450 @xmath58 m ) & 8 & & & & & & & & 0.84 & 0.97 & 0.99 & 0.95 + ngc 7538 ( 850 @xmath58 m ) & 9 & & & & & & & & & * 0.71 * & 0.87 & 0.81 + w43 & 10 & & & & & & & & & & 0.98 & 0.97 + rcw 106 & 11 & & & & & & & & & & & 0.98 + 16000 m@xmath87 ) . in the lower left corner of each panel , the regions being compared are identified by their i d numbers , @xmath88@xmath89 , from table [ tab : studies ] . the most massive clump in cmf @xmath88 ( _ thick line _ ) is normalized to mass unity and cmf @xmath89 ( _ thin line _ ) is normalized accordingly ( see text ) . the boldface number in each panel is the kolmogorov - smirnov @xmath54 value for the comparison ( see text for interpretation ) . for each region , the wavelength at which the continuum observations were made and the total mass of clumps detected are given in the legend . [ fig : ksmulti ] ] thus , without reference to the functional form of the clump / core mass function , we find that , when set on a common mass scale , all of the cmfs have a statistically significant likelihood of representing random samples drawn from the same parent distribution . in other words , the cmfs all appear to have the same ( or very similar ) _ shapes _ , despite the differences in their intrinsic mass _ scales_. because the k - s test is least sensitive to the high- and low - mass tails of the cmfs , it is less affected by the few most massive clumps in each region which , as we saw in [ sec : funcform ] , can bias the determination of the functional form of the cmf . the combined results of the parametric fits from [ sec : funcform ] with the non - parametric k - s test comparisons provide compelling evidence that the shapes of the clump and core mass distributions in all of the regions considered are strongly linked to each other and perhaps also to the stellar imf . in this paper , we have examined the issue of the functional form of the clump / core mass function derived from millimeter / submillimeter continuum observations of galactic star - forming regions . we have demonstrated that , even in the case of a strict single power - law mass function , great care must be taken in fitting the cumulative form of the mass function to accurately measure the exponent of this power law . unexpected curvature can be introduced into the cmf when the range of masses or the number of objects is small , particularly when the power law is shallow ( @xmath22 with @xmath90 ) . in some cases , an intrinsically single power - law mass function might appear better fit by two or more power laws . in practice , the mass function is unlikely to be a pure single power law , so the potential to be mislead when fitting the cumulative mass function is even greater . we have fit 11 independent measurements of the clump / core mass function extracted from millimeter and submillimeter continuum observations of 7 star - forming regions . the masses of the clumps / cores in these regions collectively span more than five orders of magnitude . we find that , in regions where the median clump / core mass is less than a few @xmath5 , the clump / core mass function is best fit by a lognormal distribution . in more massive regions , a double power law provides a better fit to the clump cmf . at intermediate masses , both the lognormal and the double power law adequately describe the cmf . this difference in the functional form of the clump mass function may represent an evolution toward a lognormal mass function at lower masses and smaller spatial scales . this interpretation is consistent with theoretical explanations for the origin of a lognormal mass function via stochastic processes such as turbulent fragmentation and gravitational n - body interactions . the shape of the clump / core mass function does not appear to be a strong function of the intrinsic mass scale . the transition from lognormal to double power - law mass functions is fairly subtle . the mean value of the power - law exponent of the high - mass end of the clump / core mass function is found to be @xmath76 , consistent with the salpeter mass function . this result suggests that the shape of the stellar imf may be apparent in the clump mass function on clump mass scales of several tens of @xmath5 or more and linear scales of several tenths of a parsec . it also suggests that molecular cloud cores may be converted into stars ( or small multiple systems ) on an essentially one - to - one basis . this scenario is consistent with theories in which massive stars form by core collapse and disk accretion . it also relaxes the need to invoke either coalescence or competitive accretion to explain the formation of massive stars , though these processes probably do occur at some level . this result deserves further investigation , especially to constrain any instrumental or post - processing effects which may affect it . the similarities among the shapes of the various clump / core mass functions are confirmed to be statistically significant using non - parametric pair - wise comparisons of the cmfs . in 44 of 55 comparisons , after normalizing away the differences in their median masses , the clump / core mass functions are found to have a high likelihood of representing random samplings from the same parent distribution . that the observational cmfs agree well with the cmf of clumps extracted from a simulation of self - gravitating turbulent gas suggests that turbulent fragmentation is one possible sufficient explanation for the shape of the clump / core mass function . m. a. r. has been supported by an ontario graduate scholarship in science and technology . both m. a. r. and c. d. w. are supported by the natural sciences and engineering research council of canada ( nserc ) . m. a. r. would like to thank e. feigelson and f. motte for helpful discussions during the preparation of this manuscript . adams , f. c. , & fatuzzo , m. 1996 , , 464 , 256 bonnell , i. a. , bate , m. r. , clarke , c. j. , & pringle , j. e. 1997 , , 285 , 201 bonnell , i. a. , clarke , c. j. , bate , m. r. , & pringle , j. e. 2001 , , 324 , 573 bonnell , i. a. , larson , r. b. , & zinnecker , h. 2006 , preprint(astro - ph/0603447 ) bonnell , i. a. , vine , s. g. , & bate , m. r. 2004 , , 349 , 735 chabrier , g. 2003 , , 115 , 763 clark , p. c. & bonnell , i. a. preprint ( astro - ph/0603578 ) elmegreen , b. g. 2004 , , 354 , 367 gammie , c. f. , lin , y. , stone , j. m. , & ostriker , e. c. 2003 , , 592 , 203 israel , f. p. 1977 , , 59 , 27 johnstone , d. j. , wilson , c. d. , moriarty - schieven , g. , joncas , g. , smith , g. , gregersen , e. , & fich , m. 2000 , , 545 , 327 johnstone , d. j. , fich , m. , mitchell , g. f. , moriarty - schieven , g. 2001 , , 559 , 307 klessen , r. s. , burkert , a. , & bate . m. r. 1998 , , 501 , l205 klessen , r. s. 2001 , , 556 , 837 klessen , r. , & burkert , a. 2000 , , 128 , 287 kramer , c. , stutzki , j. , rhrig , r. , & corneliussen , u. 1998 , , 329 , 249 kroupa , p. 2002 , science , 295 , 82 krumholz , m. r. , mckee , c. f. , & klein , r. i. 2005 , , 618 , 33 krumholz , m. r. 2006 , , 641 , l45 larson , r .b . 1973 , , 161 , 133 maz - apellniz , j. & beda l. 2005 , preprint ( astro - ph/0505012 ) mckee , c. f. , & tan , j. c. 2002 , , 416 , 6876 mckee , c. f. , & tan , j. c. 2003 , , 585 , 2 mookerjea , b. , kramer , c. , nielbock , m. , & nyman , l .- . 2004 , , 426 , 119 motte , f. , andr , p. , ward - thompson , d. , bontemps , s. 2001 , , 372 , l41 motte , f. , andr , p. , & neri , r. 1998 , , 336 , 150 motte , f. , schilke , p. , & lis , d.c . 2003 , , 582 , 277 padoan , p. & nordlund , . 2002 , , 576 , 870 press , w. h. , teukolsky , s. a. , vetterling , w. t. , & flannery , b. p. 1992 , numerical recipes in fortran , ( 2nd ed . , cambridge : university press ) reid , m. a. , & wilson , c. d. 2005 , , 625 , 891 reid , m. a. , & wilson , c. d. 2005 , , submitted salpeter , e. e. 1955 , , 121 , 161 scalo , j. 1998 , in asp conf . 142 , the stellar initial mass function proceedings , ed . g. gilmore & d. howell ( san francisco : asp ) , 201 testi , l. & sargent , a. i. 1998 , , 508 , l91 tilley , d. a. & pudritz , r. e. 2004 , , 353 , 769 tothill , n. f. h. , white , g. j. , matthews , h. e. , mccutcheon , w. h. , mccaughrean , m. j. , kenworthy , m. a. 2002 , , 580 , 285 williams , j. p. , de geus , e. j. , & blitz , l. 1994 , , 428 , 693 wynn - williams , c. g. , becklin , e. e. , & neugebauer , g. 1974 , , 187 , 473 zinnecker , h. 1984 , , 210 , 43

we present results from a deep 2 ( comoving scale of 3.7 mpc 5.5 mpc at ) survey at 1.1 mm taken with the atacama large millimeter / submillimeter array ( alma ) in the ssa22 field . we observe the core region of a protocluster , achieving a typical rms sensitivity of 60 jy beam at a spatial resolution of . we detect 18 robust alma sources at a signal - to - noise ratio ( snr ) . comparison between the alma map and a 1.1 mm map taken with the aztec camera on the atacama submillimeter telescope experiment ( aste ) indicates that three submillimeter sources discovered by the aztec / aste survey are resolved into eight individual submillimeter galaxies ( smgs ) by alma . at least ten of our 18 alma smgs have spectroscopic redshifts of , placing them in the protocluster . this shows that a number of dusty starburst galaxies are forming simultaneously in the core of the protocluster . the nine brightest alma smgs with snr have a median intrinsic angular size of ( physical kpc at ) , which is consistent with previous size measurements of smgs in other fields . as expected the source counts show a possible excess compared to the counts in the general fields at mjy due to the protocluster . our contiguous mm mapping highlights the importance of large - scale structures on the formation of dusty starburst galaxies .

1611.09857

the history of galaxy formation and evolution appears to be linked to the growth of cosmic large - scale structure . in the present - day universe , clusters of galaxies represent some of the densest environments . massive and passive elliptical galaxies preferentially reside in cluster cores ( e.g. , @xcite ; @xcite ) . recent works have suggested that brightest cluster galaxies at the center of clusters at @xmath13 are relatively evolved and hence they experienced rapid growth at earlier epochs ( e.g. , @xcite ; ) . galaxy cluster archaeology and cosmological simulations suggest the rapid growth of massive ellipticals in high density regions at @xmath14 ( e.g. , @xcite ; @xcite ) . therefore uncovering intense star - forming activity in protoclusters at such high redshifts is of great importance for understanding galaxy and cluster formation in the era when they actively assembled their stars . such actively star - forming galaxies are expected to be enshrouded by dust , which renders them difficult to observe at rest - frame uv to optical wavelengths . their spectral energy distributions ( seds ) thus should be dominated by far - infrared ( fir ) emission , and therefore observing at fir to submm / mm wavelengths may be essential to uncover these dusty galaxies ( so called submillimeter galaxies or smgs ; for reviews , see @xcite ; @xcite ) . extensive efforts have been made to search for and identify such obscured star - forming galaxies in protoclusters at @xmath15 , using submm / mm bolometer cameras onboard single - dish telescopes , such as aztec ( e.g. , @xcite ; @xcite ; @xcite ) , scuba ( e.g. , @xcite ; @xcite ) , scuba2 ( e.g. , @xcite ) , and laboca ( e.g. , ; @xcite ) as well as fir to submm / mm satellites , including _ ( e.g. , @xcite ) and _ planck _ ( e.g. , ) . while these single - dish telescopes are beneficial to cover wide area and find bright smgs , their relatively poor angular resolution ( fwhm @xmath16 ) and sensitivity limits due to source confusion have prevented us from obtaining accurate identifications for sources and/or revealing less extreme dusty galaxies . the advent of the atacama large millimeter / submillimeter array ( alma ) allows us to break through these limitations . contiguous alma mosaic imaging is able to open a window for submm / mm deep surveys with sub - arcsec resolution ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) . we utilized alma to conduct a deep imaging survey toward a well - studied protocluster at @xmath3 in the ssa22 field . the protocluster was originally discovered as a redshift spike of lyman - break galaxies ( lbgs ) and ly@xmath17 emitters ( laes ) by @xcite , ( @xcite ) and was proposed as an ancestor of present - day clusters such as coma . a remarkable lae density peak ( @xmath186 times the average surface density ) spreading over tens of comoving mpc has been found among a huge filamentary structure ( @xmath19 comoving mpc ) traced by laes ( @xcite ; @xcite ; @xcite ; @xcite ) . distant red galaxies ( drgs ) and @xmath20-band selected galaxies are found to be more abundant in the core of the protocluster than the field , which supports the elevated formation of massive galaxies there ( @xcite ; @xcite ; @xcite ) . in the ssa22 field , a number of submm / mm surveys have been conducted ( e.g. , @xcite , @xcite ; @xcite ; @xcite ; @xcite ) . @xcite discovered a statistical correlation between 15 bright 1.1 mm sources detected by aztec and @xmath21 laes and suggested that smgs preferentially formed within the cosmic structure . @xcite performed counterpart identification of aztec sources using radio and near - to - mid infrared data and derived optical to near - infrared photometric redshifts ( @xmath22 ) , which supported the trend found in @xcite and led to the conclusion that a significant fraction of the aztec sources are concentrated in the center of the protocluster . on the basis of our alma survey , in @xcite we reported the discovery of a concentration of dusty starbursts and x - ray agns at a center of the protocluster , and suggested that the large - scale environment plays a key role in the formation of these rare and active populations ( see also @xcite ) . here we present the full catalog of the alma smgs and the comprehensive results in terms of mm properties unveiled by alma . our survey design and observations are described in 2 . we present the procedures for source extraction and catalog selection in 3 . we compare the alma survey with the previous aztec survey and derive properties of the sources including source counts in 4 . we conclude with a summary in 5 . we will report the result of a line search in our survey in hayatsu et al . ( in preparation ) . throughout the paper , we adopt a cosmology with @xmath23 , and h@xmath24=70 km s@xmath5 mpc@xmath5 we observed a @xmath25 rectangular field and peripheral regions centered at @xmath26 , @xmath27 ( j2000 ) at 1.1 mm , using alma band 6 in cycle 2 ( program i d . 2013.1.00162.s , pi . h. umehata ) . hereafter we name this field `` adf22 '' , which is an acronym of `` alma deep field in the ssa22 field '' . adf22 is located at the very center of the cosmic large - scale structure ( or `` cosmic web '' ) at @xmath28 . as illustrated in fig . [ f1 ] , this area is close to the projected density peak of @xmath3 lae candidates selected by subaru / suprime - cam observations with a narrow - band filter ( nb497 ) ( @xcite ) . previous works have unveiled not only the projected density distribution but the density structure in three - dimensions ( @xcite ) . when we only focus on spectroscopically confirmed @xmath3 laes and draw a three - dimensional map , adf22 coincides with the junction of a comoving 50 mpc - scale filamentary structure ( @xcite ) . this means that adf22 covers a node of the cosmic large - scale structure and provides a unique opportunity to examine galaxy formation in such a dense field at this early epoch . another important feature is that adf22 contains a significant numbers of the most active populations of galaxies which are suggested to preferentially reside in dense environments , including smgs , qsos , and lyman alpha blobs ( labs ) . there are three 1.1 mm sources discovered by the aztec / aste survey ( @xcite ; @xcite ) . prior to the present work , one of the three sources , ssa22-aztec77 ( or smm j221735 + 001537 ) , had been proposed as a secure member of the @xmath3 structure ( @xcite ; @xcite ; @xcite ) and the other two sources had been also suggested to be at @xmath21 ( @xcite ; @xcite ; @xcite ) . one qso at @xmath3 discovered by @xcite and two @xmath21 labs listed in @xcite are also located within adf22 ( see also fig . [ f1 ] ) . adf22 is thus an excellent region for probing dusty star - formation activity in a wide variety of galaxies at the core of the protocluster . we chose the central observing frequency of 263 ghz ( 1.14 mm ) , which is similar to that of our previous aztec / aste survey , 270 ghz , and allows direct comparison with the flux densities of sources from our aztec / aste and alma surveys . since the size of primary beam of band 6 is larger than that of band 7 , the frequency set - up is also beneficial to reduce the number of pointings compared to higher frequency . in order to cover a @xmath29 rectangular field contiguously , we utilized a 103-field mosaic . the spacing between adjacent pointings was 0.72 times the primary beam ( fwhm 23@xmath30 at 263 ghz ) , which was a compromise between the homogeneity of sensitivity and the observing time required to cover as wide an area as possible . observations were divided into two campaigns ( 2014 and 2015 runs ) , as summarized in table 1 . the first run was conducted on 2014 june 610 with 3336 available 12 m antennas in the c34 - 4 configuration ( longest baseline 650 m ) and very good 1.1 mm weather conditions ( precipitable water vapor ( pwv ) of 0.301.27 mm ) . the second run was done on 2015 april 4 , 5 , and 13 with 3240 12 m antennas in c32 - 2 ( longest baseline 349 m ) and good conditions ( pwv of 0.91.9 mm ) . the exposure time per pointing was 30 sec for each scheduling block ( sb ) . these observations resulted in initial on - source time of 24.5 min per pointing ( depending on the individual pointing ) and a total on - source time of 386 min . we utilized the time division modes ( tdm ) correlator , with @xmath31 dual polarization channels over the full 8 ghz bandwidth , giving an effective bandwidth of 7.5 ghz after flagging edge channels . the correlator was set up to target two spectral windows of 1.875 ghz bandwidth each at 15.6 mhz ( @xmath18 20 km s@xmath5 ) channel spacing in each sideband . the central frequencies of the four spectral windows are 254.0 , 256.0 , 270.0 , and 272.0 ghz . we note that this frequency range enables us to search for the @xmath32co(9 - 8 ) line ( @xmath331036.912 ghz ) at @xmath21 ( we actually found one co(9 - 8 ) emitter as we describe below . details will be presented in hayatsu et al . , in preparation . ) . the quasar j2148 + 0657 with a flux 1.2 jy was observed regularly for amplitude and phase calibration . the absolute flux scale was set using j2148 + 0657 ( for the 2014 run ) and uranus and neptune ( for the 2015 run ) . we estimate that the absolute flux accuracy is within 20% . this uncertainty in the absolute flux calibration is not included in the following analyses and discussions . ccccccc sb1 & 2014 jun 06 & 36 & 20 m 650 m & 180 & @xmath34 ( @xmath35 deg ) & j2148 + 069 + sb2 & 2014 jun 07 & 34 & 20 m 646 m & 140 & @xmath36 ( @xmath37 deg ) & j2148 + 069 + sb3 & 2014 jun 07 & 34 & 20 m 646 m & 180 & @xmath38 ( @xmath39 deg ) & j2148 + 069 + sb4 & 2014 jun 09 & 34 & 20 m 646 m & 180 & @xmath40 ( @xmath41 deg ) & j2148 + 069 + sb5 & 2014 jun 10 & 34 & 20 m 646 m & 180 & @xmath42 ( @xmath43 deg ) & j2148 + 069 + sb6 & 2015 apr 04 & 33 & 15 m 328 m & 1103 & @xmath44 ( @xmath45 deg ) & neptune + sb7 & 2015 apr 04 & 33 & 15 m 328 m & 1103 & @xmath46 ( @xmath47 deg ) & neptune + sb8 & 2015 apr 05 & 35 & 15 m 328 m & 1103 & @xmath48 ( @xmath49 deg ) & uranus + sb9 & 2015 apr 13 & 40 & 15 m 349 m & 1103 & @xmath50 ( @xmath51 deg ) & neptune data reduction was performed using the common astronomy software application ( casa ; @xcite ) . the calibration and imaging was performed using casa version 4.2.2 of the alma pipeline and casa version 4.5.0 , respectively . our analysis is complicated by the fact that our observations were carried out with multiple array configurations . our observations are composed of nine sbs as summarized in table 1 . since each sb has a different antenna setup , the resulting synthesized beam sizes differ among sbs . there are broadly two array configurations ; the maximum baseline lengths of sb6sb9 ( taken in 2015 ) are systematically shorter than sb1sb5 ( taken in 2014 ) . because of the difference in the observed area in 2014 and 2015 , the western edge fields ( fields 81103 in the right panel of fig . [ f2 ] ) lack longer baseline data and have lower sensitivity ( table 1 ) . considering this situation , we first created a 1@xmath30-tapered map for the entire field ( i.e. , fields 1103 ) , which enables us to achieve nearly uniform angular resolution across the map . this larger synthesized beam is more sensitive to extended sources but suffers from an increased noise level . as we show in the maps of individual sources ( fig . [ stamp ] ) and measured sizes of bright sources ( table 3 ) , the resulting angular resolution , @xmath52 , seems to be sufficient to detect the majority of sources in this field . we also discuss the possibility that we are missing very extended components in 4.1.1 . the @xmath53-data for individual pointings were first combined into a single @xmath53data set . we then fourier - transformed the combined data to create a single `` dirty '' map with natural weighting , by setting the imager mode to `` mosaic '' and adopting the taper parameter , outertaper = 120 k@xmath54 . after we measured the rms noise level across the whole dirty map , we repeated the clean process down to 1.5@xmath55 , putting a tight clean box ( @xmath56 in size ) around each 5@xmath55 source ( in a manner similar to that reported by @xcite ) . the resulting synthesized beam is @xmath57 in size ( p.a . @xmath58 we denote this map as the full / lores map . we created another complementary map . since the five sbs , sb1 sb5 , have the longer baseline data with baselines up to 600 k@xmath54 , we can achieve a better sensitivity making the best of the longer baselines . for this purpose , we created a second map for 80 pointing field ( fields180 ) . we in general adopted the same procedure described above , but we did not apply any tapering to the map . the second map has a synthesized beam of @xmath59 ( p.a . @xmath58 hereafter we call it the deep / hires map . in the following analyses , we utilize primarily the full / lores map . the deep / hires map is used for detecting compact faint sources ( 3 ) and resolving bright sources ( 4.2 ) . the final full / lores map corrected for the primary beam response is shown in fig . [ f2 ] . in the following sections , we consider the effective area ; the area in the map within which the primary beam coverage is greater than 50% . this results in 7.0 arcmin@xmath60 area in the case of the full / lores map ( 5.8 arcmin@xmath61 for the deep / hires map ) . a sensitivity map was constructed utilizing an aips ( @xcite ) task , rmsd , calculating the rms for each @xmath62 pixel using the surrounding @xmath63 pixels ( or @xmath640 ) on an image prior to correcting for primary beam response . the correction results in a range of 1@xmath55 depth of 52 170 @xmath4jy beam@xmath5 with a median value of 75 @xmath4jy beam@xmath5 ( fig . [ area ] ) for the full / lores map . in the case of the deep / hires map , we obtained a range of 1@xmath55 depth from 50 129 @xmath4jy beam@xmath5 with a median value of 60 @xmath4jy beam@xmath5 ( fig . [ area ] ) . source extraction was performed using the image map and sensitivity map that were not corrected for the primary beam attenuation . first we utilized the full / lores map . a source - finding algorithm , aegean v1.9.5 - 56 ( @xcite ) , was used to extract positive sources above 3.5@xmath55 . in parallel , we also counted _ negative _ sources with source - like profiles above 3.5@xmath55 during the procedure to estimate the fraction of spurious sources among the detected sources . the cumulative number counts of positive and negative sources as a function of signal - to - noise ratio ( snr ) is shown in fig . we found 17 positive sources with @xmath65 and 28 sources with @xmath66 . there were no negative sources above 5@xmath55 , while we found seven negative ones with 44.6@xmath55 . the results show that the detection limit of 5@xmath55 is secure and conservative . gaussian statistics also support the validity of the threshold . since the map contains @xmath6732000 beams , we would expect less than one spurious peaks above 5@xmath55 . therefore we adopt these 17 sources as secure detections ( adf22.1adf22.17 ; see table 2 ) . we also performed the same procedures on the deep / hires map , independently . as a result , 14 sources were detected at @xmath65 , one of which was not detected in the shallower full / lores map . this source ( adf22.18 ; see table 2 ) was added to the catalog of secure detection because there were no negative sources with @xmath65 in the deep / hires map . therefore in total 18 sources were detected at snr@xmath7 . in the following analyses and discussion , we mostly focus on the 18 secure sources not affected by spurious features . we show a supplementary catalog of 14 sources detected tentatively in the full / lores and deep / hires maps in appendix a. the flux densities of the detected sources were measured with the imfit task of casa , using the map after correction for the primary beam attenuation . we adopt the integrated flux density as the flux density of a source unless it is lower than the peak flux density ( e.g. , @xcite ) . the measurements of fluxes on the two maps are in good agreement with each other . the median ratio of the flux density between the full / lores map and the deep / hires map is @xmath68 for the 13 sources detected above @xmath69 in both maps . we performed a suite of simulations to evaluate the completeness and boosting effect on our flux measurement . we briefly describe the method below ( for more detail , see @xcite ) . model sources created by scaling the synthesized beam are injected into the image map corrected for primary beam attenuation . the input position is randomly selected to be @xmath70 away from @xmath71 sources . we then repeated the same procedures for the source extraction using the artificial map as described in 3.1 . the completeness is measured as the recovery rate of the injected model sources ; when an input source is detected within [email protected] of the injected position with @xmath72 , the source is considered to be recovered . model sources with flux densities over the range @xmath730.020.6 mjy in steps of 0.02 mjy are considered and the procedures are repeated 1000 times for each flux density bin ( @xmath740.02 mjy ) . the results are summarized in fig . [ f4 ] as a function of input flux density . the completeness of our source extraction is @xmath75% at snr=4 and rises to @xmath76% at snr=5 . the completeness reaches @xmath18100% at [email protected] . we consider the completeness in calculating source counts in 4.3 , although it does not matter significantly for @xmath65 sources . in the course of the completeness measurement , we also evaluate the effect of flux boosting . it is known that the flux density of sources in a signal - to - noise limited catalog tend to be boosted as a whole due to random noise fluctuations and the shape of the source count distribution ( e.g. , @xcite ; @xcite ; @xcite ) . we plot the ratio of input and output flux density against the input flux density in fig . the results show that the effect is negligible if we consider only those sources detected at @xmath65 . hence we do not consider the flux boosting effect in the following discussion . ccccccccc adf22.1 * & almaj221732.41 + 001743.8 & adf22a & aztec1 & 11.3@xmath78 & 72 & 58.1 & 5.60 @xmath79 0.13 & 3.092@xmath80 + adf22.2 & almaj221736.11 + 001736.7 & & & & 63 & 31.8 & 2.02 @xmath79 0.02 & ... + adf22.3 & almaj221735.15 + 001537.2 & adf22b & aztec77 & 2.4@xmath81 & 66 & 27.0 & 1.89 @xmath79 0.04 & 3.096@xmath82 + adf22.4 & almaj221736.96 + 001820.7 & & aztec14 & 4.5@xmath83 & 72 & 26.6 & 1.95 @xmath79 0.05 & 3.091@xmath84 + adf22.5 & almaj221731.48 + 001758.0 & & & & 99 & 20.3 & 2.43 @xmath79 0.20 & ... + adf22.6 & almaj221735.83 + 001559.0 & adf22c & & & 69 & 19.1 & 1.45 @xmath79 0.09 & 3.089@xmath85 + adf22.7 & almaj221732.20 + 001735.6 & adf22i & aztec1 & 11.3@xmath78 & 86 & 18.7 & 1.65 @xmath79 0.07 & 3.097@xmath86 + adf22.8 & almaj221737.11 + 001712.3 & adf22d & & & 77 & 15.0 & 1.19 @xmath79 0.06 & 3.090@xmath87 + adf22.9 & almaj221736.54 + 001622.6 & adf22e & & & 60 & 12.8 & 0.82 @xmath79 0.08 & 3.095@xmath87 + adf22.10 & almaj221737.10 + 001826.8 & & aztec14 & 4.5@xmath83 & 71 & 9.8 & 0.72 @xmath79 0.04 & ... + adf22.11 & almaj221737.05 + 001822.3 & adf22f & aztec14 & 4.5@xmath83 & 77 & 9.5 & 0.79 @xmath79 0.05 & 3.093@xmath87 + adf22.12 & almaj221732.00 + 001655.4 & adf22 g & & & 71 & 8.8 & 0.63 @xmath79 0.03 & 3.091@xmath87 + adf22.13 & almaj221737.42 + 001732.4 & & & & 81 & 8.0 & 0.79 @xmath79 0.05 & ... + adf22.14 & almaj221731.34 + 001639.6 & & & & 99 & 7.5 & 0.98 @xmath79 0.13 & ... + adf22.15 & almaj221732.77 + 001727.5 & & & & 79 & 5.3 & 0.50 @xmath79 0.08 & ... + adf22.16 & almaj221736.81 + 001818.0 & adf22h & aztec14 & 4.5@xmath83 & 77 & 5.3 & 0.56 @xmath79 0.07 & 3.085@xmath87 + adf22.17 & almaj221737.69 + 001814.4 & & aztec14 & 4.5@xmath83 & 66 & 5.1 & 0.60 @xmath79 0.09 & ... + adf22.18 * * & almaj221732.23 + 001527.8 & & & & 82 & 5.3 & 0.44 @xmath79 0.05 & 2.105@xmath88 the measured properties of the adf22 smgs are summarized in table 2 . the first column lists the source ids in this paper , which are generally ranked in terms of snr . the nine smgs reported in @xcite are noted in the second column . there are three aztec sources in our field ( @xcite ) . we listed the ids and flux densities of the corresponding aztec source if a given alma smg is located within the aztec beam ( fwhm ; @xmath89 ) . we give the flux density , measured by gaussian fitting with casa , imfit . we note that these measurements are from the full / lores map except for adf22.18 , which was detected at @xmath65 only in the deep / hires map . we also report the spectroscopic redshift ( @xmath90 ) , if known . eleven of the 18 smgs have @xmath90 , ten of which have @xmath8 . in addition to eight @xmath8 smgs reported in @xcite , we adopt @xmath90 for three smgs , adf22.4 , adf22.7 , and adf22.18 . the redshifts of adf22.4 and adf22.7 is determined from our recent alma observations ( @xmath32co(9 - 8 ) at @xmath91 and [ cii ] 157.7 @xmath4 m at @xmath92 , respectively ; umehata et al . , in preparation ) . adf22.18 coincides with a @xmath93 radio source , rg j221732.22 + 001528.2 , reported in @xcite with a small projected separation ( @xmath94 ) . we show the alma 1.1 mm images of the 18 smgs in fig . [ stamp ] . one reasonable interest regarding the alma source catalog is the overlap between the previously known galaxy populations and our newly discovered alma smgs . within the survey area of the full / lores map ( fig . [ f2 ] ) , there are 19 laes ( @xcite ; @xcite ; @xcite ) , five lbgs ( @xcite ; @xcite ; @xcite ) , and ten @xmath20-band selected galaxies ( @xcite ; @xcite ) within the protocluster ( i.e. , with @xmath95 ; @xcite ; @xcite ) . none of the alma smgs ( including candidates in the supplementary source catalog ) have lae / lbg counterparts , which shows that these rest - frame uv selected galaxies are clearly separated populations compared to galaxies individually detected by alma . in contrast , five out of ten k - band selected galaxies are securely detected by alma . therefore such a relatively massive galaxy population ( stellar mass , @xmath96 ; @xcite ) selected at rest - frame optical wavelengths appears to significantly overlap with the alma population . this trend is broadly consistent with recent works in sxdf ( @xcite ) and hudf ( @xcite ; @xcite ) as well as previous studies of alma smgs ( @xcite ) . we summarize the relationship to other populations in table 6 in appendix b. several hundred submm / mm sources discovered by single - dish telescopes have been observed by submm / mm interferometers to date . some appear to resolve into multiple , individual smgs while others have a unique counterpart above a given detection limit ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; ; @xcite ; ; @xcite ) . for instance , @xcite reported that , using the sma , three of 16 scuba sources are composed of multiple objects . from alma observations , @xcite and @xcite suggested that @xmath183040 % of laboca or bright scuba2 sources ( with flux densities of @xmath97 mjy and @xmath98 mjy , respectively ) are resolved into multiple smgs brighter than @xmath991mjy , making most of components `` ulirgs '' . these results suggest that the relatively poor angular resolution of single - dish imaging ( @xmath100 ) causes significant source blending and complicate our interpretation of the nature of smgs . now alma enables us to resolve not only individual single - dish sources but also relatively faint sources spread over a wider area . the aztec / aste survey of ssa22 by @xcite and @xcite presented a 1.1 mm image of adf22 . the 30@xmath30 resolution image has a 1@xmath55 depth of 0.7 mjy beam@xmath5 , and detects three sources above a 3.5@xmath55 detection threshold within adf22 ( ssa22-aztec1 , ssa22-aztec14 , and ssa22-aztec77 ; hereafter aztec1 , aztec14 , and aztec77 , respectively ) . two of the three aztec sources , aztec1 and aztec14 , have two and five alma smgs located within the fwhm of the aztec beam , respectively . in contrast aztec77 has only one associated alma smg ( fig . [ dist_alma ] ) . therefore the profile itself should reflect the two alma smgs . however , alma6 is located at outside of the aztec beam of aztec77 and so does not contribute to the measured flux of aztec77 significantly . hence here we treat adf22.3 as an unique counterpart of aztec77 . ] . in summary , three aztec sources are resolved into eight alma smgs . the result is in line with previous alma studies following up single - dish submm / mm sources with interferometers , which reports that a significant fraction of submm / mm sources detected by single - dish telescopes are found to be intrinsically multiple smgs ( e.g. , @xcite ) to check the relative flux scales , we compared the flux density between the aztec sources and the associated alma smgs . for the aztec sources , deboosted flux densities in @xcite ( @xmath101 ) are considered here . regarding the measurements for the alma smgs , we calculate the sum of the flux densities of the smgs with @xmath102 within the fwhm of the aztec beam ( @xmath103 ) . fig . [ flux_comp ] shows the results for the three aztec sources . the ratio between the two , @xmath103/@xmath101 , is in agreement within error bars for aztec77 and aztec14 , which suggests that the alma smgs discovered from this survey account for the majority of the flux density of the aztec sources . however , the situation is not the same for aztec1 , the brightest aztec source in the ssa22 field , which has @xmath103/@xmath104 . the discrepancy could imply the existence of additional components of 1.1 mm emission not accounting for in the alma map . one possibility is that we are missing a number of fainter and/or diffuse sources ( e.g. , @xcite ; @xcite ) . if we include sources in the supplementary catalog with snr=@xmath105 , then the flux ratio @xmath103/@xmath101 gets closer to unity for all three aztec sources ( fig . [ flux_comp ] ) . but these sources is still insufficient to explain the case of aztec1 . much fainter sources might also contribute to the aztec sources for aztec1 . @xcite gave us another clue . tamura et al . reported the 860@xmath4 m flux density of adf22.1 , @xmath106 mjy , using the submillimeter array . they obtained the natural - weighted synthesized beam , @xmath107 ( p.a . 34.3 deg ) , and found that the source was likely not to be resolved . if we predict 1.1 mm flux density of adf22.1 using an averaged smg template from the aless survey ( @xcite ) scaled to the 860 @xmath4 m flux density , the expected value is @xmath1086.0 mjy . the estimate is broadly consistent with our measurement , which suggests that we are measureing the vast majority of the dust emission from adf22.1 . there are two alma smgs , adf22.5 and adf22.15 , located just outside of the beam for aztec1 ( fig . [ dist_alma ] ) . if we take into account adf22.5 and adf22.15 , the integral of the flux densities of the alma smgs approach the flux density of aztec1 ( @xmath103/@xmath109 ) . in order to investigate whether these smgs indeed contribute to the flux density of aztec1 in the aztec map , we made a `` model '' aztec image convolving the alma map with a gaussian kernel with a fwhm of 30@xmath30 . aztec1 has flux density 7.1 mjy in the model map , which is consistent with the sum of flux density of the alma smgs within the beam for aztec1 , not including adf22.5 and adf22.15 . therefore these two nearby smgs are likely not to account for the discrepancy . we do not consider a systematic error on the absolute flux accuracy for the aztec map throughout the above discussion . although it is difficult to estimate the influence for our small sample , alma surveys of a significant number of aztec sources will allow us to examine the effect . as displayed in fig . [ dist_alma ] , we compared the spatial distribution of smgs discovered by alma and the contours of the aztec 1.1 mm emission , below the aztec detection limit ( 3.5@xmath55 ) . our alma map has discovered ten smgs outside the aztec source positions . we found that nine of the ten alma smgs are located within or in the vicinity of the area where the 1.1 mm flux density is 0 2.1 mjy beam@xmath5 in the aztec map . this may show that the structure traced by the faint aztec emission , below 3@xmath55 , reflects the distribution of 1.1 mm sources to a certain degree . as we reported in @xcite , the aztec catalog is significantly incomplete at around 3.5@xmath55 ( i.e. , the detection threshold ) . the completeness of the aztec map is only @xmath110% at 2 mjy . among the adf22 sources with flux density @xmath182 mjy , two smgs , adf22.2 and adf22.5 , are found to be located outside of the aztec beam . our results therefore show that alma mosaic has capability to find such bright smgs which was missed by previous submm / mm surveys taken with single - dish telescopes . there has been a debate on whether multiple individual smgs of sources which are identified with a single - dish telescope are physically connected or instead are just a line - of sight projection ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) . for one source , @xcite proposed that it was composed of multiple , physically unrelated smgs . in contrast , @xcite showed that one aztec source fragments into two h@xmath17 emitters at @xmath111 , which supports the physical connection between the two ( see also @xcite ) . @xcite suggested that a portion of bright submm sources arise from physically related smgs since the number of detected alma smgs in the vicinity of the scuba2 sources are two orders of magnitude higher than the general field . on the other hand , recently @xcite claimed that physically unrelated smgs can also reproduce the single - dish sources in their semi - analytic model ( see also @xcite ) . among 18 smgs in adf22 , 11 smgs have @xmath90 ( @xcite ; in prep ) , 10 of which are at @xmath112 ( table 2 ) . all of the brightest alma counterparts of the aztec sources are at @xmath3 and aztec1 and aztec14 have multiple alma smgs at @xmath8 within the aztec beam ( @xmath113 kpc at @xmath2 ) . these results favor that physically connected multiple smgs appear as a single single - dish source , at least in dense environment such as center of a protocluster . the overdensity of galaxies on such scales is conductive to mergers and dissipative interactions ( e.g. , @xcite ) and hence such small scale over - densities of smgs may indicate that smgs undergo merger - induced star formation . furthermore , there are other @xmath3 alma smgs outside the aztec beam across the entire field of adf22 ( fig . [ dist_alma ] , table 2 ) . this suggests that intense dusty star - formation may also be enhanced by the environment on a large - scale ( @xcite ; see also @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) . two - dimensional gaussian fits to the alma smgs in the image plane using the deep / hires map suggest that adf22 sources are generally resolved . the results for the nine brightest smgs with snr@xmath9 are summarized in table 3 . the deconvolved major axes are [email protected] [email protected] and the median value for the nine smgs is @xmath114 ( gaussian fwhm ; @xmath11 physical kpc at @xmath3 ) . seven of the nine smgs have @xmath115 , which leads the almost same median value , @xmath116 if we only consider such robust protocluster members . the size distribution in adf22 is generally consistent with previous measurements for smgs in other fields ( fig . [ size ] ) . @xcite measured the size for 23 bright smgs observed at 870 @xmath4 m and derived a median angular size fwhm=@xmath117 and @xcite reported that 13 smgs , which were observed at 1.1 mm and thought to be at @xmath118 , have a median fwhm of @xmath119 . these results suggest that dusty star formation occurs in compact regions , a few kiloparsecs in extent . there does not appear to be a significant difference between the size of smgs within the @xmath3 protocluster and other smgs . this might suggest that the local mechanism triggering intense starbursts does not significantly depend on the large - scale environment . the radio continuum also provides a tool to measure the scales of dusty starburst galaxies like smgs ( e.g. , @xcite ; @xcite ; @xcite ) . these measurements shows relatively extended profile ( e.g. , a median of 5 kpc ; @xcite ) , which is larger than the bright alma smgs discussed above . up to date some explanation has been proposed . for instance , @xcite suggests that the relatively compact star formation is seen in smgs with higher star - formation rate ( sfr ) while main - sequence star - forming galaxies have different scales . @xcite indicates that the difference between the diffusion length of cosmic rays and far - infrared photons can account for it . deep and high angular resolution radio imaging in adf22 in the future will provide us with important clue to clarify the origin of the discrepancy between the radio and fir sizes of smgs . the flux density and measured source size allow us to derive star - formation rate surface density ( @xmath120 ) and investigate the condition of star - formation . we estimated @xmath121 [ 8 - 1000 @xmath4 m ] using sed templates of well - studied starburst galaxies ( @xcite ; @xcite ; @xcite ) in the same manner as in @xcite . the sfr is in turn derived from @xmath121 using the empirical calibration by adjusted to the kroupa initial mass function ( @xcite ) . since adf22.2 and adf22.5 do not have @xmath90 , we assume @xmath122 ( a median value for aztec sources ; @xcite ) . finally @xmath120 was calculated using the size of the 1.1 mm continuum emission and the sfr divided by a factor of two , following @xcite ( table 3 ) . the median value is @xmath120=50 @xmath123 yr@xmath5 kpc@xmath124 , significantly lower than the predicted eddington limit for radiation pressure supported disks ( e.g. , @xcite ; @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) . therefore starbursts seen in bright smgs in adf22 are not eddington limited as a whole . as some authors have noted ( e.g. , @xcite ) , if smgs have clumpy structure , the individual components might be eddington limited , as has been claimed from some recent high angular resolution alma images of the brightest smgs ( e.g. , hatsukade et al . 2015 ; iono et al . 2016 , but see also @xcite ) . further observations capable of resolving sub - kpc structures ( @xmath125 angular resolution ) are required to probe such a scenario for the smgs in adf22 . in this section , we report the 1.1 mm number counts in adf22 , which is one of the most fundamental parameters in describing the evolution of this population . previously there have been a number of studies investigating submm / mm number counts from a variety of alma data sets ; laboca / scuba2 source follow - up ( @xcite ; @xcite ) , the use of calibration field ( @xcite ) , wide range of archival data ( @xcite ; @xcite ; ; @xcite ) , and contiguous mapping ( @xcite ; @xcite ; @xcite ) . here we present the number counts in a @xmath187 arcmin@xmath60 contiguous field in the remarkable high density region in the early universe , well - suited for investigating the environmental dependence of smg source counts . to minimize the number of contaminants by false positives , we utilized only the 17 smgs detected with snr @xmath126 in the full / lores map . we excluded adf22.18 since the snr of this source is @xmath127 in the full / lores map . to investigate how the existence of the @xmath3 protocluster affects the counts , we calculated the counts in two ways . first we included all 17 smgs . second , we exclude the ten smgs with known @xmath128 ( table 2 ) and used only the remaining seven smgs . flux boosting effects were not considered ( see 3.2 ) . we corrected for the completeness of the source extraction , although the counts should not be significantly affected by the correction . we calculated the number counts and associated errors in the same way as @xcite . [ count ] shows the differential and cumulative counts ( in the left and right panels , respectively ) . we show the counts from all 17 smgs and those obtained when we remove the ten sources with @xmath128 . we also summarize the count statistics in table 4 . for comparison , we also show other alma counts in both panels of fig . [ count ] ( @xcite ; @xcite ; @xcite ; @xcite ; @xcite ) by scaling to 1.1 mm assuming a modified black body with typical values for smgs ( spectral index of @xmath129 = 1.5 , dust temperature of 35 k ( e.g. , @xcite ) and @xmath130 where necessary ) . as discussed in @xcite , the volume density of the alma smgs at @xmath3 in adf22 is at least about three orders of magnitude greater than the expected value in general fields . we now investigate whether the protocluster members affect the number counts or not , considering alma smgs at all redshifts . as shown in fig . [ count ] , we see a possible excess in both of the differential and cumulative counts ( for all of 17 smgs ) . in the case of cumulative counts , the counts in adf22 are approximately five times higher than those found in aless at @xmath131 mjy ( @xcite ) and four to five times higher than those obtained at @xmath132 mjy ( @xcite ; @xcite ) . the group of smgs at @xmath3 is undoubtedly responsible for this excess . on the other hand , we do not see an excess in the faintest bin ( i.e. , sub - mjy sources ) . a possible attribution is that the elevation of dusty star - forming activity in dense environment is significant only for relatively bright smgs ( @xmath133 mjy ) . deeper observation in the future will provide an answer . here another question naturally arises . what are the counts in typical fields ? the core of the ssa22 protocluster is an unusual environment . therefore we have to identify alma smgs which are not associated with the protocluster and utilize them in calculating the source counts in a typical field . it is , however , difficult to derive the counts correctly because seven sources with @xmath102 do not yet have @xmath90 . here we conservatively created counts using all seven alma smgs ( see fig . [ count ] ) , some of which may also be at @xmath3 . the cumulative source counts from the seven smgs and previous counts at bright fluxes ( @xcite ; @xcite ) are fitted to a double - power law function of the form , @xmath134^{-1}$ ] , which yields the best - fit parameters of @xmath135 deg@xmath124 , @xmath136 mjy , @xmath137 , and @xmath138 ( the right panel of fig . [ count ] ) . our counts excluding known protocluster members are in reasonable agreement with recent estimates which is ought to be relatively free of cosmic variance ; those from a number of calibration fields ( @xcite ) and those from a 2 arcmin@xmath61 contiguous survey ( @xcite ) . on the one hand , those counts from the seven smgs seem to be several times lower than @xcite at @xmath139 mjy . one possible explanation for this discrepancy is that previous counts using 3 - 4@xmath55 sources are seriously affected by contaminants , as pointed out by @xcite . another explanation involves clustering , although this effect may also contribute to our counts . the counts derived from serendipitously detected sources around other targeted galaxies might be biased or clustered as mentioned in previous papers ( e.g. , @xcite ; @xcite ) . our results suggest that the cluster environments are recognizable in the smg number counts . at the same time , there are still large uncertainties for submm / mm number counts , including field to field variation , the shape of dust seds , and absolute flux uncertainties . forthcoming much wider alma surveys with sufficient depth , covering a variety of environments , will significantly improve our understanding of the submm / mm counts and their dependence on environment . we have presented the results from a 103-pointing alma 1.1 mm mosaic in the ssa22 field . we covered a 7 arcmin@xmath60 area contiguously at the junction of a 50-mpc scale cosmic large - scale structure at @xmath3 containing three 1.1 mm sources discovered previously by the aztec / aste survey . observations were conducted in 2014 and 2015 with different array configurations and therefore we created two maps to handle the different angular resolution data ; a @xmath140 tapered map ( full / lores map ) , and a @xmath141 resolution map ( deep / hires map ) . these maps reach a median rms noise of 75 @xmath4jy beam@xmath5 and 60 @xmath4jy beam@xmath5 , respectively . applying our source detection algorithm to our two maps , we discover 18 robust alma smgs with snr@xmath7 , with 1.1 mm flux density @xmath142 mjy beam@xmath5 , corresponding to @xmath143 @xmath144 at @xmath2 . through a comparison between the aztec map and the alma map of adf22 , we have found that three single - dish ( aztec ) sources are resolved into eight discrete alma smgs . this suggests that multiple dusty galaxies may often contribute to one bright submm / mm source discovered by single - dishes in dense protocluster environments . the flux densities of aztec sources are consistent with the sum of the flux densities of alma smgs within errors for two sources , aztec14 and aztec77 , but not for a third ( aztec1 ) , possibly indicating that there are additional missing faint and/or diffuse dusty galaxies remaining undetected in the alma map . ten out of 18 smgs in our sample are known to lie at @xmath3 . the fact that multiple @xmath3 alma smgs are found to comprise two of three of the single - dish ( aztec ) sources suggests that at least in dense protocluster environments interactions may be responsible for a significant fraction of observed smg multiplicity . the alma smgs are generally resolved in our data . we measured the deconvolved source size for nine brightest alma smgs with @xmath145 , conducting gaussian fitting to the alma smgs . the median size of @xmath114 ( @xmath11 physical kpc at @xmath3 ) agrees well with previous measurements and there is no recognizable difference between the size of @xmath3 protocluster members and the alma smgs in other fields . we derived the 1.1 mm source counts using all 17 smgs above 5@xmath55 from the full / lores map and also the seven smgs without @xmath128 . the counts of 17 smgs are about five times higher than the counts in typical fields at @xmath133 mjy , which is caused by the @xmath3 smg concentration associated with the protocluster . on the other hands , we found that the source counts in adf22 are consistent with recent unbiased alma counts when we exclude known members of the @xmath3 structure . in conclusion , we have obtained deep and high angular resolution imaging covering a certain area using alma , which demonstrated that there is an unusual number of dusty galaxies residing at the core of the @xmath3 protocluster . the observation provides a basis for understanding the formation of dusty galaxies within cosmic large - scale structure in the early universe . cccccccc adf22.1 & 0.85 @xmath79 0.04 & 6.5 @xmath79 0.3 & 0.33 @xmath79 0.03 & 2.5 @xmath79 0.2 & 13.0 @xmath146 & 1100 @xmath147 & 40 + adf22.2 * & 0.29 @xmath79 0.05 & 2.2 @xmath79 0.4 & 0.05 @xmath79 0.11 & 0.4 @xmath79 0.8 & 12.6 @xmath146 & 400 @xmath148 & 300 + adf22.3 & 0.45 @xmath79 0.05 & 3.4 @xmath79 0.4 & 0.18 @xmath79 0.08 & 1.4 @xmath79 0.6 & 12.5 @xmath146 & 370 @xmath149 & 50 + adf22.4 & 0.25 @xmath79 0.05 & 1.9 @xmath79 0.4 & 0.04 @xmath79 0.11 & 0.3 @xmath79 0.8 & 12.6 @xmath146 & 380 @xmath150 & 420 + adf22.5 * & 0.57 @xmath79 0.06 & 4.4 @xmath79 0.5 & 0.26 @xmath79 0.06 & 2.0 @xmath79 0.5 & 12.7 @xmath146 & 480 @xmath151 & 30 + adf22.6 & 0.38 @xmath79 0.08 & 2.9 @xmath79 0.6 & 0.30 @xmath79 0.14 & 2.3 @xmath79 1.1 & 12.4 @xmath146 & 280 @xmath152 & 30 + adf22.7 & 0.32 @xmath79 0.08 & 2.4 @xmath79 0.6 & 0.22 @xmath79 0.10 & 1.7 @xmath79 0.8 & 12.5 @xmath146 & 320 @xmath153 & 50 + adf22.8 & 0.26 @xmath79 0.07 & 2.0 @xmath79 0.5 & 0.18 @xmath79 0.05 & 1.4 @xmath79 0.4 & 12.3 @xmath146 & 230 @xmath154 & 50 + adf22.9 & 0.28 @xmath79 0.14 & 2.1 @xmath79 1.1 & 0.26 @xmath79 0.17 & 2.0 @xmath79 1.3 & 12.2 @xmath146 & 160 @xmath155 & 20 cccccccccc 0.64 & 9 & @xmath156 & 5 & @xmath157 & 0.40 & 17 & @xmath158 & 7 & @xmath159 + 1.59 & 7 & @xmath160 & 2 & @xmath161 & 1.00 & 8 & @xmath162 & 2 & @xmath163 + 4.00 & 1 & @xmath164 & 0 & & 2.52 & 1 & @xmath165 & 0 & we deeply appreciate the anonymous referee for a significant number of valuable comments . we are thankful for ken mawatari , for helping us to compile redshift catalogs of @xmath3 laes and lbgs in the literatures . hu is supported by the alma japan research grant of naoj chile observatory , naoj - alma-0071 , 0131 , 140 , and 0152 . hu is supported by jsps grant - in - aid for research activity start - up ( 16h06713 ) . yt is supported by jsps kakenhi no . rji acknowledges support from erc in the form of the advanced investigator programme , 321302 , cosmicism . irs and dma acknowledge support from stfc ( st / l00075x/1 ) . irs acknowledge support from the erc advanced investigator program dustygal 321334 , and a royal society / wolfson merit award . si acknowledge the support of the netherlands organization for scientific research ( nwo ) through the top grant project 614.001.403 . this paper makes use of the following alma data : ads / jao.alma#2013.1.00162.s . alma is a partnership of eso ( representing its member states ) , nsf ( usa ) and nins ( japan ) , together with nrc ( canada ) and nsc and asiaa ( taiwan ) and kasi ( republic of korea ) , in cooperation with the republic of chile . the joint alma observatory is operated by eso , aui / nrao and naoj . _ facilities : _ . in addition to the robust samples of 18 alma smgs with @xmath65 detection , there are also some possible source candidates with slightly lower significances . considering the result from our test using negative maps shown in fig . [ snr ] and fig . [ np_d ] , we adopt 4.0@xmath55 and 4.5@xmath55 as tentative detection thresholds for the full / lores and deep / hires maps , respectively . the test suggests that about half of these sources detected tentatively may be false detections and therefore we need to be careful in utilizing the catalog . cccccccccccc adf22 . 19 & almaj221733.87 + 001646.1 & 63 & 4.3 & 0.29 @xmath79 0.06 & full / lores + adf22 . 20 & almaj221734.69 + 001635.2 & 68 & 4.3 & 0.67 @xmath79 0.09 & full / lores + adf22 . 21@xmath166 & almaj221733.09 + 001718.5 & 78 & 4.3 & 0.79 @xmath79 0.08 & full / lores + adf22 . 22 & almaj221732.96 + 001636.0 & 67 & 4.3 & 0.42 @xmath79 0.07 & full / lores + adf22 . 23 & almaj221735.47 + 001805.1 & 65 & 4.2 & 0.30 @xmath79 0.06 & full / lores + adf22 . 24 & almaj221737.43 + 001723.0 & 67 & 4.1 & 0.56 @xmath79 0.08 & full / lores + adf22 . 25 & almaj221733.81 + 001656.5 & 72 & 4.1 & 0.44 @xmath79 0.08 & full / lores + adf22 . 26 & almaj221733.19 + 001752.7 & 82 & 4.1 & 0.63 @xmath79 0.07 & full / lores + adf22 . 27 & almaj221737.18 + 001832.0 & 85 & 4.0 & 0.25 @xmath79 0.06 & full / lores + adf22 . 28 & almaj221732.50 + 001729.5 & 78 & 4.0 & 0.48 @xmath79 0.08 & full / lores + adf22 . 29 & almaj221738.35 + 001710.8 & 109 & 4.0 & 1.12 @xmath79 0.11 & full / lores + adf22 . 30 & almaj221732.19 + 001642.0 & 63 & 4.7 & 0.29 @xmath79 0.09 & deep / hires + adf22 . 31 & almaj221734.97 + 001527.6 & 64 & 4.6 & 0.41 @xmath79 0.12 & deep / hires + adf22 . 32 & almaj221734.08 + 001632.6 & 59 & 4.5 & 0.35 @xmath79 0.12 & deep / hires cccccccccccc laes + 001 & 22:17:33.10 + 00:18:29.0 & 3.090 & ly@xmath17 & ( m05 ) & + 002 & 22:17:35.61 + 00:18:00.2 & 3.091 & ly@xmath17 & 016 ( n13 ) & + 003 & 22:17:31.73 + 00:16:06.9 & 3.101 & ly@xmath17 & 023 ( n13 ) & + 004 & 22:17:34.17 + 00:16:09.7 & 3.096 & ly@xmath17 & 024 ( n13 ) & + 005 & 22:17:36.74 + 00:16:28.8 & 3.091 & ly@xmath17 & 025 ( n13 ) & + 006 & 22:17:31.80 + 00:17:17.9 & 3.088 & ly@xmath17 & 028 ( n13 ) & + 007 & 22:17:33.63 + 00:17:15.1 & 3.092 & ly@xmath17 & 031 ( n13 ) & + 008 & 22:17:34.77 + 00:15:41.3 & 3.099 & ly@xmath17 & 038 ( n13 ) & + 009 & 22:17:35.97 + 00:16:30.2 & 3.094 & ly@xmath17 & 045 ( n13 ) & + 010 & 22:17:34.70 + 00:16:33.4 & 3.090 & ly@xmath17 & 053 ( n13 ) & + 011 & 22:17:34.10 + 00:15:40.2 & 3.101 & ly@xmath17 & 061 ( n13 ) & + 012 & 22:17:31.24 + 00:17:32.1 & 3.084 & ly@xmath17 & 072 ( n13 ) & + & & 3.0845 & [ oiii]@xmath545007 & 072 ( e14 ) & + 013 & 22:17:37.68 + 00:16:48.3 & 3.090 & ly@xmath17 & 078 ( n13 ) & + & & 3.0870 & [ oiii]@xmath545007 & 078 ( e14 ) & + 014 & 22:17:35.44 + 00:16:47.6 & 3.087 & ly@xmath17 & 082 ( n13 ) & + & & 3.0873 & [ oiii]@xmath545007 & 082 ( e14 ) & + 015 & 22:17:36.14 + 00:15:40.7 & 3.095 & ly@xmath17 & 091 ( n13 ) & + 016 & 22:17:31.14 + 00:16:42.9 & 3.096 & ly@xmath17 & 111 ( n13 ) & + 017 & 22:17:32.72 + 00:15:54.2 & 3.096 & ly@xmath17 & 112 ( n13 ) & + 018 & 22:17:33.46 + 00:17:01.2 & 3.093 & ly@xmath17 & 115 ( n13 ) & + 019 & 22:17:32.84 + 00:16:48.8 & 3.092 & ly@xmath17 & 130 ( n13 ) & + + lbgs + 001 & 22:17:31.49 + 00:16:31.2 & 3.098 & ly@xmath17 & m25 ( n13 ) & + 002 & 22:17:31.66 + 00:16:58.0 & 3.094 & ly@xmath17 & m28 , 012 ( s03 ; n13 ) & + & & 3.0902 & [ oiii]@xmath545007 & 012 ( e14 ) & + 003 & 22:17:36.87 + 00:17:12.4 & 3.099 & ly@xmath17 & m31 ( n13 ) & + 004 & 22:17:33.80 + 00:17:57.2 & 3.084 & ly@xmath17 & m34 ( n13 ) & + 005 & 22:17:37.66 + 00:18:20.9 & 3.086 & ly@xmath17 ( abs ) & c50 ( s03 ) & + + @xmath20-band selected galaxies + 001 & 22:17:37.1 + 00:17:12.4 & 3.0899 & [ oiii ] @xmath545007 & ( k15 ) & adf22.8 + 002 & 22:17:37.3 + 00:16:30.7 & 3.0888 & [ oiii ] @xmath545007 & ( k15 ) & + 003 & 22:17:36.5 + 00:16:22.6 & 3.0945 & [ oiii ] @xmath545007 & ( k15 ) & adf22.9 + 004 & 22:17:31.8 + 00:16:06.3 & 3.0981 & [ oiii ] @xmath545007 & ( k15 ) & + 005 & 22:17:32.0 + 00:16:55.5 & 3.0909 & [ oiii ] @xmath545007 & ( k15 ) & adf22.12 + 006 & 22:17:37.3 + 00:18:23.2 & 3.0851 & [ oiii ] @xmath545007 & k15a ( k15 ; k16 ) & + 007 & 22:17:36.8 + 00:18:18.2 & 3.0854 & [ oiii ] @xmath545007 & k15b ( k15 ; k16 ) & adf22.16 + 008 & 22:17:37.1 + 00:18:17.9 & 3.0774 & [ oiii ] @xmath545007 & k15d ( k15 ; k16 ) & + 009 & 22:17:37.1 + 00:18:22.4 & 3.0925 & [ oiii ] @xmath545007 & k15e ( k15 ; k16 ) & adf22.11 + 010 & 22:17:36.9 + 00:18:38.0 & 3.0866 & [ oiii ] @xmath545007 & k15f ( k15 ; k16 ) & + + x - ray sources + and labs + 001 & 22:17:32.0 + 00:16:55.6 & 3.091 & [ oiii ] @xmath545007 & 114 ( l09 ) , lab12 ( m04 ; g09 ) & adf22.12 + 002 & 22:17:32.2,+00:17:36.0 & 3.097 & [ cii ] 158@xmath4 m & 116 ( l09 ) & adf22.7 + 003 & 22:17:32.4,+00:17:43.9 & 3.092 & co(3 - 2 ) & 120 ( l09 ) & adf22.1 + 004 & 22:17:35.8,+00:15:59.1 & 3.089 & ly@xmath17 & 139 ( l09 ) , lab14 ( m04 ; g09 ) & adf22.6 + 005 & 22:17:36.5,+00:16:22.6 & 3.084 & ly@xmath17 & 140 ( l09 ) , qso ( s98 ) & adf22.9 + 006 & 22:17:37.0,+00:18:20.8 & 3.091 & co(9 - 8 ) , [ cii ] 158@xmath4 m & 142 ( l09 ) & adf22.4 + 007 & 22:17:37.3,+00:16:30.7 & 3.0888 & [ oiii ] @xmath545007 & 144 ( l09 ) & + 008 & 22:17:37.3,+00:18:23.5 & 3.0851 & [ oiii ] @xmath545007 & 145 ( l09 ) & , j. p. , waters , b. , schiebel , d. , young , w. , & golap , k. 2007 , in astronomical society of the pacific conference series , vol . 376 , astronomical data analysis software and systems xvi , ed . r. a. shaw , f. hill , & d. j. bell , 127

we have carried out calculations of the triple - differential cross section for one - photon double ionization of molecular hydrogen for a central photon energy of ev , using a fully _ ab initio _ , nonperturbative approach to solve the time - dependent schrdinger equation in prolate spheroidal coordinates . the spatial coordinates and are discretized in a finite - element discrete - variable representation . the wave packet of the laser - driven two - electron system is propagated in time through an effective short iterative lanczos method to simulate the double ionization of the hydrogen molecule . for both symmetric and asymmetric energy sharing , the present results agree to a satisfactory level with most earlier predictions for the absolute magnitude and the shape of the angular distributions . a notable exception , however , concerns the predictions of the recent time - independent calculations based on the exterior complex scaling method in prolate spheroidal coordinates [ phys . rev . a * 82 * , 023423 ( 2010 ) ] . extensive tests of the numerical implementation were performed , including the effect of truncating the neumann expansion for the dielectronic interaction on the description of the initial bound state and the predicted cross sections . we observe that the dominant escape mode of the two photoelectrons dramatically depends upon the energy sharing . in the parallel geometry , when the ejected electrons are collected along the direction of the laser polarization axis , back - to - back escape is the dominant channel for strongly asymmetric energy sharing , while it is completely forbidden if the two electrons share the excess energy equally .

1101.1607

a measurement of the complete breakup of the atomic helium target by xuv radiation was achieved over @xmath3 years ago @xcite . since then , rapid developments in strong xuv light sources and momentum imaging techniques have made it possible to record all the reaction fragments , nuclei and electrons , in double photoionization of the simplest two - electron hydrogen / deuterium molecule by one - photon absorption @xcite , and most recently also for two - photon absorption @xcite . for the double ionization of h@xmath4 by single - photon absorption , only randomly oriented molecules were investigated in earlier experiments ( e.g. @xcite ) . using fixed - in - space " molecules , more recent experimental efforts include the measurements of energy- and angle - resolved differential cross sections by weber _ et al . _ for either equal energy sharing @xcite or asymmetric energy sharing @xcite , and by gisselbrecht _ @xcite for equal energy sharing at a photon energy of @xmath5 ev . these experimental studies were at least partially stimulated by the goal of understanding the similarities and differences between the hydrogen molecule and its atomic counterpart helium . however , all recorded fully differential cross sections to date suffer from some experimental uncertainties regarding the alignment angle of the molecule with respect to the polarization vector of the laser and the emission angles of the photoelectrons . from a theoretical point of view , the hydrogen molecule exhibits a significant complexity compared to helium and , therefore , provides an enormous challenge to a fully _ ab initio _ description inherent in a multi - center , multi - electron system . the single - center convergent close - coupling method was used to model the double ionization of h@xmath4 by kheifets and bray @xcite . later mccurdy , rescigno , martn and their collaborators @xcite implemented a formulation based on time - independent exterior complex scaling ( ecs ) in spherical coordinates , with the origin of the coordinate system placed at the center of the molecule , to treat the double photoionization at a photon energy of @xmath0 ev . the radial coordinates of the two electrons are measured from the center , and the radial parts of the wave function were either expanded in @xmath6-splines or using a finite - element discrete - variable representation ( fe - dvr ) . the time - dependent close - coupling ( tdcc ) method @xcite , again in spherical coordinates , was also extended to calculate the triple - differential cross section ( tdcs ) for double photoionization of the h@xmath4 molecule . while the agreement between the published tdcss from the ecs @xcite and tdcc @xcite calculations is basically acceptable , noticeable discrepancies remained for a few particular geometries . in the parallel geometry , for instance , where the molecular axis @xmath7 is chosen along the laser polarization vector @xmath8 , the coplanar tdcs predictions from the ecs and tdcc calculations differ by up to @xmath9 percent when one of the electrons ( we will refer to it as the `` fixed electron '' below ) is observed along the direction perpendicular to the @xmath8-axis . in some other cases , there exists a noticeable wing " structure in the published tdcc predictions for equal energy sharing . additional tdcc calculations @xcite suggest that the agreement can be systematically improved , albeit the above - mentioned discrepancy still exists at a somewhat reduced level . another independent approach @xcite to this problem is the very recent time - independent ecs treatment , formulated as in the current work in prolate spheroidal coordinates . quite surprisingly , the results of that calculation differed from both the earlier ecs and also the tdcc predictions , both obtained in spherical coordinates . specifically , the ecs prolate spheroidal calculations showed differences from the earlier spherical coordinate calculation for the tdcss , at a level of about @xmath10 depending on the details of the energy sharing . as will be demonstrated below , we have gone to considerable lengths in an attempt to resolve these discrepancies . however , significant differences between the present results and those of the ecs @xcite still remain . both the ecs @xcite and the tdcc @xcite calculations made some attempt to deal with the experimental uncertainties in the scattering angles . given the experimental uncertainties and the differences in the previous calculations , however , it appeared worthwhile to investigate the computational effort required to obtain accurate before averaging over any experimental acceptance angles . consequently , the present calculation represents an independent implementation of the time - dependent fe - dvr approach in prolate spheroidal coordinates . as in the other approaches mentioned above , the internuclear separation ( @xmath11 ) was held fixed at its equilibrium distance of @xmath12 bohr . the two - center prolate system , with the foci located on the nuclei , provides a suitable description for the two - center characteristics of the h@xmath4 molecule . the formulation of the schrdinger equation in prolate spheroidal coordinates for diatomic molecules is not new . the pioneering work of bates , pik , and poots @xcite for the h@xmath13 ion , which is exactly separable in prolate spheroidal coordinates , already revealed the appealing features of the prolate system . in particular , the electron - nuclear interaction is rendered benign in this coordinate system . a partial list of recent applications of prolate spheroidal coordinates to diatomic molecules can be found in @xcite . as has been demonstrated in a number of recent publications , a grid - based approach provides a very appropriate description of laser - driven atomic and molecular physics when combined with an efficient time - propagation method such as the short iterative lanczos ( sil ) method @xcite . in the present work , we employ the fe - dvr / sil approach in prolate spheroidal coordinates to study the correlated response of a two - electron molecule in the double ionization process . the remainder of this manuscript is organized as follows . after presenting the hamiltonian of the hydrogen molecule in section [ sec.2 ] and providing some details about the discretization of the system in an fe - dvr basis in sec . [ sec.3 ] , the solution of the two - center poisson equation is presented in sec . [ sec.4 ] . this is followed by a description of the procedure for extracting the cross sections of interest in sec . [ sec.5 ] . the results are presented and discussed in sec . [ sec.6 ] , before we finish with a summary in sec . [ sec.7 ] . the prolate spheroidal coordinates with the two foci separated by a distance @xmath11 are defined by @xmath14 and the azimuthal angle @xmath15 . here @xmath16 and @xmath17 are the distances measured from the two nuclei , respectively . these coordinates are specified in the ranges @xmath18 , @xmath19 $ ] , and @xmath20 $ ] . the volume element is @xmath21 . according to the asymptotic behaviors as @xmath22 , @xmath1 and @xmath2 approach @xmath23 and @xmath24 , respectively , where @xmath25 and @xmath26 are the standard spherical coordinates . consequently , @xmath1 is the quasi - radial " coordinate , while @xmath2 is quasi - angular " . the hamiltonian of a single electron , @xmath27 , ( @xmath28 for the two electrons in h@xmath4 below ) , is written as @xmath29 -\frac{4\xi_q}{r(\xi_q^2-\eta_q^2)}.\end{aligned}\ ] ] we solve the time - dependent schrdinger equation ( tdse ) of the laser - driven h@xmath4 molecule ( with two electrons ) in the dipole length gauge : @xmath30\psi(1,2,t).\ ] ] here @xmath31 is the coordinate of the @xmath32-th electron measured relative to the center of the molecule and @xmath33 is the interelectronic distance . without loss of generality , we choose the molecular axis along the @xmath34 axis , and the plane formed by the molecular axis and the polarization vector as the @xmath35 plane . generally , we can decompose the polarization vector into its two components , @xmath36 , where @xmath37 is the angle between the @xmath38 and @xmath39 axes , and @xmath40 and @xmath41 are the unit vectors along the @xmath34 and @xmath42 axes , respectively . the dipole interaction is therefore given as @xmath43,\end{aligned}\ ] ] where the rectangular coordinates @xmath42 and @xmath34 and the prolate spheroidal coordinates are related through @xmath44 we expand the wave function for the h@xmath4 molecule in the body - frame as @xmath45 here @xmath46 is the angular function , where @xmath47 and @xmath48 denote the magnetic quantum numbers of the two electrons along the molecular axis . next , @xmath49 is expanded in a product of normalized radial " @xmath50 and angular " @xmath51 dvr bases : @xmath52 note that the basis is not symmetrized with respect to the coordinates of the two electrons . since we always begin the calculation with a properly symmetrized initial state , however , that symmetry will be preserved in the calculation . to discretize this partial differential equation , we employ the fe - dvr approach for both the @xmath1 and @xmath2 variables @xcite . if we normalize the dvr bases according to @xmath53 respectively , the overall @xmath54 dvr basis is not normalized with respect to the volume element in the prolate coordinate system . this is corrected by defining the two - electron basis @xmath55 which satisfies the desired normalization @xmath56 to expand @xmath49 . specifically , we have @xmath57 introducing the normalized @xmath54 dvr basis eliminates the complexities of matrix operations related to the overlap matrix at each time , and hence makes the standard sil algorithm directly applicable to study the temporal response of the molecule to laser pulses . in our current implementation of the fe - dvr approach for the time - dependent wave function in prolate spheroidal coordinates , we have chosen to work directly in the fe - dvr basis . this differs from what is usually done for atoms in spherical coordinates , where spherical harmonics are used for the angular variables and an fe - dvr for the radial coordinates . consequently , the boundary conditions in prolate spheroidal coordinates require some more discussion . analyzing the asymptotics reveals that in the region near the boundaries of @xmath58 and @xmath59 , which correspond to the molecular axis , the single - electron wave function behaves like @xmath60 . this indicates that the physical wave function is finite for @xmath61 , whereas it goes to zero in the region close to the molecular axis for @xmath62 . more importantly , the behavior of the wave function for odd @xmath63 contains a square - root factor , giving a decidedly nonpolynomial behavior to the wave function that is impossible to capture in a straightforward fashion using a dvr basis . the former problem is readily treated by using a gauss - radau quadrature in the first dvr element for @xmath1 , where only the right - most point is constrained to lie on the boundary between the first and second finite element . the volume element ensures that the integrand is well behaved near the end points and makes it unnecessary to invoke a separate quadrature for different @xmath64 values . for all the other elements , a gauss - lobatto quadrature is employed . this allows us to make the fe - dvr basis continuous everywhere and to satisfy the @xmath63-dependent boundary condition . to overcome the nonanalytic behavior of the basis for odd @xmath64 , bachau and collaborators @xcite explicitly factored out the @xmath60 part before the wave function was expanded in terms of @xmath6-splines in the discretization approach . we have adopted a similar idea in our fe - dvr treatment of the two - center problem to achieve much faster convergence , as was also done in ref . @xcite . for the case of even @xmath63 , no changes need to be made to define the dvr basis , i.e. , the normalized basis is written as @xmath65 for odd @xmath63 , however , we define the dvr basis as @xmath66 and @xmath67 here @xmath68 and @xmath69 are the weight factors related to the dvr bases @xmath70 and @xmath71 , respectively . the goal of using a unique set of mesh points , which are @xmath63-independent , to discretize the @xmath54 coordinates has now been achieved in this scheme . the same technique was employed in recent calculations of one- and two - photon double ionization of h@xmath4 @xcite . in principle , it is also possible to introduce the factors into the dvr bases to circumvent the difficulties related to the nonanalytic behavior near the boundary . however , this results in an @xmath63-dependence of the dvr bases and quadrature points . this , in turn , leads to a number of unnecessary complications in the practical implementation of the computational methodology . one might argue that an discretization procedure could be useful for a system in which the magnetic quantum number @xmath64 is conserved . an example is the h@xmath13 ion in external magnetic fields along the molecular axis @xcite . however , that is not the situation in the current calculation . similar to the expansion of the electron - electron interaction in terms of spherical coordinates , a counterpart exists in prolate spheroidal coordinates @xcite through the neumann expansion @xmath72 where @xmath73 . the two nuclei are located at @xmath74 along the @xmath34 axis and @xmath75 . both the regular @xmath76 and irregular @xmath77 legendre functions @xcite , which are defined in the region @xmath78 , are involved in the expansion as the radial " part , while the angular " part is only related to @xmath79 . note that we chose to work in terms of an un - normalized angular " basis rather than the usual spherical harmonics . the matrix elements of @xmath80 in a traditional basis , for example , a or slater - type basis , can be computed through the well - known mehler - ruedenberg transformation @xcite . due to the discontinuous derivative along the line of @xmath81 in the neumann expansion , the straightforward computation of the matrix element of @xmath80 , using the value of this interaction potential at the mesh points , is very slowly convergent . we seek a more robust representation of the @xmath80 matrix which retains both the underlying gauss quadrature and the dvr property of all potentials being exactly diagonal with respect to the highly localized dvr basis . in the following , we use the simplified notation @xmath82 to denote the basis . essentially , we need the integral @xmath83 after integrating over @xmath84 and @xmath85 , the matrix element of @xmath80 can be written as @xmath86 where the selection rule @xmath87 has been used . hence @xmath64 is uniquely determined for a given pair of angular partial waves . above we introduced the reduced ( @xmath88 ) integral @xmath89 since @xmath63 is fixed in the above equation , we omitted it in @xmath90 and will do so in the related quantities below as well . it is now worthwhile to define the two electron densities @xcite : @xmath91 after truncating the radial integral to the edge of the box , @xmath92 , this yields @xmath93 here a convention for the volume element was made in such a way that , for any function @xmath94 , we define @xmath95 to simplify the notation . most importantly , the function @xmath96 is defined by @xmath97 instead of evaluating the above integrals directly , we solve the differential equation satisfied by @xmath96 . as will become apparent later , this equation can be shown to be the radial " poisson equation in the prolate spheroidal coordinate system . the differential equations satisfied by the legendre functions @xmath76 and @xmath77 suggests that we introduce the operator @xmath98 for given quantum numbers @xmath99 and @xmath64 . this is the one - dimensional laplacian operator in the @xmath1 coordinate . after some algebra , we obtain @xmath100 and @xmath101 here the wronskian of the legendre functions @xmath76 and @xmath77 is given by @xmath102 consequently , we obtain @xmath103 this is the second - order inhomogeneous poisson equation satisfied by @xmath104 with the source " term given by @xmath105 after carrying out the integral over @xmath106 via gauss quadrature , we recast the source term as @xmath107 as one might expect , the inhomogeneous equation reduces to the homogeneous one if @xmath108 . the solution to the poisson equations ( [ poissoneq ] ) and ( [ boundcond ] ) can be uniquely determined by enforcing the boundary conditions @xmath109 at @xmath58 and @xmath110 at @xmath111 , respectively . there are two important points to realize , namely : first , on the right - hand boundary , the function @xmath96 assumes a nonzero value , which is given by eq . ( [ bdr ] ) , for all possible @xmath63 values . its asymptotic behavior relies on the function @xmath77 at large @xmath1 , which behaves like @xmath112 . this indicates that it is _ nonzero _ generally , although it could be small at the large @xmath92 values used in practical calculations . second , on the left - hand boundary , the situation depends on the value of @xmath63 . @xmath104 takes a _ value if @xmath61 , while it becomes _ zero _ if @xmath62 . following the philosophy employed to handle the spherical case @xcite , we first seek a solution , @xmath113 , to the poisson equations ( [ poissoneq])-([bdr ] ) that satisfies the _ zero_-value boundary condition at @xmath92 by using exactly the same @xmath1 mesh points as those for the wave functions . in other words , we have @xmath114 with @xmath115 and @xmath116 . after substituting the dvr expansion @xmath117 of the solution into the differential equation , we obtain a system of linear equations for the unknown coefficients @xmath118 : @xmath119 the matrix @xmath120 is defined by its elements @xmath121f_{\mu'}(\xi).\end{aligned}\ ] ] therefore , the coefficient @xmath122 can formally be written as @xmath123^{-1}_{\mu i}}{\sqrt{\omega_\xi^i } } ( -1)^{|m|}\frac{(l\!+\!|m|)!}{(l\!-\!|m|)!}a^3\delta_{ii'}\delta_{kk ' } ( \xi_i^2-\eta_k^2)p _ { l } ^ { |m| } ( \eta_k ) , \label{cpc}\ ] ] where @xmath124^{-1}$ ] denotes the inverse of the matrix @xmath125 . in this case @xmath113 fulfills the left - hand boundary condition @xmath126 , and so does @xmath127 . recall , however , that its right - hand boundary condition differs from those of @xmath128 . this suggests that the final answer to the function @xmath96 can be constructed as @xmath129 , i.e. , we add the difference function @xmath130 to @xmath113 . the function @xmath130 is also a solution to the homogeneous poisson equation , subject to the boundary condition @xmath131 and @xmath132 . after writing it as a linear combination of @xmath76 and @xmath77 , and imposing the boundary conditions , @xmath130 takes the form @xmath133 we finally arrive at @xmath134^{-1}_{\mu i } f_{\mu}(\xi ) + \delta_{ii'}\delta_{kk'}(\xi_i^2-\eta_k^2)p_{l}^{|m|}(\xi_i ) \\ \times & p_{l}^{|m|}(\eta_k ) \frac{q_{l}^{|m|}(\xi_{\rm max})}{p_{l}^{|m|}(\xi_{\rm max } ) } p_{l}^{|m|}(\xi ) . \label{solutionu } \end{split}\end{aligned}\ ] ] at this point , the dvr version of the solution @xmath96 is ready for all possible values of @xmath63 , either @xmath62 or @xmath61 . substituting eq . ( [ solutionu ] ) into eq . ( [ intei ] ) allows us to obtain the kernel integral , @xmath135^{-1}_{ji}p_{l}^{|m|}(\eta_k)+ \\ + & p_{l}^{|m|}(\xi_i)p_{l}^{|m|}(\xi_j)p_{l}^{|m|}(\eta_k ) \frac{q_{l}^{|m|}(\xi_{\rm max})}{p_{l}^{|m|}(\xi_{\rm max})}\bigg ] . \end{split}\end{aligned}\ ] ] the matrix element of @xmath80 can finally be written as @xmath136^{-1}_{ji}+(-1)^{|m| } \frac{(l-|m|)!}{(l+|m| ) ! } p_{l}^{|m|}(\xi_i ) p_{l}^{|m|}(\xi_j ) \frac{q_{l}^{|m|}(\xi_{\rm max})}{p_{l}^{|m|}(\xi_{\rm max } ) } \bigg ] , \end{split } \label{eq:1/r12}\end{aligned}\ ] ] where we truncated the @xmath99 summation in the neumann expansion to @xmath137 . the above equation can be converted to the normalized @xmath54 bases with the help of eq . ( [ eq : overall - basis ] ) . this results in a diagonal representation of the matrix elements of the electron - electron coulomb interaction and thus considerably simplifies the fe - dvr discretization procedure . the above treatment of the @xmath80 matrix was successfully applied to the two - photon double ionization of h@xmath4 @xcite . the implementation of this representation will be illustrated below . the time - dependent laser - driven electronic wave packet in the hydrogen molecule is obtained by solving the tdse on the @xmath54 grid . launched from the previously determined ground state , the time evolution of the system is achieved by using our recently developed sil method @xcite . the ground state is determined by relaxing the system in imaginary time from an initial guess of the wave function on the grid points . at each time step we only need to generate the values of the discretized wave function on the selected grid points . if desired , the information at arbitrary points within the spatial box can be obtained from the interpolation procedure in terms of the dvr bases . a few remarks seem appropriate regarding the efficient implementation of the sil algorithm . the highest energy , @xmath138 , which essentially depends on the smallest separation between the @xmath54 mesh points and also on the maximum values of @xmath139 and @xmath140 , determines the largest time step @xmath141 for the propagation in real time . typically , @xmath138 is about @xmath142 atomic units ( a.u . ) in our calculations . although the chances of electrons populating states with such high energies are practically negligible for short time scales of the laser - molecule interaction , we generally require @xmath143 in order to resolve the most rapid oscillations in the time evolution . this means that at least a few time steps are needed during one period of @xmath144 . we refer readers to refs . @xcite for further details and discussions behind the sil method . in order to ensure that the double - ionization wave packet is sufficiently far away from the nuclei , and also that the two photoelectrons are well separated , we allow the system to evolve for a few more cycles in the field - free hamiltonian , i.e. , after the laser pulse has died off . this is the wave packet we use to extract the physical information . the ionization probabilities and the corresponding cross sections are extracted by projecting the time - dependent wave packet onto uncorrelated two - electron continuum states satisfying the standard incoming boundary conditions . the latter states of h@xmath4 are constructed from the one - electron continuum state of the h@xmath13 ion described in the following subsection . the field - free wave function @xmath145 of the one - electron molecular ion is completely separable in prolate spheroidal coordinates . for a given , and conserved , magnetic quantum number @xmath64 , the wave function takes the form @xmath146 , where the azimuthal dependence , is given by @xmath147 . the radial " part @xmath148 and the angular " part @xmath149 of the wave function satisfy the equations @xmath150t_{mq}(\xi)=0 \label{eq : xi}\ ] ] and @xmath151\xi_{mq}(\eta)=0 , \label{eq : eta}\ ] ] respectively . here @xmath152 for the continuum state whose momentum vector has the magnitude @xmath153 . in addition , we need to introduce another quantum number @xmath32 , which denotes the number of nodes of @xmath154 in the region @xmath155 $ ] , to label the states , and finally the separation constant @xmath156 . when the angular function @xmath157 is discretized in terms of the relevant dvr bases , a few spurious " solutions might be encountered . this is caused by the residual errors associated with the gauss quadratures . consequently , we expand the angular function , or spheroidal harmonics " function @xmath158 with @xmath159 instead in terms of spherical harmonics . these functions are normalized according to @xmath160 after obtaining the separation constant @xmath156 by solving eq . ( [ eq : eta ] ) , the radial " function @xmath148 is again expanded in terms of the dvr bases . the last dvr point at @xmath111 needs to be kept for the continuum state . asymptotically , the radial function behaves like @xmath161 \label{eq : asym}\ ] ] as @xmath162 . here @xmath163 is the two - center coulomb phase shift . the normalization factor on either the energy or the momentum scale and the coulomb phase shift can be determined by matching the numerical solution of @xmath148 according to its asymptotic behavior given in eq . ( [ eq : asym ] ) . the plane wave in prolate spheroidal coordinates can be written as @xcite @xmath164 where @xmath165 $ ] in the asymptotic region . note that @xmath166 and @xmath167 are related to the directions of @xmath168 and @xmath169 in spherical coordinates through @xmath170 . the partial - wave expansion of the plane wave @xmath171 reminds us that the two - center coulomb wave satisfying the incoming boundary condition can be expanded as @xmath172 this function is normalized in momentum space according to @xmath173 , provided the asymptotic solution in eq . ( [ eq : asym ] ) is satisfied . uncorrelated two - electron continuum states with total spin angular momentum @xmath174 ( @xmath175 in our case ) can generally be constructed as follows : @xmath176 . \label{eq : two - continuum}\end{aligned}\ ] ] with the help of eq . ( [ eq : con - one ] ) , its partial - wave representation can be written as @xmath177 . \label{eq : two - conti}\end{aligned}\ ] ] here we introduced @xmath178 by representing the radial and angular parts on the @xmath54 grid points : @xmath179 @xmath180 here the exchange symmetry @xmath181 is satisfied . it has been demonstrated @xcite that using uncorrelated two - electron continuum states is a good approximation in a time - dependent propagation approach , provided the two ejected electrons are well separated from each other . the probability amplitude of double ionization is then given by @xmath182 where @xmath183 here the exchange symmetry @xmath184 is satisfied . we also see that the probability amplitude formulated in prolate spheroidal coordinates takes a similar form as for the atomic case in spherical coordinates . however , a subtle difference from the atomic case is worth pointing out . strictly speaking , the spheroidal harmonics involved in the probability amplitude generally depend on the magnitude of the momenta @xmath185 and @xmath186 , in addition to their directions . the energy sharing of the two photoelectrons can be specified by introducing the hyperangle @xmath187 . this describes the double - ionization reaction with kinetic energies @xmath188 and @xmath189 for the two ionized electrons , respectively . here @xmath190 is the available excess energy above the double - ionization threshold . in the present work , specifically , @xmath191 ev for absorption of a photon . for double ionization by one - photon absorption , the triple differential cross section with respect to one of the kinetic energies and the two solid angles @xmath192 and @xmath193 can be extracted using the same formalism as in the corresponding he case @xcite , i.e. , @xmath194 here @xmath195 and @xmath196 are the central photon energy and the peak intensity of the laser pulse , respectively , while @xmath197 denotes the _ effective _ interaction time between the temporal electric laser field and the electrons in the one - photon absorption process . for a laser pulse of time duration @xmath198 with a sine - squared envelope for the field amplitude , @xmath199 . note that @xmath200 corresponds to the special case of the generalized @xmath201-photon effective interaction time @xmath202 @xcite for a one - photon reaction . generalized cross sections for two - photon double ionization of the hydrogen molecule were extracted in the same formalism @xcite . in the tdcc treatment @xcite , different strategies were employed to describe the linear one - photon and the nonlinear two - photon double ionization processes of atoms and molecules . for the one - photon case , the cross sections were obtained through the time derivative of the double - ionization probability , @xmath203 . the laser field does not need to be turned off in this case . on the other hand , a true laser pulse was used for the two - photon case and an effective time , defined as the time integral under a flat - top pulse with a smooth turn - on and turn - off , was introduced . in the present work , we employed a unified formulation through an effective interaction time for both one - photon and multi - photon ionization in laser pulses . for the one - photon double ionization initialized from the lowest @xmath204 state , the two ejected electrons can only populate the final @xmath205 and @xmath206 continuum states , with the specifics depending on the relative orientation of the molecular axis and the laser polarization vector . consequently only partial waves with _ ungerade _ parity [ i.e. , @xmath207 need to be included in eq . ( [ eq : tdcs ] ) . for the nonsequential double - ionization process induced by one- or two - photon absorption , electronic correlation plays a dominant role , as the two photoelectrons must share the available excess energy @xmath190 . double ionization by a single photon would not occur at all if the two - electron atom or molecule were approximated by an independent - electron model . therefore , the quality of the description of electron - electron correlation in a laser - driven system is crucially important for accurate results to be obtained . the coulomb interaction between the two electrons has to be described in a consistent manner for both the initial bound state and the time - evolved wave packet . before we go any further , it is worth discussing how we prepare the initial @xmath204 state at the equilibrium distance of @xmath208 bohr . as seen from eq . ( [ eq:1/r12 ] ) , the magnetic quantum number in the neumann expansion of the matrix element of @xmath80 is uniquely determined by the angular bases . however , this is not the case for the index @xmath99 , if we choose to discretize the coordinate @xmath2 , rather then expanding that part of the wave function into spherical harmonics . in practice , the summation over @xmath99 must be truncated at a finite value of @xmath137 . in principle , the higher - order expansion terms always guarantee well - converged results . however , as mentioned earlier , we approximate the relevant @xmath2-integrals by using gauss - legendre quadrature . this is the price we have to pay for making the dielectronic coulomb potential diagonal in the dvr bases . as a consequence , we need to determine how the approximation introduced in the @xmath2-integrals for the two - electron integrals affects the results for the cross sections of interest . to answer this question , we first investigate the dependence of the energy obtained for the initial @xmath209 state on the value of @xmath137 used in the neumann expansion . figure [ fig : energy ] shows the variation of the initial - state electronic energy of the hydrogen molecule with respect to @xmath137 , obtained with a @xmath1 setup of ten elements in the region of @xmath210 , five in the region @xmath211 and another five in @xmath212 . each element , in turn , contains five dvr points to further discretize the configuration space . furthermore , we employ dvr points for @xmath2 . for a given number of @xmath2 mesh points ( @xmath213 ) and @xmath214 , we observe that the resulting energy typically exhibits a plateau - like behavior with increasing @xmath137 . for given @xmath213 , when @xmath137 is relatively small , the @xmath2-integral can be computed very accurately by using gauss quadrature . however , when @xmath137 is too large , the numerical errors introduced from the gauss quadrature cause the energy value to fluctuate . this occurs when @xmath137 approaches @xmath215 and is shown by the grey stripes in fig . [ fig : energy ] . in this region of @xmath137 an unphysically _ low _ energy can be produced . beyond that point , the calculated energy increases to the next plateau . ultimately , this is not too surprising , since any gauss quadrature is only reasonably accurate up to a limited polynomial order of the integrand . consequently , if we want to keep more terms in the neumann expansion , we have to increase @xmath213 correspondingly . this finding is further substantiated by the dependence of the energy found for @xmath216 and @xmath217 . the plateaus are indeed extended to the correspondingly larger values of @xmath215 . most importantly , the amplitude of the energy fluctuation is systematically reduced with increasing @xmath213 . the error in the energy is lowered from @xmath218 to @xmath219 and finally @xmath220 a.u . , when @xmath213 increases from @xmath221 to @xmath222 and then @xmath217 for @xmath223 . we obtained the energy at @xmath208 bohr as @xmath224 a.u . for @xmath225 , @xmath226 , and @xmath223 , resulting in a double - ionization potential of @xmath227 ev . keeping the other parameters unchanged , we obtained an energy of @xmath228 a.u . for @xmath216 . the benchmark energy in the literature is @xmath229 a.u . at the same @xmath11 @xcite , after we take out the nucleus - nucleus interaction of @xmath230 a.u . to summarize : unlike for other expansion parameters , it is important to be consistent in the size of the angular quadrature and the largest @xmath137 employed in the neumann expansion in practical calculations , if we discretize the coordinate @xmath2 . however , this provides a way to examine a potential sensitivity of the physical observables of interest ( here the differential cross sections ) to the ground - state wave functions generated by varying @xmath137 and other parameters . this will be further discussed below . before we present our results for the cross sections , let us take a closer look at the survival probability @xmath231 of the aligned h@xmath4 molecule in its ground state @xmath232 . this is shown in fig . [ fig : surv ] . for homonuclear molecules , the independent alignment angle @xmath37 between the molecular axis and the polarization vector can be confined to the region from @xmath233 to @xmath234 . in the xuv regime , we observe that the hydrogen molecule shows a larger probability of being ionized or excited ( i.e. , a lower probability of staying in the initial state ) at the end of the pulse in an aligned geometry . this indicates that the perpendicular component of the temporal electric field exerts more influence on the ionization process due to the larger dipole momentum . interestingly , at the earlier stages of the time evolution ( e.g. , @xmath235 , when the ionized wave packet is driven back by the change in direction of the electric field , the tilted molecule has a larger probability of staying in its ground state . this happens near the various minima in @xmath236 . however , once the electric field has become sufficiently strong ( @xmath237 a.u . ) , the wave packet is driven out and spread into a larger space . this leads to lower minima in @xmath236 for the tilted molecule . when the wave packet is driven back to the nuclear region and therefore has a chance to recombine with the h@xmath13 ion , a maximum in @xmath236 appears . not surprisingly , the parallel geometry always has the largest probability for this to happen . although the wave packet can also be scattered for the untilted molecule in the plane perpendicular to the molecular axis , the probability is undoubtedly larger if the laser electric field is perpendicular to the molecular axis . a similar behavior of h@xmath13 in xuv pulses was observed in ref . @xcite . for most calculations performed in this study , we expose the hydrogen molecule to a laser pulse with a peak intensity of @xmath238 w/@xmath239 . looking at fig . [ fig : surv ] we see that the depletion of the initial ground state can be safely neglected for our typical interaction times . even for @xmath240 , @xmath241 remains very close to unity . the negligible depletion of the ground state suggests that the concept of cross sections is valid and applicable . on the other hand , it also presents a numerical challenge to predict the cross sections accurately from a time - dependent treatment , due to the generally small ionization probability . at first glance , a peak intensity of @xmath238 w/@xmath239 might seem very intense for most atomic and molecular targets . here , however , we consider an xuv rather than an ir pulse . for an xuv pulse with central photon energy of @xmath0 ev , such laser fields definitely fall into the weak - field " regime . the ponderomotive energy in the xuv regime is much smaller than the photon energy of interest . .[tab : parameters ] the discretization and expansion parameters of the h@xmath4 wave function in prolate spheroidal coordinates . here @xmath242 stands for the border between the inner and outer regions in the @xmath1 coordinate , while @xmath92 is the size of the @xmath1 box . in addition , @xmath243 denotes the number of @xmath1 mesh points in each element . the numbers of @xmath1 elements in the inner and outer region are @xmath244 and @xmath245 , respectively . these @xmath1 parameters produce the total number of @xmath1 mesh points @xmath246 . the @xmath1 grid i and @xmath1 grid ii are used to examine the convergence of our results . [ cols="<,^,^,^,^,^,^",options="header " , ] in this work , we are mainly interested in the triple - differential cross section , since it reveals the fine details of possible energy sharings and preferred directions of the ejected electrons in the double - ionization process . given the discrepancies between results from various calculations found in the literature , we carried out comprehensive convergence tests for our predictions of the tdcss . these tests are essentially divided into two groups . the first group concerns the laser parameters , while the second one deals with the discretization and expansion parameters . an example of two different parameter sets for the @xmath1 grid is given in table [ tab : parameters ] and will be further discussed below . in order to obtain a good handle on the sensitivity of the results to the various parameters and the resulting level of convergence " , we try to only vary a single parameter while keeping all others fixed if possible . for the dependence on the laser parameters , we use the @xmath1 grid i combined with @xmath247 . for the tests regarding the discretizations and expansions , the peak intensity of laser was fixed at @xmath238 w/@xmath239 and a time scale of `` @xmath248 '' optical cycles ( o.c . ) was used . here `` @xmath248 '' refers to a laser pulse with a sine - squared envelope for the field amplitude , followed by a field - free propagation . + + + + + + figures [ fig : tdcs - laser ] and [ fig : tdcs - discret ] show the convergence pattern of our tdcs results for asymmetric energy sharing in the parallel geometry ( @xmath249 ) . the energy sharing between electron @xmath250 ( observed at the fixed angle @xmath251 ) and electron @xmath252 ( observed at the variable angle @xmath253 ) is @xmath254 . only the electron that takes away @xmath10 of the excess energy is recorded at fixed positions either parallel or perpendicular to the polarization axis . since the laser pulse is explicitly involved in our time - dependent treatment , we first have to be sure that the extracted cross sections are essentially independent of the laser intensity and the time scales . only then are the calculations of cross sections meaningful . this also allows us to compare the physical information extracted from our time - dependent scenario to that obtained through conventional time - independent treatments , which are effectively equivalent to the weak - field approximation and infinitely " long interaction times . rather than computing the cross sections , it would be more appropriate to consider ionization rates if the cross sections were found to be sensitive to the laser parameters . in fig . [ fig : tdcs - laser ] , we display the dependences of our tdcs results upon the laser parameters . note that the tdcss extracted from @xmath255 w/@xmath239 and @xmath256 w/@xmath239 at fixed time evolution of @xmath248 " cycles are nearly identical and agree with each other to better than the thickness of the line . when we turn to the dependence of time scales at a fixed intensity of @xmath238 w/@xmath239 , we use the same pulse , but allow the system to freely evolve for a few additional cycles to extract the tdcs . this corresponds to the time scale of @xmath257 " o.c . also , we may increase the laser - molecule interaction time , but extract the tdcss at the same cycles of field - free time evolution after the pulse died off . this gives the scenario of @xmath258 " o.c . since the total time durations are the same ( @xmath259 o.c . ) , they allow us to examine the extracted tdcss from different perspectives . the increased interaction time yields a reduced bandwidth of the photon energy , while the longer field - free propagation ensures that the double - ionization wave packet is further away from the nuclear region @xcite . the calculated tdcss indeed show a slight , though in our opinion acceptable , sensitivity to the time scales . not surprisingly , the sensitivity is most visible for the smaller cross sections , when the two ejected electrons travel nearly parallel along the same direction ( c.f . [ fig : tdcs - laser]@xmath260 ) . + + + having confidence in using the current sets of laser parameters , we now turn our attention to the scheme of spatial discretization ( @xmath243 , @xmath92 , @xmath213 ) and the convergence of the expansion ( @xmath137 , @xmath261 ) . the results are displayed in fig . [ fig : tdcs - discret ] . for the discretization parameters , we obtain well - converged tdcss by increasing @xmath262 from @xmath263 to @xmath264 , @xmath265 from @xmath221 to @xmath222 , and extending the spatial box of @xmath92 from @xmath266 to @xmath267 . most importantly , however , we consider two sets of @xmath1 mesh points : @xmath1 grid i and @xmath1 grid ii ( see table [ tab : parameters ] ) . the principal motivation was to see whether or not we can reproduce the much lower tdcs values ( by about @xmath10 compared to the one - center spherical results ) that were recently obtained in an ecs calculation in two - center elliptical coordinates by tao _ et al . _ @xcite . we emphasize that these two grids in the radial " @xmath1 coordinate are completely different regarding both the distribution of the elements and the number of grid points per element . in the @xmath1 grid i , we divide the @xmath1 space into two parts , an inner and an outer region with a border at @xmath268 . we place a narrow span of elements in the inner region , and then wider elements in the outer region . in contrast to that , @xmath1 grid ii does not distinguish between inner and outer regions , i.e. , the elements uniformly span the region from @xmath250 to @xmath92 . the mesh setup in @xmath1 grid ii is the same as that used in ref . @xcite , except for the complex rotation . the @xmath1 grid i has a much denser distribution of mesh points than @xmath1 grid ii . nevertheless , the extracted tdcss from both sets of @xmath1 grids are in excellent agreement with each other , even for the smallest cross sections . this strongly suggests that the results are well converged at least with regard to the @xmath1 grid . both @xmath1 sets are good enough to capture the physics of interest . differences at the @xmath10 level are unlikely to be caused by using different sets of @xmath1 meshes . finally , we discuss the convergence of our results with respect to the expansion parameters , @xmath261 and @xmath137 . as expected for a one - photon process , @xmath223 produced well - converged results . recall the discussion above regarding the ground state , especially how the truncated neumann expansion of @xmath80 in our present fe - dvr implementation affects the initial - state energy and therefore the quality of the wave function . for consistency , we use the same @xmath137 in the real - time propagation and in the ground - state wave function . as seen from figs . [ fig : tdcs - discret]@xmath269 and [ fig : tdcs - discret]@xmath270 , our truncated neumann expansion has little effect on the calculated tdcs values . well - converged results can be obtained even with an inappropriately large value of @xmath271 , which yields a slightly higher energy of the ground state ( c.f . [ fig : energy ] ) . overall , our detailed convergence tests only reveal a very weak sensitivity of the tdcs results to both the time scales and the values of @xmath137 . well - converged tdcs results can be obtained by using either @xmath1 grid i or @xmath1 grid ii combined with @xmath247 . in the production calculations for the tdcss shown in the next subsection , we used the @xmath1 grid i to discretize the two - electron wave packet and a @xmath248 " sine - squared laser pulse with a peak intensity of @xmath238 w/@xmath239 . figures [ fig : tdcs - equal - sharing - ele1 - 00 ] , [ fig : tdcs - equal - sharing - ele1 - 90 ] , and [ fig : tdcs - nonequal - sharing - ele1 - 90 ] display the coplanar tdcss of the aligned hydrogen molecule at equal and asymmetric ( @xmath272 ) energy sharing . the two electrons are detected in the same ( coplanar ) plane defined by the @xmath7 and @xmath8 axes . the angles @xmath251 , @xmath253 , and @xmath37 are all measured with respect to the laser linear polarization axis . we compare our tdcs predictions with those obtained in the time - independent one - center spherical ecs calculation @xcite , the time - independent two - center spheroidal ecs model @xcite , and the time - dependent one - center spherical tdcc approach @xcite . the tdcc numbers were recently recalculated with a bigger box size and differ , in some cases substantially , from those published originally @xcite . except for the recent two - center prolate spheroidal ecs results of tao _ et al . _ @xcite , the agreement between the other three sets of results is very satisfactory . once again , the largest relative differences occur when the cross sections are small ( see figs . [ fig : tdcs - equal - sharing - ele1 - 00]@xmath273 and [ fig : tdcs - equal - sharing - ele1 - 90]@xmath273 ) . using spheroidal coordinates as well , as an illustrative example of their two - center ecs approach , serov and joulakian @xcite recently presented the tdcs at the same photon energy , but only for a single geometry of and for asymmetric energy sharing of @xmath274 . although not shown here , there is again good agreement between their results , vanroose _ et al . _ s one - center spherical ecs numbers @xcite , and our time - dependent fe - dvr predictions . it is also interesting to investigate the dominant escape modes for the various scenarios . these modes are strongly dependent on how the electrons share the excess energy . in an arbitrary geometry ( @xmath275 ) , for example , the back - to - back escape mode @xmath276 is forbidden for equal energy sharing . on the other hand , it becomes the dominant mode for significantly asymmetric energy sharing , including the 20%:80% scenario discussed in the present paper ( see fig . [ fig : tdcs - laser ] ) . these results can be understood from a symmetry analysis @xcite . equal - energy sharing and back - to - back emission is equivalent to @xmath277 . when we consider the exchange and parity operations simultaneously in eq . ( [ eq : two - continuum ] ) , we have @xmath278 . here @xmath279 is the parity for the gerade and ungerade states , respectively . for the singlet double - continuum state with _ parity , we therefore must have at any configuration of @xmath280 and @xmath281 . although the magnitudes of the momenta @xmath282 and @xmath283 are not exactly conserved in the time - dependent picture , the ionization events we collect must satisfy the condition @xmath284 ( because of the @xmath285 function in eq . ( [ eq : tdcs ] ) ) for the equal - energy sharing . this is the reason behind the forbidden back - to - back ( @xmath286 ) escape mode for the equal energy sharing , as we observed in figs . [ fig : tdcs - equal - sharing - ele1 - 00 ] and [ fig : tdcs - equal - sharing - ele1 - 90 ] . since the argument does not involve the relative alignment angles , it is valid for all possible values of @xmath37 . on the other hand , this is not the case when the excess energy is not evenly distributed among the two electrons . indeed , figs . [ fig : tdcs - laser]@xmath287 and [ fig : tdcs - laser]@xmath260 show maxima in the back - to - back emission , thereby illustrating the dramatic change in the dominant escape mode . for equal - energy sharing in the parallel geometry ( @xmath249 ) , the electron - electron coulomb repulsion suggests that the tdcs should be dynamically small if the two electrons travel along the same direction . this is in agreement with the numerically small cross sections ( not exactly zero , however ) at @xmath288 or @xmath289 seen in fig . [ fig : tdcs - equal - sharing - ele1 - 00]@xmath273 . recall that the one - photon double - photoionization process in helium @xcite shares the same property . the back - to - back mode is forbidden for equal - energy sharing , and this can be explained by the above argument . it is one of the similarities between the molecular hydrogen and the atomic helium targets for double photoionization . however , figs . [ fig : tdcs - equal - sharing - ele1 - 00 ] , [ fig : tdcs - equal - sharing - ele1 - 90 ] , and [ fig : tdcs - nonequal - sharing - ele1 - 90 ] also reveal significant molecular effects in the tdcs results . these are missing for the helium atom , not only in the shape of the angular distributions , but also in the magnitudes of the cross sections . depending on the relative orientation @xmath290 , there is interference between the @xmath291 and @xmath292 symmetries in h@xmath4 . a nice example of this effect was presented by reddish _ et al . _ @xcite . even without interference ( i.e. , for @xmath249 or @xmath234 ) , the perpendicular geometry shows much larger magnitudes of the tdcs than the parallel geometry . figure [ fig : tdcs - h2-helium ] shows the three cases of angular distributions : h@xmath4 @xmath293 , h@xmath4 @xmath294 , and he at equal energy sharing . interestingly , in most cases the angular distributions of the perpendicular geometry resemble those of helium . the molecular effect can definitely not be ignored in the parallel geometry for @xmath295 ( c.f . [ fig : tdcs - h2-helium]@xmath287 ) . the _ forward _ escape mode of the second electron is dominant for the h@xmath4 parallel geometry . in contrast , the _ backward _ mode is dominant for the h@xmath4 perpendicular geometry and also for helium . in fig . [ fig : tdcs - out - plane ] , we show the tdcs for noncoplanar geometries . again , all angles are defined with respect to the polarization vector . for the perpendicular geometry , [ fig : tdcs - out - plane]@xmath287 depicts the escape modes for the configuration of @xmath296 ( the fixed electron ) and at the same time @xmath297 in the plane perpendicular to the plane formed by @xmath8 and @xmath7 . figure [ fig : tdcs - out - plane]@xmath298 shows the tdcs after exchanging the directions of @xmath299 and @xmath297 in fig . [ fig : tdcs - out - plane]@xmath287 . with the same directions of @xmath299 and @xmath297 as in fig . [ fig : tdcs - out - plane]@xmath298 , fig . [ fig : tdcs - out - plane]@xmath260 is for the case of the molecular axis orientated along the polarization vector . in the parallel case ( @xmath249 ) , we observe that any escape modes of both electrons ejected in the direction perpendicular to @xmath8 are forbidden . this can be understood by analyzing the spheroidal harmonics in eq . ( [ eq : tdcs ] ) . in this case , only the @xmath205 states can be populated . hence , only partial waves with @xmath300 and @xmath301 can contribute to the cross sections , since @xmath302 at the angles of @xmath303 vanish in spherical coordinates . once again , the agreement between our fe - dvr noncoplanar tdcss and the refined tdcc results @xcite is excellent . we have presented calculations for one - photon double ionization of the hydrogen molecule at a photon energy of @xmath0 ev by solving the time - dependent schrdinger equation in prolate spheroidal coordinates . the triple - differential cross sections were extracted through the projection of the time - dependent wave packet onto uncorrelated two - electron continuum states , a few cycles of field - free time evolution after the laser pulse died off . exhaustive convergence studies of the tdcs results were performed with respect to a number of discretization and expansion parameters , as well as the details of the laser field . these tests provide a strong indication that the results for the triple - differential cross sections presented here are well converged and numerically accurate . excellent agreement was obtained between the current time - dependent results in prolate spheroidal coordinates , those obtained with the ecs approach in spherical coordinates @xcite and , finally , larger tdcc calculations @xcite than those published earlier @xcite . the present calculations do not confirm the significant reduction by about @xmath10 in the tdcs results predicted in recent ecs calculations in the two - center prolate spheroidal coordinates @xcite . furthermore , our results did not show the level of sensitivity to the description of the ground state that was also reported by tao _ et al . _ @xcite . the detailed analysis reported in this study provides a high level of confidence in the present results . we hope that they will be used as benchmarks for comparison in future investigations . tables of these results are available in electronic format from the authors upon request . we thank drs . t. n. rescigno and j. colgan for sending their results in numerical form and for helpful discussions . this work was supported by the nsf under grant phy-0757755 ( xg and kb ) and generous supercomputer resources through the nsf teragrid allocation award tg - phy@xmath304 ( kraken at nics , oak ridge national laboratory ) and by the department of energy allocation award mph@xmath305 ( jaguar at nccs , oak ridge national laboratory ) . without these computational resources it would have been impossible to study the convergence properties of our cross section results on various physical and computational parameters . 99 h. bruning , r. drner , c. l. cocke , m. h. prior , b. krssig , a. s. kheifets , i. bray , a. bruning - demian , k. carnes , s. dreuil , v. mergel , p. richard , j. ullrich , and h. schmidt - bcking , j. phys . b * 31 * , 5149 ( 1998 ) . weber , a. czasch , o. jagutzki , a. mller , v. mergel , a. kheifets , j. feagin , e. rotenberg , g. meigs , m. h. prior , s. daveau , a. l. landers , c. l. cocke , t. osipov , h. schmidt - bcking , and r. drner , phys . . lett . * 92 * 163001 ( 2004 ) . t. weber , a. o. czasch , o. jagutzki , a. k. mller , v. mergel , a. kheifets , e. rotenberg , g. meigs , m. h. prior , s. daveau , a. landers , c. l. cocke , t. osipov , r. dez muio , h. schmidt - bcking , and r. drner , nature ( london ) * 431 * , 437 ( 2004 ) . weber , phd . thesis , universitt frankfurt ( unpublished , 2003 ) . m. gisselbrecht , m. lavolle , a. huetz , p. bolognesi , l. avaldi , d. p. seccombe , and t. j. reddish , phys . lett . * 96 * , 153002 ( 2006 ) . y. h. jiang , a. rudenko , e. plsiat , l. foucar , m. kurka , k. u. khnel , th . ergler , j. f. prez - torres , f. martn , o. herrwerth , m. lezius , m. f. kling , j. titze , t. jahnke , r. drner , j. l. sanz - vicario , m. schffler , j. van tilborg , a. belkacem , k. ueda , t. j. m. zouros , s. dsterer , r. treusch , c. d. schrter , r. moshammer , and j. ullrich , phys . rev . a * 81 * , 021401(r ) ( 2010 ) . j. p. wightman , s. cvejanovi , and t. j. reddish , j. phys . b * 31 * , 1753 ( 1998 ) . a. s. kheifets , phys . a * 71 * , 022704 ( 2005 ) . a. s. kheifets and i. bray , phys . a * 72 * , 022703 ( 2005 ) . w. vanroose , f. martn , t. n. rescigno , and c. w. mccurdy , phys . a * 70 * , 050703(r ) ( 2004 ) . w. vanroose , d. a. horner , f. martn , t. n. rescigno , and c. w. mccurdy , phys . a * 74 * , 052702 ( 2006 ) . w. vanroose , f. martn , t. n. rescigno , and c. w. mccurdy , science * 310 * , 1787 ( 2006 ) . j. colgan , m. s. pindzola , and f. robicheaux , phys . rev . lett . * 98 * , 153001 ( 2007 ) . j. colgan , private communication ( 2010 ) . l. tao , c. w. mccurdy , and t. n. rescigno , phys . rev . a * 82 * , 023423 ( 2010 ) . d. r. bates , u. pik , and g. poots , proc . phys . a * 66 * , 1113 ( 1953 ) . s. barmaki , s. laulan , h. bachau , and m. ghalim , j. phys . b * 36 * , 817 ( 2003 ) . s. barmaki , h. bachau , and m. ghalim , phys . a * 69 * , 043403 ( 2004 ) . y. v. vanne and a. saenz , j. phys . b * 37 * , 4101 ( 2004 ) . g. lagmago kamta and a. d. bandrauk , phys . rev . a * 71 * , 053407 ( 2005 ) . l. tao , c. w. mccurdy , and t. n. rescigno , phys . a * 79 * , 012719 ( 2009 ) . t. j. park and j. c. light , j. chem . * 85 * , 5870 ( 1986 ) . d. j. tannor , in _ introduction to quantum mechanics , a time - dependent perspective _ , chap . 11 , p318 ( university science books , sausalito , california 2007 ) . x. guan , k. bartschat , and b. i. schneider , phys . a * 82 * , 041404(r ) ( 2010 ) . x. guan , b. li , and k. t. taylor , j. phys . b * 36 * , 3569 ( 2003 ) . p. m. morse and h. feshbach , in _ methods of theoretical physics _ , parts i and ii ( mcgraw - hill , 1953 ) . i. a. stegun , in _ handbook of mathematical functions with formulas , graphs , mathematical tables _ , chap . 8 , p331 , eds . m. abramowitz and i. a. stegun ( 1964 ) . e. l. mehler and k. ruedenberg , j. chem . phys . * 50 * , 2575 ( 1969 ) . c. w. mccurdy , m. baertschy , and t. n. rescigno , j. phys . b * 37 * , r137 ( 2004 ) . x. guan , k. bartschat , and b. i. schneider , phys . a * 77 * , 043421 ( 2008 ) . x. guan , c. j. noble , o. zatsarinny , k. bartschat , and b. i. schneider , comp . comm . * 180 * , 2401 ( 2009 ) . k. t. taylor , j. s. parker , d. dundas , k. j. meharg , l. r. moore , e. s. smyth , and j. f. mccann , in _ many - particle quantum dynamics in atomic and molecular fragmentation _ 9 , p153 , eds . j. ullrich and v. shevelko ( springer - verlag , berlin heidelberg 2003 ) . c. flammer , in _ spheroidal wave functions _ , ( dover publications , inc . , mineola , new york 2005 ) . j. colgan , m. s. pindzola , and robicheaux , j. phys . b * 34 * , l457 ( 2001 ) . l. a. a. nikolopoulos and p. lambropoulos , phys . rev . a * 74 * , 063410 ( 2006 ) . j. sims and s. hagstrom , j. chem . phys . * 124 * , 094101 ( 2006 ) . s. x. hu , l. a. collins , and b. i. schneider , phys . rev . a * 80 * , 023426 ( 2009 ) . l. b. madsen , l. a. a. nikolopoulos , t. k. kjeldsen , and j. fernndez , phys . a * 76 * , 063407 ( 2007 ) . t. n. rescigno , private communication ( 2010 ) . v. v. serov and b. b. joulakian , phys . a * 80 * , 062713 ( 2009 ) . f. maulbetsch and j. s. briggs , j. phys . b * 28 * , 551 ( 1995 ) . x. guan , unpublished ( 2008 ) . t. j. reddish , j. colgan , p. bolognesi , l. avaldi , m. gisselbrecht , m. lavolle , m. s. pindzola , and a. huetz , phys . lett . * 100 * , 193001 ( 2008 ) .

this paper presents an overview of the derivation and significance of recently derived conservation laws for the matrix moments of hermitean random matrices with dominant exponential weights that may be either even or odd . this is based on a detailed asymptotic analysis of the partition function for these unitary ensembles and their scaling limits . as a particular application we derive closed form expressions for the coefficients of the genus expansion for the associated free energy in a particular class of dominant even weights . these coefficients are generating functions for enumerating _ g_-maps , related to graphical combinatorics on riemann surfaces . this generalizes and resolves a 30 + year old conjecture in the physics literature related to quantum gravity .

1203.6123

this paper will present an overview of some recent developments in the application of random matrix analysis to the _ topological combinatorics _ of surfaces . such applications have a long history about which we should say a few words at the outset . the combinatorial objects of interest here are _ a map is an embedding of a graph into a compact , oriented and connected surface @xmath0 with the requirement that the complement of the graph in @xmath0 should be a disjoint union of simply connected open sets . if the genus of @xmath0 is @xmath1 , this object is referred to as a _ g - map_. the notion of @xmath1-maps was introduced by tutte and his collaborators in the 60s @xcite as part of their investigations of the four color conjecture . in the early 80s bessis , itzykson and zuber , a group of physicists studying t hooft s diagrammatic approaches to large n expansions in quantum field theory , discovered a profound connection between the problem of enumerating @xmath1-maps and random matrix theory @xcite . that seminal work was the basis for bringing asymptotic analytical methods into the study of maps and other related combinatorial problems . subsequently , in the early 90s , other physicists @xcite realized that the matrix model diagrammatics described in @xcite provide a natural means for discretizing the einstein - hilbert action in two dimensions . from that and a formal double scaling limit , they were able to put forward a candidate for so - called _ 2d quantum gravity_. this generated a great deal of interest in the emerging field of string theory . we refer to @xcite for a systematic review of this activity and to @xcite for a description of more recent developments related to topological string theory . all of these applications were based on the postulated existence of a @xmath2 asymptotic expansion of the free energy associated to the random matrix partition function , where @xmath3 denotes the size of the matrix , as @xmath3 becomes large . the combinatorial significance of this expansion is that the coefficient of @xmath4 should be the generating function for the enumeration of @xmath1-maps ( ordered by the cardinality of the map s vertices ) . in @xcite the existence of this asymptotic expansion and several of its important analytical properties were rigorously established . this analysis was based on a riemann - hilbert problem originally introduced by fokas , its and kitaev @xcite to study the 2d gravity problem . the aim of this paper is to outline how the results of @xcite and its sequel @xcite have been used to gain new insights into the map enumeration problem . in particular , we will be able to prove and significantly extend a conjecture made in @xcite about the closed form structure of the generating functions for map enumeration . over time combinatorialists have made novel use of many tools from analysis including contour integrals and differential equations . in this work we also introduce nonlinear partial diferential equations , in particular a hierarchy of conservation laws reminiscent of the _ shallow water wave equations _ @xcite ( see ( [ toda ] ) ) . this appears to make contact with the class of _ differential posets _ introduced by stanley @xcite ( see remark [ diffposet ] ) . the general class of matrix ensembles we analyze has probability measures of the form @xmath5\right\ } dm,\,\ , \mbox{where}\\ \label{i.001b } v_j(\lambda ; \ t_{j } ) & = & \frac{1}{2 } \lambda^{2 } + \frac{t_{j}}{j } \lambda^{j}\end{aligned}\ ] ] defined on the space @xmath6 of @xmath7 hermitean matrices , @xmath8 , and with @xmath9 a positive parameter , referred to as the _ string coefficient_. the normalization factor @xmath10 , which serves to make @xmath11 a probability measure , is called the _ partition function _ of this unitary ensemble . in previous treatments , @xcite , we have used the parameter @xmath12 instead of @xmath9 . this was in keeping with notational usages in some areas of random matrix theory ; however , since here we are trying to make a connection to some applications in quatum gravity , we have adopted the notation traditionally used in that context . this also is why we have scaled the time parameter @xmath13 by @xmath14 in this paper . for general polynomial weights @xmath15 it is possible to establish the following fundamental asymptotic expansion @xcite , @xcite of the logarithm of the _ free energy _ associated to the partition function . more precisely , those papers consider weights of the form @xmath16 with @xmath17 even . we introduce a renormalized partition function , which we refer to as a _ tau function _ representation , @xmath18 where @xmath19 . the principal object of interest is the _ large n _ asymptotic expansion of this representation for which one has the result @xcite @xmath20 as @xmath21 while @xmath22 with @xmath23 , called the t hooft parameter , held fixed . moreover , for @xmath24 for some @xmath25 , a. [ unif ] the expansion is uniformly valid on compact subsets of @xmath26 ; b. [ analyt ] @xmath27 extends to be complex analytic in @xmath28 ; c. [ diff ] the expansion may be differentiated term by term in @xmath29 with uniform error estimates as in ( [ unif ] ) the meaning of ( [ unif ] ) is that for each @xmath1 there is a constant , @xmath30 , depending only on @xmath26 and @xmath1 such that @xmath31 for @xmath32 in a compact subset of @xmath26 . the estimates referred to in ( [ diff ] ) have a similar form with @xmath33 and @xmath34 replaced by their mixed derivatives ( the same derivatives in each term ) and with a possibly different set of constants @xmath30 . recently these results were extended to the case where @xmath17 is odd @xcite . in this case one should replace the normalized partiiton function ( [ tausquare ] ) by its szeg representation in terms of eignevalues ( [ szego ] ) . to explain the topological significance of the @xmath35 as generating functions , we begin with a precise definition of the objects they enumerate . @xmath36 on a compact , oriented and connected surface @xmath0 is a pair @xmath37)$ ] where * @xmath38 is a connected 1-complex ; * @xmath39 $ ] is an isotopical class of inclusions @xmath40 ; * the complement of @xmath38 in @xmath0 is a disjoint union of open cells ( faces ) ; * the complement of the vertices in @xmath38 is a disjoint union of open segments ( edges ) . when the genus of x is @xmath1 one refers to the map as a @xmath41 . what @xcite effectively showed was that the partial derivatives of @xmath42 evaluated at @xmath43 `` count '' a geometric quotient of a certain class of _ labelled g - maps_. as a means to reduce from enumerating these labelled @xmath1-maps to enumerating @xmath1-maps , it is natural to try taking a geometric quotient by a `` relabelling group '' more properly referred to as a _ cartographic group _ @xcite . this labelling has two parts ; first the vertices of the same valence , @xmath44 have an order labelling @xmath45 and second at each vertex one of the edges is distinguished . given that x is oriented , this second labelling gives a unique ordering of the edges around each vertex . the fact that the coefficients of the free energy expansion ( [ i.002 ] ) enumerate this class of labelled @xmath1-maps is a consequence of ( [ i.002 ] ) ( i ) which enables one to evaluate a mixed partial derivative of @xmath46 in terms of the _ gaussian unitary ensemble _ ( gue ) where correlation functions of matrix coefficients all reduce to two point functions . ( a precise description of this correspondence may be found in @xcite ) . to help fix these ideas we consider the case of a @xmath17-regular @xmath1-map ( i.e. , every vertex has the same valence , @xmath17 ) of size @xmath47 ( i.e. , the map has @xmath47 vertices ) which is the main interest of this paper . the cartographic group in this case is generated by the symmmetric group @xmath48 which permutes the vertex labels and @xmath47 factors of the cyclic group @xmath49 which rotates the distinguished edge at a given vertex in the direction of the holomorphic ( counter - clockwise ) orientation on @xmath0 . the order of the cartographic group here is the same as that of the product of its factors which is @xmath50 . on the other hand the generating function for @xmath1-maps in this setting is given by @xmath51 where @xmath52 = the number of labelled @xmath17-regular @xmath1-maps on @xmath47 vertices . the factor @xmath53 perfectly cancels the order of the cartographic group , making this series appear to indeed be the ordinary generating function for pure @xmath1-maps . however , for some @xmath1-maps the cartographic action may have non - trivial isotropy and this can create an `` over - cancellation '' of the labelling . this happens when a particular relabelling of a given map can be transformed back to the original labelling by a diffeomorphism of the underlying riemann surface @xmath0 . in this event the two labellings are indistinguishable and the diffeomorphism induces an automorphism of the underlying map . in addition , the element of the cartographic group giving rise to this situation is an element of the isotropy group of the given map . hence , as a generating function for the geometric quotient , ( [ gquoti ] ) is expressible as @xmath54 where @xmath55 = the number of vertices of @xmath36 , @xmath56 = the number of faces of @xmath36 and aut(@xmath36 ) = the automorphism group of the map @xmath36 . we have included the @xmath57-dependent form , ( [ gquotiii ] ) , of @xmath46 since that will play an important role later on and also to observe that this is in fact a _ bivariate _ generating function for enumerating @xmath1-maps with a fised number of vertices and faces . moreover , in this @xmath17-regular setting , one sees that the bivariate function is self - similar . this is a direct consequence of euler s relation : @xmath58 the presence of geometric factors such as @xmath59 is not uncommon in enumerative graph theory , a classical example being that of erds - rnyi graphs @xcite . in the quantum gravity setting these factors also have a natural interpretation in terms of the discretization of the reduction to conformal structures via a quotient of metrics by the action of the diffeomorphism group . we refer to @xcite for further details on this attractive set of ideas . in @xcite , @xmath60 and @xmath61 were explicitly computed for the case of valence @xmath62 . we quote , from the same paper , the following conjecture ( some notation has been changed to be consistent with ours ) : _ it would of course be very interesting to obtain @xmath63 in closed form for any value of @xmath1 . the method of this paper enabled us to do so up to @xmath64 , but works in the general case , although it requires an incresing amount of work . we conjecture a general expression of the form _ @xmath65 _ with @xmath66 a polynomial in @xmath67 , the degree of which could be obtained by a careful analysis of the above procedure . " _ here @xmath68 is equal , up to a scaling , to the generating function for the catalan numbers ; below it will signify @xmath69 which is similarly related to the generating function for the higher catalan numbers ( [ catalan ] ) . over the years there have been a number of attempts to systematically address this question by studying the resolvent of the random matrix and associated schwinger - dyson equations @xcite . our methods take a different approach . the main purpose of this paper is to show how this conjecture can be verified and significantly extended . in particular , we will show that for the case of even valence , @xmath70 , [ thm51 ] for @xmath71 , @xmath72 for all @xmath73 . the top coefficient and the constant term are respectively given by @xmath74 \qquad\end{aligned}\ ] ] where @xmath75 ( see thm [ result ] ) is proportional to the @xmath76 coefficient in the asymptotic expansion at infinity of the @xmath77 equation in the painlev i hierarchy @xcite and @xmath78 . our methods can be extended to the case of @xmath17 odd and the derivation of the analogue to theorem [ thm51 ] is in progress ( see section [ sec:6 ] ) . the route to getting these results passes through nonlinear pde , in particular a class of nonlinear evolution equations known as conservation laws which come from studying scaling limits of the recursion operators for orthogonal polynomials whose weights match those of the matrix models . this appeal to orthogonal polynomials also motivated the approaches of @xcite and @xcite . however to give a rigorous _ and _ effective treatment to the problem of finding closed form expressions for the coefficients of the asymptotic free energy , ( [ i.002 ] ) , requires essential use of riemann - hilbert analysis on the riemann hilbert problem for orthogonal polynomials that was introduced in @xcite . though we will not review this analysis here , we will state the consequences of it needed for our applications and reference their sources . in section [ sec:2 ] we present the necessary background on orthogonal polynomials and introduce the main equations governing their recurrences operators : the _ difference string equations _ and the _ toda lattice equations_. in section [ sec:44 ] we describe how ( [ i.002 ] ) can be used to derive and solve ( in the case of even valence ) the continuum limits of these equations which relates to the nonlinear evolution equations alluded to earlier . in section [ sec:4 ] we outline the proof theorem [ thm51 ] and in section [ sec:6 ] we describe the extension of this program to the case of odd valence and briefly mention what has been accomplished in that case thus far . this will also help to illuminate the full picture behind the idea of conservation laws for random matrices . let us recall the classical relation between orthogonal polynomials and the space of square - integrable functions on the real line , @xmath79 , with respect to exponentially weighted measures . in particular , we want to focus attention on weights that correspond to the random matrix weights , @xmath80 , ( [ i.001b ] ) with @xmath17 even . ( recently this relation has been extended to the cases of @xmath17 odd @xcite , with the orthogonal polynomials generalised to the class of so - called _ non - hermitean _ orthogonal polynomials ; however , for this exposition we will stick primarily with the even case . ) to that end we consider the hilbert space @xmath81 of weighted square integrable functions . this space has a natural polynomial basis , @xmath82 , determined by the conditions that @xmath83 for the construction of this basis and related details we refer the reader to @xcite . with respect to this basis , the operator of multiplication by @xmath84 is representable as a semi - infinite tri - diagonal matrix , @xmath85 @xmath86 is commonly referred as the _ recursion operator _ for the orthogonal polynomials and its entries as _ ( when @xmath15 is an even potential , it follows from symmetry that @xmath87 for all @xmath17 . ) we remark that often a basis of orthonormal , rather than monic orthogonal , polynomials is used to make this representation . in that case the analogue of ( [ multop ] ) is a symmetric tri - diagonal matrix . as long as the coefficients @xmath88 do not vanish , these two matrix representations can be related through conjugation by a semi - infinite diagonal matrix of the form @xmath89 . similarly , the operator of differentiation with respect to @xmath84 , which is densely defined on @xmath90 , has a semi - infinite matrix representation , @xmath91 , that can be expressed in terms of @xmath86 as @xmath92 where the `` minus '' subscript denotes projection onto the strictly lower part of the matrix . from the canonical ( heisenberg ) relation on @xmath93 , one sees that @xmath94 & = & 1,\end{aligned}\ ] ] where here @xmath84 in the bracket and @xmath95 on the right hand side are regarded as multiplication operators . with respect to the basis of orthogonal polynomials this may be re - expressed as @xmath96 & = & g_s i .\end{aligned}\ ] ] the relations implicit in ( [ fl1 ] ) have been referred to as _ string equations _ in the physics literature . in fact the relations that one has , row by row , in ( [ fl1 ] ) are actually successive differences of consecutive string equations in the usual sense . however , by continuing back to the first row one may recursively de - couple these differences to get the usual equations . to make this distinction clear we will refer to the row by row equations that one has directly from ( [ fl1 ] ) as _ difference string equations_. @xmath86 depends smoothly on the coupling parameter @xmath13 in the weight @xmath80 ( see [ i.001b ] ) . the explicit dependence can be determined from the fact that multiplication by @xmath84 commutes with differentiation by @xmath13 . this yields our second fundamental relation on the recurrence coefficients , @xmath97\,,\end{aligned}\ ] ] which is equivalent to the @xmath98 equation of the semi - infinite toda lattice hierarchy . the toda equations for @xmath99 are @xmath100 one may apply standard methods of orthogonal polynomial theory @xcite to deduce the existence of a semi - infinite lower unipotent matrix @xmath101 such that @xmath102 where @xmath103 ( for a description of the construction of such a unipotent matrix we refer to proposition 1 of @xcite . ) this is related to the hankel matrix @xmath104 where @xmath105 is the @xmath106 moment of the measure , by @xmath107 with @xmath108 where @xmath6 denotes the @xmath7 principal sub - matrix of @xmath109 whose determinant may be expressed as ( see szeg s classical text @xcite ) , @xmath110 \right\ } d^{n } \lambda,\end{aligned}\ ] ] where @xmath111 . we set @xmath112 . we sometimes need to extend the domain of the tau functions to include other parameters , such as @xmath113 , as we have done here . doing this presents no difficulties in the prior constructions . the diagonal elements may in fact be expressed as @xmath114 where @xmath115 which agrees with the definition of the tau function given in ( [ tausquare ] ) . the second equality follows by reducing the unitarily invariant matrix integrals in ( [ szego2 ] ) to their diagonalizations which yields ( [ szego1 ] ) @xcite . tracing through these connections , from @xmath86 to @xmath116 , one may derive the fundamental identity relating the random matrix partition function to the recurrence coefficients , @xmath117 which is the basis for our analysis of continuum limits in the next section . ( note that @xmath118 and therefore @xmath119 . ) we will also need a differential version of this relation : ( hirota ) @xmath120 = -g_s \frac{\partial}{\partial t_1 } \log \left [ \frac{z^{(n+1)}(t_1 , t_{2\nu})}{z^{(n)}(t_1 , t_{2\nu } ) } \right]\\ \label{b } b_{n , g_s}^2 & = g_s^2 \frac{\partial^2}{\partial t_1 ^ 2 } \log \tau^2_{n , g_s } = g_s^2 \frac{\partial^2}{\partial t_1 ^ 2 } \log z^{(n)}(t_1 , t_{2\nu})\,,\end{aligned}\ ] ] ( a derivation of this lemma may be found in @xcite . ) it follows from ( [ b ] ) and ( [ i.002 ] ) that [ two - leg ] @xmath121 is a uniformly valid asymptotic asymptotic expansion in the sense of ( [ i.002 ] iii ) . in order to effectively utilize the relations ( [ fl1 ] , [ fl2 ] ) it will be essential to keep track of how the matrix entries of powers of the recurrence operator , @xmath122 , depend on the original recurrence coefficients . that is best done via the combinatorics of weighted walks on the index lattice of the orthogonal polynomials . for the case of even potentials , the relevant walks are _ dyck paths _ which are walks , @xmath123 , on @xmath124 which , at each step , can either increase by 1 or decrease by 1 . set @xmath125 then step weights , path weights and the @xmath126-entry of @xmath122 are , respectively , given by @xmath127 the _ difference string equations _ are given ( for the @xmath128-valent case ) by ( [ fl1 ] ) : @xmath129 = g_s i \,.\ ] ] by parity considerations , when the potential @xmath15 is even , the only non - tautological equations come from the diagonal entries of ( [ string - star2 ] ) : _ the @xmath130 entry _ gives @xmath131 in terms of dyck paths this becomes @xmath132 where @xmath133 denotes the lattice location of the path @xmath123 after the @xmath134 downstep and we have used the relation @xmath23 on the left hand side of the equation . we illustrate this more concretely for the case of @xmath135 . referring to ( [ motzkin ] ) , the relevant path classes here are @xmath136 note that the structure of the path classes does not actually depend upon @xmath3 . this is a reflection of the underlying spatial homogeneity of these equations . thus , for the purpose of describing the path classes , one can translate @xmath3 to @xmath137 . now applying ( [ weights ] ) the difference string equation becomes , for @xmath138 , @xmath139 where , for this example , we have set the parameter @xmath57 equal to @xmath95 . we now pass to a more explicit form of of the _ toda equations _ ( [ fl2 ] ) in the case @xmath70 : @xmath140 once agian we illustrate these equations in the tetravalent case ( @xmath141 ) . the relevant path class is : @xmath142 applying ( [ weights ] ) , the tetravalent toda equations become @xmath143 where we have again used the relation @xmath23 and then set the parameter @xmath144 . the continuum limits of the difference string and toda equations will be described in terms of certain scalings of the independent variables , both discrete and continuous . as indicated at the outset , the positive parameter @xmath9 sets the scale for the potential in the random matrix partition function and is taken to be small . the discrete variable @xmath3 labels the lattice _ position _ on @xmath145 that marks , for instance , the @xmath146 orthogonal polynomial and recurrence coefficients . we also always take @xmath3 to be large and in fact to be of the same order as @xmath147 ; i.e. , as @xmath3 and @xmath9 tend to @xmath148 and @xmath137 respectively , they do so in such a way that their product @xmath149 remains fixed at a value close to @xmath95 . in addition to the _ global _ or _ absolute _ lattice variable @xmath3 , we also introduce a _ local _ or _ relative _ lattice variable denoted by @xmath150 . it varies over integers but will always be taken to be small in comparison to @xmath3 and independent of @xmath3 . the dyck lattice paths naturally introduce the composite discrete variable @xmath151 into the formulation of the difference string and toda equations which we think of as a small discrete variation around a large value of @xmath3 . the spatial homogeneity of those equations manifests itself in their all having the same form , independent of what @xmath3 is , while @xmath150 in those equations varies over @xmath152 , the _ bandwidth _ of the @xmath153 toda / difference string equations . taking @xmath154 will insure the necessary separation of scales between @xmath150 and @xmath3 . we define @xmath155 as a _ spatial _ variation close to @xmath57 which will serve as a continuous analogue of the lattice location along a dyck path relative to the starting location of the path . we also introduce the self - similar scalings : @xmath156 that are natural given ( [ sss ] ) . in terms of these scalings , ( [ b - asymp ] ) may be rewritten @xcite as @xmath157 and , by @xcite app . a , @xmath158 we mention here that the variables @xmath159 as defined above differ slightly from their usage in related works @xcite where @xmath160 for appropriate parameters @xmath161 . we also introduce a shorthand notation to denote the expansion of the coefficients of @xmath162 around @xmath163 . [ shorthand ] for @xmath164 , @xmath165 where the subscript @xmath166 denotes the operation of taking the @xmath98 derivative with respect to @xmath167 of each coefficient of @xmath168 : @xmath169 as valid asymptotic expansions these representations denote the asymptotic series whose successive terms are gotten by collecting all terms with a common power of @xmath170 in ( [ f1k ] ) . in what follows we will frequently abuse notation and drop the evaluation at @xmath171 . in particular , we will write @xmath172 in doing this these series must now be regarded as formal but whose orders are still defined by collecting all terms in @xmath170 and @xmath9 of a common order . ( recall that @xmath173 so that @xmath174 ) . they will be substituted into the difference string and the toda equations to derive the respective continuum equations . at any point in this process , if one evaluates these expressions at @xmath171 and @xmath175 one may recover valid asymptotic expansions in which the @xmath176 and @xmath177 have their original significance as valid asymptotic expansions of the recursion coefficients . we are now in a position to study the toda lattice equations ( [ bn ] ) expanded on the formal asymptotic series ( [ f1kform ] ) : @xmath178 from now on we will take @xmath179 , since its role in determining the structure of the asymptotic expansions of the @xmath180 is now completed , and set @xmath181 . collecting terms in these equations order by order in orders of @xmath170 we will have a hierarchy of equations that , in principle , allows one to recursively determine the coefficients of ( [ bs - asymp ] ) . we will refer to this hierarchy as the _ continuum toda equations_. ( note that one has such a hierarchy for each value of @xmath182 . ) of course this is a standard procedure in perturbation theory . the equations we will derive are pdes in the form of evolution equations in which @xmath167 , now regarded as a continuous variable , is the independent _ spatial _ variable and @xmath183 is the _ temporal _ variable . one must still determine , at each level of the hierarchy , which solution of the ode is the one that corresponds to the expressions given for @xmath184 in ( [ b - shift_g ] ) . this amounts to a kind of solvability condition . this process was carried out fully in @xcite and @xcite . we will now state the results of that analysis . [ cont ] @xcite the continuum limit , to all orders , of the toda lattice equations as @xmath21 is given by the following infinite order partial differential equation for @xmath185 : @xmath186 where @xmath187 is a partition , with @xmath188 , of @xmath189 ; @xmath190 ; @xmath191 is the _ length _ of @xmath84 ; and @xmath192 is the _ size _ of @xmath84 ; and @xmath193 are coefficients to be described in the next proposition . by ( [ burgers ] ) , @xmath194 . the above result effectively reduces the determination of the hierarchy to an enumeration in terms of a pair of partitions . the first class of partitions is fairly straightforward and amounts to keeping track of the partial derivatives that enter into the expressions at a given level . thus at @xmath195 the terms are products of partial derivatives of various orders ( the _ parts _ ) which must add up to @xmath189 . these terms correspond to the tableaux at the @xmath196 level of the _ hasse - young graph _ shown in the left panel of figure [ zigzag ] . ( in this we ignore , for now , the powers of @xmath170 that are internal to the asymptotic series @xmath168 and its @xmath167-derivatives . ) the other type of partition relates to how the dyck paths enter our equations . we have already seen that a dyck path is completely determined by specifying when its downsteps occur . the toda equations depend explicitly on where these downsteps ( the @xmath133 appearing in the equations as stated at the start of this subsection ) . however these two specifications can be related and the downstep times are encoded in terms of a partition that measures deviation of the path from a standard zig - zag path . the following proposition gives an explicit closed form expression for the coefficients @xmath193 in terms of both classes of partitions . @xcite ( _ app . a.3_)[d - coeffs ] @xmath197 where @xmath198 is the set of _ restricted partitions _ ( meaning that @xmath199 ) , @xmath200 , and @xmath201 is the monomial symmetric polynomial associated to @xmath84 @xcite . the relation of inclusion between partitions , @xmath202 means that @xmath203 for all @xmath17 . the right panel of figure [ zigzag ] exemplifies , in the case when @xmath204 , a geometric realization of the partitions @xmath205 being summed over in the above formula . such a partition corresponds to a zig - zag path contained in the @xmath206 rectangle which starts at the leftmost corner with the @xmath207 step being a up - step and terminates at the rightmost corner with the @xmath153 step . the red path corresponds to the distinguished partition @xmath208 which records the downsteps at times @xmath209 so that in this case the initial ( @xmath207 ) step is a downstep . the green path illustrates a typical @xmath205 which , in this case , takes downsteps at times @xmath210 . given such a partition - path , one may project the initial point of each of its downsteps to the horizontal axis as indicated in the figure for the red and green paths . then @xmath211 equals the signed separation between the @xmath212 green point and the @xmath212 red point , reading from right to left . the symmetric polynomial @xmath213 then gets evaluated at these separation values in the above formula for @xmath193 . this description could also have been formulated in terms of the _ border strips _ associated to the _ skew tableaux _ @xmath214 but we will not elaborate on that here . [ h ] hasse - young graph ( courtesy d. eppstein ) ; right : @xmath205 partition walks , title="fig:",width=192 ] hasse - young graph ( courtesy d. eppstein ) ; right : @xmath205 partition walks , title="fig:",width=192 ] one is now in a position to deduce the form of the toda hierarchy . this is done by setting @xmath215 so that @xmath216 . one then collects _ all _ terms of order @xmath217 in the resulting expansion of ( [ burgers ] ) and this will be a partial differential equation in @xmath218 and @xmath167 that we refer to as the @xmath76 equation in the continuum toda hierarchy . at leading order in the hierarchy one observes that , for general @xmath182 , the continuum toda equation is an inviscid burgers equation @xcite @xmath219 with initial data @xmath220 . a solution exists and is unique for sufficiently small values of @xmath218 . it may be explicitly calculated by the method of characteristics , also known as the _ hodograph _ method in the version we now present . consider the ( hodograph ) relation among the independent variables @xmath221 , @xmath222 [ hodlemma ] a local solution of ( [ burgers ] ) is implicitly defined by ( [ hodograph ] ) . the annihilator of the differential of ( [ hodograph ] ) @xmath223 is a two - dimensional distribution locally on the space @xmath221 . an initial curve over the @xmath167-axis ( parametrized as the graph of a function @xmath224 ) , transverse to the locus where @xmath225 locally determines a unique integral surface foliated by the integral curves of the vector field of the _ characteristic _ vector field @xmath226 equation ( [ char1 ] ) requires that along an integral curve of the characteristic vector field , @xmath227 is constant ; i.e. , @xmath228 by ( [ char2 ] ) which is equivalent to ( [ burgers ] ) . using ( [ char3 ] ) to set @xmath229 pins down our solution uniquely . we note that the numerical coefficients appearing in these burgers equations depend only on the total number of dyck paths in @xmath230 . one finds @xcite from ( [ hodograph ] ) and the self similar form of @xmath227 , @xmath231 when @xmath141 , @xmath232 is the @xmath98 catalan number . for general @xmath182 these are the _ higher catalan numbers _ which play a role in a wide variety of enumerative combinatorial problems . the _ continuum difference string _ hierarchies may be derived from the difference string equations ( [ diff - string ] ) in a manner completely analogous to what was done with the toda equations in the previous subsection . expanding ( [ diff - string ] ) on the asymptotic series ( [ f1kform ] ) we arrive at the following asymptotic equations . @xmath233 the equations at leading order , @xmath234 , are @xmath235 or , equivalently , @xmath236 which one directly recognizes as the spatial derivative of the hodograph solution ( [ hodograph ] ) . evaluating that solution at @xmath237 yields @xmath238 which is the functional equation for the generating function of the @xmath77 higher catalan numbers , mentioned in the previous subsection . the terms of the equations at @xmath239 can be computed directly and are found to have the form @xmath240 & + & 2s\left(c_\nu \partial_w \sum_{\begin{array}{c } 0 \leq k_j < g \\ k_1 + \dots + k_{\nu } = g\\ \end{array } } f_{k_1}\cdots f_{k_{\nu}}\right)\\ \nonumber + \partial_w \sum_{k=0}^{g-1 } \frac{f_{k w^{(2g-2k)}}}{(2g-2k+1 ) ! } & + & 2\nu s \left(f_1^{(\nu - 1)}[2g-2 ] + f_2^{(\nu - 1)}[2g-4 ] + \cdots + f_{g}^{(\nu - 1)}[0]\right ) = 0 , \end{aligned}\ ] ] where @xmath241 $ ] denotes the coefficient of @xmath242 in @xmath243 in fact this equation is also @xmath167-exact ! to see this we introduce indeterminates @xmath244 and define @xmath245 we also recall that the monomial symmetric function associated to the partition @xmath84 is given by @xmath246 where @xmath247 on the right hand side denotes adding the ( minimal ) number of terms which are gotten by permuting the variables @xmath244 in the first term such that the resulting polynomial is symmetric with respect to @xmath248 . finally , define @xmath249 the following proposition shows that the @xmath250 terms of @xmath251 are an anti - derivative for the terms @xmath252 $ ] appearing in ( [ cont - string ] ) . [ higher_exact ] @xmath253 by symmetry under the action of the symmetric group , it will suffice to check this identity for the single monomial term corresponding to @xmath254 . we directly calculate @xmath255 where in the second line we have made changes of variables replacing @xmath256 by @xmath167 . using this we observe that the inner summands of @xmath257 have the form @xmath258 in the second line the two inner summations of the first line have been rewritten in terms of clusters of partitions adjacent to a given partition of size @xmath189 . to get the last line we use the fact that @xmath259 . the proposition now follows from this observation , ( [ potential ] ) and ( [ fg ] ) . as a consequence of this result one sees that the continuum difference string equation is directly integrable : @xmath260 + \widehat{f}_2^{(\nu - 1)}[2g-4 ] + \cdots + \widehat{f}_{g}^{(\nu - 1)}[0]\right ) + \frac{1}{2s}\sum_{k=0}^{g-1 } \frac{f_{k w^{(2g-2k)}}}{(2g-2k+1)!}\right\}.\end{aligned}\ ] ] setting @xmath237 and applying ( [ stringeqn ] ) to eliminate @xmath218 this reduces to @xmath261 + \widehat{f}_2^{(\nu - 1)}[2g-4 ] + \cdots + \widehat{f}_{g}^{(\nu - 1)}[0]\right)\big|_{w = 1 } \right\}.\end{aligned}\ ] ] it is immediate from this representation that @xmath262 is a rational function of @xmath67 . apriori this _ anti - derivative _ should also include a constant term ( in @xmath167 ; it could depend on @xmath218 ) . this would lead to a term of the form @xmath263 . however , in @xcite it is shown , by an independent argument , that the pole order in @xmath67 at @xmath264 is always greater than one . hence the constant of integration must be zero . with further effort this can be refined to @xcite[result ] @xmath265 where @xmath266 is a polynomial of degree @xmath267 in @xmath67 whose coefficients are rational functions of @xmath182 over the rational numbers @xmath268 and @xmath269 . [ diffposet ] a key element in the proof of proposition [ higher_exact ] is the observation that differentiation with respect to @xmath167 adjusts the multinomial labelling of partial derivatives in the expansion according to the edges of the hasse - young graph ( fig [ zigzag ] ) . this graph describes the adjacency relations between young diagrams of differing sizes . the edges describe which partitions of size @xmath189 are _ covered _ by a given partition of size @xmath270 . conversely it describes which partitions of size @xmath270 cover a partition of size @xmath189 which in the setting described here acts as an anti - differentiation operator . this kind of structure was called a differential poset by stanley and systematically examined in @xcite . recalling the basic identity ( [ hirota ] ) @xmath271 we have , by taking logarithms , @xmath272 where the initial value @xmath273 is given by the recursion relations of the hermite polynomials . as in @xcite , we can use formula to recursively determine @xmath46 in terms of solutions to the continuum equations . we use the asymptotic expansion of @xmath274 which has the form ( [ bs - asymp ] ) : @xmath275 note that the left hand side of equation ( [ tauk-2nddiff ] ) has the form of a centered second difference , @xmath276 . it follows @xcite that this expression has an expansion for large @xmath3 involving only even derivatives of the spatial variable @xmath167 . we have , at order @xmath217 , @xmath277 where @xmath278 . in @xcite it was shown that @xmath46 is rational in @xmath67 with poles located only at @xmath279 . however we will now prove the more refined result stated in theorem [ thm51 ] . the proof of this result is by induction on @xmath1 . ( the base case of @xmath280 is established by direct calculation @xcite . ) we assume that ( [ note ] ) holds for all @xmath281 . we state here , without proof , some straightforward lemmas and propositions describing the derivatives of ( [ note ] ) ( details may be found in @xcite where similar lemmas are proved for the @xmath262 ) . set @xmath282 [ lem51 ] @xmath283 [ lem52 ] for @xmath284 and @xmath285 , @xmath286 c_\ell^{k , j-1}(\nu ) + \nu(2k + \ell + ( j-3 ) ) c_{\ell -1 } ^{k , j-1}(\nu)\\ c_\ell^{(k , j)}(\nu ) & = & 0 \qquad \ell < 0 , \qquad \ell \geq 3k-3+j\\ c_\ell^{(k,0)}(\nu ) & = & c_\ell^{(k)}(\nu).\end{aligned}\ ] ] [ lem53 ] for @xmath284 and @xmath285 @xmath287 where @xmath288 . [ prop52 ] for @xmath284 and @xmath289 @xmath290 but in fact , by the following vanishing lemma [ lem54 ] @xmath291 for @xmath292 , the minimal pole order of the expansion in proposition [ prop52 ] is @xmath293 . in particular the minimal pole orders coming from terms involving @xmath294 on the right - hand side of ( [ hirota2 ] ) are all greater than @xmath295 . [ prop53 ] the @xmath217 terms of @xmath296 @xmath297 by ( [ rational ] ) . this result shows that the minimal pole order coming from the @xmath298 terms in ( [ hirota2 ] ) is once again greater than @xmath299 . the preceding lemmas and propositions provide explicit laurent expansions ( in @xmath67 ) for all terms on the right hand side of ( [ hirota2 ] ) with two exceptions : @xmath300 with a small modification ( [ e1 ] ) may be brought in line with proposition [ prop52 ] , [ prop54 ] @xmath301 with @xmath302 . all other coefficients are then specified by the corresponding recursions stated in lemmas [ lem51 ] - [ lem53 ] with @xmath150 set to @xmath95 . a variant of the vanishing lemma [ lem54 ] also holds for @xmath303 : @xmath304 for @xmath305 . it follows that the minimal pole order of the expansion in proposition [ prop54 ] is at least @xmath306 and so the corresponding contribution to the minimal pole order of ( [ hirota2 ] ) is @xmath307 . finally we observe that for @xmath308 is a rational function of @xmath227 and its @xmath167-derivatives , [ prop55 ] @xmath309\\ \nonumber & + & ( p-3 ) ! \left(-\frac{1}{w}\right)^{p-2}\end{aligned}\ ] ] each line of the above proposition can be established directly by induction starting with the base case for @xmath310 . it then follows from proposition 3.1(iii ) of @xcite that the minimal pole order contributed by ( [ e0 ] ) is @xmath311 . we are now in a position to outline the ( of theorem [ thm51 ] ) in @xcite ( theorem 1.3 ) it was shown that @xmath312 where @xmath313 denotes a polynomial of degree @xmath314 in @xmath67 . we first want to determine the relation between this degree and the pole order @xmath315 . to this end we observe from propositions [ prop52 ] , [ prop53 ] , [ prop54 ] , and [ prop55 ] that the right hand side of ( [ hirota2 ] ) , evaluated at @xmath237 , is a rational function in @xmath67 which approaches a finite constant value as @xmath316 . from the form of the left hand side of ( [ hirota2 ] ) evaluated at @xmath237 one also sees that its asymptotic order ( as @xmath316 ) is the same as that of @xmath46 . hence , @xmath317 and this shows that ( [ note ] ) is valid up to the determination of the minimal and maximal pole orders at @xmath318 . in the preceding lemmas and proposiitons we have seen that , for all terms on the right hand side of ( [ hirota2 ] ) , the minimal pole order is @xmath311 . furthermore , from these same representations together with proposition 3.1(iii ) of @xcite one sees that , with the possible exception of the genus 0 terms in ( [ e0exp ] ) , the maximal pole order of the terms on the right hand side of ( [ hirota2 ] ) is @xmath319 . the apparent maximal pole order in ( [ e0exp ] ) is @xmath320 which exceeds the stated bound when @xmath321 . this maximal order comes from terms containing the factor @xmath322 which are , specifically , @xmath323_{w=1}\\ & = & \frac{f_{0 w^{(2g+2)}}}{2f_0}|_{w=1 } \left[1 - \frac{(2\nu+1)(\nu-1)}{2\nu(\nu+1 ) } z_0 + \frac{(\nu-1)^2}{2\nu ( \nu+1 ) } z_0 ^ 2\right]\\ & = & \frac{f_{0 w^{(2g+2)}}}{2f_0}|_{w=1 } \mathcal{o}(\nu - ( \nu-1)z_0).\end{aligned}\ ] ] hence the maximal pole order contributed by the genus 0 terms is , in fact , @xmath324 which is @xmath325 for @xmath326 and @xmath327 for @xmath328 . this establishes that , for @xmath329 , the pole orders on the right hand side of ( [ hirota2 ] ) are bounded between @xmath270 and @xmath319 . moreover , for @xmath330 , the case by case checking of terms on the right hand side of ( [ hirota2 ] ) that has been carried out in this subsection , shows that the maximal pole order is realized by the term in proposition [ prop53 ] corresponding to the partition @xmath84 of @xmath270 having minimal length ( = @xmath95 ) ; i.e. , the partition whose young diagram is a single row . this implies that the residue of the maximal order pole is @xmath331 which is non - vanishing by theorem [ result ] . hence the maximal order pole is realized . now , given that @xmath46 has the form ( [ eg_ratl ] ) with @xmath317 , it follows from direct calculation that @xmath332 raises the minimum pole degree by @xmath333 and the maximum pole degree by @xmath334 with the coefficient at this order given by ( [ leadcoeff ] ) . this establishes ( [ note ] ) for @xmath330 . the cases of @xmath335 may be established separately by direct calculation ( see , for example , section 1.4.2 of @xcite ) . to establish ( [ note3 ] ) first note that by euler s relation , @xmath336 for a @xmath1-map where @xmath47 is the number of ( @xmath128-valent ) vertices and @xmath337 is the number of faces . since @xmath338 , one immediately sees that the number of vertices of such a map must satisfy the inequality @xmath339 it follows that @xmath340 for @xmath341 . ( @xmath46 must vanish at least simply at @xmath342 since since @xmath343 . ) via cauchy s theorem these conditions may be re - expressed as @xmath344 for @xmath345 where in the second line we have rewritten @xmath46 as a rational function of @xmath346 ( [ eg_ratl ] ) and employed the change of variables @xmath347 which may be deduced from the string equation ( [ stringeqn ] ) . this yields a contour integral in @xmath346 centerd at @xmath95 . now one can see that these vanishing conditions are satisfied if and only if @xmath348 for @xmath345 which in turn proves ( [ note3 ] ) . finally we turn to the determination of the constant @xmath349 . by proposition [ prop53 ] , contributions to the constant term of @xmath46 come only from the first sum on the right hand side of ( [ hirota2 ] ) . the parts of this coming from @xmath350 and @xmath351 are , by propositions [ prop55 ] and [ prop54 ] respectively , @xmath352 and @xmath353 . at higher genus , @xmath281 , the contribution to the constant term is determined by lemma [ lem51 ] to be @xmath354 . hence , by ( [ hirota2 ] ) we have @xmath355 from which ( [ note3 ] ) immediately follows . in the case when @xmath17 is odd in the weight ( [ genpot ] ) for @xmath15 , there is clearly a problem in applying the method of orthogonal polynomials as it was outlined in section [ sec:2 ] . very recently , however , a generalization of the _ equilibrium measure _ ( which governs the leading order behavior of the free energy associated to ( [ rmt ] ) ) was developed and applied to this problem , @xcite . it is based on generalizing to a class of complex valued non - hermitean orthogonal polynomials on a contour in the complex plane other than the real axis . these extensions were motivated by new ideas in approximation theory related to complex gaussian quadrature of integrals with high order stationary points @xcite . but even when the issue of existence of appropriate orthogonal polynomials has been resolved , there are still a number of significant obstacles to deriving results like theorem [ thm51 ] that are not present when the valence @xmath17 is even . for odd valence there is an additional string of recurrence coefficients , the diagonal coefficients @xmath356 of @xmath86 , whose asymptotics needs to be analyzed . this in turn requires that the lattice paths used to define and analyze the toda and difference string equations must be generalized to the class of _ motzkin paths _ which can have segments where the lattice site remains fixed rather than always taking a step ( either up or down ) as was the case for dyck paths . nevertheless , all these constructions have been carried out in @xcite to derive the hierarchies of continuum toda and difference string equations when the valence @xmath17 is odd . the recurrence coefficients again have asymptotic expansions with continuum representations given by @xmath357_{\tilde{w } = 1}\end{aligned}\ ] ] the off - diagonal coefficients have corresponding representations which are much as they were in the even valence case , @xmath358 the coefficients in these expansions have a self - similar structure given by @xmath359 at leading order the continuum toda equations are @xmath360 and the leading order continuum difference string equations are @xmath361 where the coefficients of the matrix @xmath362 in ( [ toda ] ) are specified by @xmath363 and those of the matrix @xmath364 in ( [ string ] ) by @xmath365 the index @xmath205 appearing in the trinomial coefficients corresponds to the number of flat steps in the motzkin paths giving rise to that term . it is straightforward to see that ( [ toda ] ) may be rewritten in conservation law form as @xmath366 where the coefficients in the flux vector are given by @xmath367 recently , @xcite , we have determined that the equations ( [ string ] ) are in fact a differentiated form of the generalized hodograph solution of the conservation law ( [ law ] ) . this hodograph solution is given by @xmath368 analogous to what was done in theorem [ thm51 ] we expect to determine closed form expressions for all the coefficients in the topological expansion with odd weights . the first few of these , for the trivalent case , are @xcite @xmath369 where @xmath67 is implicitly related to @xmath370 by the polynomial equation @xmath371 @xmath372 is in fact the generating function for a _ fractional _ generalization of the catalan numbers . its @xmath134 coefficient counts the number of connected , non - crossing symmetric graphs on @xmath373 equi - distributed vertices on the unit circle @xcite . nothing has been said , in this article , about the eigenvalues of the random matrix @xmath8 although this is at the heart of the riemann hilbert analysis underlying all of our results . the essential link comes through the _ equilibrium measure _ @xcite , or density of states , for these eigenvalues . when @xmath374 in ( [ i.001b ] ) , this equilibrium measure reduces to the well - known wigner semi - circle law . as @xmath13 changes this measure deforms ; but , for @xmath13 satisfying the bounds implicit in ( [ i.002 ] ) ( i ) , its support remains a single interval , @xmath375 $ ] . the _ edge of the spectrum _ @xmath376 evolves dynamically with @xmath13 . in fact we have determined @xcite that @xmath377 and @xmath378 are the riemann invariants of the hyperbolic system ( [ toda ] ) . in the case of even valence these invariants collapse to @xmath379 where @xmath67 is the generating function upon which all the @xmath46 are built , as described in theorem [ thm51 ] . so the edge of the spectrum is indeed directly related to the genus expansion ( [ i.002 ] ) . it follows from corollary [ two - leg ] and the map - theoretic interpretation of @xmath46 given by ( [ gquotii ] ) that @xmath380 is a generating function for enumerating @xmath17-regular @xmath1-maps with , in addition , two _ legs_. a _ leg _ is a univalent vertex ; i.e. , a vertex with just one adjacent edge , that is connected to some other vertex of the map . in particular @xmath381 enumerates such maps on the riemann sphere or what are more commonly referred to as two - legged _ planar maps_. in a remarkable paper , @xcite , building on prior work of cori and vauquelin @xcite , schaeffer found a constructive correspondence between two - legged @xmath128-regular planar maps and @xmath128 valent _ blossom trees_. a @xmath128 valent blossom tree is a rooted @xmath128 valent tree , with each external vertex taken to be a leaf that is colored either black or white and such that each internal ( non - root ) vertex is adjacent to exactly @xmath382 black leaves . this gives another interpretation of @xmath69 as the generating function for the enumeration of blossom trees . it seems reasonable to hope that the arithmetic data implicit in the coefficients @xmath383 in theorem [ result ] ( resp . @xmath384 in theorem [ thm51 ] ) might provide a means , such as _ sewing rules _ , for constructing two - legged @xmath128-regular @xmath1-maps ( resp . pure @xmath128-regular @xmath1-maps ) from blossom trees . in another direction bouttier , di francesco and guitter @xcite have studied the combinatorics of geodesic distance for planar maps . they define the geodesic distance of a two - legged graph to be the minimum number of edges crossed by a continuous path between the two legs and study @xmath385 , the generating function for enumerating all two - legged @xmath128 valent planar maps whose geodesic distance is @xmath386 . they find surprising and elegant closed form expressions for the @xmath385 . the statisitcs of planar maps is a natural stepping off point for the study of random surfaces . there has been a lot of recent activity in this direction by le gall and his collaborators related to the work of schaeffer and bouttier et al . see for example @xcite . a different ( from matrix models ) representation of 2d quantum gravity may be given in terms of intersection theory on the moduli space of stable curves ( riemann surfaces ) , @xmath387 and from this alternate perspective witten conjectured that a generating function for the intersection numbers of tautological bundles on @xmath387 should be `` given by '' a double - scaling limit of the differentiated free energy ( [ b - asymp ] ) for matrix models . ( a precise description of this double - scaling limit may be found in @xcite . ) he further conjectured that this intersection theoretic generating function should , with repect to appropriate choices of parameter vatriables , satisfy the korteweg - devries ( kdv ) equation . subsequently , kontsevich @xcite was able to outline a proof of witten s kdv conjecture based on a combinatorial model of intersection theory on @xmath387 . this model expresses tautological intersections in terms of sums over tri - valent graphs on a genus @xmath1 riemann surface . he was then able to recast this sum in terms of a special matrix integral involving cubic weights on which the proof of witten s kdv conjecture is based . for a readable overview on the above circle of ideas we refer to @xcite . however , the first witten conjecture , on relating the intersection - theoretic free energy to the matrix model free energy , ( [ i.002 ] ) , remains open . with the results described in this paper it may now be possible to determine if , and in precisely what sense , this conjecture might be true and to see if this leads to connections between the kdv equation in the witten - kontsevich model and the conservation laws given by ( [ toda ] ) . in addition , given the recent results on matrix models with odd dominant weights , @xcite , it may now be possible to give a rigorous treatment of kontsevich s matrix integral which , up to now , has been formal . more recently there have been other , perhaps more natural , approaches to the proof of witten s kdv conjecture , @xcite , in terms of coverings of the riemann sphere and _ hurwitz numbers _ for which the generating functions specified in theorem [ thm51 ] should also have a natural interpetation . in @xcite it was observed that the equilibrium measure ( see [ 71 ] ) for the weight @xmath15 , with @xmath388 in ( [ i.001b ] ) , may be re - expressed as @xmath389 where @xmath390 is the equilibrium measure for @xmath391 ( the _ semicircle _ law ) and @xmath392 is the equilibrium measure for @xmath393 ; i.e. , the general measure is a linear combination , over @xmath67 , of two extremal monomial equilibrium measures . for @xmath394 $ ] ( which corresponds to @xmath395 $ ] ) , this combination is convex and ( [ eqmeas ] ) is indeed a measure with a single interval of support in @xmath396 . this may be analytically continued to a complex @xmath67 neighborhood of @xmath397 $ ] so that ( [ eqmeas ] ) remains a positive measure along an appropriate connected contour ( `` single interval '' ) in the complex @xmath84-plane . for @xmath141 this continuation may be made up to a boundary curve in the complex @xmath67-plane passing through @xmath398 ( with a corresponding image in the complex @xmath399 plane ) . extension of this to more general values of @xmath182 is in progress @xcite . the mechanism for carrying out this continuation is to regard ( [ eqmeas ] ) as a @xmath67-parametrized family of holomorphic quadratic differentials . the candidate for the measure s support is then an appropriate bounded real trajectory of the quadratic differential . outside the boundary curve , the riemann - hilbert analysis used in this paper may be analytically deformed and our results extended . the boundary may be regarded as a curve of critical parameters for this deformation . this curve is precisely the locus where the riemann invariants , that determine the edge of the spectrum ( as described in [ 71 ] ) exhibit a shock . this scenario is reminiscent of that for the small @xmath400-limit of the nonlinear schrdinger equation @xcite in which the analogue of our boundary curve is the envelope of _ dispersive shocks_. in that setting it is the zakharov - shabat inverse scattering problem that shows one how to pass through the dispersive shocks and describe a continuation of measure - valued solutions with so - called _ multi - gap _ support . it is our expectation that coupling gravity to an appropriate conformal field theory ( to thus arrive at a bona fide string theory ) @xcite will play a similar role in our setting to determine a unique continuation through the boundary curve of critical parameters to a unique equilibrium measure with multi - cut support . we also hope that this will help bring powerful methods from the study of dispersive limits of nonlinear pde into the realm of random matrix theory . * acknowledgement . * the author wishes to thank msri for its hospitality and the organizers for the excellent fall 2010 program on random matrix theory . most of the new results described here had their inception during that happy period . n. m. ercolani and k. d. t .- mclaughlin , _ asymptotics of the partition function for random matrices via riemann hilbert techniques , and applications to graphical enumeration _ , int . res . not . * 14 * , 755820 , 2003 . s. kamvissis , k. d. t .- mclaughlin and p. d. miller , _ semiclassical soliton ensembles for the focusing nonlinear schrdinger equation_. annals of mathematics studies , * 154 * princeton university press , princeton , nj , 2003 .

we study the coleman - de luccia ( cdl ) instanton characterizing the tunneling from a false vacuum to the true vacuum in a semi - classical way in drgt ( derham - gabadadze - tolley ) massive gravity theory , and evaluate the dependence of the tunneling rate on the model parameters . it is found that provided with the same physical hubble parameters for the true vacuum and the false vacuum as in general relativity ( gr ) , the thin - wall approximation method implies the same tunneling rate as gr . however , deviations of tunneling rate from gr arise when one goes beyond the thin - wall approximation and they change monotonically until the hawking - moss ( hm ) case . moreover , under the thin - wall approximation , the hm process may dominate over the cdl one if the value for the graviton mass is larger than the inverse of the radius of the bubble .

1312.0709

the notion of gravitational theory with a massive graviton is nothing new . a massive gravity theory was firstly proposed by fierz and pauli ( fp ) in 1939 where the theory of general relativity ( gr ) was extended by a linear mass term @xcite . however , lack of hamiltonian and momentum constraints leads to 6 degrees of freedom in this theory , with 5 of which corresponding to those of a massive spin-2 graviton and the rest one a boulware - deser ( bd ) ghost mode @xcite . a breakthrough was achieved by recent development of de rham - gabadadze - tolley ( drgt ) nonlinear massive gravity theory @xcite , where a special form of potential was introduced to recover the hamiltonian constraint so that the sixth bd ghost mode is eliminated @xcite . one of the most remarkable consequences in this theory is that it allows self - accelerating solutions @xcite , where the universe takes the de sitter form even without a bare cosmological constant , and its hubble scale is of the order of the graviton mass . however , application of the self - accelerating solution to explain the current accelerated expansion of universe does not solve the `` cosmological constant problem '' ( ccp ) @xcite , which implies a serious contradiction between smallness of the cosmological constant and expected large quantum corrections . motivated by the proposal that ccp may hopefully be solved by the anthropic selection of the cosmological constant in the landscape of vacua @xcite , the hawking - moss ( hm ) solution @xcite in drgt massive gravity was studied in @xcite . it was found that depending on the choice of parameters in drgt massive gravity theory , the non - vanishing mass of a graviton will influence the tunneling rate of hm instanton , hence affects the stability of a vacuum in this theory . on the other hand , it was known that in gr , traditionally , the hartle - hawking ( hh ) no - boundary wavefunction @xcite exponentially prefers small number of @xmath2-foldings near the minimum of the inflaton potential , hence it does not seem to predict the universe we observe today . such situation may be drastically changed by application of the correction term in the euclidean hm action to hh no - boundary proposal . it was found that for a wide range choice of the parameters in drgt massive gravity theory , the no - boundary wavefunction can peak at a sufficiently large value of the hubble parameter , hence one may obtain a sufficient number of @xmath2-folds of inflation @xcite . inspired by these interesting achievements , as a necessary step towards understanding of tunneling process in drgt massive gravity theory , it is necessary to explore another kind of phase transition in theories of scalar fields coupled to gravity : coleman - de luccia ( cdl ) instanton that exists for a special form of potentials such that the curvature scale of the barrier is large compared to the potential energy ( in planck units ) @xcite , and gradually approaches to the hm instanton as the curvature scale shrinks . in this paper , we consider the cdl solution for a scalar field with minimal coupling to gravity in drgt massive gravity theory . we set up the model and found the bounce solutions corresponding to the cdl instantons . based on these solutions , we evaluate the cdl action in two approaches : `` thin - wall '' approximation and perturbations around hm solution , and find monotonic contributions from the graviton mass terms until the `` thick - wall '' limit : the hm case . moreover , in the `` thin - wall '' limit , comparison of the tunneling rates for the hm and cdl solutions with the same potential shows that , even when the cdl process dominates over the hm one in the case of gr , it may behave inversely in the context of drgt massive gravity , i.e. depending on the values of parameters in this model , the hm process may dominate over the cdl one . this paper is organized as follows . in sec . [ setup ] , we setup the lagrangian for our model . in sec . [ s : basic ] , we formulate the equations of motion ( eom ) and solve the constraint equation . in sec . [ s : cdlthin ] , the cdl solution is studied by using `` thin - wall '' approximation . in sec . [ s : cdlperturb ] , we study the cdl solution as perturbations around the hawking - moss ( hm ) solution obtained in ref . @xcite and clarify its relationship with the one obtained by `` thin - wall '' approximation . in sec . [ s : hmvscdl ] , we compare the tunneling rates for the cdl and hm instantons for the same potential which satisfies the condition for `` thin - wall '' approximation . in sec . [ conclusion ] , we draw the conclusions . in appendix . [ perturbhm ] , we present the detailed calculations of the cdl process as perturbations around hm one . appendix . [ cdltohm ] is devoted to the deduction of comparison of the cdl to hm probabilities . throughout the paper , the lorentzian metric signature is set to be @xmath3 , while the euclidean metric signature @xmath4 . meanwhile , we use the conventional notations where indices with greek letters @xmath5 for the spacetime indices , the latin letters @xmath6 for the space indices , while the latin indices @xmath7 for the internal space ( lorentz frame ) indices . also repeated indices imply the summation unless otherwise stated . we study the tunneling process of a minimally coupled scalar field @xmath8 which tunnels from one vacua @xmath9 to another one @xmath10 in the context of drgt massive gravity , as illustrated in fig . [ fig : potential ] . the drgt massive gravity is composed of two metrics , namely a physical metric @xmath11 and a fiducial metric @xmath12 , with the stckelberg fields @xmath13 @xcite . as usual , the whole action can be divided into two parts : the nonlinear massive gravity part @xmath14 and the minimally coupled scalar field part @xmath15 as follows : throughout this paper . ] @xmath16 , \\ i_m & \equiv -\int\mr{d}^4x~\sqrt{-g}\left [ \frac{1}{2}(\pd \sigma)^2 + v(\sigma ) \right],\end{aligned}\ ] ] where @xmath17 , @xmath18 and @xmath19 are three free parameters in this model , and @xmath20 ^ 2 -\left[{\cal k}^2\right]\right)\ , , \nonumber\\ { \cal l}_3 & = \frac{1}{6}\left(\left[{\cal k}\right]^3 -3\left[{\cal k}\right]\left[{\cal k}^2\right]+2\left[{\cal k}^3\right]\right ) , \nonumber\\ { \cal l}_4 & = \frac{1}{24 } \left(\left[{\cal k}\right]^4 -6\left[{\cal k}\right]^2\left[{\cal k}^2\right ] + 3\left[{\cal k}^2\right]^2 + 8\left[{\cal k}\right]\left[{\cal k}^3\right ] -6\left[{\cal k}^4\right]\right),\end{aligned}\ ] ] with @xmath21 with two local minima @xmath9 and @xmath10 , which correspond to the false and true vacuum respectively , while @xmath22 labels its local maximum value . in the thin - wall approximation , the difference of the potential values at these two local minima is very small , i.e. @xmath23.,height=245 ] in order to investigate tunneling process , the potential @xmath24 is assumed to have two local minima @xmath9 and @xmath10 which correspond to the false and true vacuum , respectively , with a local maximum between them , @xmath25 , as illustrated in fig . [ fig : potential ] . the euclidean action of ( [ eq : actionorig ] ) is obtained by wick rotation @xmath26 and correspondingly @xmath27 . in the semiclassical limit , the tunneling rate per unit time per unit volume can be expressed in terms of the euclidean action as follows : @xmath28 -s_{\rm e}[{g}_{\mu\nu,{\rm f}},{\phi}_{\rm f}]\ , , \label{eq : rate}\end{aligned}\ ] ] where @xmath29 is the bounce solution , which is a solution of the euclidean equations of motion with appropriate boundary conditions , and @xmath30 is the solution of false vacuum @xcite . conventionally , the bounce solution @xmath31 is explored under the assumption of @xmath32-symmetry , since an @xmath32-symmetric solution gives the lowest action for a wide class of scalar - field theories , hence gives the least value of action which dominates the tunneling process @xcite . the same assumption is also reasonable in the presence of gravity @xcite , therefore , the physical metric can be assumed to take the following form , @xmath33 where @xmath34 is the metric on a three - sphere with @xmath35 , @xmath36 in the context of drgt massive gravity , the fiducial metric @xmath12 is assumed to be non - dynamical @xcite . in order to guarantee the @xmath32-symmetry , we assume that it is given by the de sitter metric @xcite with a constant hubble parameter @xmath37 @xcite : @xmath38 where @xmath39 it should be noted that here we stick to the lorentzian signature for the fiducial metric since it is non - dynamical in drgt massive gravity theory . nevertheless , thanks to the assumption of the de sitter fiducial metric , we may adopt the @xmath32-ansatz . in this section , we present the euclidean equations of motion . more detailed deductions can be found in @xcite . inserting the @xmath32-ansatz ( [ eq : o4metric ] ) and ( [ eq : bdefinition ] ) into ( [ eq : k ] ) , one obtains the euclidean version of the action ( [ eq : action ] ) @xmath40 where the gravity action is reduced to @xmath41,\end{aligned}\ ] ] with [ eq : mass ] @xmath42 and a dot means a derivative with respect to the radial coordinate , @xmath43 . meanwhile , the action for the tunneling field is reduced to @xmath44 . \end{aligned}\ ] ] variation of the action ( [ eq : euclidean ] ) with respect to the stckelberg field @xmath45 gives the following constraint equation : @xmath46 = 0\,,\end{aligned}\ ] ] where @xmath47 . correspondingly , we obtain two branches : @xmath48 in branch i , it is known that there exists a tension between the vainstein mechanism and the higuchi bound @xcite . this situation is not improved even in the extended massive gravity theories or bigravity theory . hence , in the following , we mainly concentrate on analysis of branch ii . the solution to eq . ( [ eq : b2 ] ) is given by @xmath49 where it should be noted that we require @xmath50 for our interest . hereafter , for definiteness , the choice of @xmath51 is called branch ii@xmath52 while @xmath53 is called branch ii@xmath54 . on the other hand , variations of the action ( [ eq : action ] ) with respect to lapse function @xmath55 and @xmath8 give the `` friedman equation '' and field equation respectively : @xmath56 where we have introduced the proper radial coordinate @xmath57 and a prime means derivative with respect to the proper time : @xmath58 , @xmath59 , while @xmath60\nonumber\\ & = -\frac{m_g^2}{\left(\alpha_3+\alpha_4\right)^2 } \left[\left(1+\alpha_3\right)\left(2+\alpha_3 + 2\,\alpha_3 ^ 2 - 3\,\alpha_4\right ) \pm 2\,\left(1+\alpha_3 + \alpha_3 ^ 2-\alpha_4\right)^{3/2}\right]\,.\end{aligned}\ ] ] as explained in the previous section , the tunneling rate ( [ eq : rate ] ) is given by the euclidean action evaluated for a solution of eqs . ( [ eq : eom ] ) and ( [ eq : eom1 ] ) . in the following , we construct the cdl solution and then evaluate the action for it . in order to evaluate the cdl solution by using thin - wall approximation , one should firstly solve the euclidean friedmann equation ( [ eq : eom ] ) at the locally minimal point @xmath61 and the globally minimal point @xmath62 separately : @xmath63 where @xmath64 is the point at which the tunneling process occurs . hence , from eq . ( [ eq : friedcdl ] ) , inside and outside solutions can be obtained as @xmath65 where for convenience , we set @xmath66 in the following . moreover , we have introduced the inside / outside hubble parameter of the physical metric by @xmath67 whereas it should be noted that the phase angle @xmath68 is determined by the continuity condition on the shell @xmath69 : @xmath70 using the constraint equation ( [ eq : ii ] ) and the definition for @xmath71 eq . ( [ eq : bdefinition ] ) , the following relationship holds : @xmath72 from which the expression for @xmath73 can be obtained as follows : @xmath74\ , , \qquad \tau<\tau_0 \\ & f^{-1}\cosh^{-1}\left[\frac{fx_\pm}{h_{\rm{f}}}\cos\left(h_{\rm{f}}\tau+\theta_{\rm{f}}\right)\right]\ , , \qquad \tau>\tau_0 \end{aligned } \right.\ ] ] hence , its derivative with respect to the proper radial coordinate @xmath75 can be obtained as : @xmath76 where we note that provided with the continuity equation ( [ eq : conti ] ) , @xmath73 is continuous at @xmath69 , but its derivative @xmath77 is discontinuous on the shell . inserting eqs . ( [ eq : acdlsol ] ) and ( [ eq : fderi ] ) into the euclidian action given by eqs . ( [ eq : euclidean ] ) and ( [ eq : ematter ] ) , and using @xmath78 , the total action can be expressed by division into three parts : @xmath79 & = \int { \rm d}^3x\sqrt{\omega } \int^{\pi/(2h_{\rm f})}_{-\pi/(2h_{\rm t})}{\rm d}\tau~ a^3(\tau ) \left[2\big(v(\sigma)+\lambda_{\pm}\big)-\frac{6k}{a^2(\tau ) } + m_g^2y_\pm \sqrt{-(f^{\prime})^2}\right]\nonumber\\ & = s_{\rm{inside}}+s_{\rm{outside}}+s_{\rm{wall}}\,,\end{aligned}\ ] ] where for brevity , we have introduced the parameter @xmath80 in terms of @xmath81 as follows : @xmath82 while @xmath83 , @xmath84 and @xmath85 are defined as : @xmath86\,,\label{sin}\\ s_{\rm{outside } } & \equiv \int { \rm d}^3x\sqrt{\omega } \int^{\pi/(2h_{\rm f})}_{\tau_0(1+\delta)}{\rm d}\tau~ a^3_{\rm f } \left[2\lambda_{\pm , { \rm f}}-\frac{6k}{a^2_{\rm f } } + m_g^2y_\pm |x_\pm|\sqrt{\frac{k-\left(a_{\rm{f}}h_{\rm{f}}\right)^2}{k-\left(a_{\rm{f}}fx_\pm\right)^2}}\right]\,,\label{sout}\\ s_{\rm wall } & \equiv \int { \rm d}^3x\sqrt{\omega } \int^{\tau_0(1+\delta)}_{\tau_0(1-\delta)}{\rm d}\tau~ a^3(\tau ) \left[2\big(v(\sigma)+\lambda_{\pm}\big)-\frac{6k}{a^2(\tau ) } + m_g^2y_\pm \sqrt{-(f^\prime)^2}\right]\,,\label{sw}\end{aligned}\ ] ] with an infinitely small parameter @xmath87 . in the following , we use thin - wall approximation to evaluate eqs . ( [ sin])([sw ] ) . recalling that the tunneling rate is expressed in terms of the euclidean action as shown in eq . ( [ eq : rate ] ) : @xmath88 -s_{\rm e}[{g}_{\mu\nu,{\rm f}},{\phi}_{\rm f}]\ , , \label{eq : rateb}\end{aligned}\ ] ] where the exponential factor @xmath89 can be divided into three parts with respect to the integration boundaries for @xmath75 , in accordance to the division of action @xmath90 in eq . ( [ eq : actioncdl ] ) : @xmath91 where @xmath92 with @xmath93 the corresponding euclidean action of the false vacuum . it immediately follows that @xmath94 , since the bouncing solution outside the bubble @xmath95 coincides with that of false vacuum . so in the following , it is unnecessary to evaluate the euclidean action @xmath96 . now we turn to evaluate @xmath97 . from eq . ( [ eq : eom ] ) , one obtains the following relationship : @xmath98}\,.\end{aligned}\ ] ] inserting eq . ( [ eq : adot ] ) into ( [ sin ] ) and rewrite the integration as follows : @xmath99 with @xmath100 , then from eq . ( [ eq : bres ] ) , the exponential factor inside the bubble can be expressed as : @xmath101\nonumber\\ & -\int^{a_0}_{0}\frac{a^3{\rm d}a}{\sqrt{k - a^2\lambda_{\pm , { \rm f}}/3}}~ \left[2\lambda_{\pm , { \rm f}}-\frac{6k}{a^2 } + m_g^2y_\pm |x_\pm|\sqrt{\frac{k-\left(ah_{\rm{f}}\right)^2}{k-\left(afx_\pm\right)^2}}\right]\bigg\}\nonumber\\ & = -12\pi^2k^{-\frac{3}{2}}\int^{a_0}_{0}a{\rm d}a\left[\sqrt{k-(ah_{\rm t})^2}-\sqrt{k-(ah_{\rm f})^2}\right]\,,\label{bin}\end{aligned}\ ] ] where we have used @xmath102 in the first step and eq . ( [ eq : hubblecdl ] ) in the last step . it should be noted that though the term proportional to @xmath103 inside the bubble ( the last term in the first line of eq . ( [ bin ] ) ) eliminates with the corresponding term outside ( the last term in the second line ) , nevertheless , the mass term contributes to the effective cosmological constant @xmath104 as shown in eq . ( [ eq : friedcdl ] ) , hence appears in the corresponding hubble parameters @xmath0 and @xmath1 by eq . ( [ eq : hubblecdl ] ) . in order to evaluate the exponential factor on the wall @xmath105 , we use the thin - wall approximation @xcite , @xmath106 hence , eq . ( [ eq : eom1 ] ) can be easily solved as : @xmath107}\,.\end{aligned}\ ] ] using the relationship @xmath108 inserting eq . ( [ eq : derisig ] ) into the relationship above , the exponential factor on the wall can be evaluated as follows : @xmath109 } } \left[2\big(v(\sigma)+\lambda_{\pm}\big)-\frac{6k}{a^2_0 } + m_g^2y_\pm \sqrt{- ( f^\prime)^2}\bigg|_{\tau<\tau_0}\right]\,,\label{swall}\end{aligned}\ ] ] while the euclidean action of false vacuum on the wall can be written as : @xmath110 } } \left[2\big(v(\sigma_{\rm{f}})+\lambda_{\pm}\big)-\frac{6k}{a^2_0 } + m_g^2y_\pm \sqrt{- ( f^\prime)^2}\bigg|_{\tau>\tau_0}\right]\,,\label{swallf}\end{aligned}\ ] ] then inserting the above two equations into eq . ( [ eq : bres ] ) , under the assumption that @xmath111 one obtains the tunneling rate factor on the wall : @xmath112 } + \mathcal{o}\left(\epsilon\right)\,.\end{aligned}\ ] ] hence , combining eqs . ( [ bin ] ) and ( [ bwall ] ) , the whole tunneling rate factor can be expressed as @xmath113- a^3_0\sigma\right\}\,,\end{aligned}\ ] ] where the tension @xmath114 is defined as follows : @xmath115}\,,\end{aligned}\ ] ] and @xmath116 is determined by demanding that @xmath89 is stationary : @xmath117 thus , comparing eqs . ( [ ball ] ) and ( [ eq : bcri ] ) to the case in gr @xcite , we find that in thin - wall limit , provided with the same value of @xmath116 , @xmath0 and @xmath1 , respectively , the tunneling rate for the cdl instantons in nonlinear massive gravity is the same as the one in gr . however , investigation of hm instantons in drgt massive gravity theory shows contributions to tunneling rate coming from the graviton mass @xcite . hence , in the next section , we take another limit of solutions `` thick wall '' approximation to investigate the cdl solution . the hawking - moss ( hm ) instanton in nonlinear massive gravity has been discussed in details in ref . @xcite . in this subsection , we make a brief review on the results . a hm solution can be found by setting the tunneling field to the local maximum value , @xmath118 , as illustrated in fig . [ fig : potential ] . then the equation of motion ( [ eq : eom1 ] ) is trivially satisfied and the euclidean friedmann equation ( [ eq : eom ] ) reduces to @xmath119 setting the boundary condition @xmath120 and assuming @xmath121 , the hm solution is obtained as @xmath122 where @xmath123 then under the constraint equaiton ( [ eq : ii ] ) , one obtains @xmath124 taking derivative with respect to @xmath75 on both sides of eq . ( [ eq : bf ] ) , it immediately follows that : @xmath125 where the parameter @xmath126 is defined as @xmath127 provided that @xmath50 , then the parameter @xmath128 . moreover , we note that from eq . ( [ eq : sqfhm2der ] ) , it is clear that at range @xmath129 , singularities will appear unless @xmath130 . hence , for consistency of the theory , we derive the constraint : @xmath131 inserting eq . ( [ eq : eom2 ] ) into the euclidian action eqs . ( [ eq : euclidean ] ) and ( [ eq : ematter ] ) , and using @xmath78 , the total action can be expressed as @xmath132\ , , \end{aligned}\ ] ] where the function @xmath133 is defined as follows : @xmath134 it is clear from eq . ( [ eq : actionhm ] ) that provided with the same hm hubble parameter , the second term proportional to @xmath80 is the correction term arising from the non - vanishing graviton mass . hence , compared with the corresponding hawking - moss tunneling rate in gr : @xmath135 , one obtains the correction term arising from the mass of graviton : @xmath136\,.\end{aligned}\ ] ] since the function @xmath133 is both positive and monotonically increasing for @xmath137 , we have @xmath138 . together with @xmath139 , we find that the sign of @xmath140 is determined by that of @xmath80 , i.e. for @xmath141 ( @xmath142 ) , the correction @xmath143 ( @xmath144 ) , which implies that the tunneling rate is enhanced ( suppressed ) compared to the case of gr . the cdl solution can be also investigated as perturbations around the hm solution @xcite . firstly , we expand the potential @xmath24 around @xmath145 as follows , @xmath146 where we have introduced @xmath147 . near the hm limit where @xmath148 with @xmath149 , the regular solutions are perturbatively found to be @xmath150 + \mathcal{o}(\varepsilon_m^3)\\ & \equiv & a_0(\tau)+\delta a(\tau)\,,\nonumber\\ \sigma(\tau ) & = & \sigma_{\rm hm}+\varepsilon_m h_{\rm hm}\sin\left(\tilde{h}_{\rm hm}\tau\right)+\frac{\varepsilon_m^2m}{12}\left[1 - 2\sin^2\left(\tilde{h}_{\rm hm}\tau\right)\right]\nonumber\\ & -&\!\!\varepsilon_m^3h_{\rm hm}\sin\left(\tilde{h}_{\rm hm}\tau\right)\!\!\left[\frac{3h_{\rm hm}^2 - 4\mu}{56}\cos^2\left(\tilde{h}_{\rm hm}\tau\right)-\frac{m^2}{36h_{\rm hm}^2}\sin^2\left(\tilde{h}_{\rm hm}\tau\right)\right]\!+\!\mathcal{o}(\varepsilon_m^4),\label{eq : perturbscalar}\end{aligned}\ ] ] where @xmath151 , @xmath152 and @xmath153 , while for simplicity of symbols , we define the background value @xmath154 and the perturbations around it as @xmath155 . as in the case of gr , perturbations of euclidean hilbert - einstein action vanish up to order @xmath156 . hence , up to order @xmath157 , it is sufficient to evaluate the mass part of the action : @xmath158\right\}\nonumber\\ & = & 4\pi^2m_g^2y_\pm\int_0^{\pi/2\tilde{h}_{\rm hm } } { \rm d}\tau~a_{\rm 0}^2\left[~3\sqrt{-(f_0^\prime)^2}~\delta a+a_{\rm 0}\delta\sqrt{- ( f^\prime)^2}~\right]\,,\end{aligned}\ ] ] where in the last step , we have used the constraint equations @xmath159 and correspondingly , @xmath160 . it is shown in appendix [ perturbhm ] that by using eq . ( [ eq : perturba ] ) , the second term @xmath161 can be expressed as follows : @xmath162\,,\end{aligned}\ ] ] where @xmath163 . thus , inserting eqs . ( [ eq : perturba])and ( [ df1 ] ) into ( [ dsformal ] ) , one finally obtains the second order perturbation in action as : @xmath164 as can be seen above , since @xmath50 , the sign of perturbation depends on parameter @xmath80 defined by eq . ( [ eq : y ] ) . when @xmath165 , @xmath166 so that hm solution dominates , while if @xmath167 , @xmath168 so that the cdl solution dominates , which is in sharp difference from the case of gr where the cdl instanton always dominates over hm one , if it exists . generally speaking , it is difficult to analytically estimate the tunneling rate beyond hm or thin - wall approximation . nevertheless , in this subsection , we present an estimation of the qualitative behavior for a more general case . firstly , we rewrite the previous results in an uniform way . the mass term can be written as follows , @xmath169 where in the second step we have inserted eqs . ( [ eq : ii ] ) , ( [ eq : y ] ) and @xmath102 . on the other hand , taking derivative of eq . ( [ eq : bfa ] ) with respect to @xmath75 , one can express @xmath170 in terms of @xmath171 and @xmath172 as follows : @xmath173 hence , inserting eq . ( [ eq : fdotsol ] ) into ( [ eq : mass ] ) , one can express @xmath174 as integration of @xmath172 instead of @xmath75 as follows @xmath175_{0}^{a_{\rm max}}\,,\end{aligned}\ ] ] where @xmath176 is the largest radius of the bubble in euclidean time . for convenience of discussion , we define @xmath177 as follows @xmath178_{a_{\rm{f , max}}}^{a_{\rm max}}\,,\end{aligned}\ ] ] which is the correction term arising from the graviton mass when one calculates the tunneling rate by eq . ( [ eq : rate ] ) . now let us reconsider the conclusions drawn from thin - wall and hm limit by evaluating @xmath179 in these two cases , respectively . we recall that in the thin - wall limit , the maximum value of the scale factor is equal to that of the false vacuum : @xmath180 , as illustrated in the right panel of fig . [ fig : amax ] . hence , using eqs . ( [ eq : bres ] ) and ( [ eq : massa ] ) , one obtains a vanishing correction term in the thin - wall limit : @xmath181 which leads to the conclusion in sec . [ s : cdlthin ] that the non - vanishing mass of the graviton does not contribute to the cdl tunneling rate in the thin - wall limit . on the other hand , the hm limit corresponds to a `` thick - wall '' limit , where the maximum value of the scale factor is given by that of the local maximum between true and false vacuum , @xmath182 ( see the left panel of fig . [ fig : amax ] ) , hence leads to a non - vanishing correction term for tunneling rate : @xmath183\,,\end{aligned}\ ] ] where in the last step , we set @xmath184 while @xmath126 and @xmath133 are defined in eqs . ( [ eq : reparameter ] ) and ( [ eq : a ] ) , respectively . it is obvious that eq . ( [ eq : binamaxhm ] ) coincides with ( [ eq : deltab ] ) as expected . comparison of eq . ( [ eq : binamaxtw ] ) to ( [ eq : binamaxhm ] ) suggests the expectation that deviations from thin - wall and hm limit may lead to contributions to the tunneling rate which change monotonically in @xmath185 in nonlinear massive gravity theory . hence , in the following , we consider deviations from thin - wall and hm limit , respectively . in cases of hawking - moss instanton ( left panel ) and coleman - de luccia instanton ( right panel ) . in both cases , the evolution of the scale factor is labeled by red curves , while that of false vacuum @xmath186 and true vacuum @xmath187 are labeled by gray and blue curves , respectively . from the left panel , it is obvious that the maximum value of scale factor in hm case is different from that of false vacuum , which leads to a non - vanishing contribution to the tunneling rate as shown in eq . ( [ eq : binamaxhm ] ) . however , from the right panel , one finds that in the cdl case , the scale factor firstly evolves along that of true vacuum until some point @xmath64 , then follows along that of false vacuum so that its maximum value coincides with that of false vacuum . hence , using eq . ( [ eq : bdefine ] ) , contribution from the graviton mass vanishes in the cdl case.,title="fig:",height=207 ] in cases of hawking - moss instanton ( left panel ) and coleman - de luccia instanton ( right panel ) . in both cases , the evolution of the scale factor is labeled by red curves , while that of false vacuum @xmath186 and true vacuum @xmath187 are labeled by gray and blue curves , respectively . from the left panel , it is obvious that the maximum value of scale factor in hm case is different from that of false vacuum , which leads to a non - vanishing contribution to the tunneling rate as shown in eq . ( [ eq : binamaxhm ] ) . however , from the right panel , one finds that in the cdl case , the scale factor firstly evolves along that of true vacuum until some point @xmath64 , then follows along that of false vacuum so that its maximum value coincides with that of false vacuum . hence , using eq . ( [ eq : bdefine ] ) , contribution from the graviton mass vanishes in the cdl case.,title="fig:",height=207 ] the hm instanton corresponds to a `` thick - wall '' limit where the curvature of local maximum of the potential is flat enough : @xmath188 . using the perturbation approach in sec . [ perturb ] , from eq . ( [ eq : perturba ] ) , small deviation from hm instanton implies that @xmath189 inserting eq . ( [ eq : amaxperturb ] ) into ( [ eq : massa ] ) , the correction term for the perturbational approach around hm limit is evaluated as @xmath190+\frac{m_g^2\pi^2x_\pm y_\pm{h}_{\rm hm}^2\varepsilon_m^2}{2\tilde{h}_{\rm hm}^4\sqrt{1-\tilde{\alpha}^2}}+\mathcal{o}(\varepsilon_m^4)\nonumber\\ & = -\frac{4\pi^2m_g^2x_\pm y_\pm}{3(fx_\pm)^4 } \left[\sqrt{1-\alpha_{\rm hm}^2}\left(2+\alpha_{\rm hm}^2\right)-\sqrt{1-\alpha_{\rm f}^2}\left(2+\alpha_{\rm f}^2\right)\right]+\frac{m_g^2\pi^2x_\pm y_\pm\varepsilon_m^2}{3h_{\rm hm}^4\sqrt{1-\alpha_{\rm hm}^2}}\nonumber\\ & \quad+\mathcal{o}(\varepsilon_m^4)\,.\end{aligned}\ ] ] obviously , in the second step , the correction term of order @xmath157 coincides with eq . ( [ s2 ] ) as expected . hence , the result in sec . [ perturb ] is recovered by applying eq . ( [ eq : massa ] ) . + for an intuitive analysis of deviation from thin - wall approximation , let us consider the classical trajectory of @xmath191 , where the scalar field is driven from the true vacuum @xmath10 toward the false vacuum @xmath9 , as illustrated in fig . [ fig : potential2 ] . in the thin - wall limit , the friction term @xmath192 in eq . ( [ eq : eom1 ] ) is neglected so that @xmath8 can reach @xmath9 because of conservation of total energy . however , small deviation from the thin - wall limit implies a non - negligible friction term which causes loss of energy so that the scalar field starting from @xmath10 can only reach some point @xmath193 where @xmath194 . correspondingly , to calculate the maximum radius of the bubble , we have @xmath195 , then inserting into eq . ( [ eq : eom ] ) , one can evaluate in the following way : @xmath196 . when friction term @xmath197 is taken into account , scalar field starting around @xmath10 along the potential can not reach @xmath9 as in the thin - wall case . instead , it will stop at some point @xmath193 and correspondingly gives the maximum radius @xmath198 as shown in eq . ( [ eq : astar ] ) , hence leads to correction of tunneling rate arising from the non - vanishing graviton mass.,height=245 ] again inserting eq . ( [ eq : astar ] ) into ( [ eq : massa ] ) , the correction term can be evaluated as : @xmath199\neq0\,.\end{aligned}\ ] ] hence , small deviation from thin - wall limit makes @xmath200 , which furthermore leads to corrections to the tunneling rate for the cdl instanton when compared to gr . for a qualitative analysis , it is convenient to define a normalized function @xmath201 in the following way : @xmath202_{h_{\rm{f}}^{-1}}^{a_{\rm{max}}}\nonumber\\ & \equiv h(\alpha_{\rm max})-h(\alpha_{\rm f})\geq0\,,\end{aligned}\ ] ] where we have defined the function @xmath203 and variable @xmath126 as follows : @xmath204 which is plotted in fig . [ fig : bmass ] . it should be noted that @xmath205 coincides with the one in ref . @xcite and leads to a constraint on the height of potential at false vacuum : @xmath206 inserting eq . ( [ eq : bredefine ] ) into ( [ eq : bdefine ] ) , the correction to the tunneling rate in drgt massive gravity can be expressed as : @xmath207 provided with definite parameters @xmath103 , @xmath18 and @xmath19 , it is convenient to consider the behavior of @xmath201 with respect to @xmath185 , as plotted in fig . [ fig : bmass ] . the values of normalized factor @xmath208 defined in eq . ( [ eq : bredefine ] ) are illustrated by black , blue and green double arrow lines , corresponding to the value in cases of hm ( sec . [ hmcase ] ) , perturbations from hm ( sec . [ perturb ] ) and deviation from thin - wall limit ( sec . [ devitw ] ) , respectively . in the thin - wall limit @xmath209 ( i.e. @xmath210 ) , hence the graviton mass has no contribution to the cdl tunneling rate as shown in eq . ( [ eq : binamaxtw ] ) . provided with @xmath141 , under deviations from thin - wall limit , @xmath211 ( i.e. @xmath212 ) . hence , a non - vanishing contribution to the cdl tunneling rate arises ( as shown in the green double arrow line ) and increases gradually , as evaluated in eq . ( [ eq : binamaxstar ] ) . on the other hand , the most probable point is the hm limit which can be interpreted as `` thick - wall '' limit and correspondingly gives the largest tunneling rate as expected ( black double arrow line ) . perturbations around this limit give larger maximum radius @xmath213 ( i.e. @xmath214 ) as shown in eq . ( [ eq : amaxperturb ] ) , so the tunneling rate decreases monotonically ( blue double arrow line ) . eq . ( [ eq : bredefine ] ) implies monotonic behavior of function @xmath201 when @xmath185 changes from @xmath215 to @xmath216 . correspondingly , a monotonic behavior of the cdl tunneling rate with respect to different @xmath185 is expected . it should be noted that if @xmath167 , the above conclusion holds inversely , while @xmath217 implies vanishing contribution to the tunneling rate . since @xmath137 , the value of @xmath201 is of order unity : @xmath218 . moreover , when @xmath219 , the fiducial metric becomes minkowskian . in this case , eq . ( [ eq : bredefine1 ] ) reduces to the following form @xmath220\,.\end{aligned}\ ] ] similarly as the behavior in de sitter fiducial metric case as shown in eq . ( [ eq : bredefine1 ] ) , when @xmath165 , contribution to the cdl tunneling rate arising from the graviton mass appears when one go beyond `` thin - wall '' approximation , and increases monotonically until its maximum value at hm point . if @xmath167 , the conclusion holds inversely . defined in eq . ( [ eq : f ] ) . for a given model , @xmath221 and @xmath222 are definite , from which one obtains @xmath223 and @xmath224 . using eq . ( [ eq : bredefine ] ) , the value of normalized factor @xmath208 is illustrated by black , blue and green double arrow lines , corresponding to the value in cases of hm ( sec . [ hmcase ] ) , perturbations from hm ( sec . [ perturb ] ) and deviation from thin - wall limit ( sec . [ devitw ] ) , respectively . it should be noted that since @xmath225 $ ] , the value of @xmath226 is of order unity : @xmath227.,height=245 ] in the previous section , it is found that when compared to the situation in gr , correction to the cdl tunneling rate will appear because of the non - vanishing graviton mass , and its value will change monotonically with respect of @xmath185 until hm solution , i.e. from `` thin - wall '' to `` thick - wall '' case . correspondingly , the physical picture of the analysis is that the shape of the potential changes gradually : the rate of its typical height of the local maximum to its width decreases monotonically . on the other hand , it is interesting to consider another prospect : provided with the same shape of potential which satisfies the `` thin - wall '' condition eq . ( [ twcondition ] ) , whether cdl instanton will dominate over hm one or inversely , which may imply sharp difference from gr where the cdl one always dominates if it exists . the comparison of the probability of the cdl process to that of hm is expressed as follows ( for details of deduction see appendix [ cdltohm ] ) : @xmath228\,,\end{aligned}\ ] ] assumming @xmath229 and @xmath230 with @xmath231 , where @xmath133 is defined in eq . ( [ eq : a ] ) . in the context of gr where @xmath232 , under the thin - wall approximation , the probability of the cdl instanton always dominates over the hm one . however , in drgt massive gravity theory , there appears a term which is proportional to the mass of graviton , hence gives rise to the possibility that hm process may dominate over the cdl one when @xmath141 . it should be noted that , even when @xmath233 , where the fiducial metric reduces to minkowskian one , the function @xmath133 is finite , @xmath234 , and then the ratio ( [ bcdlhm ] ) is non - singular as has been stated in ref . @xcite . to find such a case , we note that within the range @xmath235 $ ] , the function @xmath133 of order unity . moreover , the thin - wall approximation implies that @xmath236 . hence , provided that the parameters @xmath237 , one finds the condition on the value of graviton mass for hm process dominance : @xmath238 where @xmath116 is the radius of bubble defined in eq . ( [ eq : bcri ] ) while eq . ( [ eq : bcriapp ] ) has been used in the last step . is shown for branch ii@xmath239 ( left panel ) and branch ii@xmath240 ( right panel ) solutions , respectively . the white region corresponds to @xmath241 while in the gray one @xmath242 , both of which should be excluded since the cosmological solution does not exist in these regions @xcite . in pink region , @xmath243 and @xmath167 , while green region corresponds to @xmath244 with @xmath165 . in blue region , @xmath245 and @xmath246 , while yellow region corresponds to @xmath247 and @xmath248 . along the solid lines ( defining the boundary between blue and pink regions for ii@xmath239 , while the boundary between pink and yellow regions for ii@xmath240 ) , @xmath217 so that it reduces to the case in gr . the dotted lines denote @xmath249 where the solution ceases to exist . along the dashed lines , @xmath250 diverges and thus defines another boundary of the solution space.,title="fig:",height=245 ] is shown for branch ii@xmath239 ( left panel ) and branch ii@xmath240 ( right panel ) solutions , respectively . the white region corresponds to @xmath241 while in the gray one @xmath242 , both of which should be excluded since the cosmological solution does not exist in these regions @xcite . in pink region , @xmath243 and @xmath167 , while green region corresponds to @xmath244 with @xmath165 . in blue region , @xmath245 and @xmath246 , while yellow region corresponds to @xmath247 and @xmath248 . along the solid lines ( defining the boundary between blue and pink regions for ii@xmath239 , while the boundary between pink and yellow regions for ii@xmath240 ) , @xmath217 so that it reduces to the case in gr . the dotted lines denote @xmath249 where the solution ceases to exist . along the dashed lines , @xmath250 diverges and thus defines another boundary of the solution space.,title="fig:",height=245 ] in order to see the possibilities for @xmath165 , in fig . [ fig : xbranches ] , we show the sign of @xmath80 in the parameter space @xmath251 for branch ii@xmath239 ( left panel ) and branch ii@xmath240 ( right panel ) solutions , respectively . the value of parameter @xmath250 is further constrained by the mass of tensor mode for self - accelerating solutions @xcite @xmath252 which implies the constraints on parameter @xmath250 to avoid the tachyonic instability : in branch ii@xmath239 , @xmath253 and in branch ii@xmath240 , @xmath254 . hence , as shown in fig . [ fig : xbranches ] , there exists a region for @xmath246 within the constraints ( the green region in branch ii@xmath239 ) while in branch ii@xmath240 , under the constraint @xmath254 , the parameter @xmath255 is always negative . hence , in branch ii@xmath239 , there exists the case where hm process would dominate over the cdl one , provided with eq . ( [ mgconstr ] ) . however , in branch ii@xmath240 , the cdl process always dominates . moreover , it should be noted that in the limit where @xmath256 , @xmath51 diverges so that only branch ii@xmath240 solution exists @xcite . thus , under the thin - wall approximation where the potential is very `` sharp '' at its local maximum , when the value of graviton mass is large enough so that eq . ( [ mgconstr ] ) is satisfied , the hm process may dominate over the cdl one , which is very different from the case of gr . so we conclude that in the context of drgt massive gravity theory , not only the shape of the potential , but also the values of parameters @xmath18 , @xmath19 and @xmath17 will influence the tunneling process . towards the understanding of stability of vacuum in the landscape of vacua in drgt massive gravity , in this paper , we investigated the coleman - deluccia ( cdl ) solution , under the assumption that a tunneling field minimally couples to gravity . for comparison with hawking - moss ( hm ) instanton @xcite , we choose branch - ii@xmath257 for analysis and evaluate the corresponding tunneling rate of the cdl instanton . firstly , we used the `` thin - wall '' approximation @xcite and found that the non - vanishing graviton mass term does not contribute to the tunneling rate . to compare this result with the hm case , where the non - vanishing correction arises @xcite , we derived the cdl solution as perturbations around the hm case , which corresponds to a `` thick - wall '' approximation ( or equivalently , the potential is very `` flat '' around its local maximum ) . in this approach , we found non - vanishing second - order perturbation to the tunneling rate due to the non - vanishing graviton mass , as shown in eq . ( [ s2 ] ) , which implies correction to the tunneling rate for the cdl instanton . moreover , it is found that in this approach , when the parameter @xmath165 ( defined in eq . ( [ eq : y ] ) ) , hm process will dominate over the cdl one even when the cdl solutions exist . in order to go beyond the thin - wall and thick - wall approximations , we rewrite the corrections to the tunneling rate due to the graviton mass in terms of @xmath185 , which is the largest radius of the bubble in euclidean time ( eqs . ( [ eq : massa])([eq : bdefine ] ) ) . it is found that in the thin - wall approximation , @xmath185 coincides with the scale factor for the false vacuum , hence the contributions from the graviton mass cancel with the counter term of the false vacuum . corrections to the cdl tunneling rate appear when one considers the deviations from thin - wall approximation , and its value varies monotonically with respect to @xmath185 until the hm case , as illustrated in fig . [ fig : bmass ] . moreover , provided with the same shape of potential which satisfies the condition for the thin - wall approximation , we compare the probabilities for the hm and cdl process . it is found that if the typical value for the graviton mass is larger than the inverse of bubble radius , the hm process may dominate over the cdl one , which is very different from the situation in gr . hence , in drgt massive gravity theory , not only the shape of the potential but also the value of the parameters @xmath18 , @xmath19 and @xmath17 will qualitatively influence the tunneling process . on the other hand , it is known that drgt massive gravity theory suffers from some problems @xcite . hence , as one step towards a more realistic model , it is necessary to study the tunneling issues in the extended massive gravity theories : for example , quasi - dilaton massive gravity @xcite , varying - mass massive gravity @xcite or @xmath258 massive gravity @xcite . especially , in the varying - mass massive gravity theory , due to the mass dependence , the effective cosmological constant @xmath104 of some vacuum may become larger even with a relatively smaller potential energy , which may imply a scenario of tunneling from a lower potential energy to a higher one . investigations of the corresponding tunneling process is one of the future studies . in this appendix , we present detailed calculations of perturbations in euclidean action around hm solutions . it is obvious that from in eq . ( [ dsformal ] ) , provided with eq . ( [ eq : perturba ] ) , for our calculation of the perturbations of the action up to 2nd order , we should firstly consider the perturbations in the term @xmath259 . under perturbation @xmath260 , we have : @xmath261 the constraint eq . ( [ eq : ii ] ) implies the relationship @xmath262 , so we have @xmath263 while @xmath264 combining these two equations , we then obtain the following relationship : @xmath265 on the other hand , noting that @xmath266 with @xmath267 , we have : @xmath268 from which one immediately obtains that @xmath269 inserting eqs . ( [ eq : perturba ] ) and ( [ bf ] ) into eq . ( [ f ] ) , we obtain the following expression : @xmath270 hence , from eqs . ( [ eq : perterbf ] ) and ( [ eq : fcorr ] ) , we conclude that at range @xmath271 , the the second order perturbation arising from the term @xmath259 can be expressed in the following way : @xmath272\,.\end{aligned}\ ] ] now let us calculate the two parts in eq . ( [ dsformal ] ) separately . the first part reads @xmath273\,,\end{aligned}\ ] ] where in the second step , @xmath274 , and in the third step , @xmath275 . meanwhile , the second part of eq . ( [ dsformal ] ) can be calculated by using eq . ( [ df ] ) as follows : @xmath276 thus , inserting eqs . ( [ p1 ] ) and ( [ p2 ] ) into ( [ dsformal ] ) , one finally obtains the second order perturbation in euclidean action : @xmath277 in this appendix , we derive eq . ( [ bcdlhm ] ) in details . for simplicity , let us assume @xmath184 and set the effective cosmological constant of the true vacuum @xmath278 , while that of the false vacuum @xmath279 . then from eq . ( [ ball ] ) , one obtains : @xmath280- a^3_0\sigma\right\}\nonumber\\ & = -12\pi^2\left\{\frac{a_0 ^ 2}{2}+\frac{1}{\epsilon}\left[\left(1-\frac{a_0 ^ 2}{3}\epsilon\right)^\frac{3}{2}-1\right]\right\ } + 2\pi^2a_0 ^ 3\sigma\,,\end{aligned}\ ] ] which is stationary when @xmath281 inserting eq . ( [ a0extreme ] ) into ( [ bcomp ] ) , one obtains : @xmath282\,.\end{aligned}\ ] ] in order to evaluate the absolute value term in the right - hand side of eq . ( [ babso ] ) , we note that from eq . ( [ eq : bcri ] ) , the extreme value for tension @xmath114 can be also evaluated as follows : @xmath283 since @xmath284 , eq . ( [ eq : bcriapp ] ) implies that @xmath285 . thus , from eq . ( [ babso ] ) , we finally obtain that : @xmath286\right\}\nonumber\\ & = \frac{24\pi^2}{\epsilon}\left(1+\frac{4\epsilon}{3\sigma^2}\right)^{-2}\,.\end{aligned}\ ] ] hence , the corresponding tunneling rate defined in eq . ( [ eq : rate ] ) is expressed as : @xmath287 \approx\exp\left(-\frac{24\pi^2}{\epsilon}+\frac{64\pi^2}{\sigma^2}\right)\,,\end{aligned}\ ] ] while by using eqs . ( [ eq : bredefine ] ) and ( [ eq : bredefine1 ] ) , that of thick - wall approximation can be expressed as @xcite @xmath288\nonumber\\ & \approx\exp\left(-\frac{24\pi^2}{\epsilon}+\frac{24\pi^2}{\lambda_{\pm , \rm eff}}+\frac{4\pi^2m_g^2y_\pm\mathfrak{b}^{\rm mass}}{3f^4x_\pm^3}\right)\,.\end{aligned}\ ] ] noting that from eq . ( [ tension ] ) , one can make a comparison between @xmath290 and @xmath291 as follows ( here @xmath292 is recovered for comparison ) @xmath293 where @xmath294 and we have used @xmath295 . on the other hand , expending the potential near @xmath22 , we have @xmath296 where @xmath147 . so one obtains @xmath297 we note that eq . ( [ comparresult ] ) can be justified in another way : from eq . ( [ eq : bcri ] ) , using the relationship @xmath301 , the radius of the bubble can be expressed as @xcite @xmath302 where we have used the assumption @xmath303 and @xmath304 . on the other hand , using eqs . ( [ eq : adot ] ) and ( [ eq : derisig ] ) , one finds that @xmath305 so using the relationship @xmath306 , the thickness of the wall can be approximately evaluated as @xmath307 where @xmath308 . the thin - wall approximation is valid if @xmath309 , so one obtains @xmath310 @xcite , which verifies eq . ( [ comparresult ] ) . thus , in the thin - wall approximation , eq . ( [ bcdlhmapp1 ] ) reduces to the following form : @xmath311\,,\end{aligned}\ ] ] where we used the eq . ( [ eq : a ] ) for definition of function @xmath133 . in the context of gr where @xmath232 , provided that the cdl instantons exist , the cdl process always dominates over the hm one @xcite . we thank stefano ansoldi , qing - guo huang and kazuyuki sugimura for helpful discussions . this work was supported in part by the grant - in - aid for the global coe program `` the next generation of physics , spun from universality and emergence '' from the ministry of education , culture , sports , science and technology ( mext ) of japan , and by jsps grant - in - aid for scientific research ( a ) no . rs is supported by a jsps grant - in - aid through the jsps postdoctoral fellowship no . 23@xmath3123430 . d. g. boulware and s. deser , phys . d * 6 * , 3368 - 3382 ( 1972 ) . p. creminelli , a. nicolis , m. papucci and e. trincherini , jhep * 0509 * , 003 ( 2005 ) [ hep - th/0505147 ] . v. a. rubakov and p. g. tinyakov , phys . usp . * 51 * , 759 - 792 ( 2008 ) . [ arxiv:0802.4379 [ hep - th ] ] . s. f. hassan , r. a. rosen and a. schmidt - may , jhep * 1202 * , 026 ( 2012 ) [ arxiv:1109.3230 [ hep - th ] ] . g. damico , c. de rham , s. dubovsky , g. gabadadze , d. pirtskhalava and a. j. tolley , phys . d * 84 * , 124046 ( 2011 ) [ arxiv:1108.5231 [ hep - th ] ] . a. e. gmrkolu , c. lin and s. mukohyama , jcap * 11 * , 030 ( 2011 ) . [ arxiv:1109.3845 [ hep - th ] ] . s. weinberg , rev . * 61 * , 1 ( 1989 ) . s. nobbenhuis , found . phys . * 36 * , 613 ( 2006 ) [ gr - qc/0411093 ] . l. susskind , in * carr , bernard ( ed . ) : universe or multiverse ? * 247 - 266 [ hep - th/0302219 ] . s. w. hawking and i. g. moss , phys . b * 110 * , 35 ( 1982 ) . y. zhang , r. saito and m. sasaki , jcap * 1302 * , 029 ( 2013 ) . [ arxiv:1210.6224 [ hep - th ] ] . s. f. hassan and r. a. rosen , jhep * 1202 * , 126 ( 2012 ) [ arxiv:1109.3515 [ hep - th ] ] . m. s. volkov , jhep * 1201 * , 035 ( 2012 ) [ arxiv:1110.6153 [ hep - th ] ] . d. comelli , m. crisostomi , f. nesti and l. pilo , jhep * 1203 * , 067 ( 2012 ) [ erratum - ibid . * 1206 * , 020 ( 2012 ) ] [ arxiv:1111.1983 [ hep - th ] ] . d. langlois and a. naruko , class . * 29 * , 202001 ( 2012 ) . [ arxiv:1206.6810 [ hep - th ] ] . c. de rham and s. renaux - petel , jcap * 1301 * , 035 ( 2013 ) [ arxiv:1206.3482 [ hep - th ] ] ; s. deser , k. izumi , y. c. ong and a. waldron , phys . b * 726 * , 544 ( 2013 ) [ arxiv:1306.5457 [ hep - th ] ] . a. de felice , a. e. gmrkolu and s. mukohyama , phys . lett . * 109 * , 171101 ( 2012 ) [ arxiv:1206.2080 [ hep - th ] ] . g. damico , g. gabadadze , l. hui and d. pirtskhalava , phys . rev . d * 87 * , ( 2013 ) 064037 . [ arxiv:1206.4253 [ hep - th ] ] .

we study the thermodynamic casimir force for films with various types of boundary conditions and the bulk universality class of the three - dimensional ising model . to this end we perform monte carlo simulations of the improved blume - capel model on the simple cubic lattice . in particular , we employ the exchange or geometric cluster cluster algorithm [ j.r . heringa and h. w. j. blte , phys . rev . e * 57 * , 4976 ( 1998 ) ] . in a previous work we demonstrated that this algorithm allows to compute the thermodynamic casimir force for the plate - sphere geometry efficiently . it turns out that also for the film geometry a substantial reduction of the statistical error can achieved . concerning physics , we focus on boundary conditions , where denotes the ordinary surface transition . these are implemented by free boundary conditions on both sides of the film . films with such boundary conditions undergo a phase transition in the universality class of the two - dimensional ising model . we determine the inverse transition temperature for a large range of thicknesses of the film and study the scaling of this temperature with . in the neighborhood of the transition , the thermodynamic casimir force is affected by finite size effects , where finite size refers to a finite transversal extension of the film . we demonstrate that these finite size effects can be computed by using the universal finite size scaling function of the free energy of the two - dimensional ising model .

1410.7161

in their seminal work , de gennes and fisher @xcite pointed out that the spatial restriction of thermal fluctuations should lead to an effective force . due to its analogy with the casimir effect @xcite , where the spatial restriction of quantum fluctuations leads to a force , it is called thermal , thermodynamic or critical casimir effect . here `` critical '' refers to the fact that thermal fluctuations become large in the neighbourhood of a critical point . at a second order phase transition , in the thermodynamic limit of the bulk system , the correlation length , which characterizes the spatial extent of these fluctuations , behaves as @xmath4 where @xmath5 are the amplitudes of the correlation length in the high and the low temperature phase , respectively , and @xmath6 is the critical exponent of the correlation length . the reduced temperature is given by @xmath7 , where @xmath8 is the critical temperature . note that in the following we shall use for simplicity @xmath9 , where @xmath10 . for reviews on critical phenomena see for example @xcite . owing to their simplicity , often films are studied . for films the thermodynamic casimir force per area is given by @xmath11 where @xmath12 is the excess free energy per area of the film of thickness @xmath2 , where @xmath13 is the free energy per area of the film and @xmath14 is the free energy density of the bulk system . the thermodynamic casimir force per area follows the finite size scaling law @xmath15^{1/\nu } ) \;\;,\ ] ] see for example ref . the function @xmath16 is expected to be universal , which means that it should only depend on the universality classes of the transitions of the bulk system and the surfaces . for reviews on surface critical phenomena see @xcite . the thermodynamic casimir effect has been demonstated in experiments on films of @xmath17he and @xmath18he-@xmath17he mixtures near the @xmath19-transition or the tri - critical point of the bulk system @xcite . the force obtained for different thicknesses is described quite well by a unique scaling function @xmath20 . also experiments with liquid binary mixtures near the mixing - demixing transition were performed , where either films @xcite or the sphere - plate geometry @xcite were studied . in other experiments , the thermodynamic casimir force is the driving force for colloidal aggregation @xcite . it is a theoretical challenge to compute the universal scaling function @xmath20 for different bulk universality classes and types of boundary conditions to compare with experimental data . still the mean - field approximation is used as tool that can be employed relatively easily for more complicated geometrical setups . for recent work see for example @xcite . obviously , no accurate results can be expected this way . unfortunately field theoretic methods do not allow to compute @xmath20 for all types of boundary conditions of interest or do not allow to compute @xmath20 in the full range of the scaling variable @xmath21 @xcite . for a discussion of this point see for example the introduction of @xcite . exact results can be obtained in the large @xmath22 limit for periodic and free boundary conditions @xcite . also for the two - dimensional ising model with various boundary conditions exact results were obtained @xcite . in the case of the three - dimensional ising universality class and strongly symmetry breaking boundary conditions , quite accurate results had been obtained by using the extended de gennes - fisher local - functional method @xcite . o(@xmath23)-symmetric systems with periodic boundary conditions had been studied using a functional renormalization group approach @xcite . in the last few years there has been considerable progress in the study of the thermodynamic casimir force by using monte carlo simulations of lattice spin models . at least in princible , the finite size scaling function can be determined with a controlable statistical and systematical error . in particular , in refs . @xcite the three - dimensional xy bulk universality class and a vanishing field at the boundary have been studied , which is relevant for the experiments on @xmath17he . a quite satisfactory agreement between the experimental results and the theory was found . in refs . @xcite the ising bulk universality class and various types of boundary conditions were studied . note that a continuous mixing - demixing transition of binary mixtures belongs to the ising bulk universality class . notwithstanding this nice progress , further algorithm improvements are certainly welcome to study problems with a large parameter space like structured surfaces @xcite , disorder at the surface , the crossover from the special to the ordinary surface universality class @xcite , the presence of an external bulk field @xcite , or more complicated geometrical setups @xcite . in ref . @xcite we determined the thermodynamic casimir force for the plate - sphere geometry . we studied the three - dimensional ising universality class and strongly symmetry breaking boundary conditions . a preliminary study showed that with a conventional approach and a reasonable amount of cpu - time it is impossible to get meaningful results for this problem . employing the exchange cluster algorithm , it is possible to define a variance reduced estimator for the difference of the internal energy . this allowed us to obtain the scaling functions of the thermodynamic casimir force with high accuracy . the exchange cluster algorithm is a variant of the geometric cluster algorithm of @xcite . in the geometric cluster algorithm the sites of a single lattice are organized in pairs . this is achieved for example by a reflection at a plane of the lattice . the elementary operation of the update is the exchange of the spin value within such pairs of sites . instead , we consider two independent systems . we consider pairs of sites , where one is in one lattice , while the other site belongs to the other lattice . in the present work we apply the exchange cluster algorithm to the film geometry . the relative simplicity of the film geometry allows us to study the properties of the exchange cluster algorithm and its associated improved estimators more systematically . in the present work we first study strongly symmetry breaking boundary conditions , @xmath24 and @xmath25 , then @xmath26 and finally @xmath0 boundary conditions . here the sign indicates the value of the spins at the boundary and @xmath1 the ordinary surface transition . these problems have been studied before , and the scaling functions of the thermodynamic casimir force are known fairly well . here we are mainly aiming at a better understanding of the exchange cluster algorithm before attacking more complicated problems . it turns out that , depending on the type of the surfaces of the film , large reductions of the variance can be achieved . in the case of @xmath0 boundary conditions , the problem arises that the film undergoes a second order phase transition in the universality class of the two - dimensional ising model . this leads to sizeable finite size effects , where the finite extension in the transversal directions is meant . to understand these finite size effects and the interplay of the transition with the thermodynamic casimir force on a quantitative level , we first accurately determined the critical temperature for a large range of thicknesses @xmath2 by using the method discussed in @xcite . we match the reduced temperature of the two - dimensional ising model and the films . we analyze how the temperature of the effectively two - dimensional transition approaches the bulk transition temperature as the thickness of the film increases . based on these results , we demonstrate that finite size effects of the thermodymanic casimir force due to the finite extension of the lattice in the transversal directions are governed by the universal finite size scaling function of the free energy density that is obtained by analyzing the two - dimensional ising model . the paper is organised as follows . in section [ themodel ] we define the model and discuss the boundary conditions that we study in this work . in section [ algorithm ] we discuss the exchange cluster algorithm and the variance reduced estimator for differences of the internal energy and other quantities . at the example of @xmath25 boundary conditions at the critical point of the bulk system , we carefully study how the performance of the algorithm depends on its parameters . in sections [ strongsection ] and [ mixedresults ] we present our numerical results for strongly symmetry breaking and @xmath27 boundary conditions , respectively . in section [ finite2d ] we determine the finite size scaling function of the free energy density of the two - dimensional ising model . in section [ filmtc ] we study the phase transition of films with @xmath0 boundary conditions for a large range of thicknesses @xmath2 . then in section [ casimiroo ] we determine the scaling function of the thermodynamic casimir force for films with @xmath0 boundary conditions . finally we summarize our results and give an outlook . as in previous work , we study the blume - capel model on the simple cubic lattice . the bulk system , in absence of an external field , is defined by the reduced hamiltonian @xmath28 where the spin might assume the values @xmath29 . @xmath30 denotes a site on the simple cubic lattice , where @xmath31 and @xmath32 denotes a pair of nearest neighbors on the lattice . the inverse temperature is denoted by @xmath33 . the partition function is given by @xmath34 , where the sum runs over all spin configurations . the parameter @xmath35 controls the density of vacancies @xmath36 . in the limit @xmath37 vacancies are completely suppressed and hence the spin-1/2 ising model is recovered . in @xmath38 dimensions the model undergoes a continuous phase transition for @xmath39 at a @xmath40 that depends on @xmath35 , while for @xmath41 the model undergoes a first order phase transition , where @xmath42 , see ref . @xcite . numerically , using monte carlo simulations it has been shown that there is a point @xmath43 on the line of second order phase transitions , where the amplitude of leading corrections to scaling vanishes . in @xcite we simulated the model at @xmath44 close to @xmath40 on lattices of a linear size up to @xmath45 . we obtained @xmath46 and @xmath47 . the amplitude of leading corrections to scaling at @xmath44 is at least by a factor of @xmath48 smaller than for the spin-1/2 ising model . following eq . ( 12 ) of ref . @xcite , the amplitude of the second moment correlation length in the high temperature phase at @xmath44 is @xmath49 in the high temperature phase there is little difference between @xmath50 and the exponential correlation length @xmath51 which is defined by the asymptotic decay of the two - point correlation function . following @xcite : @xmath52 for the thermodynamic limit of the three - dimensional system . note that in the following @xmath53 always refers to @xmath54 . in the present work we study the thermodynamic casimir effect for systems with film geometry . in the ideal case this means that the system has a finite thickness @xmath2 , while in the other two directions the limit @xmath55 is taken . in our monte carlo simulations we shall study lattices with @xmath56 and periodic boundary conditions in the @xmath57 and @xmath58 directions . throughout we simulate lattices with @xmath59 . the types of boundary conditions discussed here can be characterized by the reduced hamiltonian @xmath60 where @xmath61 break the symmetry at the surfaces . in our convention @xmath32 runs over all pairs of nearest neighbor sites . note that here the sites @xmath62 and @xmath63 are not nearest neighbors as it would be the case for periodic boundary conditions . in general there is ambiguity , where exactly the boundaries are located and how the thickness of the film is precisely defined . here we follow the convention that @xmath2 gives the number of layers with fluctuating spins . first we study strongly symmetry breaking boundary conditions that are given by @xmath64 , @xmath65 . there are , up to symmetry transformations , two choices . either @xmath66 and @xmath67 have the same or a different sign , which we shall denote by @xmath24 and @xmath25 , respectively . taking the limit @xmath64 , @xmath65 fixes the spins at the surface to the sign of the surface field . in order to keep @xmath2 layers of fluctuating spins , which is done to be consistent with our previous work @xcite , we actually put the surface fields @xmath68 at @xmath69 and @xmath70 . note that this is equivalent to @xmath71 at @xmath72 and @xmath73 . in a semi - infinite system , following the classification of refs . @xcite , this choice of boundary conditions corresponds to the normal or extraordinary surface universality class . next we simulated the case @xmath74 at @xmath72 and @xmath75 at @xmath70 . in a semi - infinite system , a vanishing external surface field corresponds to the ordinary surface universality class . hence , we denote this combination of boundary conditions by @xmath27 . finally we simulated systems with @xmath74 and @xmath76 at @xmath72 and @xmath73 . this set of boundary conditions is denoted by @xmath0 . in our program code we have implemented these boundary conditions by spin variables that reside at @xmath69 and @xmath70 that are fixed to either @xmath77 , @xmath78 , or @xmath57 , depending on the type of the boundary condition . in the case of @xmath27 and @xmath0 boundary conditions , we studied small @xmath66 and @xmath67 by computing the coefficients of the taylor - expansion of the quantities of interest up to second order around vanishing surface fields . given that leading bulk corrections are eliminated , the leading remaining corrections are due to the surfaces . there are theoretical arguments that these can be expressed by an effective thickness @xmath79 of the film @xcite . the value of @xmath80 depends on the precise definition of the thickness @xmath2 . concerning the physics , it depends on the model that is considered and the type of boundary conditions that are imposed . however it should be independent of the scaling variable @xmath21 and the physical quantity that is considered . it can be decomposed as @xmath81 , where @xmath82 are extrapolation lengths that depend on the type of boundary conditions at the boundary @xmath83 and not on the boundary conditions at the other boundary . for a discussion see for example section iv of @xcite or section iii of @xcite . in ref . @xcite the concept of an effective thickness has been verified with high numerical precision for the large @xmath22 limit of the three - dimensional @xmath84-symmetric @xmath85 model with free boundary conditions . in the following we shall use the numerical values @xmath86 , ref . @xcite , for strongly symmetry breaking boundary conditions , @xmath87 for @xmath27 boundary conditions @xcite . in the case of @xmath0 we take @xmath88 where @xmath89 , see eq . ( 63 ) of @xcite . the estimates of @xmath80 were obtained by analyzing the finite size scaling behavior of various quantities directly at the critical point . analyzing the numerical results for the thermodynamic casimir force below , we shall use these values as input . the reduced excess free energy per area of the film is defined by @xmath90 where @xmath91 is the reduced bulk free energy density and @xmath92 is the partition function of the film . we compute the thermodynamic casimir force by using eq . ( [ defineff ] ) . on the lattice , the partial derivative of the reduced excess free energy per area with respect to the thickness of the film is approximated by @xmath93 where @xmath94 is a small positive integer . except for a few preliminary algorithmic studies , we shall use the minimal value @xmath95 . following hucht @xcite , we compute the difference of free energies as integral over the inverse temperature of the difference of the corresponding internal energies @xmath96 where @xmath97 and @xmath98 where in our convention the energy per area is given by @xmath99 and @xmath100 is the bulk energy density . the integration is done numerically , using the trapezoidal rule : @xmath101\ ] ] where @xmath102 are the values of @xmath103 we simulated at . they are ordered such that @xmath104 for all @xmath83 . typically @xmath105 nodes @xmath106 are needed to compute the thermodynamic casimir force in the whole range of temperatures that is of interest to us . obviously , @xmath107 should be known with good accuracy . usually one chooses @xmath108 such that @xmath109 and hence @xmath110 . in the case of strongly symmetry breaking boundary conditions , we shall use a different choice of @xmath108 that is discussed in @xcite . one important aspect of the present work is to demonstrate that the exchange cluster algorithm allows to compute @xmath111 by using a variance reduced estimator . the reduction of the variance depends on the type of the boundary conditions and the parameters @xmath2 , @xmath94 and @xmath103 as we shall see below . the variance of @xmath112 , computed in the standard way , is @xmath113 at the critical point , taking @xmath114 and @xmath115 proportional to @xmath2 , the variance of the energy per area behaves as @xmath116 where @xmath117 is the specific heat of the finite system . on the other hand , the quantity we are interested in scales as @xmath118 at the critical point . hence the ratio @xmath119 which is , for a given number of statistically independent measurements , proportional to the square of the statistical error , increases with increasing thickness @xmath2 . in order to keep the statistical error small , we used in ref . @xcite @xmath120 and @xmath121 for @xmath122 and @xmath123 , respectively . this in turn makes it more difficult to control the discretization error of eq . ( [ discr ] ) . as we shall see below , the exchange cluster improved estimator of @xmath111 eliminates this problem and for strongly symmetry breaking boundary conditions , we get statistically accurate results for @xmath124 and @xmath95 . note that , with comparable numerical effort , @xmath100 can be computed more accurately than @xmath111 , even when using the exchange cluster improved estimator . here we shall mainly use numerical results for @xmath100 obtained in previous work @xcite . for a discussion , see section vii of @xcite . note that one could also simulate the geometry discussed in ref . @xcite by using the exchange cluster algorithm exactly in the same fashion as we simulated the sphere - plate geometry in ref . the layer of fixed spins , called `` wall '' by the authors , which separates two sub - systems , would take over the role of the sphere . this way the simulation allows to measure @xmath125 directly . effectively @xmath100 is provided by the larger of the two sub - systems . we performed a preliminary study that demonstrated that this indeed works . however we did not follow this line , since , as discussed above , accurate results for @xmath100 are already available from simulations of systems with periodic boundary conditions in all directions . with the exchange cluster algorithm , we simulate two systems that are defined on identical lattices . let us denote the sites of this pair of lattices by @xmath126 , where @xmath21 labels a site in a given lattice and @xmath127 denotes the lattice . the sites of these two lattices are mapped by @xmath128 one to one on each other such that the neighborhood relation of the sites is preserved . in the simplest case , @xmath128 is the identity . here we shall use random translations along the transversal directions of the film . one also could employ reflections . the basic operation of the exchange cluster algorithm is to exchange the values of the spins between corresponding sites . this operation can be described by an auxiliary variable @xmath129 : @xmath130 in order to keep the notation simple , we assume @xmath131 . for @xmath132 the exchange is performed , while for @xmath133 the old values are kept . the contribution of a pair @xmath32 of nearest neighbors to the reduced hamiltonian is given by @xmath134 note that terms linear in @xmath135 cancel . the exchange of spins is performed by using a cluster update . the construction of the clusters is characterized by the probability to delete the link between the nearest neighbors @xmath21 and @xmath136 @xcite @xmath137 \;,\ ] ] where @xmath138 which is the prefactor of @xmath139 in eq . ( [ hem ] ) . this is sufficient for the problems studied in this work . let us briefly sketch how the exchange cluster algorithm can be applied to a more general class of problems . for an enhanced coupling at the boundary , as it is required for the study of the special surface universality class , eq . ( [ hem ] ) has to be generalized to @xmath140 this leads to the embedded coupling @xmath141 and in addition to an external field that acts on @xmath135 : @xmath142 where the indices of @xmath143 indicate that it is the contribution to the field at the site @xmath21 stemming from the pair @xmath32 of sites . in case there is also an external field in the original problem we get the contribution @xmath144 in total @xmath145 this generalized problem can be simulated for example by constructing the clusters only based on the pair interaction and then taking into account the external field in the probability to flip the cluster , where here flipping a cluster means that for all sites in the cluster the spins are exchanged . for example , the cluster is flipped with the metropolis - type probability @xmath146 \;,\ ] ] where the sum runs over all sites @xmath21 that belong to the given cluster @xmath147 . here we study two films of the thicknesses @xmath148 and @xmath149 , where @xmath95 , @xmath58 , @xmath150 . in the case of system 1 , the spins at @xmath69 and @xmath151 are fixed in order to implement the boundary conditions , while for system 2 , the spins at @xmath69 and @xmath152 are fixed . in order to have the same number of sites for both systems 1 and 2 , we add in the case of system 2 auxiliary spins at @xmath153 , which assume the same value as those at @xmath154 . clusters are constructed according to the delete probability given by eq . ( [ pdsimple ] ) . this means that a link between a pair of neighbor sites is frozen with the probability @xmath155 . two sites belong to the same cluster , if there exists a chain of frozen links that connects the two sites . in order to keep the boundary conditions in place , only clusters are flipped that do not contain sites with fixed spins . the purpose of the exchange cluster algorithm is to obtain a variance reduced estimator of @xmath111 . to this end , it is optimal to exchange as many spins as possible . hence only those spins are not exchanged that belong to clusters that contain fixed spins . to this end we have to construct only those clusters that contain fixed spins . starting the cluster at @xmath69 , the cluster can not grow to @xmath72 , since @xmath156 and hence @xmath157 , which implies that @xmath158 . only starting from @xmath159 , a cluster containing fixed spins of system 2 only , might grow to @xmath160 . hence we start the construction of the frozen clusters by running through all sites @xmath161 and add the site @xmath162 to the frozen clusters with the probability @xmath163 , eq . ( [ pdsimple ] ) . note that in this initial step we have to check only this single neighbor , since the other ones are frozen anyway . then the construction of the frozen clusters is completed using a standard algorithm for the cluster search . in our c - program the spins are stored in an array ` char spins[i_d][l_z][l][l ] ; ` where ` i_d ` equals two and ` l_z ` equals @xmath164 . similar to the case of the plate - sphere geometry , it turns out that the frozen clusters usually take only a small fraction of the lattice . therefore , in order to save cpu time we do not copy all spins outside the frozen clusters from ` spins[0 ] [ ] [ ] [ ] ` to ` spins[1 ] [ ] [ ] [ ] ` and vice versa . instead , we do that for the spins that belong to frozen clusters . this way , the systems @xmath57 and @xmath58 interchange their position in the array ` spins ` . in order to keep track of where the systems are stored in the array ` spins ` , we introduce the array ` int posi[i_d ] ; ` where the index ` i_d ` equals @xmath57 or @xmath58 and ` posi[i_d ] ` indicates whether system @xmath57 is stored in ` spins[0 ] [ ] [ ] [ ] ` or ` spins[1 ] [ ] [ ] [ ] ` and system @xmath58 correspondingly . implemented this way , the cpu - time required by the cluster exchange update is essentially proportional to the size of the frozen clusters . the main purpose of the exchange cluster is to allow us to define improved estimators for the difference of observables defined in systems 1 and 2 . here this is mainly @xmath112 , however also other quantities can be computed efficiently as we shall see below . the basic idea behind these improved differences is that large parts of the configurations are swapped between the two systems . this way we get exact cancellations for most of the lattice volume . let us consider an observable @xmath165 that is defined for both systems @xmath57 and @xmath58 . we are aiming at a variance reduced estimator for the difference @xmath166 to this end we make use of the correlation of the configuration of system 1 at markov - time @xmath167 with that of system 2 at markov - time @xmath168 , and vice versa : @xmath169 + [ a_{1,t+1 } - a_{2,t}]\right ) \;,\ ] ] where the second index of @xmath165 now gives the position of the configuration in the markov chain and @xmath168 and @xmath167 are separated by a single exchange cluster update . let us work out eq . ( [ improved_general ] ) explicitly for @xmath112 : @xmath170 + [ s_{x,2}^{(t+1 ) } s_{y,2}^{(t+1 ) } - s_{x,1}^{(t+1 ) } s_{y,1}^{(t+1 ) } ] \right ) \nonumber \\ & = & \frac{1}{2 } \sum_{<xy > } \left ( [ s_{x,2}^{(t ) } s_{y,2}^{(t ) } - s_{x,1}^{(t+1 ) } s_{y,1}^{(t+1 ) } ] + [ s_{x,2}^{(t+1 ) } s_{y,2}^{(t+1 ) } - s_{x,1}^{(t ) } s_{y,1}^{(t ) } ] \right ) \nonumber \\ & = & \frac{1}{2 } \sum_{<xy > \in c_f } \left ( [ s_{x,2}^{(t ) } s_{y,2}^{(t ) } - s_{x,1}^{(t+1 ) } s_{y,1}^{(t+1 ) } ] + [ s_{x,2}^{(t+1 ) } s_{y,2}^{(t+1 ) } - s_{x,1}^{(t ) } s_{y,1}^{(t ) } ] \right)\end{aligned}\ ] ] where @xmath171 means that at least one of the sites @xmath21 or @xmath136 belongs to a frozen cluster . hence also the numerical effort to compute @xmath172 is approximately proportional to the size of the frozen clusters . note that for our choice of the update @xmath173 for all nearest neighbor pairs @xmath174 where neither @xmath21 nor @xmath136 belongs to a frozen cluster . the exchange cluster algorithm on its own is not ergodic , since it keeps the total number of spins of a given value fixed . therefore we performed in addition updates of the individual systems , using standard cluster and local updates @xcite . in all our simulations we used the mersenne twister algorithm @xcite as pseudo - random number generator . we used the standard delete probability @xmath175 $ ] in the construction of the clusters . one has to take into account that clusters that contain sites with fixed spins can not be flipped . flipped means that all spins that belong to the cluster are multiplied by @xmath77 . we have used two types of cluster - updates . in the first one , denoted by sw - cluster algorithm in the following , we flip the clusters , that do not contain fixed sites , following ref . @xcite , with the probability @xmath176 . in the second one , denoted by b - cluster algorithm in the following , clusters that do not contain fixed sites , are always flipped . this has the technical advantage , that actually only clusters that contain sites with fixed spins have to be constructed . all other spins are flipped . for @xmath0 boundary conditions , only the sw - cluster algorithm is used , since for @xmath36 or @xmath177 we get @xmath178 and hence there are no clusters that contain both sites of the interior and the boundary . the authors of @xcite have pointed out that auto - correlation times of local updating algorithms can be reduced by a significant factor , when one abstains from detailed balance and only demands the sufficient condition of balance . this idea still leaves considerable freedom for the design of the algorithm . todo and suwa suggest to order the possible values of the local spin on a cycle . then one preferentially updates in one of the two directions on the cycle . for the precise description see ref . todo and suwa have tested their algorithm for example at the 4- and 8-state potts model in two dimensions in the neighborhood of the critical point . they find a reduction of the auto - correlation time compared with the heat - bath algorithm by a factor of 2.7 and 2.6 for the 4- and 8-state potts model , respectively . in the case of the improved blume - capel model on the simple cubic lattice at the critical point one finds a reduction by a factor of about 1.7 compared with the heat - bath algorithm @xcite . since we failed to find a prove of ergodicity for the todo - suwa local update , sweeping through the lattice in type - writer fashion , we performed heat - bath sweeps in addition . note that for the heat - bath the prove of ergodicity is trivial . we initialized the spins that are not fixed by choosing one of the three possible values with equal probability . then we equilibrated the systems by performing 1000 update cycles consisting of one heat - bath sweep , one sw - cluster update , one todo - suwa sweep and one b - cluster update . in the case of @xmath0 boundary conditions , the b - cluster update is omitted . after this initial phase of the simulation we added @xmath179 exchange cluster updates to each update cycle . furthermore , since the frozen exchange clusters are very much localized at the boundary , we performed for each exchange cluster update a local update with the todo - suwa algorithm of the @xmath180 layers of the lattices that are closest to the upper boundary . only in a few preliminary tests we shall use a different sequence of updates , which will be stated below . first we tested the performance of the exchange cluster algorithm for @xmath25 boundary conditions at the critical point @xmath181 . to keep things simple , we first used the following update sequence : a global sweep with the heat - bath algorithm over both systems followed by one exchange cluster update , combined with a random translation of one system with respect to the other in the transversal directions . our results are summarized in table [ performance ] . in all cases @xmath182 update cycles and measurements were performed . in the third column we give the size of the frozen exchange clusters per area @xmath183 . the @xmath184 layers , where the spins of system @xmath58 are fixed and those of system @xmath57 are not , are taken into account in @xmath183 . this means that @xmath183 at least equals to @xmath94 . we find that @xmath183 is small compared with the thickness of the films in all cases . for given @xmath94 it depends very little on the thickness @xmath2 . as one might expect , it increases with increasing @xmath94 . we give the variance of @xmath172 and of the energies @xmath185 and @xmath186 normalized by the area @xmath187 , since this normalized number should have a finite @xmath188 limit . we find that the variance of @xmath172 is reduced compared with the sum of the variances of the energies @xmath185 and @xmath186 of the individual systems . for fixed @xmath94 , the ratio of the two variances increases with increasing lattice size . on the other hand , the advantage of the improved estimator becomes smaller with increasing @xmath94 . often variance reduced estimators have a larger integrated auto - correlation time than the basic quantity . here , in contrast we observe that the integrated auto - correlation time of @xmath172 is considerably smaller than those of the energies @xmath185 and @xmath186 of the individual systems . ._[performance ] we study the properties of the exchange cluster algorithm for @xmath25 boundary conditions at @xmath40 . the transversal extension of the lattices is @xmath189 , @xmath190 , and @xmath191 for @xmath192 , @xmath193 , and @xmath194 , respectively . for the definition of the quantities and a discussion see the text . _ [ cols=">,>,^,^,<,^,<,<",options="header " , ] finally we studied the behavior of @xmath195 at the critical point as a function of the thickness @xmath2 of the film . it behaves as @xmath196^{1/\nu-1 } \;\;.\ ] ] performing various fits , using @xmath197 and @xmath198 as input , we arrive at @xmath199 . in terms of the scaling variable @xmath200^{1/\nu}$ ] this means @xmath201^{-1/\nu } l \simeq a \xi_0^{1/\nu } \frac{l}{l_0+l_s } = 0.1074(10 ) \frac{l}{l_0+l_s } \;.\ ] ] for the transversal correlation length of the film in the high temperature phase , eq . ( [ xi2dising ] ) translates to @xmath202 ( x - x_c)^{-1 } \;\;\ ] ] using @xmath203_{2d ising } = \lim_{l \rightarrow \infty } \ ; \frac{1}{l } \left . \frac{\partial z_a / z_p}{\partial \tau } \right|_{\tau=0 } = 0.3021247100407 ... \;\;\ ] ] for the two - dimensional ising model . in fig . [ binderplot ] we plot @xmath204 as a function of @xmath205 , where @xmath204 is the binder cumulant @xmath206 at @xmath207 , where @xmath208 is the magnetization . following ref . @xcite @xmath209 . its interesting to see that already starting from @xmath210 , finite @xmath3 effects nicely scale with the effective thickness @xmath211 . we have checked that the decay of corrections with increasing @xmath3 is consistent with @xmath212 , as theoretically expected . finally we convinced ourself that @xmath213 converges to @xmath214 @xcite as @xmath215 . these observations strongly support the hypothesis that the transition of the film , for any thickness @xmath2 , belongs to the two - dimensional ising universality class . we plot @xmath204 as a function of @xmath205 with @xmath216 . for comparison we give @xmath217 as green dashed line . the data points for @xmath210 and @xmath218 are given as black circles and red squares , respectively . for all other thicknesses , the data points are shown as blue diamonds . the error bars are smaller than the size of the symbols . for the definition of the quantities and a discussion see the text . , width=548 ] the thermodynamic casimir force for @xmath0 boundary conditions has been studied for the ising model @xcite and the improved blume - capel model @xcite . we have simulated films of the thicknesses @xmath219 , @xmath220 , @xmath221 and @xmath222 . for the parameters of the update we took @xmath223 throughout and @xmath224 , @xmath58 , @xmath58 , and @xmath225 for @xmath219 , @xmath220 , @xmath221 , and @xmath222 , respectively . we simulated the transversal lattices sizes @xmath189 , @xmath190 , @xmath191 and @xmath226 for @xmath219 , @xmath227 , @xmath228 and @xmath229 for @xmath230 , @xmath231 for @xmath232 , and @xmath233 and @xmath229 for @xmath234 . for all thicknesses we simulated at slightly more than hundred values of @xmath103 in the neighborhood of the bulk critical point . the larger transversal lattices sizes were simulated at less values of @xmath103 than the smaller ones , focussing at the neighborhood of the transition of the film . we performed @xmath235 update cycles for each value of @xmath103 and most lattice sizes . exceptions are @xmath236 and @xmath237 were we performed only @xmath238 update cycles . in total we used about 5 years of cpu time on a single core of an amd opteron 2378 running at 2.4ghz . let us first discuss the performance of the exchange cluster algorithm . in fig . [ balloo16 ] we plot the average size per area of the frozen exchange clusters @xmath183 for @xmath232 and @xmath231 . for comparison we give our result for @xmath27 boundary conditions , where the exchange cluster update is performed at the @xmath1 boundary . for small @xmath103 the curves for @xmath27 and @xmath0 boundary conditions fall on top of each other . while for @xmath27 boundary conditions a maximum is reached at @xmath239 , for @xmath0 ones we find that @xmath183 is growing monotonically with increasing @xmath103 . at the inverse transition temperatures of the two films , @xmath183 is already a significant fraction of the thickness @xmath2 of the film . we find @xmath240 , @xmath241 , @xmath242 , and @xmath243 at @xmath244 , for @xmath219 , @xmath220 , @xmath221 , and @xmath222 , respectively . for those thicknesses , where we have simulated more than one value of @xmath3 , we find at @xmath245 and in a certain neighborhood below a small dependence of @xmath183 on @xmath3 . in fig . [ ball12fss ] , we plot as an example @xmath183 for @xmath230 and @xmath227 , @xmath228 and @xmath229 . looking at the simulation in the low temperature phase in detail we find that the large frozen exchange clusters grow , when the magnetization of the two systems have different sign . physically one could force the two systems to have the same sign by applying a bulk field @xmath143 , such that @xmath246 , where @xmath247 is the magnetization of the film . the larger @xmath3 , the smaller the amplitude of the bulk field @xmath143 could be chosen . at the end one would extrapolate the results obtained to @xmath248 . here instead , we break the symmetry by hand . after the sweeps with the heat - bath and the todo - suwa algorithm and the swendsen - wang cluster update of the two systems , before starting the @xmath179 exchange cluster updates , we forced the two systems to positive or zero magnetization . to this end , we multiplied all spins of a system with @xmath77 , if its magnetization is negative . this is certainly an update of the configuration that does not fulfil balance and hence we introduce a systematic error . however , we expect that this error vanishes in the limit @xmath188 and also decreases as we go deeper into the symmetry broken phase . in fig . [ balloo16 ] we also give @xmath183 for simulations with this explicit symmetry breaking ( sb ) . we find that indeed @xmath183 is much smaller than for the simulation without sb . also in the low temperature phase of the films , @xmath183 is now decreasing with increasing @xmath103 . for large @xmath103 , the curve is falling on top of that for @xmath27 boundary conditions . we plot the size of the frozen exchange clusters @xmath183 for the thickness @xmath232 . we compare @xmath27 and @xmath0 boundary conditions . in case of @xmath0 we give results for the simulation with and without breaking of the @xmath249 symmetry . the vertical lines give the inverse transition temperature of films of the thickness @xmath250 and @xmath251 . , width=566 ] we plot the size of the frozen exchange clusters @xmath183 for the thickness @xmath230 for the three transversal extensions @xmath227 , @xmath228 and @xmath229 . , width=566 ] let us briefly discuss the gain ( [ gaindef ] ) that we do not plot here . without sb , for all @xmath2 that we studied , it is almost linearly decreasing with increasing @xmath103 , until @xmath252 is reached . starting from this point it stays roughly constant with a value that is approximately equal to @xmath253 . for @xmath254 gain takes about the same value @xmath121 for all thicknesses that we study . using sb , starting from @xmath103 above @xmath255 , the gain rapidly increases with increasing @xmath103 . for example the gain reaches the value @xmath256 at @xmath257 , @xmath258 , @xmath259 and @xmath260 for @xmath219 , @xmath220 , @xmath221 and @xmath222 , respectively . for @xmath103 somewhat larger than @xmath245 we simulated with sb and without . for example for @xmath219 we find that the results for @xmath112 are consistent at the level of our statistical accuracy starting from @xmath261 , @xmath262 , @xmath263 , and @xmath264 for @xmath189 , @xmath190 , @xmath191 and @xmath226 , respectively . in our analysis of the thermodynamic casimir force below , we have used the results obtained with sb starting from slightly larger values of @xmath103 , to have a safety margin . in a first step of the analysis we check whether finite @xmath3 effects in @xmath125 are well described by the universal finite size scaling function @xmath265 , eq . ( [ defgn ] ) . in fig . [ diffs8 ] we plot @xmath266 for @xmath267 and @xmath189 , @xmath190 and @xmath191 . note that @xmath268 , since the bulk energy density cancels . our numerical results are compared with the prediction obtained from the universal finite size scaling function @xmath269 . as input we use the inverse transition temperature @xmath245 and the slope of @xmath270 at @xmath271 given in table [ betac2dall ] , and eq . ( [ slopeising ] ) : @xmath272_{predict}= - \frac{\mbox{d}}{\mbox{d } \beta } g_2(c \ ; [ \beta_{c,2d}(l_0)-\beta ] \ ; l ) \ ; l^{-2}\ ] ] where @xmath273_{film , l_0 } } { [ \bar{s}/l]_{2d ising } } \;\;.\ ] ] we find that for @xmath189 the data are quite close to the prediction obtained from the universal finite size scaling function @xmath269 . note that for @xmath274 and @xmath275 similar observations can be made . going to @xmath231 the matching between the data points and the predicted behavior becomes better . only at the minimum and the maximum of the curve a small missmatch can be observed . for @xmath230 and @xmath233 a similar observation can be made . finally , for @xmath276 , at the level of our statistical accuracy , the match between the data points and the predicted behavior is perfect . we plot @xmath266 for @xmath219 . our numerical data are given by black circles , red squares and blue diamonds for and @xmath189 , @xmath190 and @xmath191 , respectively . the solid lines give the theoretical prediction , obtained from the universal finite size scaling function of the free energy density of the 2d ising transition . the vertical dashed green line indicates the phase transition for @xmath277 and the vertical dashed - dotted violet line the phase transition for @xmath278 . , width=566 ] next we checked how the results for the thermodynamic casimir force are scaling with the thickness @xmath2 of the film . to this end we plot in fig . [ smallqplot ] our numerical results for @xmath279 as function of @xmath280^{1/\nu}$ ] for @xmath281 , @xmath282 , @xmath283 , and @xmath275 . since @xmath284 $ ] is similar for these lattices , we expect that finite @xmath284 $ ] corrections to scaling are similar . we plot @xmath285 as function of @xmath286 for @xmath281 , @xmath282 , @xmath283 , and @xmath275 , where we used @xmath287 with @xmath216 , @xmath288 and @xmath289 . the vertical dashed violet line gives the position of the phase transition of the film . , width=566 ] for @xmath290 the curves fall almost perfectly on top of each other . in contrast , for smaller values of @xmath21 the different curves can be resolved at our level of numerical accuracy . in particular the one for @xmath219 is clearly different from the others . since the difference between the results for @xmath232 and @xmath222 is rather minute , we expect that for @xmath234 deviations from the scaling limit are of a similar size as our statistical errors for @xmath234 . a more quantitative discussion of corrections will be given below , when we analyze the position of the minimum of the scaling function @xmath16 . finally , in fig . [ finaloo ] we plot @xmath279 as function of @xmath280^{1/\nu}$ ] for @xmath234 for @xmath233 and @xmath229 and our extrapolation of the @xmath291 result to @xmath188 obtained by using the universal scaling function @xmath292 . we see that the minimum of @xmath16 deepens as the lattice size increases and the position of the minimum approaches @xmath293 . the position of the minimum for @xmath188 is close to @xmath293 but definitely different from it . we extrapolated our results obtained for @xmath219 , @xmath294 and @xmath230 , @xmath291 to @xmath295 . note that for @xmath232 we have only data for @xmath231 , and therefore a reliable extrapolation is not possible . analyzing these data we find that @xmath296 @xmath297,-1.6922(4)[108])$ ] , @xmath298,-1.6924(8)[76])$ ] , and @xmath299,-1.6925(10)[40])$ ] for @xmath219 , @xmath220 and @xmath222 , respectively . again the number in @xmath300 $ ] gives the error due to the uncertainty of @xmath80 . as our final result for the limit @xmath301 we quote @xmath302 which is consistent with the results obtained for the three different thicknesses . since @xmath303 is definitely larger than @xmath304 , the correlation length of the film at @xmath303 is finite . following eq . ( [ filmxi ] ) , @xmath305 @xmath306 . for @xmath307 , finite @xmath3 effects should be small . hence for @xmath308 the features of the minimum of @xmath16 should be essentially independent of @xmath3 . this is consistent with the observations of ref . @xcite ; see in particular their fig . obviously , in an experiment no periodic boundary conditions can be applied . still @xmath309 indicates how large the transversal linear size of the system should be to avoid finite size effects . we plot @xmath285 as function of @xmath286 for @xmath234 for @xmath233 and @xmath229 and our extrapolation to @xmath188 . we used @xmath287 with @xmath216 , @xmath288 and @xmath289 . the vertical dashed violet line indicates @xmath293 . , width=566 ] our result can be compared with ref . @xcite who simulated the ising model on the simple cubic lattice and the thicknesses @xmath310 , @xmath311 , @xmath312 , and @xmath313 . throughout , they used @xmath314 . they arrive at @xmath315 and @xmath316 , depending on whether they use their eqs . ( 18,20 ) or eq . ( 21 ) to extrapolate to @xmath301 . interpolating our data to @xmath317 using the universal finite size scaling function of the free energy , we arrive at @xmath318 and @xmath319 . hence the apparently good agreement of @xmath303 with our result seems to be a coincidence . the authors of ref . @xcite give no explicit result for @xmath303 and @xmath320 in the text . from the insert of their fig . 16 we read off @xmath321 and @xmath322 . the main reason for the larger error bar of @xcite compared with us is that they use @xmath323 , @xmath324 in their notation , instead of our @xmath197 . using field theoretic methods the author of ref . @xcite arrives at @xmath325 and @xmath326 . similar to eq . ( [ generalizedfss ] ) , the thermodynamic casimir force per area as a function of the inverse temperature @xmath103 and the surface fields @xmath66 and @xmath67 follows the scaling law @xmath327 where @xmath328^{y_{h_1 } } \;\;,\;x_{h_2 } = h_2 [ l_0/l_{ex , nor,0}]^{y_{h_1}}\ ] ] where for our model @xmath329 , eq . ( 73 ) of @xcite , and the surface critical rg - exponent @xmath330 , eq . ( 52 ) of @xcite . the partial derivatives of @xmath331 with respect to @xmath66 and @xmath67 at @xmath332 are determined in a similar fashion as for @xmath27 boundary conditions . in the high temperature phase of the film , due to the @xmath249-symmetry of the problem , the first derivatives vanish . in fig . [ plotoohh ] we plot our results for @xmath333 and @xmath334 despite variance reduction , the statistical error increases rapidly with increasing thickness . our data for @xmath234 already have a quite large statistical error and we therefore did not plot them in fig . [ plotoohh ] . we plot @xmath335 and @xmath336 at @xmath332 as a function of @xmath337 for @xmath0 boundary conditions for the thicknesses @xmath219 , @xmath220 , and @xmath221 . to this end , we have used @xmath287 with @xmath216 , @xmath288 , @xmath289 , @xmath338 , and @xmath339 . to keep the figure readable , error bars are only shown for @xmath232 , where they are the largest . we use the same types of lines for @xmath340 and @xmath341 . note that @xmath342 in the whole range that is plotted . , width=566 ] in the high temperature phase of the bulk system only @xmath343 has a significant amplitude and it is negative . going to lower temperatures , towards the transition temperature of the film , both @xmath343 and @xmath344 rapidly increase . also @xmath343 and @xmath344 approach each other in this range . as a result , in this range , the thermodynamic casimir force varies much less with @xmath66 for @xmath345 than for example for @xmath346 . at the minimum of @xmath347 we have @xmath348 this means that for example for @xmath349 , already for @xmath350 the characteristics of the thermodynamic casimir force for @xmath0 boundary conditions are completely wiped out . for completeness we also give our results for temperatures below the transition temperature of the film . here we rely on our simulations with sb . since the @xmath249 symmetry is broken , the first derivative with respect to @xmath351 does not vanish . the numerical integration is started at large values of @xmath103 . our numerical data are plotted in fig . [ plotux ] . for @xmath230 , @xmath221 and @xmath222 , we find a quite good collapse of the data on a single scaling curve . the function @xmath352 is positive in the whole range @xmath353 . it rapidly increases as @xmath293 is approached . we plot @xmath354 as a function of @xmath337 for @xmath355 boundary conditions for the thicknesses @xmath219 , @xmath220 , @xmath221 , and @xmath222 for the low temperature phase of the film . to this end , we have used @xmath287 with @xmath216 , @xmath288 , @xmath356 , @xmath338 , and @xmath339 . , width=566 ] finally in fig . [ plotuxx ] we plot our results for the second derivatives of the scaling function with respect to the scaling variables . here the statistical errors are quite large and grow rapidly with the thickness of the film . therefore we give only results for @xmath219 and @xmath220 . in the whole range @xmath353 we find that @xmath357 . the functions are negative and the amplitude increases rapidly as @xmath293 is approached . we plot @xmath335 and @xmath336 as a function of @xmath337 for @xmath0 boundary conditions for the thicknesses @xmath219 and @xmath220 in the low temperature phase of the film . to this end , we have used @xmath287 with @xmath216 , @xmath288 , @xmath289 , @xmath338 , and @xmath339 . , width=566 ] our results can be compared with those of @xcite , who studied films with finite values of @xmath66 and @xmath67 . in particular in their figs . 7 and 8 they give results for @xmath358 and @xmath76 , respectively . their results for small @xmath359 are essentially consistent with ours . matching their data with ours we get @xmath360 for the relation between the scaling variables that are used . we study the thermodynamic casimir force by using monte carlo simulations of lattice models . in particular we are concerned with the bulk universality class of the three - dimensional ising model , which for example characterizes a continuous demixing transition of fluid binary mixtures . in ref . @xcite we used the exchange cluster algorithm , or geometric cluster algorithm @xcite , to study the thermodynamic casimir force between a spherical object and a plane substrate . the main point of the exchange cluster algorithm applied to this problem is that it allows to define a variance reduced estimator for the difference of the internal energy of two systems that are characterized by slightly different distances between the spherical object and the substrate . in the case of the sphere - plate geometry it turned out to be mandatory to use this variance reduced estimator to get a meaningful result for the thermodynamic casimir force by using the approach discussed by hucht @xcite . here , we go one step back and apply the exchange cluster algorithm to the film or plate - plate geometry . for this geometry , quite satisfactory numerical results were obtained already . a long list of references is given in the introduction . we simulate the improved blume - capel model on the simple cubic lattice with @xmath24 , @xmath25 , @xmath27 , and @xmath0 boundary conditions , where @xmath361 and @xmath362 are strongly symmetry breaking boundary conditions and @xmath1 stands for the ordinary surface universality class . for a discussion of these boundary conditions see the introduction and section [ themodel ] . we demonstrate that also for the film geometry , the exchange cluster algorithm allows for a considerable reduction of the variance . the only exception are films with @xmath0 boundary conditions in the direct neighborhood of the transition of the film . this allowed us to simulate films with a larger thickness than before , allowing us to consolidate previous results . our final estimates for the thermodynamic casimir force only moderately improve on previous estimates . this is due to the fact that the remaining errors mainly stem from quantities like @xmath80 , see section [ themodel ] , and @xmath363 , see eq . ( [ defxh ] ) , that were used as input . these quantities were taken from previous work and are computed by analyzing physical quantities different from the thermodynamic casimir force . in section [ algorithm ] we discuss that the exchange cluster algorithm can be applied to a larger class of boundary conditions than simulated here . in particular enhanced surface couplings or finite surface fields could be studied . quite recently the authors of @xcite computed the thermodynamic casimir force in the presence of an external bulk field . to this end , one can compute the difference in the excess free energy per area by integrating the difference in the excess magnetisation per area over the external field @xcite , where the difference is taken for films of thickness @xmath364 and @xmath365 . the integration is started at a strong external field , where the difference in the excess free energy vanishes . alternatively , one might start at a vanishing external field , where the difference in the excess free energy per area is known from previous simulations . it seems likely that the exchange cluster algorithm allows to reduce the variance of the difference in the excess magnetisation in such studies . furthermore one could think of applications different from the thermodynamic casimir force . for example one could compute the free energy of defects . it would be interesting to check whether the simulation of spin glass models could be speeded up by exchanging spins between replica . the emphasis of our physics analysis is on @xmath0 boundary conditions . films with such boundary conditions are expected to undergo a second order phase transition in the universality class of the two - dimensional ising model . this transition has been studied for the ising model on the simple cubic lattice for thicknesses up to @xmath366 in ref . @xcite . here we obtain accurate results for thicknesses up to @xmath367 using the finite size scaling approach discussed in ref . our numerical results nicely confirm the expectation that the transition belongs to the universality class of the two - dimensional ising model . we compute the finite size scaling function @xmath368 , eq . ( [ defgn ] ) , that governs the finite size scaling behavior of the free energy density in the universality class of the two - dimensional ising model for @xmath369 and @xmath370 . we show that finite @xmath3-effects in the thermodynamic casimir force , where @xmath3 is the extension of the film in the transversal directions , are described by @xmath368 . in particular using @xmath292 , our knowledge of the inverse transition temperature of the film and the numerical matching of the scaling variable , we extrapolate our results for the thermodynamic casimir force to @xmath188 . for details see section [ casimiroo ] . this approach could also be applied to other types of boundary conditions that do not break the @xmath249-symmetry of the problem , in particular to periodic boundary conditions or enhanced surface couplings that allow to study the special surface universality class . this work was supported by the dfg under the grant no ha 3150/3 - 1 . f. soyka , o. zvyagolskaya , ch . hertlein , l. helden , and c. bechinger , _ critical casimir forces in colloidal suspensions on chemically patterned surface _ , [ arxiv:0810.1866 ] , phys . * 101 * , 208301 ( 2008 ) . a. gambassi , a. macioek , c. hertlein , u. nellen , l. helden , c. bechinger , and s. dietrich , _ critical casimir effect in classical binary liquid mixtures _ , [ arxiv:0908.1795 ] , phys . rev . * e * 80 , 061143 ( 2009 ) . m. trndle , o. zvyagolskaya , a. gambassi , d. vogt , l. harnau , c. bechinger , and s. dietrich , _ trapping colloids near chemical stripes via critical casimir forces _ , [ arxiv:1012.0181 ] , molecular physics * 109 * , 1169 ( 2011 ) . daniel bonn , jakub otwinowski , stefano sacanna , hua guo , gerard wegdam , and peter schall , _ direct observation of colloidal aggregation by critical casimir forces _ , lett . * 103 * , 156101 ( 2009 ) ; andrea gambassi and s. dietrich , phys . rev . lett . * 105 * , 059601 ( 2010 ) ; bonn , wegdam , and schall reply : daniel bonn , gerard wegdam , and peter schall phys . lett . * 105 * , 059602 ( 2010 ) . o. zvyagolskaya , a. j. archer and c. bechinger , _ criticality and phase separation in a two - dimensional binary colloidal fluid induced by solvent critical behavior _ , [ arxiv:1108.5966 ] , epl * 96 * , 28005 ( 2011 ) . francesco parisen toldin , matthias trndle , and s. dietrich , _ critical casimir forces between homogeneous and chemically striped surfaces _ , [ arxiv:1303.6104 ] , phys . rev . e * 88 * , 052110 ( 2013 ) . h. w. diehl , daniel grneberg and m. a. shpot , _ fluctuation - induced forces in periodic slabs : breakdown of epsilon expansion at the bulk critical point and revised field theory _ , [ cond - mat/0605293 ] , europhys . lett . * 75 * , 241 ( 2006 ) . h. w. diehl and daniel grneberg , _ critical casimir amplitudes for n - component @xmath85 models with o(n)-symmetry breaking quadratic boundary terms _ , [ arxiv:0905.3113 ] , nucl . b * 822 * , 517 ( 2009 ) . h. w. diehl , daniel grneberg , martin hasenbusch , alfred hucht , sergei b. rutkevich , felix m. schmidt , _ large - n approach to thermodynamic casimir effects in slabs with free surfaces _ , [ arxiv:1402.3510 ] , phys . rev . e 89 , 062123 ( 2014 ) . daniel dantchev , h. w. diehl , and daniel grneberg , _ excess free energy and casimir forces in systems with long - range interactions of van der waals type : general considerations and exact spherical - model results _ , [ cond - mat/0510405 ] , phys . e * 73 * , 016131 ( 2006 ) . h. w. diehl , daniel grneberg , martin hasenbusch , alfred hucht , sergei b. rutkevich , and felix m. schmidt , _ exact thermodynamic casimir forces for an interacting three - dimensional model system in film geometry with free surfaces _ , [ arxiv:1205.6613 ] , epl * 100 * , 10004 ( 2012 ) . joseph rudnick , roya zandi , aviva shackell , and douglas abraham , _ boundary conditions and the critical casimir force on an ising model film : exact results in one and two dimensions _ , [ arxiv:1001.0994 ] , phys . e * 82 * , 041118 ( 2010 ) . o. vasilyev , a. gambassi , a. macioek , and s. dietrich , _ universal scaling functions of critical casimir forces obtained by monte carlo simulations _ , [ arxiv:0812.0750 ] , phys . rev . e * 79 * , 041142 ( 2009 ) . m. hasenbusch , _ the thermodynamic casimir effect in the neighbourhood of the lambda - transition : a monte carlo study of an improved three dimensional lattice model _ , [ arxiv:0905.2096 ] , j. stat . ( 2009 ) p07031 . m. hasenbusch , _ the thermodynamic casimir force : a monte carlo study of the crossover between the ordinary and the normal surface universality class _ , [ arxiv:1012.4986 ] , phys . b * 83 * , 134425 ( 2011 ) . m. saito and m. matsumoto , `` simd - oriented fast mersenne twister : a 128-bit pseudorandom number generator '' , in _ monte carlo and quasi - monte carlo methods 2006 _ , edited by a. keller , s. heinrich , h. niederreiter , ( springer , 2008 ) ; m. saito , masters thesis , math . , graduate school of science , hiroshima university , 2007 . the source code of the program is provided at `` http://www.math.sci.hiroshima-u.ac.jp/@xmath376m-mat/mt/sfmt/index.html '' s. todo and h. suwa , _ geometric allocation approaches in markov chain monte carlo _ , [ arxiv:1310.6615 ] , j. phys . ser . * 473 * , 012013 ( 2013 ) . f. gutsch , _ markov - ketten ohne detailliertes gleichgewicht _ , bachelor thesis , humboldt - universitt zu berlin ( 2014 ) . b. m. mccoy and t. t. wu , _ the two dimensional ising model _ , ( harvard univ . press , cambridge , 1973 ) . + b. m. mccoy , in _ statistical mechanics and field theory _ , edited by v. v. bazhanov and c. j. burden ( world scientific , singapore , 1995 ) . m. hasenbusch , k. pinn and s. vinti , _ critical exponents of the three - dimensional ising universality class from finite - size scaling with standard and improved actions _ , hep - lat/9806012 , phys . b * 59 * , 11471 ( 1999 ) .

the randomized load balancing model ( also called _ supermarket model _ ) is now being applied to the study of load balancing in data centers and multi - core servers systems . it is very interesting to analyze the general service times in the supermarket model , and specifically understand influence of the heavy - tailed service times on the doubly exponential solution . since the supermarket model is a complex queueing system , the general service times make its analysis more challenging than the exponential or ph service case . up to now , it still is an open problem whether or how the heavy - tailed service times can disrupt the doubly exponential structure of the fixed point in the supermarket model . in this paper , we provide a novel and simple approach to study the supermarket model with general service times . this approach is based on the supplementary variable method used in analyzing stochastic models extensively . we organize an infinite - size system of integral - differential equations by means of the density dependent jump markov process , and obtain a close - form solution : doubly exponential structure , for the fixed point satisfying the system of nonlinear equations , which is always a key in the study of supermarket models . the fixed point is decomposited into two groups of information under a product form : the arrival information and the service information . based on this , we indicate two important observations : the fixed point for the supermarket model is different from the tail of stationary queue length distribution for the ordinary m / g/1 queue , and the doubly exponential solution to the fixed point can extensively exist even if the service time distribution is heavy - tailed . furthermore , we analyze the exponential convergence of the current location of the supermarket model to its fixed point , and study the lipschitz condition in the kurtz theorem under general service times . based on these analysis , one can gain a new understanding how workload probing can help in load balancing jobs with general service times such as heavy - tailed service . 0.5 cm * keywords : * randomized load balancing , supermarket model , density dependent jump markov process , fixed point , doubly exponential solution , heavy - tailed distribution , exponential convergence , lipschitz condition .

1008.3788

randomized load balancing , where a job is assigned to a server from a small subset of randomly chosen servers , is very simple to implement , and can surprisingly deliver better performance ( for example reducing collisions , waiting times , backlogs ) in a number of applications , such as , data center , hash tables , distributed memory machines , path selection in networks , and task assignment at web servers . one useful model that has been extensively used to study the randomized load balancing schemes is the supermarket model . in the supermarket model , a key result by vvedenskaya , dobrushin and karpelevich @xcite indicated that when each poisson arriving job is assigned to the shortest one of @xmath0 randomly chosen queues with exponential service times , the equilibrium queue length can decay doubly exponentially in the limit as the population number @xmath1 , and the stationary fraction of queues with at least @xmath2 customers is @xmath3 , which is a substantially exponential improvement over the case for @xmath4 , where the tail of stationary queue length distribution in the corresponding m / m/1 queue is @xmath5 . the distributed load balancing strategies , in which individual job decisions are based on information on a limited number of other processors , have been studied analytically by eager , lazokwska and zahorjan @xcite and through trace - driven simulations by zhou @xcite . based on this , the supermarket model is developed by queueing theory and markov processes . most of recent research applied the density dependent jump markov processes to deal with a simple supermarket model with poisson arrival processes and exponential service times , a key result of which illustrates that there exists a unique fixed point which is decreasing doubly exponentially . that approach used in the literature relies on determining the behavior of the supermarket model as its size grows to infinity , and its behavior is naturally described as a system of differential equations , which leads to a closed form solution : doubly exponential structure , of the fixed point . readers may refer to , such as , analyzing a basic and simple supermarket model by azar , broder , karlin and upfal @xcite , vvedenskaya , dobrushin and karpelevich @xcite , mitzenmacher @xcite . certain generalization of the supermarket model have been explored , for example , simple variations by mitzenmacher and vcking @xcite , mitzenmacher @xcite , vcking @xcite , mitzenmacher , richa , and sitaraman @xcite and vvedenskaya and suhov @xcite ; and analyzing load information by mirchandaney , towsley , and stankovic @xcite , dahlin @xcite , mitzenmacher @xcite . furthermore , martin and suhov @xcite , martin @xcite , suhov and vvedenskaya @xcite studied the supermarket mall model by means of the fast jackson network , where each node in a jackson network is replaced by @xmath6 parallel servers , and a job joins the shortest of @xmath7 randomly chosen queues at the node to which it is directed . luczak and mcdiamid @xcite studied the maximum queue length of the original supermarket model with exponential service times when the service speed scales linearly with the number of jobs in the queue . li , lui and wang @xcite discussed the supermarket model with ph service times and the supermarket model with markovian arrival processes , respectively . readers may refer to an excellent overview by mitzenmacher , richa , and sitaraman @xcite . this paper is interested in analyzing the supermarket model with general service times , which is an open problem for determining whether or how the heavy - tailed service times can disrupt the doubly exponential structure of the fixed point . on the other hand , note that the supermarket model is a complex queueing system and has much different characteristics from the ordinary queueing systems , thus the general service times make its analysis more challenging than the exponential or ph service case . up to now , there has not been an effective method to be able to deal with the supermarket model with general service times . the main contributions of the paper are threefold . the first one is to provide a novel and simple approach to study the supermarket model with general service times . this approach is based on the supplementary variable method but is described as a new integral - differential structure for expressing and computing the fraction of queues efficiently . using the new approach , we setup an infinite - size system of integral - differential equations , which makes applications of the density dependent jump markov processes to be able to deal with the general distributions , such as general service times , involved in the supermarket model . the second one is to obtain a close - form solution : doubly exponential structure , for the fixed point satisfying the system of nonlinear equations , which is always a key in the study of supermarket models . furthermore , this paper analyzes the exponential convergence of the current location of the supermarket model to its fixed point , and studies the lipschitz condition in the kurtz theorem under general service times . also , this paper provides numerical examples to illustrate the effectiveness of our approach in analyzing the randomized load balancing schemes with the non - exponential service requirements . the third one is to obtain that the fixed point is decomposited into two groups of information under a product form : the arrival information and the service information . based on this , we indicate three important observations : ( a ) : : the fixed point for the supermarket model is different from the tail of stationary queue length distribution for the ordinary m / g/1 queue , because the fixed point is light - tailed but the stationary queue length is heavy - tail if the service times are heavy - tailed . note that such a difference is illustrated in this paper for the first time , while it can not be observed in the literature for the supermarket model with poisson arrivals and exponential service times , e.g. , see mitzenmacher , richa , and sitaraman @xcite . ( b ) : : the doubly exponential solution to the fixed point can extensively exist even if the service time distribution is heavy - tailed . this is an answer of the above open problem to illustrate the role played by the heavy - tailed service time distribution in the doubly exponential solution to the fixed point . ( c ) : : the doubly exponential solution to the fixed point is not unique for a more general supermarket model . note that we give three different doubly exponential solutions in the supermarket model with poisson arrivals and ph service times , thus it is very interesting to provide all the doubly exponential solutions for a more general supermarket model . based on this , one can gain the new and important understanding how the workload probing can help in load balancing jobs with general service times such as heavy - tailed service . the remainder of this paper is organized as follows . in section 2 , we first describe a supermarket model with general service times , which is always useful in the study of randomized load balancing schemes . then the supermarket model is expressed as a systems of integral - differential equations in terms of the density dependent jump markov processes . in section 3 , we first introduce a fixed point of the system of integral - differential equations , and set up a system of nonlinear equations satisfied by the fixed point . then we provide a close - form solution : doubly exponential structure , to the system of nonlinear equations . in section 4 , we provide a necessary discussion on the key parameter @xmath8 used in the doubly exponential structure , and indicate that the doubly exponential solution to the fixed point extensively exists even if the service time distribution is heavy - tailed . in section 5 , we give three methods to analyze the supermarket model with poisson arrivals and ph service times , and provide three different ways to determine the key parameter @xmath8 and compute the doubly exponential solution to the fixed point . we show that the doubly exponential solution to the fixed point is not unique for a more general supermarket . in section 6 , we study the exponential convergence of the current location of the supermarket model to its fixed point . not only does the exponential convergence indicates the existence of the fixed point , but it also explains such a convergent process is very fast . in section 7 , we apply the kurtz theorem to study the supermarket model with the general service times , and analyze the lipschitz condition with respect to general service times . some concluding remarks are given in section 8 . in this section , we first describe a supermarket model with general service times . then we provide a novel and simple approach to setup an infinite - size system of integral - differential equations based on the density dependent jump markov processes . note that this approach is based on the supplementary variable method but is described as a new integral - differential structure so that the corresponding boundary conditions are written in a different version . let us formally describe the supermarket model , which is abstracted as a multi - server multi - queue stochastic system . customers arrive at a queueing system of @xmath9 servers as a poisson process with an average arrival rate @xmath10 for @xmath11 . the service time @xmath12 of the @xmath2th customer is general with the distribution function@xmath13 where all the random variables @xmath12 for @xmath14 are i.i.d . with the mean @xmath15 = 1/\mu$ ] . each arriving customer chooses @xmath16 servers independently and uniformly at random from these @xmath17 servers , and waits for service at the server which currently contains the fewest number of customers . if there is a tie , servers with the fewest number of customers will be chosen randomly . all customers in any service center will be served in the first - come - first - served ( fcfs ) manner . figure 1 simply shows such a supermarket model . in the study of supermarket models , it is necessary for us to study general service time distributions , for example , heavy - tailed distributions . not only because the general distribution makes analysis of the supermarket models more difficult and challenging than those in the literature for the exponential or ph service case , but it also allows us to model more realistic systems and understand their performance implication under the randomized load balancing strategy . as indicated in @xcite , the process times of many parallel jobs , in particular , jobs to data centers , tend to be non - exponential . unless we state otherwise , we assume that all the random variables defined above are independent , and that the system is operating under the condition : @xmath18 . the supermarket model with general service times is stable if @xmath19 * proof : * let @xmath20 and @xmath21 be the queue lengths of the ordinary m / g/1 queue and of an arbitrary server in the supermarket model at time @xmath22 , respectively . note that in the supermarket model , each customer chooses @xmath7 servers independently and uniformily at random , and the queue length of the entering server is currently shorten , it is easy to see that for each @xmath23,@xmath24 since the ordinary m / g/1 queue is stable if @xmath18 , it follows from ( [ equ0 ] ) that the supermarket model with general service times is stable if @xmath18 . this completes the proof . ' '' '' for @xmath14 , we define @xmath25d@xmath26 as the number of queues with at least @xmath2 customers and the residual service time of each server be in the interval @xmath27d@xmath28 at time @xmath23 . clearly , @xmath29 for @xmath30 and @xmath31 . let@xmath32 which is the density function of the fraction of queues with at least @xmath2 customers and the residual service time of each server be @xmath26 . we write@xmath33 we define @xmath34 as the number of queues with at least @xmath35 customers at time @xmath23 . clearly , @xmath36 . let@xmath37 then @xmath38 for all @xmath23 and@xmath39 let @xmath40 be the fraction of queues with zero customer at time @xmath22 . then@xmath41 thus we have@xmath42 the following proposition shows that the sequence @xmath43 is monotone increasing for @xmath14 , while its proof is easily by means of the definition of @xmath44 for @xmath14 . for @xmath45@xmath46 and@xmath47 now , we setup a system of integral - differential equations by means of the density dependent jump markov process . to that end , we provide an example with @xmath48 to indicate how to derive the system of integral - differential equations . consider the supermarket model with @xmath17 queues , and determine the expected change in the number of servers with at least @xmath2 customers and the residual service time of each server be @xmath26 over a small time period of length d@xmath22 . the probability that a customer arriving during this time period is @xmath49d@xmath22 , and the probability that an arriving customer joins a queue of size @xmath50 is given by @xmath51d@xmath52d@xmath26 . thus , the probability that during this time period , any arriving customer joins a queue of size @xmath50 is given by@xmath53 .\ ] ] similarly , the probability that a customer leaves a server of size @xmath2 is given by@xmath54 .\ ] ] therefore we can obtain@xmath55 \\ & + n\left [ \int_{0}^{+\infty}\mu\left ( x\right ) s_{k , n}\left ( t , x\right ) \text{d}x-\int_{0}^{+\infty}\mu\left ( x\right ) s_{k+1,n}\left ( t , x\right ) \text{d}x\right ] , \end{aligned}\ ] ] which leads to@xmath56 \nonumber\\ & + \left [ \int_{0}^{+\infty}\mu\left ( x\right ) s_{k , n}\left ( t , x\right ) \text{d}x-\int_{0}^{+\infty}\mu\left ( x\right ) s_{k+1,n}\left ( t , x\right ) \text{d}x\right ] . \label{equ1}\ ] ] taking @xmath1 in the both sides of ( [ equ1 ] ) , we have@xmath57 \nonumber\\ & + \left [ \int_{0}^{+\infty}\mu\left ( x\right ) s_{k}\left ( t , x\right ) \text{d}x-\int_{0}^{+\infty}\mu\left ( x\right ) s_{k+1}\left ( t , x\right ) \text{d}x\right ] . \label{equ2}\ ] ] using a similar analysis to that for deriving equation ( [ equ2 ] ) , we can easily obtain a system of integral - differential equations for the fraction density vector @xmath58 as follows:@xmath59@xmath60@xmath61 and for @xmath48,@xmath62 when there are @xmath17 servers in the supermarket model , it is necessary to give a finite - size system of integral - differential equations for the fraction density vector @xmath63 as follows:@xmath64@xmath65@xmath66 and for @xmath67,@xmath68 in this section , we discuss the fixed point of the system of integral - differential equations in equations ( [ equ3 ] ) to ( [ equ6 ] ) , and set up a system of nonlinear equations satisfied by the fixed point . also , we provide a closed - form solution : doubly exponential structure , to the system of nonlinear equations . a row vector @xmath69 is called a fixed point of the fraction density vector @xmath70 if there exists a @xmath71 such that @xmath72 and @xmath73 for all @xmath74 and @xmath14 . it is easy to see that if @xmath75 is a fixed point of the fraction density vector @xmath58 for all @xmath76 , then@xmath77 and for @xmath14@xmath78 which leads to@xmath79 since for @xmath14@xmath80 using the dominated convergence theorem we obtain@xmath81 therefore , if @xmath69 is a fixed point of the fraction density vector @xmath70 for all @xmath74 , then the system of integral - differential equations ( [ equ3 ] ) to ( [ equ6 ] ) can be simplified as@xmath82@xmath83@xmath84 and for @xmath48,@xmath85 in what follows we derive a closed - form expression for @xmath86 . it follows from equations ( [ equ8 ] ) and ( [ equ9 ] ) that@xmath87 to solve equation ( [ equ12 ] ) , using the fact that @xmath88d@xmath89 we have@xmath90 based on the fact that @xmath91 and @xmath92 , it follows from equations ( [ equ10 ] ) and ( [ equ12 ] ) that@xmath93 ^{d}\text{d}x+\int_{0}^{+\infty}\mu\left ( x\right ) \pi_{2}\left ( x\right ) \text{d}x=0.\ ] ] let @xmath94 ^{d}$]d@xmath26 , and we assume that @xmath95 . then@xmath96 using a similar analysis on equation ( [ equ14 ] ) , we have@xmath97 based on @xmath98 and @xmath99 , we can compute@xmath100@xmath101 and@xmath102 ^{d}\text{d}x=\lambda\theta^{d+1}\rho^{d^{2}+d},\ ] ] thus it follows from equation ( [ equ11 ] ) that for @xmath103,@xmath104 which leads to@xmath105 based on the above analysis for the simple expressions @xmath106 for @xmath107 and @xmath108 , we can summarize the following theorem . [ the : cfs]the fixed point @xmath109 is given by@xmath110@xmath111 and for @xmath112@xmath113 or@xmath114 * proof * by induction , one can easily derive the above result . it is clear from ( [ equ15 ] ) and ( [ equ15 - 1 ] ) that equation ( [ equ16 ] ) or ( [ equ17 ] ) is correct for the cases with @xmath115 . now , we assume that equation ( [ equ17 ] ) is correct for the cases with @xmath116 . then@xmath117@xmath118 and@xmath119 it follows from equation ( [ equ11 ] ) that@xmath120 thus , for @xmath121 we have@xmath122 this completes the proof . ' '' '' let @xmath123 ^{d}$]d@xmath26 . then @xmath124 . the following corollary provides another expression for the fixed point . [ cor : decom]@xmath125 and for @xmath14@xmath126 it is easy to see from corollary [ cor : decom ] that the fixed point is decomposited into two groups of information under a product form : the arrival information and the service information . at the same time , the service information indicates that the doubly exponential solution to the fixed point must exist for @xmath127 , even if the service times are heavy - tailed . the following corollary provides an upper bound for the fixed point . for @xmath14 and @xmath128@xmath129 * proof : * note that @xmath130 , we have@xmath131 ^{d}\text{d}x<\int_{0}^{+\infty}\overline{g}\left ( x\right ) \text{d}x=\frac{1}{\mu}.\ ] ] it follows from corollary [ cor : decom ] that@xmath132 this completes the proof . ' '' '' now , we compute the expected sojourn time @xmath133 which a tagged arriving customer spends in the supermarket model . for the general service times , a tagged arriving customer is the @xmath2th customer in the corresponding queue with the following probability@xmath134 when @xmath14 , the head customer in the queue has been served , and so its service time is residual and is denoted as @xmath135 . under the stationary setting , we have@xmath136 \text{d}y\ ] ] with@xmath137 = \int_{0}^{+\infty}\int_{x}^{+\infty}\left [ \mu\overline{g}\left ( y\right ) \right ] \text{d}y\text{d}x.\ ] ] thus it is easy to see that the expected sojourn time of the tagged arriving customer is given by@xmath138 = & \left [ \pi_{0}^{\odot d}-\int_{0}^{+\infty}\pi_{1}^{d}\left ( x\right ) \text{d}x\right ] e\left [ x\right ] \\ & + \sum_{k=1}^{\infty}\left [ \int_{0}^{+\infty}\pi_{k}^{d}\left ( x\right ) \text{d}x-\int_{0}^{+\infty}\pi_{k+1}^{d}\left ( x\right ) \text{d}x\right ] \left [ e\left [ x_{r}\right ] + ke\left [ x\right ] \right ] \\ = & \left [ 1-\int_{0}^{+\infty}\pi_{1}^{d}\left ( x\right ) \text{d}x\right ] e\left [ x\right ] + \int_{0}^{+\infty}\pi_{1}^{d}\left ( x\right ) \text{d}xe\left [ x_{r}\right ] \\ & + e\left [ x\right ] \sum_{k=1}^{\infty}k\left [ \int_{0}^{+\infty}\pi _ { k}^{d}\left ( x\right ) \text{d}x-\int_{0}^{+\infty}\pi_{k+1}^{d}\left ( x\right ) \text{d}x\right ] \\ = & \left\ { e\left [ x_{r}\right ] -e\left [ x\right ] \right\ } \int _ { 0}^{+\infty}\pi_{1}^{d}\left ( x\right ) \text{d}x+e\left [ x\right ] \left\ { 1+\sum_{k=1}^{\infty}\int_{0}^{+\infty}\pi_{k}^{d}\left ( x\right ) \text{d}x\right\ } \\ = & \theta\rho^{d}\left\ { e\left [ x_{r}\right ] -e\left [ x\right ] \right\ } + e\left [ x\right ] \left [ \sum_{k=1}^{\infty}\theta^{\frac{d^{k}-1}{d-1}}\rho^{\frac{d^{k}-d}{d-1}}\right ] .\end{aligned}\ ] ] if the service times are exponential , then @xmath139 = e\left [ x\right ] $ ] , thus we obtain@xmath140 = \frac{1}{\mu}\left [ \sum_{k=1}^{\infty}\theta^{\frac{d^{k}-1}{d-1}}\rho^{\frac{d^{k}-d}{d-1}}\right ] , \ ] ] which is the same as corollary 3.8 in mitzenmacher @xcite . we consider a computational example for the expected sojourn time in the supermarket model with an erlang service time distribution @xmath141 , where @xmath142 . figure 3 shows how the the expected sojourn time depends on the arrival rate . $ ] depends on the arrival rate @xmath143,width=302 ] with the results from equation ( [ equ16 ] ) or ( [ equ17 ] ) , let us now provide some useful discussions on the _ asymptotic behavior _ of the fixed point @xmath144 . note that we express @xmath145 if @xmath146 . if the general distribution @xmath147 and its mean @xmath148 are given , then @xmath149 ^{d}$]d@xmath26 is a deterministic factor . we have@xmath150 in this case , the heavy traffic should have a bigger influence on the asymptotic behavior of the sequence @xmath151 . if @xmath152 is given , then@xmath153 in this case , the maximal value @xmath154 of the positive number @xmath155 ^{d}$]d@xmath26 should have a bigger influence on the asymptotic behavior of the sequence @xmath156 . in this section , we provide a necessary discussion for the key parameter @xmath8 in the doubly exponential solution of theorem [ the : cfs ] . based on this , for the fixed point we give a new and important observation : the doubly exponential solution to the fixed point can extensively exist for @xmath127 , even if the service time distribution is heavy - tailed . note that@xmath157 ^{d}\text{d}x\nonumber\\ & = \frac{\int_{0}^{+\infty}\left [ \overline{g}\left ( x\right ) \right ] ^{d}\text{d}x}{\left [ \int_{0}^{+\infty}\overline{g}\left ( x\right ) \text{d}x\right ] ^{d } } , \label{dequ1}\ ] ] it is easy to see that @xmath158 if @xmath4 . thus , we need to analyze the case for @xmath48 as follows . since @xmath130 , we get that @xmath159 ^{d}\leq\overline{g}\left ( x\right ) \leq1 $ ] , which leads to@xmath160 ^{d}dx\leq\int_{0}^{+\infty}\overline{g}\left ( x\right ) dx=1/\mu.\ ] ] it is easy to see that @xmath161 , and thus if @xmath127 , then @xmath95 . in what follows we analyze five simple and useful examples . in first two examples , the service time distribution is light - tailed ; while in the last three examples , the service time distribution is heavy - tailed . specifically , the examples with heavy - tailed service times illustrate two important observations : the first one indicates that the fixed point for the supermarket model is different from the tail of stationary queue length distribution for the ordinary m / g/1 queue , and the second one is to show that the doubly exponential solution to the fixed point can exist extensively if the service time mean is non - zero and finite . * example one : * exponential distribution . let @xmath162 . then @xmath163 . it is easy to see that when @xmath164{d}$ ] , @xmath165 ; when @xmath166{d}$ ] , @xmath158 ; and when @xmath167{d}$ ] , @xmath168 . if @xmath169 , then @xmath8 is a linear function of @xmath170 , and if @xmath171 , then @xmath8 is a nonlinear function of @xmath170 . figures 3 and 4 show the functions @xmath172 and @xmath173 , respectively . is a linear function of @xmath170 for @xmath169,width=302 ] is a nonlinear function of @xmath170 for @xmath171,width=302 ] * example two : * erlang distribution @xmath174 . let @xmath175 . then @xmath8 is given by@xmath176 ^{d}\text{d}x.\ ] ] let @xmath177 . table 1 lists how @xmath8 depends on the parameter pair @xmath178 . as seen from table 1 , @xmath8 is decreasing for each of the two parameters @xmath179 and @xmath7 . [ c]|c||c|c|c|c|c||c|c|@xmath180 & ( 2 , 2 ) & ( 2 , 5 ) & ( 2 , 10 ) & ( 5 , 2 ) & ( 10 , 2 ) & ( 5 , 5 ) & ( 10 , 10 ) + @xmath8 & 0.52 & 0.19 & @xmath181 & @xmath182 & @xmath183 & @xmath184 & @xmath185 + * example three : * weibull distribution @xmath186 . let @xmath187 . it is easy to check that the mean of the weibull distribution is given by@xmath188 which follows that @xmath8 is given by@xmath189 ^{d-1}},\ ] ] where @xmath190d@xmath26 . obviously , the weibull distribution @xmath186 is heavy - tailed if @xmath191 ; and the weibull distribution @xmath186 is light - tailed if @xmath192 . to indicate the role played by the heavy - tailed parameter @xmath193 for @xmath191 , taking @xmath194 and @xmath169 we have@xmath195 table 2 indicates how @xmath8 depends on the heavy - tailed parameter @xmath193 , such as , @xmath168 if @xmath196 ; @xmath165 if @xmath197 . this example , together with theorem [ the : cfs ] , illustrates an important observation that the fixed point @xmath198 is doubly exponential ( clearly , it is light - tailed ) even if the service time distribution is heavy - tailed . based on this , the the fixed point is different from the tail of stationary queue length distribution of the ordinary m / g/1 queue , since for the ordinary m / g/1 queue , the stationary queue length distribution is heavy - tailed if the service time distribution is heavy - tailed , e.g. , see adler , feldman and taqqu @xcite . [ c]|c||c|c|c|c|c||c|c|c|@xmath193 & 0.2 & 0.3 & 0.4 & 0.5 & 0.6 & 0.7 & 0.8 & 0.9 + @xmath8 & @xmath199 & @xmath200 & 0.27 & 0.63 & 1.05 & 1.47 & 1.86 & 2.19 + * example four : * power law distribution . let @xmath201 . if @xmath202 , then the power law distribution does not exist the finite mean . in this case , we can not setup the system of integral - differential equations for the fraction density vector @xmath70 which leads to the analysis for the fixed point . thus we only deal with the case with @xmath203 . note that for each @xmath203 @xmath204 and@xmath160 ^{d}\text{d}x=\frac{1}{\mu},\ ] ] thus we obtain @xmath205 . it is easy to see that when @xmath206 , @xmath165 ; when @xmath177 , @xmath158 ; and when @xmath207 , @xmath168 . it follows from theorem [ the : cfs ] that for @xmath14@xmath208 this indicates that the fixed point @xmath209 is doubly exponential ( of course , it is light - tailed ) if the service time distribution is power law . * example five : * almost exponential distribution . let @xmath210 . then it is easy to see that the almost exponential distribution is heavy - tailed if @xmath211 @xmath212 ^{d}}\ ] ] table 3 lists how @xmath8 depends on the parameter pair @xmath213 . as seen from table 3 , @xmath8 is decreasing for each of the two parameters @xmath7 and @xmath214 . [ c]|c||c|c|c|c|@xmath215 & ( 2 , 2 ) & ( 4 , 2 ) & ( 2 , 4 ) & ( 4 , 4 ) + @xmath8 & @xmath216 & @xmath217 & @xmath218 & @xmath219 + in this section , as an important example we provide three methods to analyze a supermarket model with poisson arrivals and ph service times . our purpose is to provide three different ways to determine the key parameter @xmath8 and compute the doubly exponential solution to the fixed point . also , we indicate that the doubly exponential solution to the fixed point is not unique for a more general supermarket model . the supermarket model with poisson arrivals and ph service times is described as follows . customers arrive at a queueing system of @xmath9 servers as a poisson process with arrival rate @xmath10 for @xmath11 . the service times of these customers are of phase type with irreducible representation @xmath220 of order @xmath179 . each arriving customer chooses @xmath16 servers independently and uniformly at random from these @xmath17 servers , and waits for service at the server which currently contains the fewest number of customers . if there is a tie , servers with the fewest number of customers will be chosen randomly . all customers in any service center will be served in the fcfs manner . for the ph service time distribution , we use the irreducible representation @xmath220 of order @xmath179 , where the row vector @xmath214 is a probability vector whose @xmath221th entry is the probability that a service begins in phase @xmath221 for @xmath222 ; and @xmath223 is a matrix of order @xmath179 whose @xmath224 entry is denoted by @xmath225 with @xmath226 for @xmath227 , and @xmath228 for @xmath229 and @xmath230 . let @xmath231 , where @xmath232 is a column vector of ones with a suitable dimension in the context . when a ph service time is in phase @xmath233 , the transition rate from phase @xmath233 to phase @xmath221 is @xmath225 , the service completion rate is @xmath234 . at the same time , the mean service rate is given by@xmath235 unless we state otherwise , we assume that all the random variables defined above are independent , and that the system is operating at the stable region : @xmath18 . we introduce some useful notation . let @xmath236 be the number of queues with at least @xmath2 customers and the service time in phase @xmath233 at time @xmath23 . clearly , @xmath237 for @xmath227 and @xmath238 . we define@xmath239 which is the fraction of queues with at least @xmath2 customers and the service time in phase @xmath233 . we write@xmath240 and for @xmath14,@xmath241@xmath242 we now introduce hadamard product of two matrices @xmath243 and @xmath244 as follows:@xmath245 specifically , for @xmath48 we have@xmath246 let @xmath247 . we write@xmath248 using a similar analysis to that in equations ( [ equ3 ] ) to ( [ equ6 ] ) , we can obtain the following systems of differential vector equations for the fraction density vector @xmath249.@xmath250@xmath251@xmath252 and for @xmath48,@xmath253 if @xmath254 is a fixed point of the fraction density vector @xmath255 , then the system of differential vector equations ( [ eq0 ] ) to ( [ eq3 ] ) can be simplified as@xmath256@xmath257@xmath258 and for @xmath48,@xmath259 in what follows we provide three methods to solve the system of nonlinear equations ( [ eq3 - 1 ] ) to ( [ eq6 ] ) , and give three different doubly exponential solutions to the fixed point . the first method is based on theorem [ the : cfs ] given in this paper . for the ph service time distribution@xmath260 let @xmath94 ^{d}$]d@xmath26 , and we assume that @xmath95 . then the fixed point @xmath261 is given by@xmath110 and for @xmath14@xmath262 the second method is proposed in li , wang and liu @xcite , and the key parameter @xmath8 is based on the stationary probability vector @xmath263 of the irreducible markov chain @xmath264 , that is , @xmath265 . it follows from equation ( [ eq4 ] ) that@xmath266 note that@xmath267@xmath268 thus , we obtain@xmath269 based on the fact that @xmath91 and @xmath270 , it follows from equation ( [ eq5 ] ) that@xmath271 which leads to@xmath272 note that @xmath273 and @xmath274 , we obtain@xmath275 let @xmath276 . then it is easy to see that @xmath277 , and@xmath278 using a similar analysis on equation ( [ eq8 ] ) , we have@xmath279 based on @xmath280 and @xmath281 , it follows from equation ( [ eq6 ] ) that for @xmath103,@xmath282 which leads to@xmath283 thus we obtain@xmath284 using a similar analysis on equation ( [ eq8 ] ) , we have@xmath285 now , we assume that @xmath286 is correct for the cases with @xmath116 . then it follows from equation ( [ eq6 ] ) that for @xmath121 , we have @xmath287 which leads to@xmath288 thus we obtain@xmath289 by a similar analysis to ( [ eq8 ] ) , we have@xmath290 therefore , by induction the fixed point @xmath291 is given by@xmath110 and for @xmath14@xmath292 the third method is based on the matrix computation for the system of nonlinear equations ( [ eq3 - 1 ] ) to ( [ eq6 ] ) , and shows that the key parameter @xmath8 is based on the initial probability vector @xmath214 in the ph service time distribution , that is , @xmath293 it follows from ( [ eq3 - 1 ] ) to ( [ eq6 ] ) that @xmath294{ccccc}-\lambda & \lambda & & & \\ & -\lambda & \lambda & & \\ & & -\lambda & \lambda & \\ & & & \ddots & \ddots \end{array } \right ) + \left ( \pi_{1},\pi_{2},\pi_{3},\ldots\right ) \left ( \begin{array } [ c]{cccc}t & & & \\ t^{0}\alpha & t & & \\ & t^{0}\alpha & t & \\ & & \ddots & \ddots \end{array } \right ) \\ & = -\left ( \lambda\alpha,0,0,\ldots\right ) , \end{aligned}\ ] ] which leads to@xmath295{ccccc}r & v & & & \\ & r & v & & \\ & & r & v & \\ & & & \ddots & \ddots \end{array } \right ) \\ & + \left ( \lambda\alpha\left ( -t\right ) ^{-1},0,0,\ldots\right ) , \end{aligned}\ ] ] where@xmath296 and@xmath297 thus we obtain@xmath298 \label{equ29}\ ] ] and for @xmath48@xmath299 + \pi_{k}^{\odot d}\left [ \lambda\left ( -i+e\alpha\right ) \left ( -t\right ) ^{-1}\right ] . \label{equ30}\ ] ] to omit the term @xmath300 $ ] for @xmath14 , we assume that @xmath301 has the following expression@xmath302 in this case , we have@xmath303 = r^{d}\left ( k\right ) \alpha\left [ \lambda\left ( -i+e\alpha\right ) \left ( -t\right ) ^{-1}\right ] = 0,\ ] ] thus it follows from ( [ equ29 ] ) and ( [ equ30 ] ) that@xmath304 and for @xmath48@xmath299 . \label{equ32}\ ] ] it follows from ( [ equ31 ] ) that@xmath305 which follows that@xmath306 where@xmath307 it follows from ( [ equ32 ] ) that@xmath308 , \ ] ] which follows that@xmath309 therefore , we can obtain@xmath125 and for @xmath14@xmath310 based on the above three methods , we can summarize the key parameter and the doubly exponential solution to the fixed point in the following table . [ c]|c||c|c| & key parameter & fixed point + method 1 & @xmath311 ^{d}\text{d}x}{\left [ -\alpha t^{-1}e\right ] ^{d}}$ ] & @xmath312 + method 2 & @xmath276 & @xmath286 + method 3 & @xmath313 & @xmath314 + when the ph service time is an @xmath179-order erlang distribution with the irreducible representation @xmath315 , where@xmath316 and@xmath317{ccccc}-\eta & \eta & & & \\ & -\eta & \eta & & \\ & & \ddots & \ddots & \\ & & & -\eta & \eta\\ & & & & -\eta \end{array } \right ) , \text { \ \ } t^{0}=\left ( \begin{array } [ c]{c}0\\ 0\\ \vdots\\ 0\\ \eta \end{array } \right ) .\ ] ] we have@xmath318 and@xmath319 thus the doubly exponential solution by the third method is given by@xmath320 it is clear that@xmath321{ccccc}-\eta & \eta & & & \\ & -\eta & \eta & & \\ & & \ddots & \ddots & \\ & & & -\eta & \eta\\ \eta & & & & -\eta \end{array } \right ) , \ ] ] which leads to the stationary probability vector of the markov chain @xmath264 as follows : @xmath322@xmath323@xmath324 and@xmath325 thus the doubly exponential solution by the second method is given by @xmath326 it is clear that the three doubly exponential solutions ( [ eq7 ] ) , ( [ equ37 ] ) and ( [ equ38 ] ) are different for @xmath327 . for the supermarket model with poisson arrivals and ph service times , we have obtained three different doubly exponential solutions to the fixed point . it is interesting but difficult how to be able to find another new doubly exponential solution . we believe that it is an open problem how to give all the doubly exponential solutions to the fixed point for a more general supermarket model including the case with map arrivals , ph service times or general service times . in this section , we study the exponential convergence of the current location @xmath58 of the supermarket model to its fixed point @xmath75 for @xmath23 and @xmath30 . not only does the exponential convergence indicates the existence of the fixed point , but it also explains such a convergent process is very fast . for the supermarket model , the initial point @xmath328 can affect the current location @xmath58 for each @xmath329 , since the service process in the supermarket model is under a unified structure . here , we provide notation for comparison of two vectors . let @xmath247 and @xmath330 . we write @xmath331 if @xmath332 for all @xmath14 ; @xmath333 if @xmath334 for all @xmath14 . now , we can obtain the following useful proposition whose proof is clear from a sample path analysis and is omitted here . [ prop1]if @xmath335 , then @xmath336 . based on proposition [ prop1 ] , the following theorem shows that the fixed point @xmath75 is an upper bound of the current location @xmath58 for all @xmath23 and @xmath30 . [ the : upperb]for the supermarket model , if there exists some @xmath2 such that @xmath337 , then the sequence @xmath43 has a upper bound sequence which decreases doubly exponentially for all @xmath23 and @xmath30 , that is , @xmath338 for all @xmath23 and @xmath30 . * proof * let@xmath339@xmath340 then @xmath341 , and for each @xmath14 , @xmath342 for all @xmath23 and @xmath30 , since @xmath343 is a fixed point in the supermarket model . if @xmath337 for some @xmath2 , then @xmath344 and @xmath345 for all @xmath346 and @xmath347 , thus @xmath348 . it is easy to see from proposition [ prop1 ] that @xmath349 for all @xmath14 , @xmath23 and @xmath30 . thus we obtain that for all @xmath14 , @xmath23 and @xmath30 @xmath350 .\ ] ] this completes the proof . to show the exponential convergence , we use theorem [ the : upperb ] to define a potential function ( or lyapunov function ) @xmath351 as follows:@xmath352 \text{d}x,\ ] ] where @xmath353 is a positive scalar sequence with @xmath354 for @xmath48 . note that @xmath355 . it is easy to see from proposition [ prop1 ] that @xmath356 for all @xmath23 . when @xmath357 $ ] d@xmath358 for @xmath14 , we write@xmath359 \text{d}x}=c_{k}\left ( t\right)\ ] ] and@xmath360 \text{d}x}=d_{k}\left ( t\right ) .\ ] ] the following lemma provide a method to determine the positive scalar sequence @xmath353 with @xmath354 for @xmath48 . this proof is easy by means of some simple computation . [ lem : positivec]if @xmath361 is a positive constant , @xmath362@xmath363 and for @xmath48@xmath364 then@xmath365 and for @xmath366@xmath367 the following theorem measures the distance @xmath351 of the current location @xmath58 for @xmath23 and the fixed point @xmath75 for @xmath30 , and illustrates that the distance @xmath351 to the fixed point from the current location is very go to zero with exponential convergence . hence , it shows that from any suitable starting point , the supermarket model can be quickly close to the fixed point , that is , there always exists a fixed point in the supermarket model . for @xmath23,@xmath368 where @xmath369 and @xmath361 are two positive constants , and they possibly depend on time @xmath23 . in this case , the potential function @xmath370 is exponentially convergent . * proof * note that@xmath352 \text{d}x,\ ] ] we have@xmath371 \text{d}x\\ & = -\sum_{k=1}^{\infty}w_{k}\frac{\text{d}}{\text{d}t}\int_{0}^{+\infty}s_{k}\left ( t , x\right ) \text{d}x\end{aligned}\ ] ] by means of the dominated convergence theorem . it follows from ( [ equ3 ] ) to ( [ equ6 ] ) that@xmath372 and using ( [ eq4 - 1 ] ) we obtain@xmath373\\ & -\sum_{k=2}^{\infty}w_{k}[\lambda\int_{0}^{+\infty}s_{k-1}^{d}\left ( t , x\right ) \text{d}x-\lambda\int_{0}^{+\infty}s_{k}^{d}\left ( t , x\right ) \text{d}x\\ & -\int_{0}^{+\infty}\mu\left ( x\right ) s_{k}\left ( t , x\right ) \text{d}x+\int_{0}^{+\infty}\mu\left ( x\right ) s_{k+1}\left ( t , x\right ) \text{d}x]\\ = & -w_{1}[-\lambda\int_{0}^{+\infty}s_{1}^{d}\left ( t , x\right ) \text{d}x+\int_{0}^{+\infty}\mu\left ( x\right ) s_{2}\left ( t , x\right ) \text{d}x]\\ & -\sum_{k=2}^{\infty}w_{k}[\lambda\int_{0}^{+\infty}s_{k-1}^{d}\left ( t , x\right ) \text{d}x-\lambda\int_{0}^{+\infty}s_{k}^{d}\left ( t , x\right ) \text{d}x\\ & -\int_{0}^{+\infty}\mu\left ( x\right ) s_{k}\left ( t , x\right ) \text{d}x+\int_{0}^{+\infty}\mu\left ( x\right ) s_{k+1}\left ( t , x\right ) \text{d}x],\end{aligned}\ ] ] which follows @xmath374 \text{d}x\\ & + \sum_{k=2}^{\infty}\left [ \lambda\left ( w_{k}-w_{k+1}\right ) c_{k}\left ( t\right ) + \left ( w_{k}-w_{k-1}\right ) d_{k}\left ( t\right ) \right ] \\ & \cdot\int_{0}^{+\infty}\left [ \pi_{k}\left ( x\right ) -s_{k}\left ( t , x\right ) \right ] \text{d}x.\end{aligned}\ ] ] using lemma [ lem : positivec ] we can easily choose a parameter @xmath375 and a suitable positive scalar sequence @xmath353 with @xmath354 for @xmath48 such that@xmath376 and for @xmath48@xmath377 thus we can obtain@xmath378 which leads to @xmath379 this completes the proof . we have provided an algorithm for computing the positive scalar sequence @xmath353 with @xmath380 for @xmath48 as follows : step one:@xmath381 step two:@xmath382 step three : for @xmath48@xmath367 this illustrates that @xmath383 is a function of time @xmath22 . note that @xmath384 , it is clear that for @xmath48@xmath385 in this section , we apply the kurtz theorem to study the supermarket model with general service times , and analyze the lipschitz condition with respect to general service times . the supermarket model can be analyzed by a density dependent jump markov process , where the density dependent jump markov process is a markov process with a single parameter @xmath17 which corresponds to the population size . kurtz s work provides a basis for the density dependent jump markov processes in order to relate the infinite - size system of differential equations to the corresponding finite - size system of differential equations . readers may refer to kurtz @xcite for more details . in the supermarket model , the states of density dependent jump markov process can be normalized and interpreted as measuring population densities , so that the transition rates depend only on these densities . hence , the infinite - size system of differential equations can be regarded as the limiting model of the corresponding finite - size system of differential equations as the population size grows arbitrarily large . when the population size is @xmath17 , we write@xmath386 for @xmath14 and @xmath30 , we write@xmath387 where @xmath26 is the residual service time of each server , and@xmath388 let@xmath389 and for @xmath14@xmath390 let @xmath391 be a density dependent jump markov process on the state space @xmath392 whose transition rates are given by@xmath393 in this supermarket model , @xmath394 is the unscaled process which records the number of servers with at least @xmath2 customers for @xmath238 . let @xmath395 and @xmath396 denote an arrival and a service completion , respectively . hence taking @xmath397 or @xmath396 for @xmath398 , we write @xmath399@xmath400@xmath401 ^{d}\text{d}x,\]]@xmath402 and for @xmath403@xmath404 ^{d}\text{d}x-\lambda\int_{0}^{+\infty}\left [ s_{k}^{\left ( n\right ) } \left ( x\right ) \right ] ^{d}\text{d}x,\]]@xmath405 using chapter 7 in kurtz @xcite or subsection 3.4.1 in mitzenmacher @xcite , the markov process @xmath406 with transition rates @xmath407 is given by@xmath408 where @xmath409 for @xmath397 and @xmath396 are two independent standard poisson processes . clearly , the jump markov process by equation ( [ eq7.1 ] ) at time @xmath22 is determined by the starting point and the transition rates which are integrated over its history . let@xmath410 taking @xmath411 which is an appropriate scaled process , we have@xmath412 where @xmath413 is a poisson process centered at its expectation . note that in ( [ equ7.5 ] ) , the function @xmath414 given in ( [ equ7.4 ] ) is for @xmath415 . taking @xmath416 and @xmath417 , we obtain@xmath418 due to the fact that@xmath419 by means of the law of large numbers . note that in ( [ equ7.6 ] ) , the function @xmath414 given in ( [ equ7.4 ] ) is for @xmath420 . in the supermarket model , the deterministic and continuous process @xmath421 is described by the infinite - size system of integral - differential equations ( [ equ3 ] ) to ( [ equ6 ] ) , or simply in the below@xmath422 with the initial condition@xmath423 now , we consider the uniqueness of the limiting deterministic process @xmath421 with ( [ equ7.2 ] ) to ( [ equ7.3 ] ) , or the uniqueness of solution to the infinite - size system of integral - differential equations ( [ equ3 ] ) to ( [ equ6 ] ) . to that end , a sufficient condition is lipschitz , that is , for some constant @xmath424@xmath425 in general , the lipschitz condition is standard and sufficient for the uniqueness of solution to the finite - size system of differential equations ; while for the countable infinite - size case , readers may refer to theorem 3.2 in deimling @xcite and subsection 3.4.1 in mitzenmacher @xcite for some generalization . to check the lipschitz condition , as @xmath1 we have @xmath426@xmath427@xmath428 ^{d}\text{d}x,\]]@xmath429 and for @xmath112@xmath430 ^{d}\text{d}x-\lambda\int_{0}^{+\infty } \left [ s_{k}\left ( x\right ) \right ] ^{d}\text{d}x,\]]@xmath431 let@xmath432 ^{d}\text{d}x}{\int_{0}^{+\infty}s_{k}\left ( x\right ) \text{d}x}\ ] ] and@xmath433 then @xmath434 for @xmath14 the following theorem shows that the supermarket model with general service times satisfies the lipschitz condition for the infinite - size system of integral - differential equations ( [ equ3 ] ) to ( [ equ6 ] ) . [ the : lip]the supermarket model with general service times satisfies the lipschitz condition . * proof * let@xmath435 for two arbitrary entries @xmath436 , we have@xmath437 now , we analyze the following four cases for the function @xmath438 , while the function @xmath439 can be analyzed similarly . case one : @xmath440 . in this case , we have@xmath441 ^{d}\text{d}x|\\ & = \lambda|2-\zeta_{1}\int_{0}^{+\infty}s_{1}\left ( x\right ) \text{d}x|\\ & = \lambda\left [ 2-\zeta_{1}\int_{0}^{+\infty}s_{1}\left ( x\right ) \text{d}x\right ] , \end{aligned}\ ] ] since @xmath442d@xmath443 . taking@xmath444 it is clear that@xmath445 \\ & = m_{a}\left ( 0,1\right ) |y - z|.\end{aligned}\ ] ] case two : @xmath446 for @xmath48 . in this case , we have@xmath447 ^{d}\text{d}x+\lambda\int_{0}^{+\infty}\left [ s_{k}\left ( x\right ) \right ] ^{d}\text{d}x|\\ & = \lambda|-1-\zeta_{k-1}\int_{0}^{+\infty}s_{k-1}\left ( x\right ) \text{d}x+\zeta_{k}\int_{0}^{+\infty}s_{k}\left ( x\right ) \text{d}x|\\ & = \lambda\left [ 1+\zeta_{k-1}\int_{0}^{+\infty}s_{k-1}\left ( x\right ) \text{d}x-\zeta_{k}\int_{0}^{+\infty}s_{k}\left ( x\right ) \text{d}x\right]\end{aligned}\ ] ] due to that @xmath448d@xmath443 . let@xmath449 then@xmath450 \\ & = m_{a}\left ( 0,k\right ) |y - z|.\end{aligned}\ ] ] case three : @xmath451 for @xmath48 . in this case , we have@xmath452 ^{d}\text{d}x-\lambda\int_{0}^{+\infty}\left [ s_{k-1}\left ( x\right ) \right ] ^{d}\text{d}x+\lambda\int_{0}^{+\infty}\left [ s_{k}\left ( x\right ) \right ] ^{d}\text{d}x|\\ & = \lambda|1-\zeta_{1}\int_{0}^{+\infty}s_{1}\left ( x\right ) \text{d}x-\zeta_{k-1}\int_{0}^{+\infty}s_{k-1}\left ( x\right ) \text{d}x+\zeta _ { k}\int_{0}^{+\infty}s_{k}\left ( x\right ) \text{d}x|\end{aligned}\ ] ] let@xmath453 then@xmath454 case four : @xmath455 for @xmath456 . in this case , we have@xmath457 ^{d}\text{d}x-\lambda\int_{0}^{+\infty}\left [ s_{l}\left ( x\right ) \right ] ^{d}\text{d}x\\ & -\lambda\int_{0}^{+\infty}\left [ s_{k-1}\left ( x\right ) \right ] ^{d}\text{d}x+\lambda\int_{0}^{+\infty}\left [ s_{k}\left ( x\right ) \right ] ^{d}\text{d}x|\\ = & \lambda|\zeta_{l-1}\int_{0}^{+\infty}s_{l-1}\left ( x\right ) \text{d}x+\zeta_{l}\int_{0}^{+\infty}s_{l}\left ( x\right ) \text{d}x\\ & -\zeta_{k-1}\int_{0}^{+\infty}s_{k-1}\left ( x\right ) \text{d}x+\zeta _ { k}\int_{0}^{+\infty}s_{k}\left ( x\right ) \text{d}x|.\end{aligned}\ ] ] let@xmath458 then@xmath459 based on the above four cases , taking@xmath460 we obtain that for two arbitrary entries @xmath461@xmath462 similarly , we can choose a positive number @xmath463 such that for two arbitrary entries @xmath461@xmath464 let @xmath465 . then for two arbitrary entries @xmath461@xmath425 this completes the proof . based on theorem [ the : lip ] , the following theorem easily follows from theorem 3.13 in mitzenmacher @xcite . [ the : kurtz]in the supermarket model with general service times , @xmath466 and @xmath467 are respectively given by ( [ equ7.5 ] ) and ( [ equ7.6 ] ) , we have@xmath468 * proof * it is seen from that in the supermarket model with general service times , the function @xmath414 for @xmath469 satisfies the lipschitz condition . at the same time , it is easy to take a subset @xmath470 such that@xmath471 and@xmath472 thus , this proof can easily be completed by means of theorem 3.13 in mitzenmacher @xcite . this completes the proof . using theorem 3.11 in mitzenmacher @xcite and theorem [ the : kurtz ] , we can obtain the following theorem for the expected sojourn time that a customer spends in an initially empty supermarket model with general service times over the time interval @xmath473 $ ] . in the supermarket model with general service times , the expected sojourn time that a customer spends in an initially empty system over the time interval @xmath473 $ ] is bounded above by@xmath474 -e\left [ x\right ] \right\ } + e\left [ x\right ] \left [ \sum_{k=1}^{\infty}\theta^{\frac{d^{k}-1}{d-1}}\rho^{\frac{d^{k}-d}{d-1}}\right ] + o\left ( 1\right ) , \ ] ] where @xmath475 is understood as @xmath1 . in this paper , we provide a novel and simple approach to study the randomized load balancing model with general service times , which is described as an infinite - size system of integral - differential equations . this approach is based on the supplementary variable method , which is always applied in dealing with stochastic models of m / g/1 type , e.g. , see li and zhao @xcite and li @xcite . we organize an infinite - size system of integral - differential equations by means of the density dependent jump markov process , and obtain a close - form solution : doubly exponential structure , for the fixed point satisfying the system of nonlinear equations , which is always a key in the study of supermarket models . since the fixed point is decomposited into two groups of information under a product form , we indicate three important observations : 1 . the fixed point for the supermarket model is different from the tail of stationary queue length distribution for the ordinary m / g/1 queue ; 2 . the doubly exponential solution to the fixed point can exist extensively for @xmath127 even if the service time distribution is heavy - tailed ; and 3 . the doubly exponential solution to the fixed point is not unique for a more general supermarket model . furthermore , we analyze the exponential convergence of the current location of the supermarket model to its fixed point , and study the lipschitz condition in the kurtz theorem under general service times . finally , we present numerical examples to illustrate the effectiveness of our approach in analyzing the randomized load balancing schemes with the non - exponential service requirements . based on this analysis , one can gain a new and important understanding how workload probing can help in load balancing jobs with general service times such as heavy - tailed service . the approach of this paper is useful in analyzing the randomized load balancing schemes in resource allocation in computer networks . we expect that this approach will be applicable to the study other randomized load balancing schemes with general service times , for example , generalizing the arrival process to non - poisson : the renewal arrival process or the markovian arrival process . the work of q.l . li was supported by the national science foundation of china under grant no . 10871114 and the national grand fundamental research 973 program of china under grant no . 2006cb805901 . y. azar , a.z . broder , a.r . karlin and e. upfal ( 1999 ) . balanced allocations . _ siam journal on computing _ * 29 * , 180200 . a preliminary version of this paper appeared in _ proceedings of the twenty - sixth annual acm symposium on the theory of computing _ , 1994 . m. mitzenmacher , a. richa , and r. sitaraman ( 2001 ) . the power of two random choices : a survey of techniques and results . in _ handbook of randomized computing : volume 1 _ , edited by p. pardalos , s. rajasekaran and j. rolim , pp . 255 - 312 . m. mitzenmacher and b. vcking ( 1998 ) . the asymptotics of selecting the shortest of two , improved . in _ proceedings of the 37th annual allerton conference on communication , control , and computing _ , pages 326327 . a full version is available as harvard computer science tr-08 - 99 .

the framework of network equivalence theory developed by koetter et al . introduces a notion of channel emulation to construct noiseless networks as upper ( resp . lower ) bounding models , which can be used to calculate the outer ( resp . inner ) bounds for the capacity region of the original noisy network . based on the network equivalence framework , this paper presents scalable upper and lower bounding models for wireless networks with potentially many nodes . a channel decoupling method is proposed to decompose wireless networks into decoupled multiple - access channels ( macs ) and broadcast channels ( bcs ) . the upper bounding model , consisting of only point - to - point bit pipes , is constructed by firstly extending the `` one - shot '' upper bounding models developed by calmon et al . and then integrating them with network equivalence tools . the lower bounding model , consisting of both point - to - point and point - to - points bit pipes , is constructed based on a two - step update of the lower bounding models to incorporate the broadcast nature of wireless transmission . the main advantages of the proposed methods are their simplicity and the fact that they can be extended easily to large networks with a complexity that grows linearly with the number of nodes . it is demonstrated that the resulting upper and lower bounds can approach the capacity in some setups . capacity , channel decoupling , channel emulation , equivalence , wireless networks

1401.4189

a theory of network equivalence has been established in @xcite by koetter et al . to characterize the capacity of a ( large ) memoryless noisy network . the original noisy network is first decomposed into many independent single - hop noisy channels , each of which is then replaced by the corresponding upper ( resp . lower ) bounding model consisting of only noiseless bit pipes . the capacity region of the resulting noiseless network serves as an outer ( resp . inner ) bound for the capacity region of the original noisy network , whose capacity is otherwise difficult to characterize . a noisy channel and a noiseless bit pipe are said to be equivalent if the capacity region of any arbitrary network that contains the noisy channel remains unchanged after replacing the noisy channel by its noiseless counterpart . the equivalence between a point - to - point noisy channel and a point - to - point noiseless bit pipe has been established in @xcite as long as the rate of the latter equals the capacity of the former . for independent single - hop multi - terminal channels , such as the multiple - access channel ( mac ) , the broadcast channel ( bc ) , and the interference channel ( ic ) , operational frameworks for constructing upper and lower bounding models have been proposed in @xcite . the constructive proofs presented in @xcite are based on a notion of channel emulation over a stacked replicas of the channel , where the lower bounding models are established based on channel coding arguments and the upper bounding models are constructed based on lossy source coding arguments . the bounding accuracy , in terms of both multiplicative and additive gaps between capacity upper and lower bounds , has been outlined in @xcite for general noisy networks . explicit upper and lower bounding models for mac / bc / ic with two sources and/or two destinations have been constructed in @xcite . for networks consisting of only point - to - point channels , macs with two transmitters , and bcs with two receivers , the additive gap for gaussian networks and the multiplicative gap for binary networks have been specified in @xcite . the bounds obtained from network equivalence tools @xcite can be tight in some setups , as shown in @xcite for a multiple unicast network consisting of noisy two - user bcs , and in @xcite for a frequency - division awgn relay network in the wideband regime when the bc is physically degraded or when the source treats the stochastically degraded bc as physically degraded . a class of `` one - shot '' upper bounding models proposed in @xcite by calmon et al . introduces an auxiliary node for each bc / mac to facilitate separate characterization of the sum rate and the individual rates . the rate of a bit pipe from / to the auxiliary node can be characterized either by channel emulation over infinite number of channel uses as in @xcite , or by emulating the transmission over each channel use ( hence named `` one - shot '' ) . although it has been established in @xcite that separate source and channel coding incurs no capacity loss in transmitting a memoryless source over a memoryless point - to - point channel , similar separation results on memoryless networks are not known until recently . the optimality of channel and network coding separation for general communication demands over networks consisting of independent memoryless point - to - point channels is established in @xcite based on the network equivalence theory , where the networks may be cyclic or acyclic and alphabets can be discrete or continuous . feedback and cooperation among nodes are also accommodated , but the sources have to be independent . for single - source multicast connections over networks consisting of point - to - point discrete memoryless channels ( dmcs ) with finite alphabets , separation of channel and network coding has been established in @xcite for acyclic synchronized networks and in @xcite for asynchronized networks with and without cycles . while both @xcite and @xcite rely on the max - flow min - cut theorem @xcite to show the converse ( cf . upper bounding model ) , similar results on channel - network separation have been established in @xcite by using normalized entropy vectors . for point - to - point channels , the same upper bounding models established in @xcite have also been developed in @xcite for dmcs with finite - alphabet , and in @xcite under the notion of strong coordination , where total variation ( i.e. , an additive gap ) is used to measure the difference between the desired joint distribution and the empirical joint distribution of a pair of sequences ( or a pair of symbols as in _ empirical coordination _ ) . the concept of channel emulation @xcite , on the other hand , focuses on the set of jointly typical input - output pairs and the difference between the empirical joint distribution ( averaged over ensembles of channel emulators ) and the desired joint distribution is quantified by a multiplicative gap to ensure a small probability of error events . as we focus on characterizing capacity ( bounds ) rather than reconstructing ( exact ) common randomness for multi - terminal channels , we shall follow the channel emulation framework @xcite when constructing bounding models for bcs and macs . it is , however , non - trivial to apply the network equivalence tools @xcite onto wireless networks owing to the broadcast nature of wireless transmission . on one hand , the bounding models proposed in @xcite for macs / bcs with @xmath0 transmitters / receivers contain up to @xmath1 bit pipes , leading to computational inefficiency when @xmath0 is large ( as in a wireless hot - spot which may contain potentially many users ) . on the other hand , the received signal at a terminal may contain several broadcasted signals , which creates dependence / interference among several transmitter - receiver pairs . although such dependence has been partially incorporated into ics , the whole family of non - layered multi - hop channels ( e.g. , relay channels ) have been excluded from consideration since the channel emulation techniques are developed for single - hop channels . in this paper , we present simple but efficient methods to construct upper and lower bounding models for wireless networks with potentially many nodes , at a complexity that grows linearly with the number of nodes . we propose a channel decoupling approach to decompose a memoryless wireless network into decoupled bcs and macs . in our bounding models , the constraints on the sum rate and on the individual rates are characterized by different channel emulation techniques as inspired by @xcite . our upper bounding models , which consist of only point - to - point bit pipes , are constructed by first extending the one - shot models for macs / bcs to many - user scenarios and then integrating them with the channel emulation techniques . our lower bounding models , which contain both point - to - point and point - to - points ( hyper - arc ) bit pipes , are obtained based on a two - step update of the decoupled bcs and macs by taking their dependence into account . the main advantage of our proposed bounding models are their simplicity and the fact that they can be easily extended to large networks . we demonstrate by examples that the resulting upper and lower bounds can approach the capacity in some setups . throughout this paper , we assume memoryless independent sources as in @xcite , where the optimality of channel - network separation is established for networks with only point - to - point channels . extension to memoryless correlated sources can be found in @xcite where channel - network separation is established in the context of lossy @xcite and lossless @xcite source coding problems over networks with only point - to - point dmcs . extension to awgn channels is also established in @xcite . we also assume that the distortion components ( e.g. , noise ) are independent from the transmitted signals . this assumption can be relaxed in scenarios where the noise power is dependent on the power of input signals , and in such scenarios we take the smallest ( resp . largest ) noise power when constructing upper ( resp . lower ) bounding models . we further assume that the distortion components at receiving nodes within a coupled . ] bc are mutually independent , and the scenario of coupled bc with correlated noise will be investigated in our future work . note that in this paper our focus is on constructing noiseless bounding networks that can serve as a basis to compute capacity bounds , rather than finding the capacity of a noiseless network , which itself is a very difficult problem @xcite . we refer to @xcite for various computational tools available to characterize ( bounds on ) the capacity of noiseless networks . there are some other methods aiming at characterizing the capacity of wireless networks . a deterministic approach proposed in @xcite can approximate the capacity of gaussian networks within a constant gap in the high signal - to - noise ratio ( snr ) regime , where amplify - and - forward ( af ) has been proved to approach the capacity in multi - hop layered networks @xcite . a layering approach with a global information routing proposed in @xcite for wireless networks with non - coupled bcs and macs can provide lower bounds within a multiplicative gap from the capacity upper bound . capacity approximations for multiple unicast transmissions over coupled wireless networks developed in @xcite combines polymatroidal network analysis with carefully designed coding schemes for each independent single - hop interference channel , and the approximation accuracy is characterized for bidirectional networks with symmetric fading coefficients . since we aim at approaches that can be used in all snr regions and for all communication tasks , we do not follow the methods developed in @xcite . the rest of this work is organized as follows.we first introduce in sec . [ sec : eqv : background ] a few important definitions and a brief introduction of the network equivalence theory and the one - shot method . then in sec . [ sec : eqv : noncoup ] we present our improvement on the bounding models for independent bcs and macs . in sec . [ sec : eqv : coup ] we describe the network decoupling method for coupled networks and demonstrate how the upper and lower bounding models are constructed by taking the coupled structure into account . we illustrate our bounding models in sec . [ sec : coupled : illust ] by constructing upper and lower bounding models for coupled networks and conclude this work in sec . [ sec : eqv : conc ] . in this section we present a few basic definitions that are frequently used in our paper and give a brief introduction of the stacked as well as the one - shot channel emulation techniques . we inherent the setups for bounding models and channel emulation arguments from @xcite , but our notation is slightly different to minimize the number of sub-/sup - scripts . @xmath2 represents a random variable drawn from an alphabet @xmath3 with cardinality @xmath4 , and @xmath5 is a realization with probability @xmath6 . we use subscripts ( @xmath7 ) to differentiate random variables and use superscripts ( @xmath8 ) to indicate the number of realizations drawn independently from the same random variable . @xmath9}$ ] refers to @xmath10 and @xmath11}$ ] is the corresponding set of realizations , one from each random variable . we use @xmath12 to represent the transition function of a memoryless network where @xmath13 is a vector containing one realization from every channel input alphabet and @xmath14 is the vector for all channel outputs . the dimension of @xmath13 and @xmath14 depends on the specific network and therefore will not be specified unless necessary . we represent a memoryless channel / network by a triplet consisting of the input alphabets , the output alphabets , and a conditional probability distribution ( i.e. , the transition function ) that can fully characterize its behavior . a point - to - point memoryless channel with input alphabet @xmath15 , output alphabet @xmath16 , and transition probability @xmath17 is therefore denoted by @xmath18 in wireless networks , a node may receive signals from multiple transmitters via mac or parallel channels , and/or transmit to multiple receivers via bc or parallel channels . therefore a node may associate with multiple input / output alphabets . for a given node @xmath19 , letting @xmath20 be the number of its incoming parallel channels and @xmath21 be the number of its outgoing parallel channels , we denote the alphabets associated with its outgoing channels by the cartesian product @xmath22 and the alphabets associated with its incoming channels by @xmath23 . a memoryless network @xmath24 with the set of nodes @xmath25 and the transition function @xmath12 can be fully represented as @xmath26 note that the structure and behavior of a memoryless network are explicitly characterized by the triplet in , which specifies the association of input / output alphabet(s ) to a specific node in the network . the triplet in , on the other hand , focuses on each individual channel / sub - network and highlights its associated input / output alphabets , where the underlying network structure is implicitly assumed . unless necessary , hereafter we will focus on the model without specifying @xmath27 as the solo alphabet of a node or as one of its outgoing alphabets . [ dfn : indchannel ] a point - to - point channel @xmath28 within a memoryless network @xmath29 is said to be independent if the network transition function @xmath12 can be partitioned as @xmath30 where @xmath31 denotes the vector of @xmath13 without element @xmath32 , and similarly for @xmath33 . to highlight the independence , we emphasize the notation for network @xmath24 as @xmath34 a sub - network @xmath35 within @xmath24 is said to be independent , denoted by @xmath36 if the network transition probability can be partitioned as @xmath37 an independent sub - network may be further partitioned into several independent channels and/or sub - networks . [ dfn : ncoupled ] a network is said to be coupled if any of its channels is part of a mac and a bc simultaneously . that is , it contains a sub - network @xmath38 with @xmath39 , @xmath40 , and its transition function @xmath41 ca nt be partitioned into non - trivial product format . otherwise , the network is non - coupled . a mac and a bc are said to be coupled if they share a common link . for example , the classical three - node relay channel @xmath42 is coupled and the two - hop diamond network is non - coupled . wireless networks , as expected , are in general coupled owing to the broadcast nature of microwave propagation . [ dfn : bit - pipe ] a point - to - point bit pipe of rate @xmath43 is a noiseless channel that can reliably transmit @xmath44 bits during @xmath45 channel uses for any positive integer @xmath45 . it is represented as @xmath46 where @xmath47 is the kronecker delta function . a point - to - points bit pipe ( hyper - arc ) is denoted by @xmath48 where @xmath49 is the number of heads of this hyper - arc . [ dfn : cbound ] given two independent ( multi - terminal ) channels @xmath50 and @xmath51 , @xmath50 is said to upper bound @xmath51 , or equivalently @xmath51 lower bounds @xmath50 , if the capacity ( region ) of @xmath52 is a subset of that for @xmath53 for any network @xmath54 . we denote their relationship by @xmath55 . @xmath50 and @xmath51 are said to be equivalent if @xmath56 . for an independent noisy channel @xmath51 , we construct channels @xmath57 and @xmath58 consisting of only noiseless bit pipes , such that @xmath57 is the upper bounding model and @xmath58 is the lower bounding model for @xmath51 , i.e. , @xmath59 in @xcite , the equivalence between an independent point - to - point noisy channel @xmath51 and a noiseless point - to - point bit pipe @xmath50 has been established , as long as the with capacity of @xmath51 equals the rate of @xmath50 , by showing that any code that runs reliably on a network @xmath52 can also operate on @xmath53 with asymptotically vanishing error probability . the argument is based on a channel emulation technique over a stacked network where @xmath60 parallel replicas of the network have been put together to run the code . as illustrated in fig . [ fig : neq : model ] , for noisy @xmath51 with capacity @xmath61 and rate-@xmath62 bit pipe @xmath50 , the proof can be divided into three steps . in step i , a network and its @xmath60-fold stacked network ( consisting of @xmath60 parallel replicas of the network ) are proved to share the same rate region by showing that any code that can run on the network can also run on its stacked network , and vice versa . therefore we only need to show the equivalence between the stacked network for @xmath51 and that for @xmath50 , as illustrated in fig . [ fig : neq : model](a ) . [ ] [ ] @xmath63[][]@xmath64 [ ] [ ] @xmath65 [ ] [ ] @xmath66 [ ] [ ] @xmath67 [ ] [ ] @xmath68[][]@xmath69 [ ] [ ] @xmath70[][]@xmath71 [ ] [ ] @xmath72[][]@xmath73 [ ] [ ] @xmath74[][]@xmath75 [ ] [ ] @xmath76[][]@xmath77 [ ] [ ] @xmath78[][]@xmath79 with capacity @xmath80 and a noiseless point - to - point bit pipe @xmath50 of rate @xmath62 are said to be equivalent if @xmath81 , where the equality comes from the continuity of the capacity region . the input / output of their corresponding stacked networks are @xmath82 , @xmath83 , and @xmath84.,title="fig:",width=321 ] + ( a ) capacity regions of a network and its stacked network are identical + with capacity @xmath80 and a noiseless point - to - point bit pipe @xmath50 of rate @xmath62 are said to be equivalent if @xmath81 , where the equality comes from the continuity of the capacity region . the input / output of their corresponding stacked networks are @xmath82 , @xmath83 , and @xmath84.,title="fig:",width=321 ] + ( b ) channel coding argument to prove @xmath85 for all @xmath86 with capacity @xmath80 and a noiseless point - to - point bit pipe @xmath50 of rate @xmath62 are said to be equivalent if @xmath81 , where the equality comes from the continuity of the capacity region . the input / output of their corresponding stacked networks are @xmath82 , @xmath83 , and @xmath84.,title="fig:",width=321 ] + ( c ) lossy source coding argument to prove @xmath87 for all @xmath88 in step ii , the proof of @xmath89 employs a channel coding argument over the stack of @xmath60 channel replicas as illustrated in fig . [ fig : neq : model](b ) : a message @xmath72 of @xmath90 bits is mapped by the channel encoder @xmath74 onto a codeword @xmath70 of length @xmath60 , and then transmitted over the @xmath60-stack noisy channels , with one symbol on each replica , such that reliable transmission over the noisy stacked network can be realized with arbitrary small error probability as @xmath60 goes to infinity for all @xmath91 . in step iii , the proof of @xmath92 is based on a lossy source coding argument as illustrated in fig . [ fig : neq : model](c ) : the input sequence @xmath70 to the noisy stacked network is first quantized / compressed by a lossy source encoder @xmath76 into @xmath90 bits , represented by the message @xmath72 , which is then transmitted through the noiseless stacked network . the reconstructed sequence @xmath71 is selected in such a way that it is jointly typical with the transmitted sequence @xmath70 , in contrast to the usual distortion measure . the existence of a good lossy source coding codebook for any @xmath93 is proved by a random coding argument , i.e. , by showing that the average error probability over the randomly chosen ensemble of codebooks is small . finally , the equivalence between @xmath51 of capacity @xmath61 and @xmath50 of rate @xmath62 can be established when @xmath81 based on the continuity of the capacity region . readers are referred to @xcite for a rigorous and thorough treatment . note that the jointly typical pairs @xmath94 that are used to construct the channel emulation codebooks are taken from a `` restricted '' typical set @xmath95 where the associated decoding error probability ( assuming @xmath70 is transmitted through the original noisy channel ) is smaller than a threshold . sequences that do not satisfy this condition are expurgated . that is , given the classical typical set @xmath96 the restricted typical set is defined @xcite as @xmath97 the concept of capacity upper and lower bounding models developed in @xcite has been extended to independent multi - terminal channels in @xcite following similar arguments as illustrated in fig . [ fig : neq : model ] , and multiplicative and additive gaps between lower and upper bounding models for independent multi - terminal channels have been established . illustrative upper and lower bounding models for macs / bcs / ics involving two transmitters and/or two receivers have been demonstrated in @xcite . given a noisy network composed by independent building blocks whose upper and lower bounding models are available , we can replace these building blocks with their corresponding upper ( resp . lower ) bounding models and then characterize an outer ( resp . inner ) bound for its capacity region based on the resulting noiseless network models . for wireless networks , however , it may be difficult in general to apply directly the channel emulation technique to construct bounding models , as the coupled components may involve many transmitting / receiving nodes . for coupled single - hop networks which can be modeled as ics , the bounding models are difficult to characterize even for the simplest @xmath98 setup @xcite . for coupled multi - hop non - layered networks , it is unclear how the channel emulation technique can be extended to incorporate the interaction among different transmitting - receiving pairs across different layers . although one may apply the cut - set bound @xcite to construct upper bounds for wireless networks , the resulting analysis may become quite involved , as illustrated in @xcite , for characterizing upper bounds for small size relay networks . moreover , even if we manage to construct bounding models for a specific coupled network , we have to create new bounding models for each different network topology , which makes it unscalable for wireless networks that have diversified communication scenarios and topologies . instead of using emulation with channel coding or lossy source coding to construct bounding models as in @xcite , a class of one - shot upper bounding models have been proposed in @xcite for independent macs / bcs , where an auxiliary operation node is introduced for each mac / bc to facilitate separate characterization of the sum rate and the individual rates . as illustrated in fig . [ fig : neq : oneshot ] , all the channels in the one - shot upper bounding models are point - to - point bit pipes which can be fully characterized by the rate vector @xmath99 , where @xmath100 is the rate of the bit pipe @xmath101 . the rate constraint @xmath100 can either be constructed by channel emulation over the stacked network as in @xcite , or by channel emulation that is realized in each instance corresponding to a channel use ( hence referred to as `` one - shot '' approach ) . for example , we can construct the sum rate constraint by stacked emulation whilst bound individual rates by one - shot emulation , or the other way around , which results in two types of upper bounding models @xcite . to highlight the specific channel emulation method used on each link , we illustrate in fig . [ fig : neq : oneshot - models ] the explicit operation of the auxiliary node by specifying its input / output alphabets . [ ] [ ] @xmath102[][]@xmath103[][]@xmath104 [ ] [ ] @xmath105[][]@xmath106[][]@xmath107 [ ] [ ] @xmath65 [ ] [ ] @xmath108[][]@xmath109[][]@xmath110[][]@xmath111 transmitters / receivers . the white nodes indicated by @xmath112 are auxiliary operation nodes to specify the rate constraints on the sum rate and on individual rates . all the channels are noiseless bit pipes and independent from others . the one - shot upper bounding models are fully characterized by the rate vector @xmath99 , where @xmath100 is the rate of the noiseless bit pipe @xmath101.,title="fig:",width=321 ] [ ] [ ] @xmath102[][]@xmath103[][]@xmath104[][]@xmath3 [ ] [ ] @xmath105[][]@xmath106[][]@xmath65 [ ] [ ] @xmath102[][]@xmath104[][]@xmath113 [ ] [ ] @xmath105[][]@xmath106[][]@xmath114 [ ] [ ] @xmath115[][]@xmath116[][]@xmath65 [ ] [ ] @xmath3[][]@xmath3[][]@xmath3 -user macs and for two - user bcs : @xmath117 ( upper left ) , @xmath118 ( lower left ) , @xmath119 ( upper right ) , and @xmath120 ( lower right ) . the operation of the auxiliary operation nodes are explicitly specified by the associating input / output alphabets , where @xmath121 and @xmath122 are compound input / output alphabets , and @xmath123 are auxiliary random variables to synthesize @xmath124 through a predefined function @xmath125 as described in .,title="fig:",width=321 ] for macs with @xmath0 transmitters , say @xmath126 , the one - shot emulation for the channel from transmitter with alphabet @xmath15 sends exactly the corresponding source symbol @xmath127 to the auxiliary operation node @xmath112 , hence requiring a noiseless bit pipe of rate @xmath128 . the auxiliary node then combines all the inputs together and formulates a super alphabet @xmath121 . the channel from @xmath112 to the receiver is then simply a point - to - point channel with input alphabet @xmath113 and output alphabet @xmath65 . this setup is illustrated in fig . [ fig : neq : oneshot - models ] ( upper left ) . by the network equivalence theory for point - to - point channels @xcite , successful channel emulation requires @xmath129 hence , we can construct the upper bounding model as @xmath130 similarly , for a bc with two receivers @xmath131 , the channel from the source to the auxiliary operation node @xmath112 has a super output alphabet @xmath122 , which results in a sum rate constraint @xmath132 after successfully receiving @xmath133 $ ] , the auxiliary node @xmath112 sends exactly the corresponding symbol @xmath134 to the receiver with alphabet @xmath135 , hence requiring a noiseless bit pipe of rate @xmath136 . this setup is illustrated in fig . [ fig : neq : oneshot - models ] ( upper right ) . the corresponding upper bounding model is therefore @xmath137 alternatively , one may bound the sum rate by one - shot channel emulation and then specify individual rates . for the two - user bc @xmath138 , the one - shot emulation sends every @xmath2 from the source to the auxiliary node , which requires a sum rate constraint of @xmath139 . then the auxiliary node sends @xmath2 to the two receivers through independent channels , each requires a rate constraint @xmath140 this setup is illustrated in fig . [ fig : neq : oneshot - models ] ( lower right ) and the corresponding upper bounding model is @xmath141 which is valid only if the noise at two receivers are independent , i.e. , the transition probability can be factorized as @xmath142 for the @xmath0-user mac @xmath143 , the setup for channel emulation is illustrated in fig . [ fig : neq : oneshot - models ] ( lower left ) where @xmath144 is an auxiliary random variable for the channel from @xmath15 such that @xcite @xmath145 where @xmath146 is a predefined deterministic function @xmath147 to reproduce the channel output @xmath148 at the auxiliary node @xmath112 . from stacked channel emulation we can get the individual rate constraints as follows @xmath149 therefore the corresponding one - shot model can be written as @xmath150 here we give two examples to show how to construct the auxiliary random variables as specified by . for gaussian macs , auxiliary random variables can be constructed based on a noise partitioning approach , i.e. , the additive noise at the destination is partitioned into independent parts and allocated to each of the individual channels . for a two - user gaussian mac with noise power @xmath151 and the received power constraint @xmath152 for @xmath153 , the corresponding upper bounding model is @xmath154 where @xmath155 is the noise partitioning parameter chosen to minimize the total input rate @xmath156 for binary additive mac @xmath157 with bernoulli distortion @xmath158 , the corresponding distortion parameter @xmath159 for channel @xmath101 should satisfy @xmath160 although @xmath117 and @xmath119 are tight on sum rate in the sense that there are some kind of networks where the sum rate constraint @xmath161 ( @xmath162 ) is tight , the constraints on individual rates are somewhat loose . @xmath118 and @xmath120 , on the other hand , have tighter bounds on individual rates but looser on sum rate . the gap between the upper and lower bounding models for gaussian channels and for binary symmetric channels have been examined in @xcite , where a gap less than @xmath163 bit per channel use has been established for macs with two transmitters and bcs with two receivers . for non - coupled networks , which can be decomposed into independent macs / bcs and point - to - point channels , we first construct upper and lower bounding models for macs / bcs , which can then be used to substitute their noisy counterparts in construction of noiseless bounding models for the original noisy network . to give a full description of all the rate constraints on any subset of transmitters / receivers , the upper and lower bounding models for independent macs / bcs with @xmath0 transmitters / receivers need to consist of @xmath1 rate constraints . however , such an approach is not scalable as @xmath0 can be quite large in many practical scenarios . instead , we introduce a rate vector of length up to @xmath164 to specify our upper and lower bounding models . for independent macs / bcs with @xmath0 transmitters / receivers , our upper bounding models inherent the one - shot model structure as illustrated in fig . [ fig : neq : oneshot ] and therefore only contain constraints on each of the maximum allowed individual rate @xmath165 , @xmath166 , and the total sum rate . all the constraints on subsets of individual rates , i.e. , @xmath167 are omitted , which results in a looser but simpler upper bound . the benefits of keeping the one - shot structure are two fold : they can be extended to macs / bcs with @xmath0 transmitters / receivers at low complexity ; they facilitate our proposed channel decoupling method in a natural way for constructing the upper and lower bounding bounds for coupled networks . for macs with @xmath0 transmitters , the one - shot bounding model @xmath117 defined in focuses solely on the sum rate and @xmath118 defined in focuses solely on individual rates . to facilitate a tradeoff between the sum rate constraint and each of the individual rate constraints , we propose here a new upper bounding model , @xmath168 where @xmath169 are auxiliary random variables introduced to account for the noise `` allocation '' to each individual rate constraint through @xmath170 and its conditional probability @xmath171 such that @xmath172 noise allocation to the sum rate constrain is through @xmath169 and the conditional probability @xmath173})$ ] such that @xmath174 } ) p(v_{[1:m]}|x_{[1:m ] } ) , \label{eqn : pyvx}\end{aligned}\ ] ] i.e. , @xmath148@xmath175}$]@xmath176}$ ] forms a markov chain . for the new upper bounding model @xmath177 , we introduce a parameter @xmath178 $ ] to quantify the proportion of noise distortion that has been `` allocated '' to the sum rate constraint @xmath179 . it is determined by the type of the channel / noise via its channel transition function @xmath180})$ ] and by the auxiliary random variables @xmath181 via the conditional probability function @xmath173})$ ] . when all the noise distortion is put into @xmath179 , i.e. , @xmath182 and thus @xmath173})=p(y|x_{[1:m]})$ ] , we have @xmath183 and the corresponding sum rate constraint @xmath179 is the tightest . when no noise is put into @xmath179 , i.e. , when @xmath148 is fully determined by @xmath181 and thus @xmath184 , we have @xmath185 and the sum rate constraint is relaxed to the one - shot upper bound @xmath186 . the explicit dependence of @xmath178 $ ] and the noise allocation is determined by the type of the noise . for example , for additive noise with average power constraint , we may use the proportion of noise power to evaluate @xmath187 . another option is to define @xmath187 as the ration between @xmath188 and @xmath189 , provided that the latter is finite . the parameterized bounding model @xmath177 includes the two one - shot models as special cases : putting all distortion into @xmath179 ( @xmath183 ) will generate @xmath117 , which give us a tighter bound @xmath161 on the sum rate but looser constraints on all individual rates ; taking no noise into @xmath179 ( @xmath185 ) will produce @xmath118 , which leads to looser bound on the sum rate but tighter bounds on individual rates . [ ] [ ] @xmath102[][]@xmath103[][]@xmath65 [ ] [ ] @xmath191[][]@xmath192[][]@xmath193 [ ] [ ] @xmath194[][]@xmath195[][]@xmath196[][]@xmath197 ( left ) and the model developed in ( * ? ? ? * theorem 6 ) with @xmath198 ( right ) . the label on each bit pipe is the minimum rate requirement for any given @xmath199.,title="fig:",width=325 ] for @xmath190-user macs , it is interesting to compare the new upper bounding model @xmath177 with the model developed in ( * ? ? ? * theorem 6 ) , as illustrated in fig . [ fig : neq : mac2vs ] . on one hand , setting @xmath200 and @xmath201 will put all the distortion components into the sum rate constraint ( i.e. , @xmath202 ) , which results in an upper bounding model @xmath203 it is the same as the upper bounding model in ( * ? ? ? * theorem 6 ) when @xmath204 . on the other hand , if we choose @xmath205 and a deterministic function @xmath206 such that @xmath207 and @xmath208 , we can generate another bounding model @xmath209 compared to the bound in ( * ? ? ? * theorem 6 ) , it is clear that @xmath210 has a tighter bound on rate @xmath211 . however , as we will show below , @xmath210 has a looser bound on rate @xmath212 . [ lemma : worser2 ] given @xmath208 and @xmath207 , we have @xmath213 @xmath214 where the second equality is due to @xmath207 and the fact that @xmath215@xmath216@xmath217 and @xmath218@xmath219@xmath217 are markov chains , and the first inequality comes from the fact that condition reduces entropy , with equality if and only if @xmath220 ( and thus @xmath221 ) are independent . on the other hand , we can also rewrite as @xmath222 combining and , we have . furthermore , if we are only interested in a tight bound on sum rate , by lemma [ lemma : betterrs ] below , we can see that choosing a non - trivial auxiliary random variable @xmath215 in ( * ? ? ? * theorem 6 ) can not improve the sum rate constraint . [ lemma : betterrs ] given @xmath223 , we have @xmath224 with equality if and only if @xmath225 . @xmath226 where the first two equalities are due to the chain rule , the third equality comes from the markov chain @xmath215@xmath218@xmath219 ( thus @xmath227 ) , and the last in equality comes from the fact that mutual information is non - negative , with equality if and only if @xmath225 . given noise power @xmath151 and the received power constraint @xmath152 for @xmath153 , @xmath228 , the sum rate is upper bounded by @xmath229 which can be achieved only if all the transmitters can fully cooperate . the parameterized new upper bounding model @xmath230 defined in can be constructed as follows . let @xmath231 , @xmath232 , be independent gaussian random variable with zeros mean and variance @xmath233 such that @xmath234 and @xmath235 . by choosing the auxiliary random variables @xmath236 , @xmath166 , we have @xmath237 and all the rate constraints in @xmath230 can be written as @xmath238 if we are interested in a tighter bound on any individual rate , say on @xmath239 , we can partition the noise by setting @xmath240 and @xmath241 , which leads to a tighter constraint @xmath242 on rate @xmath239 but unbounded constraints on all other individual rates and the sum rate . if we are only interested in a tighter bound on the sum rate , setting @xmath202 will give us a tight bound @xmath161 on the sum rate but unbounded constraints on all individual rates . for given @xmath243 , one way to determine the noise partitioning parameters @xmath244 , @xmath166 is to solve the following optimization problem @xmath245 this is a convex optimization problem whose solution can be explicitly found by lagrangian methods @xcite as follows ( see appendix [ app : noise ] for details ) , @xmath246 where @xmath247 satisfies @xmath248 although solving this problem in closed - form is challenging , as the specific value of @xmath247 depends both on @xmath187 and the relative magnitude of all @xmath152 , its upper and lower bounds can be determined as shown by lemma [ lemma : mac : mu ] below . since the lhs of is monotonously increasing with respect to @xmath247 , it is simple to find @xmath247 numerically by evaluating within the region defined by lemma [ lemma : mac : mu ] . [ lemma : mac : mu ] given @xmath178 $ ] and @xmath249 , @xmath250 , the @xmath247 defined by is bounded by @xmath251 where both equalities hold if and only if @xmath252 . see appendix [ proof : lemma : mac : mu ] . from lemma [ lemma : mac : rs ] below , we can see that the freedom of adjusting @xmath178 $ ] in the optimized noise partition can not improve the sum rate constraint @xmath161 . this is intuitive as @xmath161 is achievable when all the source nodes can fully cooperate . [ lemma : mac : rs ] given @xmath253 , for any @xmath178 $ ] , we have @xmath254 with equality when @xmath202 . see appendix [ proof : lemma : mac : rs ] . the upper bounding model for independent bcs with @xmath0 receivers can be generalized straightforwardly from @xcite as follows @xmath255 where @xmath256 note that @xmath119 is a valid upper bound for any channel transition function @xmath257 whereas @xmath120 is only valid for bc with independent noise components at receivers , i.e. , when the transition probability can be factorized as @xmath258 we construct a new upper bounding model by combining the point - to - point channel emulation technique developed in @xcite with the _ covering lemma _ , _ conditional typicality lemma _ , and the _ joint typicality lemma _ @xcite . let @xmath259 $ ] denote a permutation of the @xmath0 receivers and @xmath260 $ ] be their corresponding channel outputs , whose dependence is characterized by the channel transition function @xmath261 . the new bounding model is represented by @xmath262 where @xmath179 is the sum rate constraint from the transmitter to the auxiliary node @xmath112 , and @xmath263 is the rate constraint from @xmath112 to receiver @xmath264 . compared to the upper bounding model @xmath119 specified in , the new model @xmath265 maintains the tight sum rate constraint @xmath162 as specified in and meanwhile improves all the individual rate constraints . below we show step - by - step how to construct the new bounding model given in . to simplify notation , for @xmath266 , let @xmath267}$ ] represent the sequence of @xmath268 random variables @xmath269 and @xmath270}$ ] be the sequence of @xmath268 integers @xmath271 . the channel emulation is done over a stacked network with @xmath60 replicas of the original bc . * step i * : fix a channel input distribution @xmath272 . as defined in @xcite , let @xmath273 be a subset of the classical typical is defined based on the strong typicality to support the _ conditional typicality lemma _ and the _ joint typicality lemma_. ] set @xmath274@xcite such that for any @xmath275 as the input to the bc , the probability that the corresponding output sequences @xmath276 are not jointly typical with @xmath8 is smaller than a predefined threshold . furthermore , let @xmath277}}(y_1,\ldots , y_k)$ ] , @xmath266 , be marginal distributions obtained from @xmath278 , and define a series of conditional distributions as follows @xmath279 } ) \triangleq \left\ { \begin{array}{cc } 0 , & \mbox { if } p_{y_{[1:k]}}(y_{[1:k]})=0,\\ \frac{p_{y_{[1:k+1]}}(y_{[1:k{+}1]})}{p_{y_{[1:k]}}(y_{[1:k ] } ) } , & \mbox{otherwise}. \end{array } \right.\end{aligned}\ ] ] * step ii * : generate independently at random @xmath280 sequences @xmath281 , each according to @xmath282 . for any sequence @xmath283 , by the _ covering lemma _ @xcite , @xmath284 , \mbox{s.t . } ( x^n , y_1^n(w_1)){\in } t^{(n)}_{\epsilon_1}\right ) = 1,\end{aligned}\ ] ] if @xmath285 for some @xmath286 and @xmath287 that goes to zero as @xmath288 . following the channel emulation argument @xcite , we define a mapping function @xmath289 as @xmath290 if more than one sequence are jointly typical with @xmath8 , @xmath289 chooses one of them uniformly at random . * step iii * : for each sequence @xmath291 , generate independently @xmath292 sequences @xmath293 , each according to @xmath294 , where @xmath295 is defined in . given @xmath296 , according to the _ conditional typicality lemma _ @xcite , we have @xmath297 and according to the _ joint typicality lemma _ @xcite , for all @xmath298 $ ] , @xmath299 for @xmath300 and some @xmath301 that goes to zeros as @xmath302 . since sequences @xmath303 are i.i.d given @xmath304 , we have @xmath305 , ( x^n , y_1^n(w_1 ) , y_2^n(w_1,w_2))\notin t^{(n)}_{\epsilon_2}\right ) \nonumber\\ & = \left ( pr\left((x^n , y_1^n(w_1 ) , y_2^n(w_1,w_2))\notin t^{(n)}_{\epsilon_2 } \right ) \right)^{2^{n r'_2 } } \nonumber\\ & \leq \left ( 1- 2^{-n(i(x;y_2|y_1 ) + \delta_2(\epsilon_2 ) ) } \right)^{2^{n r'_2}}\label{eqn : bc : u1c}\\ & \leq exp\left(-2^{n r'_2}\cdot 2^{-n(i(x;y_2|y_1 ) + \delta_2(\epsilon_2 ) ) } \right)\nonumber\\ & = exp\left(- 2^{n(r'_2 - i(x;y_2|y_1 ) - \delta_2(\epsilon_2 ) ) } \right),\nonumber\end{aligned}\ ] ] where the first inequality comes from and the second inequality comes from the fact that @xmath306 $ ] . therefore if @xmath307 , the probability that none of the sequences in @xmath308\}$ ] is joint typical with @xmath309 tends to @xmath310 as @xmath311 . we now define a mapping function @xmath312 as follows @xmath313 if there is more than one candidate that satisfies the joint typicality condition , @xmath314 chooses one of them uniformly at random . * step iv * : for @xmath315 , we treat the set of sequences @xmath316})\}$ ] together as one unit and repeat step iii , which generates the corresponding sequences @xmath317 } , w_k ) : w_k{=}1,\ldots,2^{n r'_{k}}\}$ ] , the mapping function @xmath318})$ ] , and the rate constraint @xmath319 } ) + \delta_k(\epsilon_k),\end{aligned}\ ] ] where @xmath320 , and @xmath321 that goes to zeros as @xmath322 . * step v * : define a channel emulation codebook @xmath323 and the associated encoding function @xmath324 $ ] and the decoding function @xmath325})$ ] for receiver @xmath264 , @xmath266 . for any input @xmath8 , @xmath326 generates a sequence @xmath327 of @xmath328 bits that are transmitted from the transmitter of the bc to the auxiliary node @xmath112 , which then forwards them to receiver @xmath264 . at receiver @xmath264 , the decoding function @xmath325})$ ] selects a sequence from the codebook @xmath329 based on the received information bits , i.e. , @xmath330 note that the rate constraints in should be satisfied for @xmath266 , and for all @xmath272 . let @xmath331 and @xmath332 , letting @xmath332 implies that all of them go to zero . ] , we can specify all the rate constraints in as follows @xmath333 the second equality in comes from the fact that @xmath334 } ) { = } i(x;y_1 ) { + } i(x;y_2|y_1 ) { + } \ldots { + } i(x;y_k|y_{[1:k{-}1]}),\end{aligned}\ ] ] and the first equality in comes from our emulator design as specified in step v. there are in total @xmath335 different permutations of @xmath336 , each leading to a different upper bounding model following our construction method . for each of these upper bounding models , the sum rate constraint and one of the individual rate constraints are tight . depending on the needs , we can select a specific permutation to design the upper bounding model . for bc with @xmath337 receivers , the proposed upper bounding model has two different layouts @xmath338 and @xmath339 , where the latter turns out to be equivalent to the upper bounding model developed in ( * ? ? ? * theorem 5 ) . this is not surprising as the channel emulation codebook @xmath340 used in our construction is generated in the same way as in ( * ? ? ? * theorem 5 ) : superposition encoding . note that the proof in ( * ? ? ? * theorem 5 ) , restricted for bc with @xmath341 receivers , provides explicit error analysis . in contrast , our construction is valid for @xmath342 but only claims that the error probability can be made arbitrarily small with the help of the _ covering lemma _ , the _ conditional typicality lemma _ , and the _ joint typicality lemma_. the error analysis in our construction of new upper bounding models for @xmath0-user bcs in sec . [ sec : eqv : bc - new ] relies on the validity of the _ covering lemma _ , the _ conditional typicality lemma _ , and the _ joint typicality lemma _ , which hold for discrete - alphabet channels under strong / robust typicality notions . there are several possible approaches to extend our results in sec . [ sec : eqv : bc - new ] to continuous alphabets . one way is to apply the standard discretization procedure ( * ? ? ? 3.4.1 ) to continuous alphabets and then apply the results derived based on discrete alphabets . such process has been demonstrated in ( * ? ? ? 3.4.1 ) to prove the achievability of awgn channel capacity , and in @xcite to extend separation results from dmc to awgn channels with a per - symbol average power constraint . another way is to use generalized typicality definitions such that the _ covering lemma _ ( which essentially depends on the _ conditional typicality lemma _ ) and the _ joint typicality lemma _ can be extended to continuous alphabets . for example , joint typicality properties associated with strong typicality have been extended to countably infinite alphabets in @xcite by a notion of unified typicality . the _ generalized markov lemma _ has been extended in @xcite to gaussian sources with a modified notion of typicality and by the fact that the asymptotic equipartition property ( aep ) holds for gaussian memoryless sources . other attempts in this direction can be found , for example , in @xcite where the _ covering lemma _ and _ packing lemma _ are extended to continuous alphabets ( memoryless source ) and in @xcite where the likelihood encoder @xcite with the soft - covering lemma @xcite is applied to the lossy source compression with continuous alphabets . we may also construct our new upper bounding models in sec . [ sec : eqv : bc - new ] by iteratively applying the method developed in ( * ? ? ? * theorem 5 ) : we first only focus on two channel outputs @xmath343 , and contract a valid upper bound for this 2-user bc @xmath344 as in ( * ? ? ? * theorem 5 ) ; then we treat @xmath345 \in \my_1\times\my_2 $ ] as a single compound receiver and group it with @xmath346 to formulate a new @xmath190-user bc , whose upper bound can also be constructed as in ( * ? ? ? * theorem 5 ) ; we can repeat this process until there is no more receivers to be included . however , the associated error analysis can be very involved . our lower bounding models for macs / bcs are constructed directly based on some operating points within the achievable rate region assuming no transmitter / receiver cooperation . however , the structure of the lower bounding models are quite different for macs and bcs . the main difference between macs and bcs is the encoding process . when there is no transmitter / receiver cooperation , distributed encoding is performed in macs while centralized encoding is done in bcs . as a consequence , in mac setups , only one rate constraint is needed for each point - to - point bit pipe to fully describe any operation point within the rate region . in bcs , each of the private messages dedicated for one specific receiver may also be decoded by other receivers . such `` overheard '' messages ( the common messages ) should be reflected in the rate region , which requires the usage of point - to - points bit pipes ( hyper - arc ) in the lower bounding model . the lower bounding models can be constructed by choosing an operating point in the capacity region of the mac assuming independent sources . we can choose any point in the capacity region that can be achieved by using independent codebooks at transmitters and successive interference cancellation decoding at the receiver . for gaussian mac with @xmath0 transmitters , each with received snr @xmath152 , @xmath166 , the following sum rate is achievable @xmath347 .all possible rate constraints in the lower bounding model for broadcast channels with @xmath0 receivers . [ cols="^,^,^,^,^,^",options="header " , ] the lower bounding model in in fig . [ fig : netequv_bcs](c ) contains point - to - points bit pipes ( hyper - arcs ) of rate @xmath348 , for @xmath349 . by forcing successful decoding at macs , we will have the following constraints @xmath350 therefore we can conclude from the lower bounding model that a rate @xmath351 is achievable if for all @xmath352 @xmath353 after applying the mac constraint , we have @xmath354 therefore the lower bounding models can provide a rate equals the capacity @xmath355 when the the snr @xmath356 is small or when the number of nodes is large . we now illustrate the construction of bounding models for a multiple multicast relay network shown in fig . [ fig : s1s2dn ] , where two source nodes @xmath357 and @xmath358 multicast information @xmath359 at rate @xmath211 and @xmath360 at rate @xmath212 , respectively , to all destinations @xmath361 over gaussian channels . the effective snrs of these gaussian channels are @xmath362 where @xmath363 are parameters such that @xmath364 the transmission of @xmath357 is aided by @xmath358 via a @xmath365-ary symmetric channel @xmath366 such that for all @xmath367 @xmath368 we can first decompose the original network into a point - to - point channel , two decoupled bcs originating from @xmath357 and @xmath358 , and @xmath45 decoupled macs ending at each destination node , and then construct upper and lower bounding models as described in sec [ sec : eqv : coup ] . an illustration of the resulting lower bounding model for @xmath369 is presented in fig . [ fig : s1s2dn : low ] , where the rate constraints of poin - to - point(s ) bit pipes are determined following the process in sec . [ sec : eqv : lower ] . from fig . [ fig : s1s2dn : low ] we can see that multicast of @xmath360 can only be achieved via the hyper - arc of rate @xmath370 , while multicast of @xmath359 can be achieved either via the hyper - arc of rate @xmath371 , or via the collaboration with @xmath358 at rates @xmath372 and/or @xmath373 . the collaboration from @xmath358 is subject to the rate constraint @xmath374 which is the capacity of the @xmath365-array symmetric channel from @xmath357 to @xmath358 . [ ] [ ] @xmath357 [ ] [ ] @xmath358 [ ] [ ] @xmath375[][]@xmath376 [ ] [ ] @xmath377 [ ] [ ] @xmath218 [ ] [ ] @xmath219 [ ] [ ] @xmath378[][]@xmath379 [ ] [ ] @xmath380 [ ] [ ] @xmath381 [ ] [ ] @xmath382 [ ] [ ] @xmath359 [ ] [ ] @xmath360 [ ] [ ] @xmath383 [ ] [ ] @xmath384[][]@xmath385 [ ] [ ] @xmath386 [ ] [ ] @xmath387[][]@xmath388 [ ] [ ] @xmath389 [ ] [ ] @xmath390[][]@xmath391 and @xmath358 multicast @xmath359 and @xmath360 respectively to all destinations @xmath361 through gaussian channels , where @xmath392 denotes the snr of the link from @xmath393 to @xmath394 . the link @xmath366 from @xmath357 to @xmath358 is a @xmath365-ary symmetric channel orthogonal to all the other gaussian channels.,title="fig:",width=325 ] [ ] [ ] @xmath357 [ ] [ ] @xmath358 [ ] [ ] @xmath375[][]@xmath376 [ ] [ ] @xmath395 [ ] [ ] @xmath359 [ ] [ ] @xmath360 [ ] [ ] @xmath396 [ ] [ ] @xmath397 [ ] [ ] @xmath398 [ ] [ ] @xmath399 [ ] [ ] @xmath400[][]@xmath401 [ ] [ ] @xmath374 [ ] [ ] @xmath402[][]@xmath403 [ ] [ ] @xmath404[][]@xmath405 [ ] [ ] @xmath406[][]@xmath407 destinations where @xmath374 is the channel capacity of the @xmath365-ary symmetric channel @xmath366 . @xmath408 and @xmath371 are the rates of hyper - arcs from @xmath357 to @xmath409 and @xmath410 , respectively , and @xmath411 denotes the multicast rate from @xmath358 to @xmath412 . rate constraints @xmath413 , @xmath414 and @xmath415 , come from the lower bounding models of macs.,title="fig:",width=325 ] bounds on the sum rate for the scenario with @xmath416 destinations , @xmath417db , and the @xmath365-ary symmetric channel @xmath366 with @xmath418 and @xmath419.,width=325 ] the bounds on sum rate obtained from upper and lower bounding networks with respect to varying channel quality have been illustrated in fig . [ eqv : fig : pn10 ] for a scenario with @xmath420 destinations . the @xmath365-ary symmetric channel @xmath366 has parameters @xmath418 and @xmath419 , which results in a capacity of @xmath421 bits per channel use [ bpcu ] . we also plot three benchmarks as references : the rate achieved by transmitting identical signals from @xmath357 and @xmath358 ( denoted by `` s1-s2 coop '' ) , the rate achieved by transmitting independent signals from @xmath357 and @xmath358 ( denoted by `` s1-s2 mac '' ) , and the cut - set upper bound following the method developed in @xcite . our upper bound obtained from noiseless bounding networks is very good and it even approaches the capacity ( meeting the lower bound provided by @xmath357-@xmath358 cooperation ) in low to medium snr regions . our lower bounding models discard the possibility of source cooperation and therefore suffers some performance degradation ( less than 0.4 bits and @xmath45 increase , the gap from cut - set bound converges to a constant that depends only on @xmath422 . ] from the capacity ) . in high snr region , it outperforms the two benchmarks since our lower bounding models can make use of the overhead messages to increase the multicast rate of @xmath359 : extra bits of @xmath359 can be transmitted via the collaboration with @xmath358 at rate @xmath423 if such operation is permitted by the link from @xmath357 to @xmath358 , i.e. , when @xmath424 . in this work we have presented capacity upper and lower bounding models for wireless networks , where the upper bounding models consist of only point - to - point bit pipes while the lower bounding models also contains point - to - points bit pipes ( hyper - arcs ) . we have extended the bounding models for two - user macs / bcs to many - user scenarios and established a constant additive gap between upper and lower bounding models . for networks with coupled links , we have proposed a channel decoupling method to decompose coupled networks into decoupled macs and bcs , and proposed strategies to construct step - by - step upper and lower bounding models for the originally coupled networks . we have demonstrated by examples that the gap between the resulting upper and lower bounds is usually not large , and the upper / lower bound can approach capacity in some setups . the proposed methods for constructing upper and lower bounding models , effective and computationally efficient , can be easily extended to large networks with complexity grows linearly with the number of nodes . they therefore , combined with methods calculating the capacity of noiseless networks , provide additional powerful tools for characterizing the capacity region of general wireless networks . following the lagrangian method @xcite , the optimal noise partitioning for the optimization problem can be obtained by taking partial derivatives of its lagrangian @xmath425 with respect to @xmath244 , @xmath166 , and setting them to zero . denoting @xmath426 the optimal noise power for @xmath244 , we have @xmath427 which leads to ( omitting the negative root as @xmath428 ) @xmath429 the exact value of @xmath247 is determined by the condition @xmath430 , which yields . the upper bound is obtained by contradiction . assuming @xmath431 , the lhs of is evaluated as follows @xmath432 which contradicts to the equality constraint stated in . therefore we have @xmath433 where the equality holds if and only if @xmath434 for all @xmath435 , i.e. , when @xmath252 . the lower bound is obtained as follows . from we have @xmath436 where the inequality in is due to cauchy - schwarz inequality with equality holds if and only if @xmath252 . by expanding the lhs and removing common items at both sides , we can easily obtain the lower bound . from it is straightforward to observe that @xmath437 is a monotonously decreasing function with respect to @xmath187 , with its minimum @xmath438 therefore we only need to prove @xmath439 . from and it is easy to see that @xmath440 is a monotonously increasing function with respect to @xmath243 , which leads to @xmath441 on the other hand , we have @xmath442 where holds by the optimal solution @xmath443 . hence we have proved the lemma . let @xmath444 be the lagrangian , by taking partial derivative of @xmath445 w.r.t . @xmath446 and setting them to zero , we get @xmath447 by introducing auxiliary variables @xmath448 we can derive from the following equations @xmath449 where comes from and the fact that @xmath450 , is obtained by substituting into , and is the unique feasible solution to . therefore the equivalent snrs @xmath451 for decoupled bcs are uniquely determined by where the optimal value of @xmath452 and @xmath453 can be easily obtained by iterating and . the convergence to the global optimum is guaranteed by observing the fact that @xmath453 is a monotonically decreasing function of @xmath454 via and @xmath452 is a monotonically increasing function of @xmath455 , via . c. bennett , p. shor , j. smolin , and a. thapliyal , `` entanglement - assisted capacity of a quantum channel and the reverse shannon theorem , '' _ ieee transactions on information theory _ , vol . 48 , pp . 26372655 , oct . 2002 . c. tian , j. chen , s. diggavi , and s. shamai ( shitz ) , `` optimality and approximate optimality of source - channel separation in networks , '' _ ieee transactions on information theory _ , vol . 60 , pp . 904918 , 2014 . x. yan , j. yang , and z. zhang , `` an outer bound for multisource multisink network coding with minimum cost consideration , '' _ ieee transactions on information theory _ 52 , pp . 23732385 , jun . 2006 . j. du , m. xiao , and m. skoglund , `` capacity bounds for backhaul - supported wireless multicast relay networks with cross - links , '' in _ proceedings ieee international conference on communications ( icc ) _ , jun . 2011 . j. du , m. xiao , m. skoglund , and m. mdard , `` wireless multicast relay networks with limited - rate source - conferencing , '' _ ieee journal on selected areas in communications _ 31 , pp . 13901401 , aug .

the cluster distributions of different systems are examined to search for signatures of a continuous phase transition . in a system known to possess such a phase transition , both sensitive and insensitive signatures are present ; while in systems known not to possess such a phase transition , only insensitive signatures are present . it is shown that nuclear multifragmentation results in cluster distributions belonging to the former category , suggesting that the fragments are the result of a continuous phase transition . + pacs number(s ) : 25.70 pq , 64.60.ak , 24.60.ky , 05.70.jk

nucl-ex0002006

beginning in the 1970 s significant advances in the understanding of nuclear multifragmentation were made possible with the advent of high statistics inclusive experiments . typically , only one intermediate mass fragment ( @xmath0 ) was detected per event . from these inclusive studies came the first evidence that intermediate mass fragments ( imfs ) were associated with a simultaneous multi - body breakup of a system which had undergone expansion . a study of the fragment mass yield distribution obtained in an inclusive gas jet experiment conducted at fermilab contained the first indication that nuclear multifragmentation might be related to critical phenomena normally observed in macroscopic systems @xcite . the purdue group was the first to make the suggestion that the observed power law in the fragment yield distribution might result from a system whose excitation energy was comparable to its total binding energy @xcite . the exponent of the power law was @xmath1 , within the range expected for a system near its critical point . the presence of the power law and the value of the exponent , coupled with the strong similarity of the nuclear and van der waals potentials , led the purdue group to suggest that multifragmentation of nuclei might be analogous to a fluid undergoing a continuous phase transition from a liquid to a gas . furthermore , the fisher droplet model ( fmd ) @xcite-@xcite , used to describe condensation in a fluid system near its critical point , after modification for nuclear physics effects , was capable of describing the isotopic yields of 50 fragments with one set of parameters @xcite , @xcite . the temperature of the system was determined to be about 5 mev @xcite , a reasonable value considering that the average binding energy per nucleon in a nucleus is approximately 8 mev . the success of this approach reinforced the notion that multifragmentation was both a thermal process and that it was related to critical phenomena . with the advent of exclusive experiments capable of detecting all of the charged reaction products , the possibility of studying multifragmentation on an event - by - event basis became a reality . high statistics exclusive experiments in which the fragmenting system is characterized according to its nucleon number and excitation energy permit both the correlation of dynamical and statistical information and the study of fluctuations in experimental observables . fluctuations are central to all critical phenomena , and indeed , such fluctuations are apparent in exclusive multifragmentation data . in this paper , the focus will be on the statistical signals of multifragmentation data observed in the eos experiment @xcite-@xcite . comparisons will be made with two other systems , one of which exhibits critical behavior and one of which does not . much of the pioneering work in understanding the statistical aspects of multifragmentation has been performed by campi @xcite-@xcite and mekjian @xcite-@xcite . both efforts have compared multifragmentation data to model systems in order to gain some insight into the nuclear breakup process . in this paper , many of the ideas suggested by these authors are followed and applied to both the eos data and the model systems in order to demonstrate which of the many suggested signals are useful for the identification of critical behavior . a major goal of this paper is to present a comprehensive review of several methods proposed for detecting signals of critical phenomena in multifragmentation . it is tempting to compare the experimental data to dynamical models that attempt to describe nuclear multifragmentation . however , the task of modeling multifragmentation from the initial collision phase of the reaction to freeze - out has proven to be a daunting task . models that adequately describe the initial stage of the reaction @xcite-@xcite do not satisfactorily describe the fragment formation stage , in either statistical or dynamical aspects . likewise , the most successful models in describing the statistical properties of nuclear multifragmentation @xcite-@xcite , assume thermodynamic equilibrium , yet fail to adequately match the dynamical features of the data . molecular dynamical approaches , which have enjoyed considerable success in describing critical behavior in classical systems @xcite-@xcite , have not been conclusive in describing nuclear multifragmentation and at times have yielded contradictory results @xcite , @xcite . later studies suggested flaws in the application of molecular dynamical models to nuclear multifragmentation , therefore calling into question the conclusions drawn from the earlier studies @xcite . the most striking of the early theoretical efforts came from campi s analysis of a few hundred completely reconstructed emulsion multifragmentation events @xcite and the comparison of these data to clusters generated from a percolation calculation @xcite , @xcite . in this series of papers it was shown that the fragment distributions from multifragmentation bore a striking similarity to the cluster distributions from percolation lattices . this analysis provided strong evidence that multifragmentation was a statistical process which appeared to be related to critical phenomena . in this analysis another estimate of the exponent @xmath2 was made which agreed with the first measurements from the purdue group and several later analyses of various fragment distributions . in the early 1990 s the aladin group from gsi performed several multifragmentation experiments @xcite-@xcite . of particular importance was the `` rise and fall '' of multifragmentation . in one analysis the aladin group plotted the `` rise and fall '' curve of the production of imfs versus an observable related to the excitation energy of the reaction for several multifragmenting systems . with the appropriate scaling the data collapsed to a single curve suggesting that the multifragmenting systems retained no memory of the reaction entrance channel . this is expected for an equilibrated system . the results of some statistical analyses of multifragmentation data could be interpreted to suggest that multifragmentation is a sequential decay @xcite in contrast to the phase transition picture . the same sort of statistical analysis has also been applied to explicitly simultaneous models @xcite and produced results that were similar to those of multifragmentation data . thus those signals could be interpreted as evidence for either sequential or simultaneous multifragmentation @xcite . this last effort puts into focus the main question in this work : what type of analysis of the statistical aspects of a cluster distribution can provide the most insight into the nature of the mechanism which created the clusters ? specifically , can those systems which contain critical behavior be distinguished from those which do not ? it will be argued that this question has two answers . analysis of the _ insensitive _ features of the cluster distribution can not make the above mentioned distinction @xcite . however , an analysis of the _ sensitive _ features of the cluster distribution will be shown to provide deeper insight into the cluster production mechanism . this type of analysis has been previously reported for clusters resulting from nuclear multifragmentation @xcite-@xcite . note that the more generic term _ cluster _ will be used to refer to any composite of constituents , whether these be molecules of a fluid , nuclear fragments or percolation clusters . the method employed to address the above question is as follows . the same analysis is performed on the cluster distributions produced by three different systems . in one case , clusters are generated by randomly partitioning an integer . such one - dimensional partitioning does not posses critical behavior indicative of a continuous phase transition . in the second case , three - dimensional bond building percolation is used to produce clusters . percolation is well - known mathematical construct that possesses a continuous phase transition , _ i.e. _ a critical point . finally , the cluster distributions resulting from the multifragmentation of gold nuclei are analyzed . although it is not known , a priori , whether the nuclear multifragmentation bears any relation to critical phenomena , it will be seen that the analysis presented in this work yields suggestive results . this paper is organized as follows . in section ii a description of each system is presented . in section iii the fisher droplet model is reviewed . in section iv - a the insensitive signatures of the cluster distributions for all systems are examined . in section iv - b the sensitive signatures are examined . sections v and vi present possible corrections to the analysis of the multifragmentation data . finally , section vii discusses the conclusions reached upon the completion of the analyses in sections iv and v. throughout this paper the term _ continuous phase transition _ will be used instead of _ second order phase transition _ , the latter from the outdated ehrenfest theory of phase transitions @xcite . approximately 40,000 fully reconstructed events ( @xmath4 ) were collected with the eos experimental apparatus discussed in ref . @xcite . in the collision of the projectile gold nucleus ( @xmath5 , @xmath6 ) and the target carbon nucleus , so - called prompt nucleons are knocked out of the gold nucleus by quasi - elastic and inelastic collisions between projectile and target nucleons @xcite . immdeiately following the collision , he gold projectile remnant is in an excited state with fewer than 197 nucleons . the excited remnant cools and expands , evolving to the neighborhood of the critical point in the temperature - density plane @xcite , where clusters condense from a high temperature low density vapor of nucleons . the charge and mass of the projectile remnant , @xmath7 and @xmath8 , were determined for each event by subtracting the charge and mass of the prompt particles from the charge and mass of the gold nucleus @xcite . prompt particles have @xmath9 , @xmath10 and @xmath11 and are removed from the cluster distributions analyzed in this work . only clusters created from the excited gold projectile are considered in the ensuing analysis . for events with the lowest charged particle multiplicities , @xmath12 , the remnant had @xmath13 , @xmath14 and @xmath15 mev@xmath16nucleon , while for events with the highest multiplicities the remnant had @xmath17 , @xmath18 and @xmath19 mev@xmath16nucleon @xcite . clusters of a given charge , @xmath20 , were counted on an event by event basis to determine the cluster charge distribution , @xmath21 . in this analysis , although the mass number of the clusters is of interest , a cluster s charge will be used as an index . mass numbers for clusters of charge one and two were measured in the eos time projection chamber . clusters with @xmath22 were assigned a mass number , @xmath23 , by multiplying the cluster charge by the mass to charge ratio of the excited gold projectile remnant ; for low @xmath12 events @xmath24 and for high @xmath12 events @xmath25 . this procedure provided an estimate of a cluster s mass number prior to any secondary decay effects . it is assumed that on average @xmath26 . finally , it is the normalized cluster distribution , @xmath27 , that is used in the analysis presented in this paper . bond building percolation calculations were performed on three dimensional simple cubic lattices of 216 sites . cluster distributions for @xmath28 lattice realizations were generated in the standard fashion by forming bonds between sites . bonds were either active ( on ) or inactive ( off ) according to the following algorithm . the control parameter ( _ e.g. _ temperature in thermodynamic systems ) for percolation is the lattice probability , @xmath29 . a single value of @xmath29 was chosen for the entire lattice . all probabilities were between 0 and 1 . next , a bond probability , @xmath30 , was randomly chosen from a uniform distribution on ( 0,1 ) for the @xmath31 bond . if @xmath30 was less than @xmath29 , then the @xmath31 bond was active and two sites were joined into a cluster . this process was performed for each bond in the lattice . at low values of @xmath29 , few bonds were formed resulting in a high multiplicity , @xmath12 , of small clusters , a distribution analogous to the gaseous phase of a fluid . at high values of @xmath29 , many bonds were formed resulting in a low multiplicity of mostly large clusters , analogous to the liquid phase of a fluid . in an infinite lattice the phase transition occurs at a unique value of the lattice probability , @xmath32 , when the probability of forming a percolating cluster changes from zero to unity . to examine the behavior of the average cluster distribution , the number of clusters of size @xmath23 per lattice site was calculated by histogramming the @xmath28 lattice realizations into 100 bins from @xmath33 to @xmath10 . the use of @xmath12 as a control parameter and the ensuing effects on signatures of continuous phase transition were investigated by calculating the average number of clusters of size @xmath23 with the @xmath28 lattice realizations histogrammed in units of @xmath12 . random partitions were generated from 79 total system constituents , chosen to approximate the number of charges in the gold multifragmentation system . the algorithm is as follows . first a random choice of @xmath12 was made from a uniform distribution on ( 1,79 ) . next the maximum size of a cluster , @xmath34 , for an event with @xmath12 was determined ; this depended on the constraints of the system size , @xmath35 and the choice of @xmath12 . the size of the first cluster , @xmath36 , was then randomly chosen from a uniform distribution on @xmath37 . there were then @xmath38 clusters to be generated from @xmath39 constituents . the maximum size of a cluster for an @xmath38 event from a @xmath40 constituent system was determined : @xmath41 . the size of the second cluster , @xmath42 , was then randomly chosen form a uniform distribution on @xmath43 . this process was repeated until all constituents belonged to a cluster . @xmath28 partitions were generated in this manner . this particular weighting results in a power law cluster distribution @xcite . the focus of most studies of phase transitions is on standard thermodynamical variables such as a system s temperature , density , compressibility , etc . these quantities are difficult or impossible to measure directly in present nuclear multifragmentation experiments . thus a theory which addresses quantities accessible to mf experiments is needed . to that end fisher s gas - to - liquid phase transition model , based on mayer s condensation theory , is followed @xcite , @xcite , @xcite . fisher begins his model , called the fisher droplet model ( fdm ) hereafter , by writing the free energy for the formation of clusters of size @xmath23 as : @xmath44 where @xmath45 is the boltzmann constant and the @xmath46-term is the bulk formation energy , or volume term and : @xmath47 , \label{volume_term}\ ] ] where @xmath48 is the chemical potential and @xmath49 is the chemical potential along the coexistence curve . the @xmath50-term is related to the surface free energy of cluster formation . it s a form given by fisher is : @xmath51 , \label{surface_term_fisher}\ ] ] where @xmath52 is a critical exponent and is related to the ratio of the dimensionality of the surface to the dimensionality of the volume , @xmath53 is a constant of proportionality relating the average surface area of a droplet to its number of constituents and @xmath54 is the surface entropy density ; @xmath55 is a measure of the distance from the critical point . for usual thermodynamic systems @xmath56 , in the percolation treatment @xmath57 and for multifragmentation @xmath58 will be used . all formulations of @xmath55 are such that @xmath59 ( @xmath60 ) corresponds to the _ liquid _ ( _ gas _ ) region . this form of the surface free energy is applicable on only one side of the critical point , the single phase side . a more general form suggested by efforts from percolation theory @xcite-@xcite that can be applied on both sides of the critical point and leads to a power law which describes the behavior of the order parameter is : @xmath61 , \label{surface_term_leath}\ ] ] where the scaling variable , @xmath62 , is @xmath63 the physical interpretation of the parameters @xmath64 , @xmath65 and @xmath66 is an open question . finally @xmath67 is another critical exponent depending principally on the dimensionality of the system and has its origins in considerations of a three dimensional random walk of a surface closing in on itself . for three dimensions @xmath68 @xcite . in eq . ( [ free_energy ] ) , @xmath69 is a normalization constant which will be shown to depend solely on the value of @xmath67 @xcite . from the free energy of cluster formation the average cluster distribution normalized to the size of the system is : @xmath70 at the critical point , @xmath71 , both @xmath50 and @xmath46 are unity and the cluster distribution is given by a pure power law : @xmath72 if the first moment of the normalized cluster distribution is considered at the critical point then @xcite : @xmath73 when the sum runs over all clusters . from eq . ( [ normalization ] ) it is obvious that the value of the overall cluster distribution normalization constant , @xmath69 , is dependent on @xmath67 via a riemann @xmath74-function : @xmath75 the above is true only if the scaling assumptions in the fdm apply to all clusters . for finite size systems even at the critical point this is only approximately true . however , it will be seen that eq . ( [ zeta_function ] ) holds reasonably well at the critical point for systems with a continuous phase transition over some range in cluster size . in the fdm it is assumed that all clusters of size @xmath23 can be treated as an ideal gas , so that the total pressure of the entire cluster distribution can be determined by summing all of the partial pressures : @xmath76 is is clear from eq . ( [ pressure_a ] ) that the pressure of the system is related to the zeroth moment of the cluster distribution . the density is then : @xmath77 the density is given by the first moment of the cluster distribution . it is now a simple matter to derive the power law which describes the divergence of the isothermal compressibility , @xmath78 . by definition : @xmath79 noting that @xmath80 , eq . ( [ compressibility_a ] ) can be rewritten as : @xmath81 which leads to : @xmath82 the sum in the second term illustrates the relation of the second moment of the cluster distribution , @xmath83 , to the isothermal compressibility . the sums in eq . ( [ pressure_a ] ) , ( [ density ] ) and ( [ compressibility_c ] ) run over all clusters in the gas and exclude the bulk liquid drop . in percolation and multifragmentation the largest cluster on the liquid side of the critical point will be considered as the liquid drop and will thus be excluded from the sum . on the gas side of the critical point , the sum runs over all clusters as there is no longer a liquid drop . in the thermodynamic limit , large @xmath23 dominate the sum so that it may be treated as an integral giving : @xmath84 working along the liquid - gas coexistence curve so that @xmath85 eq . ( [ compressibility_d ] ) reduces to : @xmath86 a change of variables from @xmath23 to @xmath62 shows that near the critical point : @xmath87 this is the so - called @xmath88-power law which describes the divergence of the isothermal compressibility and the second moment of the cluster distribution near the critical point . the scaling relation between the exponents @xmath88 , @xmath52 and @xmath67 is : @xmath89 the absolute normalization constants of the @xmath83 power law depend on the scaling function , @xmath90 , the exponent @xmath52 and the overall normalization of the cluster distribution , @xmath69 , which in turn depends on the exponent @xmath67 : @xmath91 the second moment is related to the isothermal compressibility by the temperature and density of the system . the derivation of the @xmath88-power law demonstrates one way to arrive at the scaling relations between the critical exponents . in addition it illustrates the existence of only two independent exponents and shows the relation of the moments of the cluster distribution to familiar thermodynamic quantities . fisher s framework here illustrated and tempered by percolation theory will be used in the analysis of the cluster distributions of the three systems discussed above . it will be seen that in the case of systems which exhibit a continuous phase transition , the framework of fisher is well followed , while for systems with no such phase transition , the framework fails , as it should . in this section the insensitive features of the cluster distribution for each system are examined . it will be demonstrated that on this level of analysis each system exhibits behaviors that are consistent with systems which undergo a continuous phase transition . the conclusion is inescapable that this sort of analysis can yield necessary , but not sufficient , signals and no further insight to the mechanism behind multifragmentation . a deeper analysis will be necessary to distinguish those systems which undergo such a phase transition from those which do not . one of the most striking characteristics of systems undergoing continuous phase transitions is the occurrence of fluctuations that exist on all length scales in a small range of the control parameter . in fluid systems this was observed as critical opalescence , first noted by andrews in the latter half of the @xmath92 century @xcite . fluctuations in cluster size and the density of the system arise because of the disappearance of the latent heat at the critical point . this is illustrated in the fdm when the isothermal compressibility diverges at the critical point and small changes in pressure gives rise to great changes in the density . in the fdm as the volume and surface contribution to the free energy of cluster formation vanishes the power law dominates and clusters of all length scales are observed @xcite . in a cluster distribution the most readily observed fluctuations are those in the size of the largest cluster . for each system the root mean square ( rms ) fluctuations in the size of the largest cluster normalized to the size of the system , @xmath93 , have been calculated as a function of the system s control parameter . this measure of the fluctuations in the cluster distribution was first studied by campi for gold multifragmentation and percolation @xcite . those results are replicated here for those two systems . figure 1a shows @xmath94 , as a function of @xmath29 for percolation . as expected for a system known to exhibit a continuous phase transition , the rms fluctuations peak over a narrow range in the control parameter . the location of this peak provides a first estimate of the critical point ; @xmath95 . see table i. [ h ] [ cols="<,<,<,<,<",options="header " , ] the scaling function for percolation ( @xmath29 and @xmath12 ) determined here is _ the _ scaling function for percolation in three dimensions , _ i.e. _ it is universal for three dimensional percolation independent of size . the scaling functions for @xmath29 and @xmath12 determined above agree well with the scaled cluster distributions of different size lattices , see figure 23 , and can be used to predict the behavior of the second moment for any size lattice @xcite . in the same spirit , the scaling function determined here for gold multifragmentation is the scaling function for charged nuclear matter which describes the cluster distributions produced in the multifragmentation of any nucleus , not just the excited gold remnant discussed in this work . with the knowledge of the form of the scaling function various other quantities can be determined as illustrated in section iii and shown below . [ h ] the cluster distributions for the random partitions is fit , by eye , with the same empirical parameterization as in eq . ( [ scaling_fit ] ) see figure 19c . the random partitions can not be described by eq . ( [ scaling_fit ] ) . the solid curve in figure 19c will be used in the following section to demonstrate the failure of the scaling analysis , as is also seen here , when applied to a system where a continuous phase transition is absent . finally , a consistency check in this analysis is the agreement between the location of the peak in the scaling functions and the values of @xmath96 determined in the @xmath52 analysis , see table iii . the behavior of @xmath97 or @xmath98 can be _ derived _ from the functional form of the scaling function and the critical parameters via eq . ( [ compressibility_f ] ) . performing the integration in eq . ( [ compressibility_f ] ) using the functional form of the scaling function determined above yields a direct calculation of the critical amplitudes , @xmath99 via eq . ( [ compressibility_g ] ) . the critical exponent @xmath100 is calculated from the values of @xmath2 and @xmath101 via a scaling relation in eq . ( [ scaling_1 ] ) . combining these two , @xmath99 and @xmath100 , it is possible to calculate the @xmath100-power law that describes the behavior of the second moment . this calculated @xmath100-power law can then be compared to the behavior of @xmath98 as measured from the cluster distribution . figure 24a , b and d shows the agreement between the measured @xmath98 data ( largest cluster omitted in the liquid region ) and the calculated @xmath88-power law curves for percolation ( @xmath29 and @xmath12 ) and gold multifragmentation , respectively and tables ii and iii list the results . the values of @xmath100 determined via the scaling relation in eq . ( [ scaling_1 ] ) for percolation ( @xmath29 and @xmath12 ) show approximate agreement with the accepted value of 1.8 . the high value of @xmath52 extracted above leads to a low value of @xmath88 here . figures 24a and b also show the behavior of the second moment of a @xmath102 site lattice . the power law predicted using the scaling function determined with a 216 site lattice shows rough agreement with the measured @xmath98 of the larger lattice in both the amplitude ( @xmath103 ) and exponent ( @xmath88 ) . there is approximate agreement between the predicted power law and the measured @xmath98 of the smaller lattice over some region in @xmath55 that is neither too near to , nor too far from the critical point , @xmath71 . it is this region that will be determined , independently , in the following section . for the percolation ( @xmath29 and @xmath12 ) system , the disagreement between the measured @xmath98 data and the calculated curves is due to two well known reasons : far from the critical point , the assumptions of scaling are no longer valid and the analytic background overwhelms the singular behavior . near the critical point finite size effects dominate @xmath98 , limiting the sizes of the large clusters which make the most significant contribution . in contrast , the @xmath2-power law was observed at the critical point because it is determined by smaller clusters which suffer the least from the finite size effects . [ h ] figure 24c shows the results when this analysis was applied to random partitions . the power law predicted from the scaling function analysis applied to the cluster distribution of the random partitions fails to reflect the behavior of the measured second moment . this is not surprising as the random partitions presented here are not the result of a system undergoing a continuous phase transition . the disagreement observed in figure 24c then serves as an indication of how this particular analysis probes for the presence of a continuous phase transition . this figure shows the results of this analysis for a system with no phase transition , while figures 24a and b show the results of this analysis on a system where such a phase transition is present . the results of this analysis when applied to nuclear multifragmentation are shown in figure 24d . in this case , the comparison to the predicted @xmath88-power law is neither as good as that for percolation nor as poor as that for the random partitions . it shall be shown in section vi that considerable improvement can be achieved if account is taken of the changing system size , @xmath104 , and finite size scaling effects . the approximate agreement between the predicted @xmath88-power law and the measured @xmath98 behavior is in keeping with the behavior expected for small systems undergoing a continuous phase transition , _ e.g. _ the percolation system . the multifragmentation results are clearly different that then results of a system without a continuous phase transition , _ e.g. _ random partitions . in the previous works the procedure for determining critical exponent values and the location of the critical point from the cluster distribution was based on a method of matching exponent values on both sides of the critical point @xcite , @xcite . the idea was to find the region on either side of the critical point where the power law behavior predicted by the scaling function holds . as is seen in figure 24 there is some intermediate @xmath55 region where the second moment data are described by a power law , a region where the @xmath98 behavior is dominated by the @xmath88-power law and all other effects are small in comparison . in earlier percolation studies @xcite general guidelines based on the correlation length and size of the fluctuations were used to find the boundaries in @xmath55 of the regions to be fit . in nuclear multifragmentation analyses @xcite it was impossible to use such guidelines . instead a method was developed that searched for regions best fit by power laws and determined the location of the critical point and exponent values simultaneously . the values of the critical exponents and the normalizations associated with power laws were obtained from the best fit power laws in those regions . as with the previous analyses presented in this paper , this method of exponent matching does not select a particular value of a critical exponent or the critical point . instead the values found are the outcome of an unbiased procedure . the method is as follows . a choice of the critical point , @xmath32 or @xmath105 was made . from this choice plots such as those shown in figure 24 were made . then fitting boundaries in @xmath55 were chosen . the fitting range was defined by @xmath106 and @xmath107 . for example , on the gas side of the critical point a fit of @xmath108 versus @xmath109 was made for all data with @xmath110 . the slope of the resulting linear fit was recorded as @xmath111 , the offset as @xmath112 and the goodness of fit as @xmath113 . the same procedure was applied to the liquid side of the chosen critical point , recording @xmath114 , @xmath115 and @xmath116 . for each choice of the critical point , several choices of fitting regions , @xmath106 and @xmath107 were made and results recorded . five parameters were chosen for each set of power law regions examined : @xmath106 , @xmath107 and @xmath32 or @xmath105 . the @xmath88-power law fit regions and critical point locations were evaluated by demanding that : ( 1 ) they yield @xmath117 and @xmath118 values that matched each other to within the error bars on those values returned by the fitting routine and ( 2 ) that the @xmath119 of the fits were in the lowest quarter of the distribution resulting from all the fits which satisfy condition ( 1 ) . the results from the power law fit regions that passed these two criteria were then histogrammed and average values for all quantities concerned were determined . the results are summarized in tables i , ii and iii and shown in figure 25 . [ h ] the lines plotted in figure 25 do not result from any single fit , but display the average results for @xmath120 and @xmath103 that have satisfied conditions ( 1 ) and ( 2 ) . the points in figure 25 are the measured second moment for the particular cluster distribution in questions plotted against @xmath55 , which depends on the average value of @xmath32 or @xmath105 that satisfies conditions ( 1 ) and ( 2 ) . therefore the lines in figure 25 should not be interpreted as a fit to the data points shown in the same figure , but as the average results from the @xmath88-matching procedure . full circles in figure 25 show the average fitting regions that satisfy conditions ( 1 ) and ( 2 ) . for percolation @xmath29 the value of @xmath88 determined in this manner is within a few percent of the value determined in @xcite and the infinite lattice value . the ratio of @xmath121 determined by this method , a ratio that depends on the universality class of the system in question , is also in agreement with the infinite lattice value and the @xmath103 values predicted by the scaling function , see table iii . the value of @xmath32 determined here is within 15@xmath122 of the value determined in a previous analysis of the @xmath123 lattice @xcite and the value determined above in the fisher @xmath67-power law analysis , see table ii and figure 25a . the results for the analysis of percolation with @xmath12 as a measure of the control parameter are worse that the results when the natural control parameter @xmath29 is used , the difference in @xmath111 and @xmath114 was : @xmath124 compared to @xmath125 . this is to be expected because for each value of @xmath29 there is some spread in the resulting values of @xmath12 , so that binning in @xmath12 groups together _ events _ with different values of @xmath29 . there is also a non - linear relation between the average values of @xmath29 and @xmath12 @xcite . in spite of these two effects the results of the analysis in section iv - b-4 suggests that vestiges of the signature of a phase transition are still present even when @xmath12 is used as the control parameter . that is also the case in the present analysis . table ii shows that the @xmath88 value agrees , within error bars , with the infinite lattice value . the values of the critical amplitudes , @xmath103 , do not yield a ratio that agrees with the infinite lattice value . this is due to the non - linear mapping of @xmath29 onto @xmath12 and is discussed in @xcite . when the @xmath88-matching procedure was applied to the cluster distributions from random partitions a very limited amount of trial fits passed the combined tests of ( 1 ) and ( 2 ) . the results compared poorly to the percolation results . at best the values of @xmath117 and @xmath118 match to within 20@xmath122 of the average value of @xmath88 , compared to perfect matching for percolation @xmath126 and matching within 5@xmath122 for percolation @xmath12 . the value of the critical point , @xmath105 , returned from this analysis also compared poorly to other outcome of previous analyses , see table i. finally , while fit regions for all systems were limited , the fit regions are the smallest and the poorest of quality for the random partitions . the results of the @xmath88-matching analysis applied to multifragmentation data has been published in ref . @xcite . in that work the data were contaminated by the inclusion of prompt nucleons ; prompt nucleons are excluded from consideration in this work . in that work the second moment of the cluster distribution was determined based on the charge of a cluster rather than its mass as is done in this work . previously , the second moment was generated from a cluster distribution that was not normalized to the changing size of the system as is done here . furthermore the prior analysis consisted of only one quarter of the total number of events used in the present analysis . thus the current analysis has higher statistics , has been freed of prompt nucleons , has a second moment that has been constructed with the masses from the cluster distribution and a cluster distribution that has been normalized to the changing system size . the exclusion of prompt nucleons and normalization to the changing system size are an effort to address the criticisms raised in @xcite and rebutted in @xcite . when the @xmath88-matching procedure was applied to the data presented in this paper essentially the same results as presented in ref . @xcite were recovered . see table ii and figure 25d . one difference observed is in the value of the critical point returned , @xmath127 reported in ref . @xcite and @xmath128 reported in this work . the difference is not as great as it appears to be . the origin of the published value of @xmath129 lies in picking the peak of the distribution of @xmath105 values that satisfied conditions ( 1 ) and ( 2 ) as the location of the critical point . the value was estimated based on the height of the peak and the error based on the width of the peak . the mean and rms of the @xmath105 distribution in ref . @xcite suggest a value of the critical point of @xmath130 . this value agrees , to within error bars , with the value of @xmath131 presented here . the relatively small shift in @xmath105 can then be understood to arise from the differences in the data sets . noting this it is clear that the present @xmath88-matching analysis is in agreement with the previous work . the results of the present work are , again , in keeping with the expected results of a small system undergoing a continuous phase transition . there is some region where matching @xmath88 values can be obtained , some regions in @xmath132 where th @xmath100-power law overwhelms all other effects . the fits in figure 25d are of quality than those for random partitions in figure 25c and cover a greater range . when compared to the percolation @xmath12 results the multifragmentation data compare favorably in terms of overall goodness of fits , width of fit region and matching of @xmath120 . see table ii . the location of the critical point returned by this analysis also compares well with the location from other analyses . see table i. in the last section it was seen that the @xmath88-power law and the data for the second moment in all systems have agreed over only a limited area . to some degree this is to be expected . near the critical point , assumptions valid for thermodynamic systems are invalid for the finite systems discussed in this work . for that reason , finite size effects dominate at the critical point and the second moment merely peaks instead of diverging . far from the critical point other effects come into play . the scaling assumptions inherent in the fdm are valid only in the neighborhood of the critical point . the size of this neighborhood is somewhat ill defined and seems to depend on many factors , _ e.g. _ the quantity in question , the nature of the system , the size of the system and so on . scaling behavior in physical systems can be observed over a wide range in temperatures and densities . this is most elegantly illustrated in the guggenheim plot @xcite of scaled temperature ( @xmath133 ) as function of scaled density ( @xmath134 ) for several different gases ( ne , ar , kr , xe , n@xmath135 , o@xmath135 , co and ch@xmath136 ) . in that plot the data collapse onto a curve that is well described by a power law with an exponent of @xmath137 . the range in validity of this agreement between data and power law is shown on the guggenheim plot to be over a range of @xmath138 and @xmath139 . however , another system , the combination of isobutyric acid and water , shows the guggenheim type of scaling only very near the critical point @xcite . already when the range considered is @xmath140 and @xmath141 corrections to scaling can be observed . to that end , higher order corrections to scaling are now examined in order to determine if fits such as those shown in previous sections can be improved . however , any improvement comes at the expense of more fit parameters and assumptions . to fully explore corrections to scaling in the context of the present systems where the cluster distributions serve as the main observable the fdm is revisited in a fashion employed in references @xcite and @xcite . assuming coexistence eq . ( [ cluster_distribution ] ) is then re - written as @xmath142 where @xmath143 is the correction - to - scaling function and @xmath144 is the correction - to - scaling exponent . the form of eq . ( [ corrected_cluster_distribution ] ) anticipates the presence of a second function of @xmath62 . in section iv-3 it was found empirically that both the scaled percolation and multifragmentation cluster distributions ( @xmath145 ) could be reasonably well described by the sum of two gaussians , eq . ( [ scaling_fit ] ) . in that treatment , the amplitude of each gaussian was a constant , @xmath146 and @xmath147 . if @xmath23 is restricted to a single value , the prescription give by eq . ( [ corrected_cluster_distribution ] ) is equivalent to that of eq . ( [ scaling_fit ] ) . ( [ corrected_cluster_distribution ] ) predicts that there should be an ordering to the scaled cluster distributions , _ i.e. _ smaller cluster sizes should lie above the larger clusters due to the correction term . this can be observed in figures 19a and b in the neighborhood of the maximum of the scaled data for the percolation systems . in the tails of the distribution , either large or small cluster production is suppressed . in the case of multifragmentation data , figure 19d , the ordering is generally observed where the statistics are adequate , namely , near @xmath148 . the ordering of the random partitions implies that @xmath149 . from eq . ( [ corrected_cluster_distribution ] ) it possible to derive the _ corrected _ isothermal compressibility ( second moment ) power law . following the method in section iii leads to : @xmath150 which is usually simplified and written as @xmath151 now the overall amplitude , @xmath103 is given by the first integral , and the correction - to - scaling amplitude is given by the second integral divided by the first . the correction - to - scaling exponent is @xmath152 . using eq . ( [ corrected_compressibility_2 ] ) to fit the second moment distribution would lead to determining four fit parameters : two amplitudes and two exponents . to explore the effects of corrections to scaling an assumption was made as to the universality class of the system in question and thus the choice of exponent values . for the three dimensional percolation universality class @xmath153 @xcite , @xcite and for the three dimensional ising universality class @xmath154 @xcite-@xcite . the amplitudes were left as free parameters and the second moment of the cluster distributions were fit . the value of the critical point determined from the @xmath88-matching analysis of section iv - b-5 was used for each system . figure 26 shows the results . [ h ] for the percolation system with the corrections - to - scaling a better fit to the second moment data was possible over a greater range in @xmath55 , up to twice the range of the average fitted region in the @xmath88-matching analysis , see figures 26a and b. the fits still failed to reproduce the behavior of @xmath98 near the critical point where finite size effects dominate the system . table iii lists the results for the critical amplitudes , @xmath103 . the agreement in the critical amplitudes determined in this analysis and the amplitudes from the @xmath88-matching analysis is due to the agreement of the behavior of eq . ( [ corrected_compressibility_2 ] ) and the @xmath88-power law from the @xmath88-matching analysis over the region in @xmath55 determined by @xmath88-matching . thus the @xmath88-matching analysis finds regions that are the least affected by higher - order corrections to scaling . for the random partitions an improvement is only observed for the high multiplicity region where a better fit over a larger range was obtained for both choices of universality class , see figure 26c . the low multiplicity events showed no such improvement partly due to the limited range in @xmath55 available . both the three dimensional percolation and three dimensional ising exponents were used in this analysis for the random partitions and the multifragmentation data . both choices of universality classes showed similar results . the lack of effect of corrections to scaling is to be expected in a system that does not follow fdm like scaling laws . the multifragmentation data also showed improvement resulting in a better agreement between the fits and the @xmath98 data points over a larger range in @xmath55 , see figure 26d . the improvement was observed for both choices of universality classes thus indicating this analysis is insensitive to the differences @xcite , @xcite , though the goodness of fit was better for the choice of the three dimensional ising exponents over the same fit regions ( 3d ising : @xmath155 and @xmath156 ; 3d percolation : @xmath157 and @xmath158 ) . at this level of analysis it appears that corrections to scaling improves the fits for the @xmath88-power law . whether this is due to the presence of a continuous phase transition in nuclear multifragmentation , or merely extra terms in the fitting function remains an open question . it has been pointed out that the previous statistical analysis of gold multifragmentation @xcite ignored the changing size of the system @xcite . to first order this may have been a reasonable procedure @xcite as many statistical signatures of a continuous phase transition have been observed both before and after the scaling to account for the changing system size has been performed ; _ e.g. _ the @xmath88 power law shown here and in previous works agree well . however , the data collapse in figure 19d is qualitatively not as great as that shown by the percolation ( @xmath12 ) system in figure 19b and the agreement between the calculated and measured @xmath98 behavior in gold multifragmentation in figure 24d is qualitatively not as good as that shown by percolation ( @xmath12 ) in figure 24b . in this section , the effect of the changing size of the system is explored and accounted for . the size of the multifragmenting system is shown in figure 27a after ref . @xcite . an approximately linear relation between the system size , @xmath159 , and @xmath12 was found : @xmath160 , see figure 27a . the functional form of @xmath104 was used in the following analysis to account for the changing system size in an average way , _ i.e. _ not on an event - to - event basis . if the multifragmenting system is assumed to be a system undergoing a phase transition , then the theory of finite size scaling of the critical point @xcite-@xcite asserts that the effective critical temperature , @xmath161 , changes as a function of the system size . coupling this with the changing size of the system indicates that at each value of @xmath12 the value of @xmath162 is different . [ h ] the value of @xmath162 can be determined in the following manner . first a relation between the multifragmenting system s temperature , @xmath163 , and @xmath12 must be determined . again from ref . @xcite a relation can be found , see figure 27b . two estimates of @xmath163 were made , one from a fermi gas ( uncorrected for the effect of expansion energy ) , @xmath164 , and the other from an isotopic yield ratio thermometer , @xmath165 , @xcite . these temperatures give an approximate indication of the initial and final temperatures of the system . figure 27b shows that @xmath164 is well fit by a quadratic function , while @xmath165 is well fit by a linear function . another linear function reproduces the average of @xmath164 and @xmath165 : @xmath166 , this was used for the following analysis . the critical temperature of infinite nuclear matter was assumed to be @xmath167 mev @xcite . from the fisher @xmath67-power law analysis the value of the multiplicity at the critical point was determined to be @xmath168 . the system size at that point is then @xmath169 and the temperature is @xmath170 mev . this indicates that the critical temperature for a nuclear system with 164 nucleons , @xmath162 , is approximately @xmath171 mev . according to theory , to first order the critical point scales with system size as : @xmath172 where @xmath173 is the euclidian dimension of the system and @xmath174 is the so - called hyperscaling exponent . at the smallest of system sizes higher order correction terms may play an important role in the scaling of the critical point @xcite . it is assumed in this procedure that the above mentioned finite size scaling dominates all other effects , including those which arise due to the charge of the protons present in the excited remnant . this is in keeping with the general philosophy of the theory of critical phenomena where near the critical point the precise details of the system are irrelevant . if it is assumed that gold multifragmentation is the result of a continuous phase transition and that transition occurs in three dimensions , @xmath175 , then using the _ hyperscaling relation _ @xcite @xmath176 with the extracted values of @xmath52 and @xmath67 , gives @xmath177 . the coefficient @xmath178 in eq . ( [ fss ] ) can be determined using @xmath179 , @xmath180 , @xmath173 and @xmath174 ; resulting in @xmath181 . note that this value of @xmath178 suggests that @xmath182 for @xmath183 . this is a result of the form of eq . ( [ fss ] ) and the notion that the critical temperature lowers as the size of the system decreases . presumably higher order effects not taken into account in eq . ( [ fss ] ) will affect the location in system size , @xmath159 , where the effective critical temperature vanishes . for the form of finite size scaling corrections shown in eq . ( [ fss ] ) , only @xmath184 yields an effective critical temperature that vanishes at @xmath185 . ( [ fss ] ) can be used to solve for @xmath186 and thus determine the scaled control parameter : @xmath187 the analysis to extract the exponent @xmath52 was performed with this @xmath188 by finding the peak in @xmath23 sized cluster production as a function of @xmath189 versus @xmath188 . previously , it was argued that the largest cluster should be excluded for all values of @xmath55 to account for finite size effects in the analysis to extract @xmath52 . the analysis in this section directly accounts for finite size effects , thus the standard fdm formalism with respect to the largest cluster is followed . this results in a value of @xmath190 and @xmath191 , figure 28a shows the resulting power law . the scaling function for gold multifragmentation was then plotted using the above corrections for the changing system size and finite size scaling , see figure 28b . the data collapse is qualitatively better than in figure 19d . the two gaussian parameterization of @xmath192 was fit to the scaled scaling function and is shown in figure 28b with parameters listed in table iv . [ h ] using the fitted parameterization of the scaling function and other quantities , the @xmath88-power law can be determined as before . figures 28b and c show the measured @xmath98 of gold multifragmentation plotted as a function of @xmath188 . a @xmath88-power law was plotted on , not fitted , figures 28b and c with @xmath193 ( from @xmath67 , @xmath52 and eq . ( [ scaling_1 ] ) ) and offsets determined via the scaled scaling function . this @xmath88-power law agrees with the measured @xmath98 over nearly all of the @xmath188-range , the exception being near @xmath194 where finite size of the system limits the maximum of @xmath98 . as stated above , this analysis makes no attempt to account for the coulomb energy of the system in the formulation of finite size scaling . the effect of the coulomb energy may evidence itself in the degree of collapse of the data . the fact that , to a large extent , the data do collapse without any explicit adjustment to the theory ( _ i.e. _ formalation of the scaling variable , @xmath195 ) may then be an indication that the coulomb energy is not a major perturbation . hence , it is perhaps justifiable to adhere to the coulomb - free theory when scaling @xmath196 . the focus in the present paper has been on cluster distributions and the types of analyses which can shed light on their creation mechanism . in particular , attempts were made to identify procedures that can distinguish those distributions which are related to critical behavior from those which are not . while this question is easily answered for systems containing large numbers of constituents , it is much more difficult to address the case of interest here , namely systems with at most only a few hundred particles . in such systems , finite size effects play a large role . therefore , two different computational models , bond building percolation and a random partitions , have been used . it is well known that in the macroscopic limit , the former system possesses a continuous phase transition characterized by a unique scaling function and set of critical exponents while the latter system does not . in addition , data arising from the multifragmentation of gold nuclei has been studied using the same procedures . for this system , it is not known , a priori , whether a critical point is present . many cluster properties have been proposed as being suitable measures of critical behavior . among these are the fluctuations in the size of largest fragment ( figure 1 ) , peaking in the quantity @xmath197 ( figure 3 ) , peaking behavior in @xmath98 ( figure 5 ) , campi plots ( figure 6 ) and simple power law behavior in the cluster mass distribution for a particular value of the appropriate control parameter . it was seen that none of these measures , taken alone or together , was sufficient to distinguish a system possessing critical behavior from one which does not . the first procedure which produced different results for critical and noncritical systems was the single parameter power law fit to the cluster mass distribution ( figures 12 through 15 ) . for the percolation systems and for the multifragmentation data , it was shown that the one parameter power law fit describes the data well only over a very limited range of the control parameter . the value of the control parameter where the power law fit is best is very close to where the abovementioned peaks occur . however , for the random partitions , this was not the case . if a system possesses a critical point , it is also expected to possess a scaling function that describes its behavior away from the critical point . phase transition theory specifies how the argument of this function depends on cluster mass and distance from the critical point . if such a function exists , the theory permits the determination of the critical exponent @xmath52 . this determination was done for the percolation system yielding satisfactory agreement with its known value . the same procedure was applied to the random breakup model and to the multifragmentation data . it was clear from this analysis that percolation and multifragmentation were very similar in many features , while the random system was distinctly different . see figures 17 and 18 . again , if a system possesses a critical point , it is expected to possess a scaling function that describes its behavior away from the critical point . therefore , when the data is properly scaled , it should collapse onto a single curve . figure 19 shows the amount of data collapse for the systems discussed here . the quality of the data collapse ( figures 19 - 21 ) reinforces the notion that the random breakup system is different from the others . although the precise form of the scaling function is not dictated by phase transition theory , both the percolation system and multifragmentation data were satisfactorily described by a sum of two gaussians . the random breakup model was not . the issue of finite size scaling was discussed . unlike the other systems examined here , which had a fixed number of constituents , the nuclear multifragmentation data originated from systems whose size varied monotonically with observed charged particle multiplicity . see figure 27a . phase transition theory makes a prediction , eq . ( [ fss ] ) , as to how the value of the control parameter at the critical point changes as a function of system size . applying eq . ( [ fss ] ) produced an improvement in the quality of the data collapse for the scaling function and yielded a better prediction for the behavior of @xmath98 . see figure 28 . critical exponent values have been determined in an unbiased manner for each system . for both sets of analyses on the percolation clusters , the standard percolation exponents were recovered to within error bars . for the random partitioning , exponents could be extracted , but none that fulfilled well known scaling laws . the exponent values determined from the gold multifragmentation cluster distributions fulfill the scaling laws , to within error bars , and fall near the three dimensional ising universality class . the effect of secondary decays from hot initial fragments on the critical exponents has not been explicitly considered in this paper . in the smm @xcite-@xcite such effects become significant above @xmath198 mev / nucleon . thus @xmath67 and @xmath88 , which are determined at lower excitation energies , will be unaffected in the smm s fragment distributions . the exponent @xmath52 is determined by the multiplicities at which individual light fragment yields attain their peak values . as shown in figure 17d , the lightest fragments peak at large multiplicities , corresponding to excitation energies for which secondary decay are important in the smm . an smm calculation @xcite indicates that the value of @xmath52 from the smm s fragment distribution was increased by about @xmath199% due to this effect . however , it is unclear from that calculation that the effects of secondary decay are as great in the experimental data as they are in the smm . the smm calculation over predicts the yield of light fragments which could indicate that the smm estimates of secondary decay are too severe . corrections to the model independent quantities determined in this paper based on the smm calculations are premature . see appendix c for discussions of another set of model based interpretations . although the multifragmentation data possess many of the gross and detailed characteristics that the percolation system does , it is not at all obvious why this should be the case . after all , real nuclei obey quantum mechanics , have varying binding energies per particle , and , most significantly , are charged . on the other hand , it is well known that near a critical point the details of the interaction become unimportant and only the dimensionality of the system and the dimension of the order parameter are important . as noted in the introduction , the attractive nuclear force bears a similarity to a van der waals force . however , the coulomb force is a long range force and imposes a natural limit to the size of stable nuclei . thus , it is not clear to what extent a finite charged system can exhibit critical behavior when the macroscopic system can not exist . the exact role of the coulomb force in physical systems undergoing a change of phase is currently of great interest @xcite , @xcite and is , at this point , an open question . the philosophy of this paper has been to make use of phase transition theory as it applies to uncharged systems . what results for the analysis of the gold multifragmentation data bears great similarity to the results of the same procedures applied to a system known to possess a critical point . it is tempting then to conclude that multifragmentation is related to critical behavior occurring in a finite nuclear system . as a demonstration of the power of the sort of scaling analysis presented above it is shown that the random partitions follow a simple power law of @xmath200 . figure a1 shows the scaled cluster distribution as a function of @xmath12 from clusters with @xmath201 . the data nearly collapses to unity along the horizontal axis over the multiplicity range for @xmath202 . the deviations are due to the constraints of @xmath12 and finite size . with a simple scaling analysis the underlying power law describing the cluster distribution becomes clear . the value of @xmath69 used in this work based on the @xmath74-function was generated with vax fortran code using double precision and letting the sum run from @xmath10 to @xmath203 . the sum was terminated at this point in order to keep computing times within reason . for a value of @xmath204 summing to @xmath203 gives a value of @xmath69 that is within 10@xmath122 of the value when the sum is terminated at @xmath205 , see figure a2 . increasing the upper limit of the summation in the @xmath74-function causes no significant changes in the analysis presented in this work . in a recent paper @xcite , gulminelli and chomaz ( gc ) report on calculations made using a lattice gas model . their results can be divided into two different categories : ( 1 ) an analysis of the thermodynamic quantities of the system ; ( 2 ) an analysis of the statistical properties of the droplet or cluster distribution . in their paper gc performed canonical calculations on a three dimensional lattice of @xmath206 sites ( @xmath207 or @xmath208 ) with periodic boundary conditions at a given temperature , @xmath163 , and density , @xmath209 . the density was varied by changing the number of particles in the system , @xmath159 , with @xmath210 . for a choice of @xmath163 , @xmath209 was varied and the system s chemical potential , @xmath48 , was determined . at some values of @xmath163 the @xmath211 vs. @xmath212 ( @xmath211 vs. @xmath159 ) isotherm exhibited a back bend . as @xmath163 increased the back bend disappeared . this was taken as evidence that a first order phase transition ( back bending @xmath211 vs. @xmath159 ) had culminated in a continuous phase transition ( flat @xmath211 vs. @xmath159 ) . the critical point , ( @xmath196 , @xmath213 ) , was thus defined . via a maxwell construction , gc also determined the boundary of the coexistence zone . the theoretical ideas behind the procedure followed by gc are sound ( with the exception of the questionable definition of volume in a system with periodic boundary conditions ) , however the execution is flawed . the error in execution lies in the method by which @xmath209 was varied . gc varied @xmath209 by holding @xmath206 , the volume of the system , fixed and varying @xmath159 , the number of constituents or `` size '' of the system . therefore , at each value of @xmath209 gc have a system of different size . it is well known that the effective @xmath196 of a system varies with its system s size @xcite-@xcite . thus each time gc varied @xmath159 to vary @xmath209 they were dealing with a system with a different @xmath196 . with this understanding it is clear that a plot of @xmath48 versus @xmath159 as shown by gc in their figure . 1a , which serves as the basis for their determination of the critical point , is difficult to interpret . @xmath196 can not be deduced in the manner followed by gc . similarly , it is impossible to determine @xmath213 for a system of @xmath214 using information from a system of @xmath215 without accounting for the finite size scaling of the critical point . this makes the thermodynamically extracted values of the critical point suspect and calls into question the validity of the coexistence line shown by gc in the lower part of their figure 3 . via an analysis of the droplet distributions gc determined the value of the critical exponent @xmath216 based on the peak in the production of different size droplets . the value of @xmath196 was determined coupling the @xmath217 results with a two parameter power law fit the droplet distribution at various temperatures , @xmath218 , at @xmath196 @xmath219 . the use of two parameter fits is generally an improper method to determine @xmath67 @xcite-@xcite . in this analysis gc determine @xmath196 on a system - by - system basis , avoiding the error of grouping systems of different size together . gc also extracted the exponent @xmath52 and then used the above results to collapse the droplet data onto a single curve illustrating the scaling function that modifies the power law away from the critical point . while the data collapses onto a single curve , the value of the curve at @xmath196 is problematic . from gc s work the value of the scaling function at @xmath196 is @xmath220 which contradicts the @xmath67 values extracted by gc according to the riemann @xmath74-function relation between @xmath67 and @xmath221 @xcite ; _ e.g. _ @xmath222 gives @xmath223 , while gc s @xmath220 gives @xmath224 . finally gc combine the results of their analyses and assert that the krtesz line of @xmath67 power laws continues into the coexistence region of the lattice gas and that this behavior is observed due to the finite size of the system and is not observed in systems of much greater size . however , due to the problems in gc s method of varying @xmath209 to find the coexistence curve , such a claim is premature . instead , their estimates of @xmath196 from the droplet analysis , which are shown as a string of points in their figure 3 , may reflect the finite size scaling of the critical point and not the krtesz line extending into the coexistence zone . that each different size system appears to be at @xmath213 as well as @xmath196 , illustrated by the data collapse , may be explained by the relation @xmath225 . for a 3d ising system @xmath226 . thus , systems of finite size near their critical density are nearly on the coexistence curve and show the sort of data collapse illustrated in gc s figure 2 @xcite . this work was supported in part by the u.s . department of energy contracts or grants no . de - aco3 - 76f00098 , de - fg02 - 89er-40513 , de - fg02 - 88er-40408 , de - fg02 - 88er40412 , de - fg05-er40437 and by the u.s . national science foundation under grant no . phy-91 - 23301 .

we have developed a new stellar population synthesis model designed to study early - type galaxies . it provides optical and near - infrared colors , and line indices for 25 absorption lines . it can synthesize single age , single metallicity stellar populations or follow the galaxy through its evolution from an initial gas cloud to the present time . the model incorporates the new isochrones of the padova group and the latest stellar spectral libraries . we have applied our model to new data for a set of three early - type galaxies , to find out whether these can be fitted using single - age old metal - rich stellar populations , as is normal practice when one uses other stellar models of this kind . the model is extensively compared with previous ones in the literature to establish its accuracy as well as the accuracy of this kind of models in general . using the evolutionary version of the model we find that we can not fit the most metal - rich elliptical galaxies if we keep the imf constant and do not allow infall of gas . we do however reproduce the results of arimoto & yoshii ( 1986 ) for the evolution of the gas , and produce colors , and , for the first time with this type of models , absorption line - strengths . it is in fact possible to fit the data for the elliptical galaxies by varying the imf with time . our numerical model is in good broad agreement with the analytical _ simple model_. we prefer however to calculate the evolution of the gas numerically instead of using the _ simple model _ , since it offers more flexibility , and even improved insight , when comparing with observations . in the present paper we describe the model , and compare a few key observables with new data for three early - type _ standard _ galaxies . however the data , as well as our fits , will be discussed in much more detail in a second paper ( vazdekis _ et al . _ 1996 ) , where some conclusions will be drawn about elliptical galaxies on the basis of this model . 23truecm

astro-ph9605112

the information that is available about galaxies primarily relates to their morphology , internal kinematics , and spectral energy distribution . the further away one goes , the less important become the first two , compared to the third . clearly , if one wants to study galaxy evolution , study of the electromagnetic spectrum is of maximum importance . since the spectrum of a galaxy generally consists of a combination of stellar spectra , emission from gas , and possibly non - thermal radiation , partly extinguished by dust , and since also stellar spectra exist in a very large number of varieties , it is clear that understanding the spectrum of a galaxy is very difficult . to study complicated spectra it is necessary to first understand the spectral energy distributions of relatively simple objects . for this reason we will discuss here a model that analyzes the spectra of early - type galaxies . these objects appear to contain relatively little dust extinction and gaseous interstellar medium , and to have little recent star formation . they have , for these reasons , been the most studied objects in the current literature on population synthesis . there have been a number of accepted ways to attack the problem of understanding the spectrum of an elliptical . to understand why we adopt our current method , we will summarize shortly some of the population synthesis methods most often used in the literature . a stellar population synthesis program tries to find a combination of stars for which the integrated spectrum agrees with the observed spectrum of the object under study . in practise the problem is often underconstrained , i.e. a number of combinations of stars can be found which are able to fit the spectrum . to overcome this problem one generally forces the solution to obey certain constraints . these range from simple continuity requirements ( e.g. the luminosity function should decrease monotonically ) to the requirement that the distribution of stars is determined completely by stellar evolution calculations . models with very few physical constraints are generally called _ empirical _ population synthesis models , as opposed to _ evolutionary _ models . empirical models have been used with some success by spinrad & taylor ( 1971 ) , faber ( 1972 ) , oconnell ( 1976 , 1980 ) and pickles ( 1985 ) . these papers often make use of linear programming to obtain their results . some workers , notably bica ( 1988 ) , have attempted to take into account evolutionary effects by using , as units of population , distributions of stars observed in clusters of our galaxy , instead of individual stars . evolutionary models use a theoretical isochrone or hr diagram , convert isochrone parameters to observed spectra in some way and , finally , integrate along the isochrone . they all need to make an assumption about which _ initial mass function _ imf to use . also , the models need a recipe prescribing when the stars have been formed . since the imf is not very well known at the present time , its treatment is not very different from one model to another . however , as far as the _ star formation rate _ ( sfr ) is concerned , some models assume that all stars are formed at the same time , others prescribe that the sfr has to decrease exponentially with time , while still others explicitly try to describe the whole formation of a galaxy from a gas cloud and form stars when the physical conditions in the gas are adequate . examples of this kind of evolutionary models can be found in tinsley ( 1968,1972,1978a,1978b,1980 ) , searle _ et al . _ ( 1973 ) , tinsley & gunn ( 1976 ) , turnrose ( 1976 ) , whitford ( 1978 ) , larson & tinsley ( 1978 ) , wu _ et al . _ ( 1980 ) , vandenberg ( 1983 ) , bruzual ( 1983,1992 ) , stetson & harris ( 1988 ) , renzini & buzzoni ( 1986 ) , rocca - volmerange & guiderdoni ( 1987,1988,1990 ) , guiderdoni & rocca - volmerange ( 1987,1990,1991 ) , yoshii & takahara ( 1988 ) , rocca - volmerange ( 1989 ) , buzzoni ( 1989 ) , charlot & bruzual ( 1991 ) , lacey _ et al . _ ( 1993 ) and bruzual & charlot ( 1993 ) . some of these models not only predict colors but also line - strengths notably those of peletier ( 1989 ) and worthey ( 1994 ) . there are also models which combine evolutionary population predictions with considerations of chemical evolution . these models follow the evolution of the gas and make use of isochrones of more than one metallicity ( solar ) . examples of these _ chemo - evolutionary _ population synthesis models are arimoto & yoshii ( 1986,1987 ) , casuso ( 1991 ) , bressan _ et al . _ ( 1994 ) as well as the model presented in this paper . note that for this type of models only the global metallicity , z , is normally taken into account to determine the stellar populations . however it is in principle possible to follow the abundance distribution separately for each of several important elements in the gas . examples of these so called chemical evolution models can be found in larson ( 1972 ) , matteucci & tornambe ( 1987 ) , tosi ( 1988 ) and there are many more . however , calculating colors and especially absorption line - strengths in such models has not been attempted up to now . finally , among this kind of models there are a few which combine chemical evolution with dynamics , e.g. theis , burkert & hensler ( 1992 ) . one can not say that one type of model is better than another . in general , more observables and physical parameters can be calculated if one makes more assumptions . if no assumptions are made about the physics , as in the empirical models , one may end up with solutions that are unphysical . if however wrong assumptions are made , one will not learn anything about the stellar evolution history either . we show later in this paper that our results can be reproduced using the _ simple _ analytical model and possibly with infall . this means that we could have replaced the part of the model that deals with the evolution of the gas by some analytical calculations . however , we have preferred to build up the numerical machinery , since it offers much more flexibility , and even improved insight . it is clear that our understanding of all these aspects is improving with time , which means that the models that are to be applied can legitemately be more and more complicated , and in this context we have developed the evolutionary population synthesis code presented in this paper . for the reasons mentioned above it should never be used as a static black box out from which a theoretical fit to the data is to be taken , but as an evolving tool , which might help in disentangling stellar populations in a composite system . if one wants to calculate the final spectrum of a galaxy that has evolved from a gas cloud , one has to integrate over time the spectra of all stars that are still _ living _ at the current time . the number of stars formed at a certain epoch is determined by the star formation rate . little is known about this , so many models give it a prescribed form , and let it decrease e.g. exponentially . in this paper we assume that the sfr is proportional to the gas density ( schmidt 1959 ) . the gas density itself and the chemical evolution of the gas is calculated taking into account the original gas , and the metal - enriched gas that is ejected by stars . the yields for the various elements needed for this calculation are not especially well known , and better knowledge would significantly improve the model . other factors that may affect the sfr , such as inflow or outflow , are not considered in the context of the present models . the stars living at the current epoch contain stellar populations with a mixture of ages and metallicities . to calculate the final spectrum we decompose the stars into single stellar populations ( ssp ) each of a single age and metallicity , and calculate their spectra . to do this , one needs in the first place theoretical isochrones . the parameters of the isochrones depend , amongst others , on opacities of ions and molecules , and are reasonably well known for the early stages of stellar evolution . however for later phases such as the agb and the post - agb , evolutionary calculations are very complicated , and could still benefit from significant improvement . the isochrones are much better for solar and sub - solar metallicity than for metal rich stars , because the former can be , and have been , tested observationally on globular clusters . in general , the relative composition of the elements heavier than helium in the models is close to solar ; only recently have people included for example oxygen - enhancement ( vandenberg 1992 ) or @xmath0-element enhancement ( weiss _ et al . _ 1995 ) . the following step is to obtain a spectrum for a star with physical parameters given by the isochrones . model atmospheres needed for this ( e.g. kurucz 1992 ) appear to be reasonably reliable in the blue . in the red , molecular opacities make them , for the time being , much less reliable . for that reason several authors ( faber _ et al . _ 1985 , gorgas _ et al . _ 1993 and worthey _ et al . _ 1994 ) have developed a method that depends much more on observations . they use a grid of observed stars with various theoretical parameters to determine fitting functions that can be used to calculate an absorption line index for any combination of theoretical isochrone parameters . the problem however is that these fitting functions at the moment are available only for a limited number of absorption lines , since one needs many stars of various types and metallicities to make them . if one wants to carry out population synthesis , one needs to cover as large a wavelength range as possible , and also at as high a spectral resolution . this is to cover many colors and line indices , since in principle every color and absorption line is affected in a different way by the parameters above : the sfr , the imf , the metallicity and abundance ratios etc . in practice a model with complete coverage from uv to near - ir is hard to implement , because of lack of fitting functions , color calibrations etc . , and the observations are hard to obtain . in this paper we are concentrating on early - type galaxies , for which high spectral resolution is less important , due to their large velocity dispersions . we have produced model output for colors and lines that are relatively easy to obtain and reproduce , and that allow us to separate effects due to various relevant physical parameters . to some extent we are limited by the availability of libraries of stars , especially for the fitting functions , so this aspect too , can be significantly improved in the future . the layout of the paper is as follows . in section 2 we will describe how we obtain colors and absorption line indices for a single stellar population , emphasizing differences from previous studies , and especially any improvements . in section 3 we introduce our chemical evolutionary model . in section 4 we apply our model to a limited set of data from three standard galaxies , for which the data is presented in an accompanying paper ( vazdekis _ et al . _ 1996 , * paper ii * ) . we perform the fits for an ssp ( the _ static _ option ) and for a full evolutionary case . in section 5 the conclusions are presented . finally , in paper ii , apart from presenting the observational basis of the data - set used here , we carry out a comprehensive fit of the whole set of indices and colors on the basis of the scenarios suggested in this paper . in particular , we compare the fits obtained assuming a single - age single - metallicity stellar population or using the full chemical evolution model . we also discuss the relations between elements . the stellar content of a model galaxy calculated by our full evolutionary model consists of stars of various ages and metallicities . in practice its output spectrum is the integrated output of a large number of spectra of stellar populations each with a single age and metallicity ( ssps ) . in effect , the ssps can be thought of as the building blocks of our stellar population model , although in an evolutionary model these ssps are not independent but linked by physical parameters governing star formation , which makes the evolutionary method intrinsically more meaningful . many authors , e.g. worthey ( 1994 ) ( hereafter w94 ) , restrict themselves to calculating a single ssp , which is less general , but whose simplicity can give insights in the stellar populations of a galaxy . we will refer to a single age single metallicity model as the _ static _ case of our stellar population model . in this section we will explain in detail how integrated colors and line strengths are calculated for this _ static _ case . this model works as follows . assuming an age and a metallicity it calculates the present distribution of the stars according to the required isochrone ( see next subsection ) and weights the light of the stars on this isochrone with the assumed imf ( as defined in section 2.2 ) . then it calculates the integrated colors and absorption lines , weighting each individual contributing star by its luminosity ( as explained in sections 2.3 and 2.4 ) . we present the results of our model for a grid of ages and metallicities and compare them with the literature . our models are based on stellar evolution theory and require a set of isochrones to predict the distribution of the stars in the hr diagram at a given time . in this work we have adopted the large grids of theoretical isochrones provided by bertelli _ ( 1994 ) ( hereafter bbcfn ) . their initial compositions have been chosen according to the empirical law of pagel ( 1989 ) @xmath1 and are [ z=0.0004,y=0.23 ] , [ z=0.001,y=0.23 ] , [ z=0.004,y=0.24 ] , [ z=0.008,y= 0.25 ] , [ z=0.02,y=0.28 ] and [ z=0.05,y=0.352 ] . all these sets of isochrones were computed using the most recent radiative opacities by iglesias _ ( 1992 ) ( opal ) except those of @xmath2 $ ] which are based on models with the radiative opacities by huebner _ et al . _ ( 1977 ) ( laol ) . these isochrones cover a wide range of ages and of masses , from approximately @xmath3 to @xmath4 and @xmath5 to @xmath6 respectively . another main advantage in the application of these sets of isochrones is that they cover all the stellar evolutionary stages : from the zams , beyond the red giant tip until the white dwarf stage after the planetary nebula phase . for more details about these isochrones we refer the reader to bbcfn . the set of isochrones with @xmath7 can be safely used because the effect of varying from laol to opal opacities at low metallicities has been estimated to be small by alongi _ ( 1993 ) ( see also bbcfn ) . in order to be used in our spectrophotometric stellar population synthesis models ( see next sections ) , the isochrones must contain the following quantities : the age , the initial and the present mass along the isochrone , the effective temperature and the gravity . because the electronic file with the isochrones provided by the padova group , instead of the initial mass contains the indefinite integral of the imf by number over the mass ( flum ) , as defined in bbcfn , we first obtain the initial mass by inverting the equation @xmath8 where @xmath9 . now , from the initial stellar mass and the mass at a given point on the isochrone , we obtain the gas mass fraction ejected by a star of mass @xmath10 and metallicity @xmath11 until time @xmath12 : @xmath13 ( see section 3.2.1 ) . since the gravities are not in the electronic data base of the isochrones we have calculated them using : @xmath14 where @xmath15 and @xmath16 are the stefan - boltzmann and the gravitational constants respectively and @xmath17 is the mass along the isochrone . as the low - mass limit of the isochrones used is @xmath18 , we decided to complete them with stars of lower masses . the reason for this is that , even if these stars do not individually contribute substantially , their integrated light may not be negligible ( see table 5 , above all in the near ir for old systems , in those cases where we require a population model with an imf with high slope . to complete the lower - mass ms of the isochrones of composition z=0.008 , z=0.02 and z=0.05 we used the results for low - mass stars , for z=0.004 , z=0.01 , z=0.02 and z=0.03 , generously provided to us by pols ( private communication ) . the corresponding tracks for solar composition are described in pols _ these tracks have been calculated using a renewed version ( han , podsiadlowski & eggleton 1994 ) of the evolution program of eggleton ( 1973 ) , incorporating an improved equation of state and making use of radiative opacity tables from alexander & ferguson ( 1994 ) for the low temperatures . since the stars with masses below @xmath18 are always on the zams and show almost no evolution , it is straightforward to incorporate them into the lower ms via the appropriate isochrones . however , for the isochrones with metallicity z=0.008 we first linearly interpolated the low - mass stars between z=0.004 and z=0.01 . finally , to complete the low - ms of the super metal rich isochrones ( z=0.05 ) we were forced to extrapolate linearly from the low - mass stars of z=0.02 and z=0.03 . as a test we see that in the theoretical hr - diagram there is good agreement between stars above ( from bbcfn ) and below @xmath18 . in this work we adopt two imf shapes : \i ) * unimodal : * @xmath19 where @xmath20 for the solar neighborhood is equal to 1.35 ( @xmath21 ) , and @xmath22 is a constant . \ii ) * bimodal : * @xmath23 where @xmath24 is a spline for which we calculate the corresponding coefficients solving the following system obtained by the boundary conditions : @xmath25 this bimodal imf is based on observational results of scalo ( 1986 ) and kroupa _ et al . _ ( 1993 ) . normalizing our imf to unity we have @xmath26 where @xmath27 and @xmath28 are , respectively , the lower and upper mass limits of the stellar range considered ( in this paper we choose @xmath29 and 72 @xmath30 for these limits ) . we obtain the constant @xmath22 via numerical integration . in fig . 1 we plot our imfs for @xmath31 , comparing them with the data of scalo ( 1986 ) . the integrated flux of an ssp at a wavelength @xmath32 , @xmath33 can be calculated using : @xmath34 where @xmath35 is the mass of a star whose lifetime is @xmath36 , @xmath37 the flux of a star with mass @xmath10 and age @xmath36 , and @xmath38 is the number of these stars . this final distribution of stars is calculated via @xmath39 where @xmath40 is the mass of the galaxy ( or of a given region of the galaxy studied ) . for the wavelengths we chose to use the johnson - kron - cousins ubvri system in the visible and the bessell - brett system in the near - infrared . definitions of the ubv passbands are given in buser & kurucz ( 1978 ) , for r and i in bessell ( 1990 ) and for jhk in bessell & brett ( 1988 ) . to obtain @xmath37 from the theoretical luminosities , effective temperatures and gravities along the isochrones for each metallicity we preferred to use , whenever possible , empirical relations from observed stars , rather than theoretical spectra . we performed this using bolometric corrections ( bc ) and @xmath41-color and color - color relations . this was possible for the isochrones with metallicities not very different from solar : z=0.008 , z=0.02 and z=0.05 . for lower metallicities , due to the lack of calibration stars , we had to take the computed colors of bbcfn , obtained by convolving the seds of the library of stellar spectra of kurucz with the theoretical passbands . [ [ dust - analysis . ] ] dust analysis . + + + + + + + + + + + + + + to be able to investigate the presence of dust , we have implemented in the code the galactic reddening law of rieke & lebofsky ( 1985 ) which allows us to correct for small amounts of reddening . for early - type stars we calculated the bcs following code _ for late - type stars we used the formula of bessell & wood ( 1984 ) which relates the bc to the @xmath42 color . finally , we used the metal - dependent formulae given in alonso _ et al . _ ( 1995 ) to determine the bcs of the low - ms in the range of temperatures : 4000 - 8000k . because the resulting bcs , from different sources , do not match exactly , we have interpolated linearly between them to avoid discontinuities . 2 summarizes the bc for solar composition . where possible we have used empirical studies for which the effective temperatures had been determined using reliable stellar angular diameter measurements ( using the lunar occultation technique ) . for color - color conversions we preferred studies which include a full set of colors rather than those which give a more restricted set , and require the inclusion of subsidiary information to complete them . in this way we avoid , as far as possible , problems derived from the use of different sets of filters . nowadays the various observatories are quite well standardized . it is rather a question of different filter sets which were used at different times . separate conversions were performed for giants and dwarfs : * dwarfs in the range 4000 - 8000 k : except for u - v ( for which we used the color - color relations of johnson 1966 ) , we inverted the polynomial fitting functions of alonso _ et al . _ ( 1996 ) to obtain each color as a function of the @xmath41 and the metallicity . to obtain their relations they observed a large sample of ms stars ( @xmath43 ) with spectral types from f0 to k5 covering a wide range in metallicity ( @xmath44 < 0.5 $ ] ) . to transform from their j and k filters to our system we used the relations given in alonso _ et al . _ * stars hotter than 8000 k : we first obtained the @xmath45 from the @xmath41-(@xmath45 ) table of code _ et al . _ ( 1976 ) and then used the color - color relations of johnson ( 1966 ) . we took a smoothly interpolated transition from alonso _ et al . _ ( 1996 ) for the boundary at 8000 k. this method works since line blanketing for these hot stars is not very important . for the case of stars hotter than 34000 k we used the colors given in bbcfn ( up to 50000 k obtained by convolving with the kurucz stellar spectra library and above 50000 k by assigning pure black - body ( bb ) spectra , see bbcfn for details ) . finally , to convert from the r and i broad - band filters in the johnson ( 1966 ) system to those in the _ johnson , kron , cousins _ system ( bessell 1990 ) we used the linear relations of bessell ( 1983 ) . * dwarfs cooler than 4000 k : due to lack of reliable observations we took the bb effective temperatures of johnson ( 1966 ) . the same paper , which does not list any metallicity dependence , was used to obtain the different colors by shifting them to match the well - calibrated metal - dependent colors of alonso _ et al . _ ( 1996 ) at 4000 k. in this way we have in fact obtained the dependence on metallicity , which becomes more important as the temperature decreases , as can be seen in alonso _ et al . _ * giants in the range 3350 - 4930 k : using the empirical calibration of ridgway _ et al . _ ( 1980 ) we first obtained the ( @xmath42 ) color from @xmath41 , after which we applied the color - color conversions for giants of bessell & brett ( 1988 ) to obtain the other colors . for the cooler giants we also used the stellar library of fluks _ et al . _ ( 1994 ) . * giants cooler than 3350 k : unfortunately there is a lack of observations for these stars . especially for old stellar populations , the contribution from giants with spectral types later than m5 is not negligible in red filters especially for steep imfs . for these stars we used the metal - dependent model colors of bessell _ et al . _ ( 1989 , 1991 ) . since these models do not contain the u and b filters , we obtained the @xmath46 and @xmath45 colors from the observational stellar library of fluks _ et al . _ ( 1994 ) , taking stars with the corresponding @xmath42 . finally , to avoid problems in matching these stars with hotter giants we have interpolated the colors between 3200 and 3350 k. we strongly encourage observers to carry out programs to obtain good empirical data for cool giants as well as for cool dwarfs . it is also important to study empirically the dependence on the metal content . in this respect observations in baade s window such as those obtained by terndrup _ ( 1990 and 1991 ) are very useful . [ [ the - low - mass - metal - poor - stars . ] ] the low - mass metal - poor stars . + + + + + + + + + + + + + + + + + + + + + + + + + + + + + + for stars with masses lower than @xmath47 and low metallicities ( z= 0.0004 , z=0.001 and z=0.004 ) , we obtained their parameters by extrapolating linearly from the low - ms range of the isochrones for which we took the colors given by the bbcfn . we were forced to this choice due to the lack of empirical data to transform from the theoretical to the observational plane . as well as the colors , we have also calculated absorption line strengths . to make it possible to compare observations with the line strengths from the models , we have worked in the extended lick - system ( worthey _ et al . this has the additional advantage that many observed spectra are available for calibration : stars in that paper and also in faber _ et al . _ ( 1985 ) and gorgas _ et al . _ we used the fitting functions of worthey _ et al . _ ( 1994 ) to relate each line index to its corresponding three atmospheric parameters : @xmath48 $ ] , @xmath49 and @xmath41 ( or @xmath42 ) . since for the cn1 and cn2 indices these authors do not give fitting functions for metallicities below @xmath48 = -1 $ ] we have obtained second order polynomial relations , which are given in table 1 . for this we used around @xmath50 stars of their sample with low metallicities and for which they measured the two indices . we also have included some features in the near - ir : the caii triplet and the mgi line . we have used the stellar spectra of daz _ et al . _ ( 1989 ) to calibrate second order polynomial relations between each line index and its corresponding atmospheric parameters . all these fitting functions have been tabulated in table 1 . as pointed out by daz _ ( 1989 ) it is useful to distinguish between two ranges , metal - poor and metal - rich ( separated at @xmath48=-0.3 $ ] ) . in the first the index is strongly dependent on @xmath48 $ ] , while in the second it depends much more strongly on gravity . we assigned a luminosity class to each _ live _ star by adopting a rather broad set of criteria to distinguish between dwarfs and giants . one of the best discriminators for this is clearly @xmath49 and we have used as limits : @xmath51 because in the gravity range @xmath52 , the value of @xmath53 alone can not discriminate between the two classes of stars , we decided to use the @xmath42 criterion cited to divide the subgiant region into two parts . finally , stars with @xmath54 or @xmath55 are not classified , because they are not important for line - strength computations . this equation was mainly used when calculating the contributions of the m stars to the integrated absorption features . the code yields a stellar distribution for a given age of the galaxy and metallicity ( @xmath11 ) , converted to [ m / h ] by the relation @xmath56=\log(z / z_{\odot})\ ] ] to calculate an integrated line index the code integrates the contributions of all the stars ( obtained using eq . 11 ) to calculate the total line index for the resulting ssp . an integrated line - index @xmath57 is a flux - weighted mean flux and could be expressed either as an equivalent width ( ew ) or in magnitudes . we have decided to maintain the convention of burstein _ et al . _ ( 1984 ) , used also in worthey _ ( 1994 ) of expressing the molecular - band features in magnitudes and the atomic - line features in equivalent width ( ew ) in . the nir features are expressed in ew as well . we calculate @xmath57 in the following way : @xmath58 where @xmath59 is the flux in the continuum corresponding to the central wavelength of the spectral feature of the star of mass @xmath10 at @xmath36 , obtained by a linear interpolation between the closest broad - band filter fluxes @xmath60 and @xmath61 @xmath62 where @xmath63 , @xmath64 and @xmath65 are the central wavelengths of the index and the two broad - band filters respectively . finally , to convert the indices expressed in ew to magnitudes we use the following relation @xmath66 where @xmath67 is the wavelength width ( in ) of the feature . to model a single stellar population the main input parameters which must be defined are : * the imf shape ( unimodal or bimodal as defined in section 2.2 ) , the imf slope @xmath20 and the lower and upper cutoff limits @xmath27 , @xmath28 ( we kept them fixed at 0.0992 and 72 m@xmath68 ) . * the age @xmath36 of the galaxy ( or of the observed zone of the galaxy ) . * the metallicity @xmath11 . in tables 2 and 3 we present the resulting colors and line indices for single stellar populations for unimodal and bimodal imfs with two slopes @xmath20 : 1.35 and 2.35 . for each @xmath20 we present the results for different metallicities and ages . the differences between the two imfs mainly affect the reddest colors or indices . these tables , as well as others which also include data for models with other @xmath20 values can be obtained from the authors . -1 cm l|rrrrrrrrrr|l index&@xmath69&@xmath70&@xmath71&@xmath72&@xmath11&@xmath41&@xmath73&@xmath74&@xmath75&constant&validity + + cn1&-0.0428&0.4230&-2.6264&-7.6603&16.2199&4.5248&-0.1189&2.1176&-3.9428&19.9478&@xmath76 + cn1&-0.0089&0.0818&-0.0558&-2.8505&0.0702&-2.9541&-0.0002&0.7670&0.0945&12.1739&@xmath77 + + cn2&-0.0196&-0.1867&0.1706&-0.0520&-10.5277&-0.1924&0.0541&0.0734&2.6526&-2.5165&@xmath76 + cn2&-0.0148&0.0170&0.1875&-2.6252&-0.4784&-4.2434&-0.0485&0.7027&0.2019&13.5325&@xmath77 + + + ca@xmath781&-0.0065&0.0024 & 0.4254 & 5.5058 & -6.2485 & 1.1452&-0.1162&-1.5381 & 1.8636 & -7.9084&@xmath79 + ca@xmath781&0.1269 & -0.8879 & 0.6840 & 4.9417 & 4.2020 & 0.5141&-0.3062&-1.6106&-0.7240 & -7.7763&@xmath80 + + ca@xmath782&0.1669 & -0.1111 & 0.6564 & 9.9490 & -8.8215 & 3.8129&-0.1269&-3.0427 & 2.6914&-17.0107&@xmath79 + ca@xmath782&0.2208 & -2.3469 & 0.0913 & 1.1574 & -9.3570 & 1.0685&-0.3788&-0.7946 & 3.2014 & 1.9801&@xmath80 + + ca@xmath783&0.1335 & -0.1789 & 1.6471 & 9.8307 & -2.5977&-4.5822&-0.0627&-2.9351 & 0.8503 & -0.6660&@xmath79 + ca@xmath783&0.1801 & -1.6372 & 0.6482&-1.1326&-20.1575&-3.4841&-0.6283&-0.1215 & 6.2543 & 10.3075&@xmath80 + + mg@xmath81 & 0.0425 & 0.0453 & 0.5193 & 3.9973 & 1.9597&-3.2429&-0.0278&-1.1445&-0.4173 & 5.9036&@xmath79 + mg@xmath81 & 0.0302 & 0.2442 & -0.8184 & 0.9179 & 1.3805 & 4.0668 & 0.0241&-0.2869&-0.3096 & -3.1040&@xmath80 + .the model observables for ssps with a unimodal imf with slopes 1.35 and 2.35 . the age is in gyr . [ cols="<,>,>,>,>,>,>,>,>,>,>,>,>,>,>,>",options="header " , ] in figs . 13 and 14 we have plotted the models , together with our three galaxies , in two important diagnostic diagrams : @xmath45 vs. mg@xmath82 , and of @xmath42 vs. mg@xmath82 . in the first place , one sees that for each galaxy there are models which fit the data . in the two diagrams our models seem to fit better than , for example , those by w94 . looking at the two diagrams one can also see that solar metallicities or larger values are required to fit the observational data for the three galaxies , whatever their ages or their imf slopes . this result was inferred by ourselves ( casuso _ et al . _ 1996 ) from an analysis of the mg@xmath82 index alone . comparing models by various authors ( see section 2.6 ) we have seen that @xmath42 is a very difficult color to model , due to uncertainties in the color - temperature and color - color transformations for low - mass stars and in our knowledge of advanced stages of stellar evolution . on the other hand it is also a very sensitive color , due to its large wavelength baseline . in fig . 14 we see that @xmath42 depends considerably on the imf slope @xmath20 . we consider imf slopes of 1.35 and 2.35 and the two shapes of imf proposed in section 2.2 . in this plot we see that low - mass stars are important in determining @xmath42 , since there is a substantial difference between the predictions for the unimodal and bimodal imfs , especially for high imfs slopes and greater ages . therefore the stars with masses below @xmath47 have a non negligible effect , and can make the @xmath42 color redder by @xmath83 for old populations ( see for details section 2.7.1 ) . finally we see that if one includes a small amount of extinction , which reddens the colors but does not change the line indices , the quality of the fits degrades . in paper ii we will discuss the fit to the other colors and absorption lines in more detail . in an exploratory attempt to find an acceptable evolutionary fit to our galaxies we will focus our attention on various very important observables : the colors @xmath42 and @xmath46 , and the mg@xmath82 and h@xmath22 indices . we first run models varying the two main parameters ( @xmath20,@xmath84 ) and also the ages assigned to the galaxies . here we also study the two forms of the imf : the unimodal and the bimodal cases . in all these cases we use an imf slope constant in time in a _ closed - box _ approximation . figs . 15 , 16 and 17 show the model output for the selected observables when we sweep our parameter space . in these plots one can easily see how our evolutionary scheme of stellar population synthesis can not fit the real data . in particular we see that the mg@xmath82 index of the models is always too low for given colors . there is an exception when using extremely steep unimodal imfs for which we can obtain @xmath85 ( independently of the sfr that we use ) but which can not increase as rapidly as the @xmath42 color , due to the coolest very low ms stars . it is didactically instructive to compare the results of our models with the distribution of stars as a function of the metallicity with those resulting from analytical models . in fig . 18 we show the contributions at the present time of the stars as a function of metallicity and age for some of the best fitting ( although not acceptable ) models , inferred from figs . 15 , 16 and 17 . we have also plotted some of the so - called _ simple analytical models _ which assume a closed system , initially all metal - free gas and a constant imf ( see the review of audouze & tinsley 1976 ) . we see that both , the simple models and ours , yield distributions which contain important fractions of low - metallicity stars , explaining the mismatch in the @xmath42 vs. mg@xmath82 . we are clearly dealing with a version of the g - dwarf problem in which we have to account for the absence of low metallicity stars . this is why , with our invariant imf models , we can not obtain higher values for mg@xmath82 . if we choose a low value for @xmath84 to decrease the importance of the first ssps , then we get more or less continuous star formation , and the youngest ssps will be the brightest , and since they are young , their mg@xmath82 will be rather low , even if they are metal - rich . all this is thoroughly discussed in casuso _ ( 1996 ) in which we focused our attention exclusively on the mg@xmath82 index . if one looks at the results presented in ay86 , one sees that the authors have the same problem , and that they can barely find a solution that fits a typical @xmath45 and @xmath42 for a galaxy . they would not have been able to fit its mg@xmath82 index with their model . the previous conclusions are more solid if we look at the h@xmath22 vs. @xmath46 diagram , which can not be fitted either . finally , in figs . 15 and 17 there are a few models , with very low imf slopes , which almost fit the @xmath42 vs. mg@xmath82 diagrams of this set of galaxies . these models however do not fit the corresponding h@xmath22 vs. @xmath46 diagrams ( see fig . -1.5 cm -0.5 cm the fact that the metallicity distribution of our model agrees with the predictions of a simple analytic model does not mean however that the latter are adequate for handling in detail the variations which have in fact occurred in stellar populations , even in less complex systems such as ellipticals . the present model is much more flexible in including the evolutionary histories of stars ( such as mass loss ) and allows us to understand better the star formation history . the fact that the stellar frequency distribution with metallicity can be reproduced with an analytical model does however give some support to our procedure . the key question is now how to build a scenario that produces a dominant old but metal - rich population in our observed galaxies . to tackle this problem , in this paper we introduce a time dependence in the imf slope as one of the ways which are able to yield a metallicity distribution which is skewed toward higher metallicities . another would be to include some gaseous infall , but we considered this to be less lively for ellipticals . in the beginning , for a period of time ( @xmath86 ) of @xmath87 , or even less , we favor the production of massive stars ( using an imf with low slope @xmath88 ) , while later mainly low - mass stars are produced ( this imf slope is termed @xmath20 ) . the first to propose this variable imf scenario was schmidt ( 1963 ) . it also was analyzed by arnaud _ ( 1992 ) , worthey _ et al . _ ( 1992 ) and more recently by elbaz _ ( 1995 ) among others . some authors tackled the problem by keeping the imf slope constant ( with the salpeter index ) , but changing the lower - mass cutoff . in our scenario , at the beginning short - lived massive stars quickly enrich the ism but these stars , at the same time , do not contribute to the light we are observing now . this light comes essentially from stars created later from metal - rich gas and a steeper imf . this scenario is plausible physically because : * it is likely that the central density and temperature were initially , high , which favored high - mass star formation ( see starbursts references ) in these massive galaxies , so that the imf slope could well have been smaller than 1 . * later , conditions tend towards those of the solar neighborhood , so that @xmath20 becomes larger . if we do not want to include a specific treatment for the galactic equilibrium between sn rates and the galactic mass - dependent binding energy to predict when gas densities are able to form stars , we have to introduce a phenomenological parameter , which is the duration of this short period in which massive stars were favored . for this purpose we first plot in fig . 19 the chemical evolution during the first 2 gyr . for flatter imf slopes ( @xmath89 ) for different sfr regimes ( characterized by the @xmath84 coefficients ) . the results here do not depend very much on whether we choose a unimodal or bimodal imf ( in the latter case the metallicities that are reached are slightly higher ) . because we are limited by the isochrones , which are available only for solar and 2.5 times solar metallicities , we will analyze those chemical evolution parameters for which the two metallicities are reached in different periods of time : 0.2 , 0.5 1 and 2 gyr . table 8 summarizes these results . the main difference for two pairs of chemical evolution models which reach the same metallicity level at this initial period is that for lower @xmath88 the contribution of the earliest ssps to the present light is less important . the ( @xmath42)-mg@xmath82 diagram is a useful diagnostic , and helps us to check which solutions are reasonable . however , other diagrams such as @xmath46 vs. h@xmath22 serve the same purpose ( see figs . 15 and 16 or figs . 21 and 22 ) . in figs . 20 , 21 and 22 we plot the diagrams for the models with @xmath84 and @xmath88 which reach a metallicity of 2.5 times solar at @xmath86 following table 8 . focusing our attention on the @xmath42 vs. mg@xmath82 plots we clearly see that with a bimodal imf our data can be fitted better . in these plots we also find that better fits can be obtained for @xmath90 . the reason for that is clear : if we extend the initial period of time we need lower sfr coefficients implying that the most recent stellar populations become relatively more important and therefore the mg@xmath82 decreases . the highest values for the mg@xmath82 index are in fact obtained for @xmath91 . in fig . 23 we detail the different contributions to the present time of some of the most representative fits . we see that the contribution of the stars with metallicities lower than solar , formed during the very short early stage of the galaxy evolution , is almost negligible ( lower than @xmath92 ) . to illustrate this , we have also compared these results with a simple analytic model which yields the same final metallicity as our variable imf model . in the same figure we also plotted an _ analytical infall model _ , in which metal - free gas enters at a rate assumed to equal the stellar birthrate ( larson 1972 , audouze & tinsley 1976 for a review ) . we see that this model yields the same results as our numerical variable imf model . therefore there are two main ways to obtain the required metallicity distributions . we have chosen a variable imf scenario rather than infall in our numerical models because this would seem to be a more reasonable physical assumption for ellipticals and bulges , whereas long - term infall would clearly be a reasonable assumption for the evolution of the disk population such as that of the solar neighborhood . finally , since the mg@xmath82 should decrease if we choose models ( @xmath84,@xmath88 ) which reach solar metallicity at @xmath86 we also took @xmath91 to make mg@xmath82 as high as possible . 24 shows that for these models we can not obtain the observed high values . the sombrero is the only galaxy which seems to be fitted by this level of metallicity for a remaining @xmath20 lower than 1.7 , but this is not a real fit , because with these imf slopes the metallicity continues increasing to reach values higher than z=0.1 and even more , and therefore the resulting observables are not fitted . ngc 3379 seems to require higher metallicity than solar but not as high as 2.5 times solar , while ngc 4472 seems to be well fitted by a value of 2.5 times solar . here we should direct our attention to the recent paper of gibson ( 1996 ) in which he criticizes this variable imf scenario . he constructs a coupled photometric and chemical evolution package based on the predictions for the bimodal star formation / imf scenario of elbaz _ et al . _ ( 1995 ) implementing both the isochrones of w94 and those of bbcfn for single stellar population models . his models produce @xmath42 colors @xmath93 magnitudes too red compared to data from ellipticals . we have shown in this section that we can fit the data very well . in our opinion the reason for the discrepancy found by gibson ( 1996 ) is not due to a variable imf scenario but to the following effects : i ) our method of converting the isochrone parameters to colors gives a bluer @xmath42 , as can be seen for example in fig . 3 for a single stellar population model of 12 gyr and metallicity higher than solar . ii ) for his calculations gibson ( 1996 ) used a ssp while we have followed the chemical evolution during the whole history of the galaxy using our evolutionary model . we have developed a stellar population model to apply to early - type galaxies . it produces optical and near - infrared colors , and absorption line strengths in lines from 4100 - 8800 on the lick system , is applicable to systems of intermediate or old age , and metallicity larger than 0.1 z@xmath94 . the model is chemo - evolutionary , i.e. it calculates the properties of a stellar system , starting from a primordial gas cloud . however , it can easily be used to predict the properties of systems with a single age and metallicity . the model colors and line strengths are determined by integrating stellar observables along theoretical isochrones , in this way obtaining single stellar populations ( ssps ) , and then integrating these ssps over time . the model uses isochrones with solar metal abundance ratios . as far as possible empirical calibrations have been used to convert theoretical isochrone values to observables for individual stars . some properties of the models are : * extensive comparisons with models from other authors show that there is broad overall agreement in the colors and line strengths . we have also made independent estimates of the errors in our observables . * our chemical evolutionary model is in good agreement with arimoto & yoshii ( 1986 ) . we present optical and ir colors which are likely to be improved as a result of new calibration relations . also , for the first time , we give integrated line strengths for an evolutionary model . * we find that for a closed box approximation the most metal rich elliptical galaxies can not be fitted with a single imf that is constant in time . to solve this problem , we propose a scenario invoking an imf skewed towards high - mass stars during a short , initial period ( smaller than 1 gyr ) , followed by preferential low - mass star formation in the remaining time . * we have briefly tested the model here by fitting it to a few key colors and line strength indices , using a new data set for 3 standard early - type galaxies . in general , satisfactory results are reported . a much more comprehensive comparison of theory with observations is given in paper ii . to conclude , we need to make a few statements about the applicability of this model . given the nature of this type studies , the numbers given in the text will soon cease to be the most accurate possible , because better isochrones become available , or better calibrations linking one parameter to another . therefore , we will try to update the model as time goes on , and interested people can always obtain the most recent version electronically from the authors . there are however a few areas in which we think that further effort by the astronomical community is needed to improve models of this kind . these are : * we show that absorption line strengths are generally more accurate than integrated colors . to change this situation better color - color relations are needed , especially for very high and very low metallicities . * inclusion of more absorption line observations will make this kind of models more useful . especially in the near - uv or near - ir very little work has been done , except in the region of the ca ii triplet . more libraries like the one of worthey _ ( 1994 ) are urgently needed . * it looks as if a substantial part of the uncertainties are due to incorrect treatment of the later stages of stellar evolution , such as the agb . theoretical work in this area would be very valuable . we are grateful to franco fagotto for discussions and for his valuable comments and suggestions on the manuscript , which have greatly improved the final version . we also would like to thank the referee , guy worthey , for improving substantially the quality of this paper . we are grateful to the padova group for providing us with the set of theoretical isochrones . we are indebted to onno pols who kindly gave us the calculations for the very low - mass stars , and to angel alonso , santiago arribas and carlos martnez who have made available their observational results before publication . we also thank gustavo bruzual and stphane charlot who provided us with their latest calculations of integrated colors and absorption line indices . we also thank nobuo arimoto , marijn franx and bianca poggianti for useful comments on the manuscript . this work was partially supported by grants pb91 - 0510 and pb94 - 1107 of the spanish dgicyt . thanks the kapteyn institute in groningen for financial assistance during a working visit . alexander , d.r . & ferguson , j.w . , 1994 , , 473 , 879 alongi , m. , bertelli , g. , bressan , a. , chiosi , c. , fagotto , f. , greggio , l. , nasi , e. , 1993 , , 97 , 851 alonso , a. , arribas , s. & martnez - roger , c. , 1994 , , 107 , 365 alonso , a. , arribas , s. & martnez - roger , c. , 1995 , , 297 , 197 alonso , a. , arribas , s. & martnez - roger , c. , 1996 , , in press arimoto , n. , & yoshii , y. , 1986 , , 164 , 260 ( * ay86 * ) arimoto , n. , & yoshii , y. , 1987 , , 173 , 23 arnaud , m. , rothenflug , r. , boulade , o. , vigroux , l. & vangioni - flam , e. , , 254 , 49 audouze , j. & tinsley , b.m . , 1976 , , 14 , 43 bertelli , g. , bressan , a. , chiosi , c. , fagotto , f. & nasi , e. , 1994 , , 106 , 275 ( * bbcfn * ) bessell , m.s . , 1983 , , 95 , 480 bessell , m.s . , 1990 , , 102 , 1181 bessell , m.s . , wood , p. r. , 1984 , , 96 , 247 bessell , m.s . & brett , j.m . , 1988 , , 100 , 1134 bessell , m.s . , brett , j.m . , scholz , m. & wood , p.r . , 1989 , , 77 , 1 bessell , m.s . , brett , j.m . , scholz , m. & wood , p.r . , 1991 , , 89 , 335 bica , e. , 1988 , , 195 , 76 bica , e. , pastoriza , m. , maia , m. , da silva , l. & dottori , h. , 1991 , , 102 , 1702 bica , e. , clara , j.j & dottori , h. , 1992 , , 103 , 1859 bressan , a. , chiosi , c. , fagotto , f. , 1994 , , 94 , 63 bressan , a. , chiosi , c. & tantalo , r. , 1996 , , submitted bruzual , g. , 1983 , , 273 , 205 bruzual , g. , 1992 , in the stellar populations of galaxies , ed . b. barbuy & a. renzini ( dordrecht : kluwer ) , 311 bruzual , g. & charlot , s. , 1993 , , 405 , 538 bruzual , g. & charlot , s. , 1996 , in preparation burstein , d. , faber , s. m. , gaskell , c. m. & krumm , n. , 1984 , , 287 , 586 buser , r. & kurucz , r. , 1978 , 70 , 555 buzzoni , a. , 1989 , , 71 , 817 caldwell , n. , kennicutt , r. & schommer , r. , 1994 , , 108 , 1186 casuso , e. , 1991 , ph . d. thesis , univ . of la laguna casuso , e. , vazdekis , a. , peletier , r. & beckman , j.e . , 1996 , , 458 , 533 charlot , s. & bruzual , g. , 1991 , , 367 , 126 charlot , s. , worthey , g. & bressan , a. , 1996 , , 457 , 625 clayton , d.d . , 1985 , in nucleosynthesis , univ . chicago press , w. d. arnett & truran , eds . , p. 65 clayton , d.d . , 1986 , , 98 , 968 code , a.d . , davis , j. , bless , r.c . , & hanbury brown , r. , 1976 , , 203,417 covino , s. , galletti , s. & pasinetti , l.e . , 1995 , , 303 , 79 daz , a. , terlevich , e. & terlevich , r. , 1989 , , 239 , 325 eggleton , p.p . , 1973 , , 163 , 279 elbaz , d. , arnaud , m & vangioni - flam , e. , 1995 , , 303 , 345 faber , s.m . , 1972 , , 20 , 361 . faber , s.m . , friel , e.d . , burstein , d. , gaskell , c.m . , 1985 , , 57 , 711 fluks , m.a . , plez , b. , th , p.s . , de winter , d. , westerlund , b.e . & steenman , h.c . , 1994 , , 105 , 311 frogel , j.a . , persson , s.e . & cohen , j.g . , 1981 , 246 , 842 gibson , b.k . , 1996 , , submitted girardi , l. , chiosi , c. , bertelli , g. & bressan , a. , 1995 , , 298 , 87 gorgas , j. , faber , s.m . , burstein , d. , gonzalez , j.j . , courteau , s. & prosser , c. , 1993 , , 86 , 153 guiderdoni , b. , rocca - volmerange , b. , 1987 , , 186 , 1 guiderdoni , b. , rocca - volmerange , b. , 1990 , , 227 , 362 guiderdoni , b. , rocca - volmerange , b. , 1991 , , 252 , 435 han , z. , podsiadlowski , p. & eggleton , p.p . , 1994 , , 270 , 121 he s , r. & peletier , r. , , 268 , 539 huebner , w.f . , merts , a.l . , magee , n.h . & argo , m.f . , 1977 , los alamos sci . lab . la-6760-m houdashelt , m.l . , frogel , j.a . & cohen , j.g . , 1992 , 103 , 163 iglesias , c.a . , rogers , f.j . , wilson , b.g . , 1992 , , 397 , 717 johnson , h.l . , 1966 , , 4 , 193 kennicutt , r.c . , 1989 , , 344 , 685 kroupa , p. , tout , c.a . & gilmore , g. , 1993 , , 262 , 545 kurucz , r.l . , 1992 , in the stellar populations of galaxies , ed . b. barbuy & a. renzini ( dordrecht : kluwer ) , 225 lacey , g. , fall , s.m . , 1985 , , 290 , 154 lacey , c.g . , guiderdoni , b. , rocca - volmerange , b. & silk , j. , 1993 , , 402 , 15 larson , r.b . , 1972 , nature phys . , 236 , 7 larson , r.b . & tinsley , b.m . , 1978 , , 219 , 46 maeder , a . , 1992 , , 264 , 105 matteucci , f. , tornambe , a. , 1987 , , 185 , 51 oconnell , r.w . , 1976 , , 206 , 370 oconnell , r.w . , 1980 , , 236 , 430 pagel , b.e.j . , 1989 , in evolutionary phenomena in galaxies , ed . j. beckman & b.e.j . pagel ( cambridge : cambridge univ . press ) , 201 peletier , r.f . , 1989 , d. thesis , univ . of groningen peletier , r.f . , davies , r.l . , illingworth , g.d . , davis , l.e . & cawson , m. , 1990a , , 100 , 1091 peletier , r.f . , valentijn , e.a . & jameson , r.f . , 1990b , , 233 , 62 pickles , a.j . , 1985 , , 296 , 340 pols , o.r . , tout , c.a . , eggleton , p.p . & han , z. , 1995 , , accepted reimers , d. , 1975 , mem . liege , ser . 6 , vol . 8 , 369 . renzini , a. & voli , m. , 1981 , , 94 , 175 renzini , a. & buzzoni , a. , 1986 , in spectral evolution of galaxies , ed . c. chiosi & a. renzini ( dordrecht : reidel ) , 195 ridgway , s.t . , joyce , r.r . , white , n.m . & wing , r.f . , 1980 , , 235 , 126 rieke , g.h . & lebofsky m.j . , 1985 , , 288 , 618 rocca - volmerange , b. , 1989 , , 236 , 7 rocca - volmerange , b. & guiderdoni , b. , 1987 , , 175 , 15 rocca - volmerange , b. & guiderdoni , b. , 1988 , , 75 , 93 rocca - volmerange , b. & guiderdoni , b. , 1990 , , 247 , 166 scalo , j.m . , 1986 , fund . cosmic phys . , 11 , 1 . schmidt , m. , 1959 , , 129 , 243 schmidt , m. , 1963 , , 137 , 758 searle , l. , sargent , w.l.w . & bagnuolo , w.g . , 1973 , , 179 , 427 searle , l. , wilkinson , a. & bagnuolo , w.g . , 1980 , , 239 , 803 spinrad , h. & taylor , b.j . , 1971 , , 22 , 445 stetson , p.b . & harris , w.e . , 1988 , , 96 , 909 tantalo , r. , chiosi , c. bressan , a. & fagotto , f. , 1995 , , submitted terndrup , d.m . , frogel , j.a . & withford , a.e . , 1990 , , 357 , 453 terndrup , d.m . , frogel , j.a . & withford , a.e . , 1991 , 378 , 742 tinsley , b.m . , 1968 , , 151 , 547 tinsley , b.m . , 1972 , , 20 , 383 tinsley , b.m & gunn , j.e.,1976 , , 203 , 52 tinsley , b.m . , 1978a , , 220 , 816 tinsley , b.m . , 1978b , , 222 , 14 tinsley , b.m . , 1980 , fund . cosmic phys . , 5 , 287 theis , c. , burkert , a. , hensler , g. , 1992 , , 265 , 465 tosi , m. 1988 , , 197 , 33 turnrose , b.e . , 1976 , , 210 , 33 twarog , b.a . , 1980 , , 242 , 242 vandenberg , d.a . , 1983 , , 51,29 vandenberg , d.a . , 1992 , , 391 , 685 vazdekis , a. , peletier , r. f. , beckman , j. e. & casuso , e. , 1996 , , submitted ( * paper ii * ) weiss , a. , peletier , r.f . & matteucci , f. , 1995 , , 296 , 73 whitford , a.e . , 1978 , , 226 , 777 worthey , g. , faber , s.m . & gonzalez , j.j . , 1992 , , 398 , 69 worthey , g. , faber , s. , gonzlez , j. & burstein , d. , 1994 , , 94 , 687 worthey , g. , 1994 , , 95 , 107 ( * w94 * ) wu , c.c . , faber , s.m . , gallagher , j.s . , peck , m. & tinsley , b.m . , 1980 , , 237 , 290 yoshii , y. & takahara , f. , 1988 , , 299 , 593

we apply the concept of helicity from classical hydrodynamics to elucidate two problematical issues in cosmic string physics . helicity , the space integral of the scalar product of a velocity - like field with its vorticity field ( curl ) , can be defined for a complex scalar field in analogy with fluids . we dwell on the topological interpretation of helicity as related to the linking of field lines of the vorticity field . earlier works failed to fully implement this interpretation for cosmic strings by missing a term connected with the linking of these lines inside the strings . as a result paradoxical conclusions were drawn : global cosmic string loops may not take on certain simple shapes , and baryon number is not quantized in integers in the presence of local cosmic strings in gauge theory . we show that both paradoxes are removed when internal contributions to helicity are properly taken into account . in particular , quantization of baryon number can be understood within a special case of the glashow - weinberg - salam model if cosmic strings are the unique mechanism for baryosynthesis . in addition , we find a new constraint on the permitted linkages of cosmic strings in a string tangle .

hep-th9908194

cosmic strings are a special case of the vorticity phenomenon , which appears in many forms - in the flow of rivers and streams , in smoke rings , in hurricanes and tornadoes , etc . a special case of vorticity are the _ quantum vortices _ @xcite which include vortices in superfluids , flux tubes in superconductors and cosmic strings . global cosmic strings are analogues to vortices in superfluids while local cosmic strings are analogues to flux tubes in superconductors @xcite . all these vortices are described by a complex wave function or scalar field with a multivalued phase function @xmath0 , such that @xmath1 over a closed contour surrounding the string . as a result different physical quantities get quantized : the circulation of a superfluid vortex , the magnetic flux in a superconductor , the magnetic flux in a local cosmic string . vorticity was first investigated by helmholz and kelvin who set down the foundations for the study of vorticity in hydrodynamics . in this work we try to implement concepts born in the field of classical hydrodynamics to cosmic strings . as we shall see the interaction between these two fields can be very rewarding . turbulent fluid is regarded as full of vortex filaments @xcite . quantum turbulence in superfluids has likewise been defined as a superfluid state featuring a tangle of quantum vortex filaments @xcite . this paper is concerned with helicity of a tangle of cosmic strings . the concept of _ fluid helicity _ was introduced by moffatt @xcite as a useful measure of the degree of linkage of ordinary fluid vortice loops . it is defined as the volume integral of the scalar product of a velocity field @xmath2 and its curl , the vorticity @xmath3 . following this , moffatt @xcite and others @xcite considered helicity integrals in general , i.e. the space integral over the scalar product of a vector field and its curl @xmath4 and found them to have a topological interpretation as the linking of the field lines of the divergence free field @xmath5 . thus the conservation of helicity can be explained for systems in which the field lines may not cross each other , such as vortex lines in an inviscid fluid and the magnetic field lines in a perfectly conducting fluid @xcite . for these flows the field lines are said to be frozen into the flow , and the helicity is a topological invariant mathematically related to the hopf invariant @xcite classifying the nontrivial homotopy classes of maps from @xmath6 to @xmath7 . , which are isomorphic to circles in @xmath6 . the hopf invariant is equivalent to the helicity integral of a vector field who s field lines lie tangent to the preimage circles . in this sense the hopf invariant is a helicity integral term , but the inverse is usually not the case . ] superfluids and cosmic strings are both described by field theories of a complex field @xmath8 which obeys a nonlinear field equation @xcite , e.g. the nonlinear ginzburg - pitaevskii equation @xcite or the higgs equation . cosmic strings and superfluid vortices both occur in the presence of spontaneous symmetry breaking @xcite . a vortex filament or cosmic string is a configuration of @xmath9 which approaches asymptotically the broken symmetry vacuum , and is characterized by a phase @xmath0 which changes by @xmath10 ( @xmath11 ) when one goes once around the filament or string axis . @xmath12 , is termed the _ winding number of the string_. single valuedness of the field @xmath9 on the filament axis requires that @xmath13 along it . the curves along which @xmath13 define the position of the strings . a vortex filament or cosmic string must either be infinite in both directions , close on itself ( cosmic string loop , vortex ring ) , or terminate on a boundary of the system . otherwise the behavior of the phase just beyond the filament s free ends would be ambiguous . we assume throughout that all strings are closed un - knotted loops confined to a finite region . the analogy between vortices in superfluids and global cosmic strings has inspired the construction of a _ helicity _ for global cosmic strings . bekenstein @xcite defined a helicity for global strings , and vachaspati and field @xcite , sato and yahikozawa @xcite and sato @xcite did this for local strings . these works tried to relate helicity for cosmic strings with the topological and geometrical structure of these strings ( such as linking , knotting , writhing and twisting ) . in all these works bizarre physical conclusions were drawn . bekenstein concluded that an isolated unknotted loop of global string is restricted to a plane @xcite i.e. , only planar configurations may exist . this is strange since dynamically there seems to be no constraint on a loop accumulating a distortion continuously in an arbitrary direction . furthermore simulations of the formation and evolution of cosmic strings show the formation of loops which are not planar @xcite . bekenstein also claimed that the contortion of a single knot is quantized in integers . this is also strange because it says a single knot may not wiggle freely about , but is frozen in a configuration of integer contortion . for local cosmic strings there is a relation between the electroweak magnetic helicity of a tangle of strings and the violation of baryon number conservation via what is known as the chiral anomaly @xcite . vachaspati and field , sato and yahikozawa and sato found the helicity to change continuously with the shape of the strings , which implies that baryon number changes continuously too . this clashes with the expectation that baryon number is quantized in integers . we show here that the topological interpretation of helicity as the linking of field lines themselves is not fully incorporated in these works , and hence an important term was left out , which accounts for all the mentioned unacceptable conclusions . helicity arises from internal structure within a string , determined by the twist and writhe of the string , in addition to the external relations between strings , i.e. linking and knotting . the missing term arises from the internal structure of field lines within the strings which we name _ internal helicity_. it was previously believed @xcite that since the strings length is much greater than its width , the internal structure of the string becomes unimportant and physical quantities of interest can be averaged over the string cross - section . however , as we will see this is not always the case . in order to clarify the contribution of internal helicity , the relation of helicity to the topological structure of field lines will be further developed here . field lines of vector fields which are divergence - free do not have endpoints . this property allows us to examine such field structures in terms of the topology of closed curves . the link between helicity and topological invariants of curves was first conjectured by moffatt in 1981 @xcite and then developed by berger and field in 1984 @xcite . in 1992 moffatt and ricca @xcite managed to derive the extra term , arising from the internal structure of magnetic flux tubes , directly from the field equations of motion for a magnetohydrodynamic fluid . this is not possible when calculating the helicity of cosmic strings since their equations are second order non - linear equations , and therefore the calculation is too complicated . however , using formulas describing curves , developed by mathematical biologists investigating the structure of dna @xcite , the internal helicity may be calculated . by adding it to the external helicity we correct the mentioned results . in section [ sec_2 ] we review the concept of _ helicity _ of a solenoidal ( divergence free ) vector field , and dwell on the relation of the helicity to the linking of the field lines of the solenoidal field . in section [ sec_3 ] we review the construction and calculation of the helicity for a tangle of global cosmic string loops , as a function of different geometrical and topological properties of the strings @xcite . next we correct the previously mentioned conclusions of bekenstein by means of the internal helicity . in section [ sec_4 ] we deduce new constraints on the permitted linkages of string loops . we find that the linking possibilities of string loops depends on their winding numbers . in section [ sec_5 ] we review how the electroweak helicity of local cosmic strings in the weinberg - salam model is related to the baryon number , and clarify how baryon number conservation is violated . the baryon number is found to be quantized , in contrast to previous results . again , we show how internal helicity accounts for the new results . the term _ helicity _ is used in particle physics for the scalar product of the momentum and spin of a particle . moffatt @xcite adopted this term to describe the volume integral over the scalar product of a velocity field @xmath2 and its curl , the vorticity @xmath3 @xmath14 which he named the _ helicity of the flow_. this idea stemmed from the essentially kinematical interpretation of the quantity @xmath15 . the streamlines of the flow passing near any representative point @xmath16 in a small volume element @xmath17 are ( locally ) helices about the streamline through @xmath16 , and the contribution @xmath18 to @xmath19 from @xmath17 is positive or negative according as the screw of these helices is right - handed or left - handed . more generally , the concept of helicity describes the volume integral over the scalar product of any general vector and its curl . in particular , the magnetic helicity of magnetohydrodynamic flow @xmath20 had already been shown by woltjer @xcite to be conserved for a perfectly conducting fluid . later moffatt @xcite showed that the _ fluid helicity _ of an ideal barotropic fluid is conserved as well . both these cases share the property that the divergence free field lines ( vortex lines or magnetic field lines ) are frozen into the fluid , and that in consequence knots and linkages of the field lines are inevitably conserved . it was through struggling to understand the physical meaning of this result that moffatt was led to the topological interpretation of helicity in terms of links and knots in convected vector fields generally , and to the characterization of helicity as a _ topological invariant _ for these cases . moreau @xcite observed that conservation of fluid helicity can be deduced from noether s theorem . helicity thus appears to have status comparable to energy , momentum and angular momentum in this sense . as we shall see , unlike these quantities helicity also has a strong topological character . to demonstrate this we review the calculation of helicity for the simple case of two linked flux tubes and show that the helicity counts the gauss linking number of the tubes @xcite . consider two flux tubes linked once ( fig . [ flux ] ) , carrying fluxes @xmath21 and @xmath22 respectively , of the vector field @xmath23 ( the flux is zero everywhere outside the filaments ) . 4.5 cm within each filament , the @xmath24-lines are unlinked curves which close on themselves after just one passage round the filament . the helicity @xmath19 of the two flux tubes has the same form as the pseudo - scalar quantity described in ( [ b_helicity ] ) . does not depend on the gauge of @xmath25 ; for if @xmath26 is replaced by @xmath27 , then @xmath19 is unchanged . ] adopting the coulomb gauge for @xmath26 ( i.e. @xmath28 ) , and imposing the further condition @xmath29 as @xmath30 , @xmath25 is given by the biot - savart law : @xmath31 so that from ( [ b_helicity ] ) , @xmath32 each tube may be built of many infinitely small flux tubes carrying fluxes @xmath33 and @xmath34 . replacing @xmath35 by @xmath36 : allowing for the fact that each flux filament is integrated over twice once as @xmath37 and once as @xmath38 , we find that @xmath39 with @xmath40 which is the gauss formula for the linking of two curves . the sign of @xmath41 depends on the relative orientations of the field in the two filaments ( according to the right hand rule).-lines within either tube on its own are unlinked closed curves . ] it is simple to obtain the total helicity of the two flux tubes . each pair of filaments ( one from each tube ) make a contribution @xmath42 to the total helicity , so that summing over all such pairs , this is now given by @xmath43 helicity has been shown to be an invariant for flows in which the divergence free field lines are frozen into the flow ( woltjer @xcite for a perfectly conducting fluid and moffatt @xcite for ideal barotropic flow ) . combining this knowledge with ( [ link_hell ] ) , the interpretation of the invariant in terms of conservation of linkage of the ( vortex / magnetic ) field lines which are frozen in the fluid is immediate . states of a higgs field containing a tangle of cosmic strings resemble a tangle of quantum vortex filaments which arise in quantum turbulent flow in superfluids . they are both described by a complex wave function @xmath8 and a winding number @xmath12 . the phase @xmath0 changes by @xmath10 when one goes once around the filament or string axes . the curve @xmath13 defines the position of the strings and the passage from @xmath13 to the vacuum value of @xmath44 takes place over a certain lengthscale denoted by @xmath45 . this lengthscale is very small in comparison with the strings lengths and distances between the strings , i.e. the strings are very thin . motivated by this resemblance bekenstein @xcite designed a helicity for cosmic strings in analogy with fluid helicity of a superfluid . we first review the original calculation @xcite of the helicity of a tangle of strings , and then correct it for the left - out contribution of internal helicity . in a superfluid the velocity is proportional to the gradient of the phase of the wave function ; bekenstein defines the vector @xmath46 where @xmath47 is taken to be a function which rises rapidly from its value at a string axis @xmath48 to near its asymptotic value , @xmath49 , over an interval of @xmath50 which corresponds to a distance of order @xmath45 . @xmath51 is a kind of regularised velocity in that the singularity of @xmath52 on the axis is defused by the @xmath53 function ( far from the axis @xmath51 will drop inversely with the distance from it as befits a vortex velocity field ) . we have @xmath54 where @xmath55 is the derivative of @xmath47 with respect to @xmath56 . @xmath55 has a rather sharp peak at a distance @xmath57 from a string axis and the integral of @xmath58 over all @xmath56 is unity . the helicity of the field @xmath51 is defined as @xmath59 unlike ordinary fluid helicity , this one vanishes . this is clear since the vectors @xmath51 and @xmath60 are perpendicular by construction . nevertheless , string loops ( and vortex rings ) can link , and knot , and their linkage is correlated with other topological properties of the strings to be introduced later . throughout it is assumed that linked and knotted loops can occur over some periods with no intersections . situations where there are intersecting strings are viewed as representing singular moments , not generic ones . ( simulations of the development of superfluid vortices @xcite and global cosmic strings @xcite confirm this assumption ) . bekenstein develops from the zero helicity a quantitative relation between the linkage and other properties of the loops . in order to evaluate the helicity explicitly for a tangle of cosmic string loops , it is convenient to rewrite ( [ h ] ) as a two - point functional of @xmath51 ( this is valid if the loops are confined to a finite region @xcite ) @xmath61 = \int d^3r \int d^3r'\ , \frac{(\nabla \times { \bf v } ) \times(\nabla \times { \bf v } ) ' \cdot ( { \bf r}-{\bf r}')}{4\pi|{\bf r}-{\bf r}'|^3}.\ ] ] we follow bekenstein in the following discussion of the reduction of @xmath62 $ ] , and the conclusions that may be drawn from it . for field configurations of size large compared to @xmath45 , and where all the string axis are always well apart on scale @xmath45 , @xmath19 receives a contribution only when each of @xmath63 and @xmath64 lies very close to a string axis . this stems from the fact that the @xmath58 function in ( [ curl_v ] ) confines the integrals to very near the axis . hence , bekenstein assumes @xmath19 is equivalent to a double line integral taken along the string axis . in addition the cross section of each loop is very little distorted by the presence of the other loops and nearby portions of itself and therefore is assumed to possess a line symmetry . under these assumptions and a few manipulations bekenstein concludes that @xmath65 for the contribution of a little piece of string , where @xmath66 here is the vectorial line element along the piece of string considered . inserting ( [ c_flux ] ) into ( [ helic ] ) we obtain a double contribution to helicity from each pair of loops , because @xmath66 may progress through one loop ( call it @xmath67 ) while @xmath68 does so through the other ( @xmath69 ) and vice - versa . in addition we obtain a contribution from each string loop alone so that @xmath70 = \frac{1}{2}\sum_{i \neq j } 2{\cal l}_{ij}(2\pi|n_i|)(2\pi|n_j| ) + \sum_i { \cal w}_i(2\pi|n_i|)^2.\ ] ] where @xmath71 is the _ gauss linking number _ ( [ gauss_link ] ) of the two loops @xmath67 and @xmath69 and @xmath72 is the _ writhing _ number of loop @xmath73 ( fuller @xcite ) . ) differs from the original one in @xcite by a factor of @xmath74 which was missed in the previous calculation . ] while @xmath75 is a topological invariant of the linking of the string loops and must be an integer , @xmath76 is not a topological invariant . it changes with the shape of the loop and can obtain any value . @xmath76 measures the deviation of a loop from its plane ; the writhe of a planar loop is zero . when all loops are confined to a finite region , @xmath77 , and ( [ hel_eq ] ) gives @xmath78 we thus obtain a constraint connecting the linking of a collection of loops with their geometrical shapes . particularly simple is the case of an isolated loop . then @xmath79 for such a loop we must have @xmath80 too . according to this result _ an isolated un - knotted loop must be a planar loop at any instant _ this is hard to accept since dynamically there seems to be no constraint on a loop accumulating a distortion continuously in an arbitrary direction . furthermore simulations of the formation and evolution of cosmic strings @xcite show the formation of loops which are not planar ! similarly , according to ( [ const ] ) two loops linked once may never be planar since the contribution of their writhe to the helicity must cancel out the contribution from their linkage . this restriction is again hard to accept . to obtain another constraint bekenstein introduces the relation @xcite @xmath81 @xmath82 measures the knotting of the curve and can jump by increments of @xmath83 , and @xmath84 is the _ contortion _ of the loop defined by @xmath85 which integrates the torsion over the whole loop.the axis of a string can be treated as a curve in differential geometry . a curve @xmath86 ( @xmath87 is the arc length ) has @xmath88 as unit tangent vector , has a unit normal @xmath12 , and a unit binormal @xmath89 . they are connected by the frenet - serret equations @xmath90 where @xmath91 , a positive scalar , is called the curvature of the curve , and @xmath92 , which may take on either sign , is called the torsion . ] inserting @xmath93 from ( [ knotting ] ) into ( [ const ] ) gives the following constraint connecting the contortion , knotting and linking of a collection of loops : @xmath94 this relation says that the sum of contortions of a collection of unit winding number loops is quantized in integers @xcite . this implies that the contortion of any single knot is quantized in integers : the knot can not wiggle freely about , but is frozen in a configuration of integer contortion . this is a very strong constraint on the dynamics of such a configuration , and together with our previous observations , makes us suspect that the calculation has missed a term . the first clue of the missing term was presented by moffatt @xcite , who notes that close inspection of an expression of the form @xmath95 integrated over a single flux loop @xmath67 ( of an arbitrary divergence - free vector field ) , shows that the helicity in this limit has an additional contribution from pairs of points @xmath63 , @xmath64 separated by a distance comparable to the cross - sectional span of the loop . in addition to the writhe @xmath93 ( [ writh ] ) , which we introduced earlier as the sole contribution to helicity from a single loop , moffatt conjectures that the close points contribute an extra term namely @xmath96 where @xmath97 is the contortion of loop @xmath73 defined earlier , and @xmath98 is the flux of the loop . actually it is easy to see that bekenstein s calculation in sec . [ global_helicity ] ignores the thickness of the string , i.e. the contribution from close pairs of points @xmath63 and @xmath64 . equation ( [ c_flux ] ) reduces a volume integral to a line integral along the core of the string and thus treats the string as though it where infinitely thin . moffatt speculates that the additional contribution , from close pairs of points @xmath63 and @xmath64 , is related to the contortion so that the helicity will be a topological invariant of the loop , i.e. , @xmath99 by ( [ knotting ] ) . but , thinking this over we realize that the contortion is also just a property of the flux tube s ( string s ) axis and not of its internal structure hence we should expect a different term to appear . later moffatt @xcite explicitly calculated the helicity of a magnetohydrodynamic system of flux tubes to find that his earlier speculation was not exact , and that @xmath82 is not a topological invariant after all since its value jumps discontinuously at inflexion points ( points of zero curvature ) . in this later calculation moffatt manipulates the magnetohydrodynamic equation of motion of a configuration of field lines moving with the flow . this calculation is much too complicated to carry over to the case of cosmic strings since the field equations for these are non - linear . hence we will follow in the footsteps of berger and field @xcite who also calculate the internal helicity due to the internal structure of a flux tube , but we will present a slightly different calculation , and a slightly different result . berger and field s result was that @xmath100 where @xmath101 is any real number , while we obtain the same result as moffatt @xcite with @xmath102 and @xmath103 , an integer , is the linking number of the field lines in the flux tube . we first calculate exactly the contribution to the helicity of a single vortex loop @xmath104 . first we consider the inner structure of the string to consist of nested toroidal vorticity surfaces ( a peace of the loop is shown in fig . [ f2 ] ) . then we may build up the string by increments of vorticity flux @xmath105 composed of many thin flux tubes @xmath106 that wind around the inner flux tube @xmath98 @xmath103 times . the increment in helicity for each such flux tube @xmath106 is @xmath107 where @xmath103 is the gauss linking number of element @xmath108 with @xmath98 ( the inner flux tube ) . the increment in helicity for the whole of @xmath105 is simply @xmath109 . ( the flux @xmath105 simply counts the number of field lines intersecting the increment in the cross section of the string , and hence the number of field lines wrapped around the flux tube in the center . ) we thus have according to ( [ helic ] ) @xmath110 where @xmath111 is along the axis of the string and @xmath112 is along one of the field lines in @xmath106 wrapping @xmath98 , and vice - versa . finally , integrating over all the nested flux tubes @xmath113 where @xmath12 is the winding number of the cosmic string ( the flux @xmath98 is easily calculated using the definition of @xmath51 , stoke s theorem and ( [ q_phase ] ) ) . throughout this calculation we have assumed that @xmath106 and @xmath98 have the same linking number as a single field line in @xmath114 with the axis of the string . we have also assumed that all field lines have the same linking number with the axis of the string ( and with each other ) . this is plausible for a continuous field . as we see , the internal helicity is due to the linking of the @xmath60 field lines inside the string . this linking occurs when the field @xmath60 is twisted : if the field lines do not link the internal helicity is zero . we would now like to express the linking number of the field lines @xmath103 as a sum of @xmath93 and some other term or terms , so that we may reconcile this calculation with bekenstein s result ( [ hel_eq ] ) . the linking number @xmath103 of @xmath108 with @xmath98 is equal to the linking number of a single field line in @xmath108 with the axis of the string . since these are two closed curves at a very short distance apart , the axis of the string can be treated as a curve in differential geometry , forming the edge of a ribbon whose other edge is traced by the field line . this type of problem appears first to have been addressed by clugreanu @xcite , who considered two neighboring closed curves @xmath115 , @xmath116 forming the boundaries of a ( possibly knotted ) ribbon of small spanwise width @xmath117 , and showed that for sufficiently small @xmath117 the linking number @xmath103 of @xmath115 and @xmath116 can be expressed in the form @xmath118 where @xmath93 and @xmath84 are respectively the writhe and the contortion of @xmath115 , and @xmath119 is an integer representing the number of rotations of the unit spanwise vector @xmath120 on the ribbon relative to the frenet pair ( the unit normal and unit binormal . ] ) in one passage round @xmath115 . ( the contortion @xmath84 actually measures the rotation round the curve of the frenet pair itself . ) 8.5 cm 2.5 cm the frenet pair has discontinuous behavior in going through a point of inflexion ( zero curvature ) . as a result @xmath84 and @xmath119 are not well defined if @xmath115 has inflexion points . if @xmath115 is continuously deformed through an inflexional configuration , then @xmath84 is discontinuous by @xmath121 , but @xmath119 is simultaneously discontinuous by an equal and opposite amount @xmath122 , so that the sum @xmath123 does vary continuously through inflexion points @xcite . this sum is known as the _ twist _ of the ribbon @xcite @xmath124 thus , as we remarked earlier , @xmath125 is not a topological invariant while @xmath103 ( [ calig ] ) is . with the help of ( [ calig ] ) and ( [ twist ] ) we are now ready to complete the calculation of the helicity for a single ( un- knotted ) cosmic string loop @xmath104 . we take @xmath115 above to be the string s axis and @xmath116 to be along one of the field lines twisted around the axis . then for the @xmath73-th loop @xmath126 our final expression for the helicity of a tangle of cosmic string loops is therefore @xmath127= \sum_{i \neq j}{\cal l}_{ij}(2\pi|n_i|)(2\pi|n_j| ) + \sum_i{\cal w}_i(2\pi n_i)^2 + { \cal t}_i(2\pi n_i)^2.\ ] ] contrast this ( correct ) result with ( [ hel_eq ] ) . the twist @xmath128 is the contribution from the close pairs of points @xmath63 and @xmath129 inside the string ; it is not a property of the axes of the string alone , but rather of the internal structure of the field lines inside the string core . in the previous sections we learned that helicity measures the linkage of the field lines of @xmath130 , and that the total helicity of global cosmic strings is zero whatever their configuration may be . these properties must put some constraint on the possible configurations of the strings . in this section we show that the knotting and external linkage of a cosmic string determines the internal structure of the field lines inside the string , and thus external and internal helicity cancel each other out so that the total helicity is zero . since the field lines in a string must close on themselves , and their twist depends on external knotting and linking , we will get constraints on the possible configurations of string loops . these constraints are weaker than those proposed by bekenstein , and are free from the latter s paradoxical aspects . let us start by examining @xmath52 ( the gradient of the phase ) of the most simple configuration , a single loop . as we know the phase varies when one goes once around the string by @xmath10 , where @xmath12 is the winding number of the string . 2.5 cm what happens if the phase varies along the string too ? because of periodicity this is only possible if @xmath131 where @xmath66 is along the string axis . but in this case there will be another topological requirement . because the phase varies by @xmath132 around the perimeter of the loop , for every surface spanning the loop there must be at least one place where the amplitude @xmath13 so that the field @xmath9 will be single - valued . this is actually the requirement that the loop be threaded by another @xmath133 unit winding - strings or a single string with winding number @xmath133 . hence an isolated loop must have @xmath134 ; for a totally symmetric loop @xmath52 will have only one component , in the direction around the string . along the loop must be zero for an isolated loop . however , any configuration where the average of the component of @xmath52 along the loop is zero can be continuously deformed into a configuration where the component of @xmath52 along the loop equals zero at every point . hence , as we shall see the linking of the @xmath130 field lines is not affected . ] how does this affect the configuration of the field lines of @xmath130 ? we recall that by definition ( [ vec_v ] ) , @xmath2 is parallel to @xmath52 and @xmath130 ( [ curl_v ] ) is perpendicular to @xmath2 . hence if @xmath2 has a component only around the string , @xmath130 will have a component only along the string , and its field lines will not be twisted or linked . since linkage of the field lines is a topological property , it will not change for continuous deformation of the loop ; therefore the field lines of any isolated loop with arbitrary shape will be unlinked . the immediate result is that _ the total helicity of any isolated loop with arbitrary shape is zero_. this corrects bekenstein s conclusion ( [ const ] ) that an isolated loop must be a planar loop at any instant so that the helicity will be zero . we next investigate a configuration of two linked loops @xmath135 and @xmath136 with winding numbers @xmath137 and @xmath138 respectively . we now have a component of @xmath52 in the direction along the strings too . the phase will vary around the perimeter of loop @xmath135 by @xmath139 and around the perimeter of @xmath136 by @xmath140 ( fig . now that @xmath141 has components along the string and around it , we must introduce a new concept , the _ chirality _ ( handedness ) of a string loop . a loop is right - handed ( and has positive chirality ) when the component of @xmath52 along the string is in the direction of the thumb of the right hand , whose fingers point along the component of @xmath52 around the string . if the component of @xmath52 along the string is in an opposite direction to the thumb , the loop is left - handed ( and the chirality is negative ) . 8.5 cm 3.5 cm along with the definition of chirality comes the definition of the sign of the directions along and around the string . we define the positive direction around the string to coincide with the direction of @xmath52 around it , and the positive direction along the string to coincide with the direction of the thumb of the right hand whose fingers follow the direction of @xmath52 around the string axis . in a right - handed ( left - handed ) string loop whose phase changes around the string by @xmath10 and along the string by @xmath142 , @xmath12 and @xmath133 will have equal ( opposite ) signs . it is easy to see that _ two strings may link only if they have the same chirality _ ; otherwise there would be a discontinuity in the phase . what will be the structure of the @xmath130 field lines in such a configuration ? now that @xmath52 has components in two directions , @xmath130 will also have at least two components , one along the string and one around it . the field lines will be twisted and linked ( fig . [ f6 ] ) . in such a case we expect to have two contributions to helicity , a contribution from the external linkage of the string loops , and a contribution from the internal linkage of the field lines inside the strings . these two contributions must cancel each other so that this configuration may exist . the external helicity for such a configuration with @xmath143 and winding numbers @xmath137 and @xmath138 is simply @xmath144 where the sign of @xmath41 is determined by the relative direction of @xmath130 along the perimeter of both loops . we will show below that @xmath41 is positive for right - handed configurations and negative for left - handed ones . ) is always positive by our current definition , there is no need to take the absolute of the winding numbers as in ( [ twist_hel ] ) . ] but computing the internal helicity @xmath145 is not straightforward . the internal helicity of magnetic flux tubes in a magnetohydrodynamic fluid or of vortices in an ordinary fluid is arbitrary since there is no restriction on the linking of the field lines inside the flux tube . in those systems the internal structure of the field lines is not determined by the topology of the flux tubes or vortex filaments and , therefore , the total helicity can not be determined exactly from their configuration . by contrast we shall now show that the internal helicity of cosmic strings is solely determined by the winding numbers of the strings and their topological configuration , so that the total helicity for a configuration of linked vortices can be expressed as a function of their linkage and winding numbers alone . 3.6 cm we now attempt to derive the internal linking number @xmath103 , that measures the linkage of the field lines inside a cosmic string loop . the loop is characterized by winding numbers @xmath12 and @xmath133 around and along the perimeter of the loop respectively . any ( un - knotted ) cosmic string loop may be deformed continuously into a planar circle loop , and the twist of the field lines inside it uniformly distributed , without changing the linking of the field lines . this is possible since the linking is a topological property of the field lines , and does not change by continuous deformations . it would further simplify our calculation if we could perform it in cylindrical symmetric coordinates , by virtually cutting the loop and standing it upright , such that it acquires cylindrical symmetry ( fig . [ f7 ] ) . we may convince ourselves that this is possible since very close to the axis of the string the field lines have cylindrical symmetry around the string , and for reasons of continuity the linkage of the field lines can not change as we move away from the string . the linking number @xmath103 is simply the number of times @xmath146 is wrapped around the @xmath147 axis ( which coincides with the axis of the string ) . this must be an integer for the string to form a loop . the vectors @xmath2 and @xmath130 are computed in cylindrical symmetric coordinates @xmath148 as follows : first @xmath149 where @xmath47 in ( [ vec_v ] ) is taken to be @xmath150 , a function of @xmath37 only . the gradient of @xmath0 must obey @xmath151 where @xmath152 is the length of the string . hence for uniformly distributed twist @xmath153 the calculation of @xmath130 gives @xmath154 we are interested in the number of times each field line encircles the @xmath147 axis . often we associate flow lines with vector fields by imagining how a test particle would move if the vector field were its velocity field . let us assume the field line traces out the trajectory of a point particle traveling with constant velocity during a period of time @xmath155 . the velocity components of the particle are equal to the components of @xmath130 . the following simple equations must apply so that the field lines may reconnect when the string forms a loop @xmath156 by eliminating @xmath155 between the two equations and using @xmath130 from ( [ sym_curl_v ] ) we obtain the result @xmath157 thus the internal linking number @xmath103 depends exclusively on the winding numbers of the loop . to evaluate the total helicity for two linked un - knotted cosmic string loops , we first determine the sign of the gauss linking number @xmath41 for left - handed and right - handed configurations , which depends on the direction of @xmath158 along the perimeter of both loops . by our definition above , the winding numbers around the axis of the string are always positive , since we have defined the positive direction around the string to coincide with the direction of @xmath52 around the axis of the string . hence from ( [ sym_curl_v ] ) the component of @xmath130 along the axis is always in the positive direction i.e. along the thumb of the right hand whose fingers follow the positive direction around the axis of the string . [ f8 ] makes it clear that a right - handed configuration will have @xmath159 and the left - handed configuration will have @xmath160 . the sign of the internal linking number @xmath103 also depends on whether the configuration is left - handed or right - handed . as we explained earlier , in a right - handed configuration @xmath12 and @xmath133 have equal signs while in a left - handed configuration , they have opposite signs . therefore , by ( [ int_link ] ) @xmath103 is negative for a right - handed pair and positive for a left - handed one . _ the signs of the external and internal linkages are opposite_. we are now able to write down the total helicity for a pair of linked strings . for two right - handed linked strings with winding numbers @xmath137 and @xmath138 , we have @xmath161 and @xmath162 ( since the axis of each loop encircles the other loop ) and by ( [ h_ext ] ) and ( [ h_int ] ) the total helicity is @xmath163 for two left - handed linked loops the total helicity is @xmath164 the external and internal helicities thus cancel each other out without constraining the shapes of the loops . we have thus removed the constraint on the geometrical shape of a single loop , a pair of linked loops or any other configuration . the contributions to helicity are purely topological . but ( [ int_link ] ) introduces a new topological constraint . the internal gauss linking number @xmath103 of the field lines in the string must be an integer . hence the winding number @xmath133 along the perimeter of the string is not free to take on just any value , i.e. , a string loop with winding number @xmath12 can not be linked arbitrarily with other loops . for the right handed pair of linked loops @xmath135 and @xmath136 , for example , the internal linking number of @xmath135 is @xmath165 and the internal linking number of @xmath136 is @xmath166 . therefore we must have @xmath167 so that both @xmath168 and @xmath169 be integers . this means that _ only string loops with the same winding number may link once_. in the next section we generalize this result to a generic configuration of linked un - knotted loops . examining a couple more examples will help us formulate a more general result . 3.5 cm fig . [ f9 ] depicts a right - handed cosmic string loop @xmath135 with winding number @xmath137 linked twice with a cosmic string loop @xmath136 with winding number @xmath138 such that their gauss linking number is @xmath170 . as explained earlier the change in phase along the perimeter of a loop is determined by the winding numbers of the strings threading it . since @xmath136 links @xmath135 twice the change in phase along @xmath135 is @xmath171 so that @xmath172 ; correspondingly the phase along @xmath136 changes by @xmath173 so that @xmath174 . hence the internal linking number of @xmath135 is @xmath175 and that of @xmath136 is @xmath176 . thus the values of @xmath138 must be restricted to the integers that divide @xmath177 and are greater than or equal to @xmath178 . for a few examples see table [ tab1 ] , only such loops may link twice . the total helicity for this configuration is @xmath179 @xmath180 our second example in fig . [ f10 ] depicts a left - handed string loop with winding number @xmath137 linked once with two other left - handed string loops , one with winding number @xmath138 and the other with winding number @xmath181 . 3.5 cm the external gauss linking numbers are @xmath182 and @xmath183 and @xmath184 . the change in phase around the perimeter of the strings , and the internal linking numbers are determined by the same arguments as above , such that @xmath185 , @xmath186 and @xmath187 . the total helicity will thus be @xmath188 as we would expect . by generalizing from these examples we see that the internal linking number @xmath189 for a loop @xmath67 linked with an arbitrary number of other loops is @xmath190 where @xmath191 is equal to + 1 for right handed loops and -1 for left handed ones . this equation is also a constraint on the winding numbers and the gauss linking numbers of the loops linking @xmath67 , since @xmath189 must come out an integer . the total helicity of loops linked in an arbitrary manner thus vanishes : @xmath192 local cosmic strings are solutions to lagrangians invariant under local gauge transformations . the simplest model for a local cosmic string is the abelian higgs model ; it has @xmath193 local symmetry . nielsen and olesen @xcite found straight string ( cylindrically symmetric ) solutions to this model characterized by a magnetic field confined to the string s core whose flux is quantized in proportion to the winding number @xmath12 of the string . another important property of these strings is that they have finite energy per unit length ( in contrast to the global strings ) since the gauge field @xmath194 can cancel out the gradient of the phase at spatial infinity . the nielsen - olesen string is actually a two dimensional vortex solution with energy density localized around a point in space , which can trivially be embedded into the extra dimension to produce an infinite string - like object whose energy density is localized along a line . by embedding the vortex along a closed curve , rather than a straight line , the ends of the string can be joined to produce a closed string of finite length . thus we may have a configuration that is a tangle of such local cosmic strings which are both knotted and linked with each other . since these strings have quantized magnetic flux tubes in their cores , we actually have a tangle of knotted and linked magnetic flux tubes , and each configuration may be characterized by a magnetic helicity quantifying the linking of the cosmic strings , and the internal linkage of the magnetic field lines inside each string . in this section we refer to the result @xcite that the change in the electroweak magnetic helicity of a tangle of local cosmic strings , in the standard model , is related to baryon number violation . previous works indicated a continuous change in baryon number . in contradiction to these works , we show here , by using our knowledge of the topological interpretation of helicity as the linking of field lines that , _ baryon number on local strings is quantized _ in this model . sato and yahikozawa @xcite scrutinize the geometrical and topological properties of the abelian higgs model with vortex strings by calculating the topological term @xmath196 @xmath197 which originates from the chiral anomaly , and is the cp violating @xmath0 term in the action . the definition of the _ dual _ is the following : @xmath198 . @xmath196 is evaluated for cases where all vortex strings are closed , with winding number @xmath199 , and the number of the vortex strings is conserved i.e. vortex string reconnection does not occur . using a _ topological formulation _ which they developed , and via a complicated calculation , sato and yahikozawa show that @xmath200_{t = -\infty}^{t = + \infty}.\ ] ] the first term is the linking number between each pair of strings and comes from the _ mutual interactions _ between different vortex strings . the second term is the writhing number ; it comes from the _ self - interactions _ of each vortex string , and can vary with the shape of the string . it is plain that the value of @xmath196 changes continuously with the shape of the strings as time elapses . equation ( [ sato_res ] ) strongly resembles bekensteins result ( [ hel_eq ] ) for the helicity of global cosmic strings . we shall now show that @xmath196 is just the total change in time of the magnetic helicity of the abelian higgs model with local vortex strings , and hence equal to the change in the external and internal linking numbers of the magnetic field lines each weighed by the square of the quantized flux in the corresponding vortex . by construction @xmath201 , so we may write @xmath202 using maxwell s equations @xmath203 and @xmath204 and integration by parts , @xmath205 where we have used the relation @xmath206and the assumption that all fields die off at spatial infinity . our final result is thus @xmath207 where @xmath19 stands for the magnetic helicity . as we already know , the magnetic helicity may be written as the sum of the external linkages of the flux tubes and the internal linkage of the field lines in each flux tube , times the square of the flux ( [ twist_hel ] ) @xmath208 the two terms @xmath209 add to @xmath189 , the internal linking number ( [ n_hel ] ) , and the magnetic flux in a local string of unit winding number in this model is @xmath210 . as @xmath211 . hence , @xmath212 and via stoke s theorem and ( [ q_phase ] ) the magnetic flux is @xmath213 . ] therefore , according to ( [ i_rel ] ) and ( [ mag_hel ] ) the term @xmath196 is simply equal to @xmath214_{t = -\infty}^{t = + \infty}.\ ] ] comparing this with ( [ sato_res ] ) we see that sato and yahikozawa have missed out the twist term . this resulted from neglecting the internal structure of the vortex strings in their calculation . our result is crucially different from theirs : according to their result the term @xmath196 may have a continuous range of values , while according to our result @xmath196 is quantized because @xmath215 is quantized by the internal and external linking numbers of the magnetic flux tubes threading the strings cores . this fact has an important physical consequence , as we shall see presently . the origin of the excess of matter over antimatter in our universe remains one of the fundamental problems . a widely discussed model @xcite is baryogenesis in the course of the broken symmetry electroweak transition , where the baryonic asymmetry is induced by the quantum chiral adler - bell - jackiw anomaly @xcite . in models of fermions coupled to gauge fields , certain current - conservation laws are violated by this anomaly . the term @xmath216 calculated on the background of the nielsen - olesen strings has been shown to cause violation of baryon number conservation via this anomaly . the anomaly arises when a classical symmetry of the lagrangian does not survive the process of quantization and regularization : the symmetry of the lagrangian is not inherited by the effective action . an axial current that is conserved at the level of the classical equations of motion can thus acquire a nonzero divergence through one - loop diagrams that couple this current to a pair of gauge boson fields . the feynman diagram that contains this anomalous contribution is a triangle diagram with the axial current and the two gauge currents at its vertices . left and right handed fermions contribute terms with opposite signs to the anomaly ; however , in chiral theories in which the gauge bosons do not couple equally to right- and left - handed species , the sum of these terms can give a non - zero contribution . in theories such as qed or qcd in which the gauge bosons couple equally to right- and left - handed fermions , the anomalies automatically cancel . within the standard model of weak interactions ( the glashow - weinberg - salam @xmath217 invariant theory ) , the requirement from experiment that weak interaction currents are left - handed forces us to choose a chiral gauge coupling . the coupling of the @xmath218 boson to quarks and leptons can be derived by assigning the left - handed components of quarks and leptons to doublets of an @xmath219 gauge symmetry , and then identifying the @xmath218 bosons as gauge fields that couple to this @xmath219 group , while making the right - handed fermions singlets under this group . the restriction of the symmetry to left - handed fields leads to the helicity structure of the weak interactions effective lagrangian , and thus causes the chiral anomaly of the baryon current to arise . the baryon number current in this model is @xmath220 and the anomalous baryonic current conservation equation is @xmath221 where @xmath222 and @xmath223 are the field strengths for the three @xmath219 and one @xmath193 gauge fields @xmath224 ( @xmath225 ) and @xmath226 respectively , and @xmath227 and @xmath228 are their charges . @xmath229 is the number of quark families . the baryon number current can be integrated to yield the change in the baryon number between two different times , if we assume that the baryonic flux through the surface of the three - volume of interest vanishes : @xmath230 equations ( [ current_anomaly ] ) and ( [ delta_charge ] ) relate the change in the baryon number to terms of the form @xmath231 which we calculated in the previous section for an abelian model on the background of nielsen - olesen vortices . this term was found to be related to the change in the sum of the linking of the strings from their initial to final configurations ( [ fin_i ] ) . this leads us to an exciting idea @xcite : if chiral fermions are coupled to electroweak vortices , baryon number may be violated because of the anomaly as vortices link and de - link . for this to happen there must be a stable string solution in this model , and the change in the baryon number must be related to the helicity of an electroweak field whose flux is confined to the string . it is generally believed that the standard electroweak model is free from topological defects . the reason is that the first homotopy group of @xmath232 can be shown to be trivial i.e. @xmath233 @xcite . however , this does not mean that the model is free from non - topological defects . it has been shown by manton @xcite that the configuration space of classical bosonic weinberg - salam theory has a non - contractible loop . hence an unstable , static , finite - energy solution of the field equations can exist . vachaspati @xcite showed that exact vortex solutions exist , which are stable to small perturbations for large values of the weinberg angle @xmath234 and small values of the higgs boson mass . this expectation has been confirmed by the numerical results of james , perivolaropoulos and vachaspati @xcite . defining @xmath235 where @xmath236 , the static vortex solution that extremizes the energy functional was found to be : @xmath237 and @xmath238 the subscript @xmath239 on the functions @xmath240 and @xmath194 means that they are identical to the corresponding functions found by nielsen and olesen for the usual abelian - higgs string . in other words , the weinberg - salam theory has a vortex solution which is simply the @xmath239 vortex of the abelian higgs model embedded in the non - abelian theory . substituting this solution into ( [ current_anomaly ] ) we have @xmath241 where @xmath242 and @xmath243 . the change in the baryon number is now @xmath244 via equations ( [ delta_charge ] ) , ( [ z_current ] ) and ( [ i_rel ] ) and with @xmath245 playing the role of baryon number charge . we see that the change in the baryon number is proportional to the change in the @xmath246 field s helicity . for a configuration of tangled strings with arbitrary winding numbers @xmath247 @xmath248 where the flux of each vortex is @xmath249 . ] and hence the factor of @xmath250 for the flux . ] the final result for a configuration of strings with unit winding number is @xmath251 and baryon number is quantized . earlier works failed to fully implement the topological interpretation of helicity as the measure of the linkage of field lines of the divergence free field . since field lines may be twisted and linked inside a vortex core , the inner structure of the field inside the core may not be ignored when calculating helicity terms . neglecting internal helicity had led to peculiar results : unacceptable constraints on string configurations and continuous violation of baryon number . once we add the contribution from the internal structure of the strings , we find new constraints on the configurations of linked global cosmic strings , which are physically more pleasing , and we also find that , the baryon number is quantized as we would expect . our work is another demonstration of the fact that helicity counts the linkage of the divergence free field lines . we explain how there is no contradiction between the fact that helicity counts the linkage of field lines , and that the total helicity of any configuration of knotted linked loops always manages to vanish . the vector @xmath252 ( [ vec_v ] ) constructed for the purpose of obtaining a helicity term for global cosmic strings is proportional to @xmath141 . hence it is perpendicular to @xmath253 ( [ curl_v ] ) and the helicity must vanish . on the other hand the helicity counts the linkage of field lines of @xmath253 , and these lines must have external and internal linkages for linked configurations of cosmic string loops . what we have discovered is that the field lines arrange themselves in such a manner that the external and internal helicities exactly cancel each other . the field lines inside a string may not link in an arbitrary manner , their linkage is constrained by the topological configuration of the string and by its linking relations to other strings . further , we have discovered a constraint on the permitted configurations of un - knotted linked loops , in the form of ( [ gen_int_link ] ) which must produce integer values . this constraint results from the combination of the topological character of helicity integrals with the topological nature of cosmic strings as topological defects . in this work we confined ourselves to the problem of un - knotted linked loops . the problem of knotted configurations is far more complicated since it is hard to separate the external contribution to helicity ( depending on the topology of the knot ) from the internal contribution depending on the twist of the field lines inside the knot . moffatt @xcite proposed a method for calculating the external helicity , but it turns out that for some knotted configurations this method is ambiguous ( it may be that it works only for _ chiral _ knots , i.e. knots that are not isotopic to their mirror images ) . hence , we leave the treatment of knots to later work . for local cosmic strings in the standard model , the change in the helicity of the electroweak magnetic field confined to the strings core is related to the change in baryon number over a period of time ( [ z_helic ] ) . thus , baryon number conservation is violated as cosmic strings link and de - link . the idea that violation of baryon number may be related to the change in the linking of electroweak strings has been developed by cohen and manohar @xcite , vachaspati and field @xcite and garriga and vachapati @xcite . however , they neglected the contribution of the writhe of the strings to helicity . sato @xcite adds the writhe term , but misses out the twist term . this led to his conclusion that baryon number conservation is violated as vortices change their shape , which implies that baryon number may change continuously and , therefore , is not quantized . only by including both the writhe and the twist terms , which together give the internal linkage of the field lines , do we recover a quantized baryon number ! charge quantization has previously been related to flux quantization . it was suggested by dirac in 1931 @xcite to explain the quantization of the electric charge . this idea was developed further by jehle @xcite . in general , quantization of charges is not a direct consequence of standard ( non - gut ) field theories . particularly , the quantization of electric charge has no widely accepted explanation within the standard model . during the last decade it has been suggested @xcite that the electric charges can be heavily constrained within the framework of the standard model . this is achieved partly by constraints related to the classical structure of the theory ( such as the requirement that the lagrangian be gauge - invariant ) , and from the cancelation of anomalies , at the quantum level . the novel idea in this work is that quantization of the baryon number may arise from the topological nature of helicity integrals calculated on the background of local cosmic strings and from the flux quantization in the strings . i thank prof . d. bekenstein for his guidance during the course of this work and for many discussions . i am also grateful to dr . l.sriramkumar for many stimulating discussions . this research is supported by a grant from the israel science foundation , established by the israel academy of sciences and humanities . 99 l. onsager , nuovo cim . * 6 * , suppl . 2 , 246 ( 1949 ) . feynman , in : progress in low temperature physics , vol . 1 , ed . c.j . gorter ( north - holland , amsterdam , 1955 ) . a. vilenkin , and e .. p.s . shellard , in : cosmic strings and other topological defects , ( cambridge u.p . , cambridge , 1994 ) h.k . moffatt , [ 1981 ] , j. fluid mech . * 106 * , 27 ( 1955 ) . donnelly and c.e . swanson , j. fluid mech . * 173 * , 387 ( 1986 ) ; k.w . schwartz , phys . rev . * b 38 * , 2398 ( 1988 ) . moffat , j. fluid mech . * 35 * , 117 ( 1969 ) . moffatt and r.l . ricca , proc . . lond . * a 439 * , 411 ( 1992 ) . berger and g.b . field , j. fluid . 147 * , 133 ( 1984 ) . arnold , ( in russian ) , proc . summer school in differential equations , ereran , armenian ssr acad . ( trans . in sel . * 5 * , 327 ( 1986 ) . ) kuznetsov and a.v . mikhailov , phys . lett . * a 77 * , 37 ( 1980 ) . h. hopf , math . annalen * 104 * , 637 ( 1931 ) . whitehead , proc . n. a. s. * 33 * , 117 ( 1947 ) . davis and e.p.s . shellard , phys . lett . * 63 * , 2021 ( 1989 ) . ginzburg , and l.p . pitaevskii , zh . * 24 * , 1240 ( 1958 ) . jetp * 7 * , 858 ( 1958 ) ) kibble , j. phys . * a9 * , 1387 ( 1976 ) . bekenstein , phys . lett . * b 282 * , 44 ( 1992 ) . t. vachaspati and g.b . field , phys . lett . * 73 * , 373 ( 1994 ) . m. sato , and s. yahikozawa , nucl . b436 * , 100 ( 1995 ) . m. sato , _ shapes of cosmic strings and baryon number violation _ , _ hep - ph_/9508375 . t. vachaspati and a. vilenkin , phys . rev . * d 30 * , 2036 ( 1984 ) shellard and b. allen , in : the formation and evolution of cosmic strings , eds . g. gibbons , s.w . hawking and t. vachaspati ( cambridge u.p . , cambridge , 1990 ) . white , am . j. maths * 91 * , 693 ( 1969 ) fuller , proc . usa * 68 * , 815 ( 1971 ) . l. woltjer , proc . 44 * , 489 ( 1958 ) . l. woltjer , proc . 44 * , 833 ( 1958 ) . moreau , sminaire danalyse convexe montpellier , expos no . 7 g. clugreanu , rev . math . pures appl . * 4 * , 5 ( 1959 ) . g. clugreanu , czechoslovak . * 11 * , 588 ( 1961 ) . fuller , proc . usa * 75 * , 3557 ( 1978 ) . nielsen and p. olesen , nucl . phys . * b61 * , 45 ( 1973 ) . cohen , and a.v . manohar , phys . b265 * , 406 ( 1991 ) . j. garriga , and t. vachaspati , nucl . phys . * b438 * , 161 ( 1995 ) . s. weinberg , the quantum theory of fields , ( cambridge u.p . , cambridge , 1996 ) . peskin and d.v . schroeder , an introduction to quantum field theory , ( addison - wesley , new york , 1995 ) . dolgov , rep . . phys . * 222 * , 309 ( 1992 ) . n. turok , in : formation and interaction of topological defects , p.283 , eds . davis and r. brandenberger ( plenum press , new york , 1995 ) . hindmarsh and t.w.b . kibble , rep . 58 * , 477 ( 1995 ) . adler , phys . rev . * 177 * , 2426 ( 1969 ) . j.s bell and r. jackiw , nuovo cim . * 60a * , 47 ( 1969 ) . s. weinberg , phys . * 19 * , 1264 ( 1967 ) . a. salam , elementary particle theory , ed . n. svarthholm , ( almquist and forlag , stockholm , 1968 ) . n.s . manton , phys . * d 28 * , 2019 ( 1983 ) . t. vachaspati , phys . rev . lett . * 68 * , 1977 ( 1992 ) . t. vachaspati , nucl . b397 * , 648 ( 1993 ) . y. nambu , nucl . phys . * b130 * , 505 ( 1977 ) . m. james , l. perivolaropoulos and t. vachaspati , nucl . b395 * , 534 ( 1993 ) . dirac , proc . ( london ) * a133 * , 60 ( 1931 ) . h. jehle , phys . rev . * d 3 * , 306 ( 1971 ) . r. foot , h. lew and r.r . volkas , j. phys . * g 19 * , 361 ( 1993 ) .

we monitored the chromospheric activity in the ca ii h & k lines of 13 solar - type stars ( including the sun ) ; 8 of them over three years at the cfht and 5 in a single run at the vlt . ten of the 13 targets have close planetary companions . all of the stars observed at the cfht show long - term ( months to years ) changes in h & k intensity levels . four stars display short - term ( days ) cyclical activity . for two , hd 73256 and ceti , the activity is likely associated with an active region rotating with the star , however , the flaring in excess of the rotational modulation may be associated with a hot jupiter . a planetary companion remains a possibility for ceti . for the other two , hd 179949 and and , the cyclic variation is synchronized to the hot jupiter s orbit . for both stars this synchronicity with the orbit is clearly seen in two out of three epochs . the effect is only marginal in the third epoch at which the seasonal level of chromospheric activity had changed for both stars . short - term chromospheric activity appears weakly dependent on the mean k - line reversal intensities for the sample of 13 stars . also , a suggestive correlation exists between this activity and thesin of the star s hot jupiter . because of their small separation ( 0.1 au ) , many of the hot jupiters lie within the alfvn radius of their host stars which allows a direct magnetic interaction with the stellar surface . we discuss the conditions under which a planet s magnetic field might induce activity on the stellar surface and why no such effect was seen for the prime candidate , boo . this work opens up the possibility of characterizing planet - star interactions , with implications for extrasolar planet magnetic fields and the energy contribution to stellar atmospheres .

astro-ph0411655

in the past decade , @xmath6 130 extrasolar planets have been discovered as companions to late - type dwarfs mostly from precise radial velocities ( prv ; e.g. walker et al . 1995 ) from which the keplerian orbital parameters and minimum planetary masses can be derived . from transiting planets , of which five are known , true planetary masses , radii , densities and albedo - dependent estimates of surface temperature are available . to date , a handful of experiments have lead to further information about these planets , including lyman @xmath7 ( vidal - madjar et al . 2003 ) and sodium detections ( charbonneau et al . 2002 ) in the atmosphere of the transiting planet around hd 209458 . a current review of follow - up techniques to probe planetary characteristics has been presented by charbonneau ( 2003 ) . the presence of a giant planet likely influences its parent star beyond the dynamical perturbations measured by the prv method . robinson & bopp ( 1987 ) caught the signatures of superflares " with energies of @xmath8 10@xmath9 ergs ( schaefer et al . 2000 ) on nine solar analogs which have no otherwise unusual properties such as very rapid rotation or very high chromospheric activity . rubenstein & schaefer ( 2000 ) suggested that these anomalous superflares were stimulated by an unseen close - in extrasolar giant planet ( cegp , also known as a ` hot jupiter ' ) . they explored the possibility of magnetic reconnection events occurring between both the star s and the planet s entangled magnetic fields . cuntz et al . ( 2000 ) suggested a more consistent observable interaction between a parent star and its hot jupiter in the form of enhanced stellar activity of the star s outer atmosphere . this interaction can take the form of tidal and/or magnetic heating of the plasma . both the star and its planet experience strong tidal forces as well as repeated magnetic reconnection with a large planetary magnetosphere . if such planet - induced heating of the star is confined to a narrow range in stellar longitude , the heated regions likely track the planet as it orbits . this implies that the period of any observed activity would be correlated with the planet s orbital period @xmath10 such that tidally induced activity has a period of @xmath11 and magnetic activity , a period of @xmath10 . in the simplest configuration of magnetic interaction , the expected enhancement would occur near the sub - planetary point , which defines orbital phase @xmath12 at the time when the planet is in front of the star relative to the line - of - sight . a partial analogy can be made to the jupiter - io system ( zarka et al . 2001 ) where the volcanically active moon of jupiter , orbiting at a distance of 5.9 @xmath13 , constantly couples with jupiter s magnetosphere , leaving two footprints at high positive and negative latitudes on the planet s surface . plasma flows along the magnetic field lines , making up the io flux tube , as these footprints follow io in its orbit . even though the analogy is limited , a similar phenomenon may occur between cegps and their stars such that coupling between the magnetic fields of the planet and the star may cause footprints or hotspots " on the star s surface which follow the planet and have a period close to the planet s orbital period similarly , auroral emission may be stimulated on the planet s atmosphere ( zarka et al . 2001 ) but no searches for planetary radio emissions have yet been successful ( e.g. bastian et al . 2000 , farrell et al . 2003 , ryabov et al . 2003 ) . more recently , santos et al . ( 2003 ) observed cyclic photometric variations with a period very similar to that of the radial velocity ( rv ) curve for the k dwarf hd 192263 . the stability of the 24.4-day periodic radial velocity through almost 4 years of data ( santos et al . 2000 , henry et al . 2002 ) rules out the interpretation that stellar activity alone is the cause of the rv curve and supports the existence of a planetary - sized companion around the star . however , they question what might cause a quasi - stable photometric period that coincides with the planetary orbit . they offered planet - induced magnetic activity off - set by 90@xmath14 from the sub - planetary point as an explanation . even though the interpretations of rubenstein & schaefer ( 2000 ) and santos et al . ( 2003 ) are uncertain , there exists ample observational evidence of such tidal and magnetic interactions in the exaggerated case of the rs canum venaticorum ( rs cvn ) stars , which are tightly - orbiting binary systems consisting of two chromospherically active late - type stars . for example , catalano et al . ( 1996 ) found as many starspots and plages within 45@xmath14 of the sub - binary point as on the rest of the stellar surface for several rs cvn systems . @xmath15 and , a relatively long - period system ( @xmath10 = 20.1 d ; walker 1944 ) , shows modulation of the mg ii uv chromospheric emission lines with a period of 10 days , half the orbital period . glebocki et al . ( 1986 ) interpreted this as a tidal heating of the primary by its companion , possibly a brown dwarf ( donati et al . 1995 ) . and , in our own ca ii h & k observations of er vul , an rs cvn system with two g v dwarfs and @xmath10 = 0.69 d , we see clear enhancements near the sub - binary longitudes ( shkolnik et al . 2004 ) . with these scenarios in mind , we searched for periodic chromospheric heating by monitoring the ca ii h & k emission in stars with giant planets within a few stellar radii . we chose to study the tightest observable systems since the tidal and magnetic interactions depend on the distance from the planet to the star as @xmath16 and @xmath17 , respectively . of the known extrasolar planets , about 20% have semi - major axes of less than 0.1 au and masses comparable to jupiter s ( schneider 2004 ) . it is expected that these planets have magnetic fields also similar to jupiter s ( 4.3 g ) . it is also reasonable to assume that any magnetic interaction would be greatest in the outermost layers of the star , namely the chromosphere , transition region and the corona due to their proximity to the planet , low density , and nonradiative heat sources . the broad , deep photospheric absorption of the ca ii h & k lines allows the chromospheric emission to be seen at higher contrast . because of this and the accessibility from the ground , the h & k reversals are an optimal choice with which to monitor chromospheric heating of these sun - like stars . our program stars have orbital periods between 2.5 and 4.6 days , eccentricities @xmath6 0 and semi - major axes @xmath18 au . these systems offer the best chance of observing upper atmospheric heating . the first five systems we observed were @xmath5 boo , hd 179949 , hd 209458 , 51 peg and @xmath1 and from the canada - france - hawaii telescope ( cfht ) . the first results from 2001 and 2002 observations , including the first evidence of planet - induced magnetic heating of hd 179949 , were published in shkolnik et al . we later extended the experiment at the very large telescope ( vlt ) to include five southern targets , hd 46357 , hd 73256 , hd 75289 , hd 76700 , and hd 83443 . the system parameters for the program stars are listed in table 1 along with our two standards , @xmath5 ceti and the sun . in this paper , we compile our 2003 cfht data with those of previous years and include the recent vlt observations . a broader understanding of stellar activity , its cycles , and planet - induced chromospheric heating emerges . we also observed @xmath0 ceti , a young ( 650 @xmath19 750 myr old ) , chromospherically active , solar analog . it was one of the nine stars reported to have anomalous superflare activity by robinson & bopp ( 1987 ) possibly caused by magnetic interactions with a close - in giant planet ( rubenstein & schaefer 2000 ) . of their sample , only @xmath0 ceti has been looked at by the prv planet searches of walker et al . ( 1995 ) , cumming et al . ( 1999 ) and halbwachs et al . ( 2003 ) ; none of which has detected a planet . interestingly though , walker et al . ( 1995 ) observed a rapid rv change of 80 m s@xmath20 in 1988 which could have been caused by a planet in highly elliptical orbit . however , it coincided with an equally sharp increase in the ca ii chromospheric activity indicator at 8662 implying the rv jump was likely due to changes intrinsic to the star . we may have observed this same phenomenon in our ca ii data . the details of our cfht and vlt observations are outlined in section [ spectra ] . we briefly discuss the precise differential radial velocities which yielded updated ephemerides for the planetary systems such that orbital phases could be determined to better than 0.02 . in section [ caii ] we discuss our analysis and results of the ca ii h & k measurements including long - term , short - term and rotational modulation . a theoretical discussion of the physical requirements for magnetic interactions between stars and their hot jupiters is presented in section [ theory ] with future experiments suggested in section [ summary ] . ccccccccccl @xmath5 boo & f7 iv & [email protected] & 3.2 & 3.313 & 3.87 & 0.046 & 0.323 & 0.177 & 0.0020 + hd 179949 & f8 v & [email protected] & @xmath229 & 3.092 & 0.98 & 0.045 & 0.399 & 0.202 & 0.0051 + hd 209458 & g0 v & [email protected] & 16 & 3.525 & 0.69 & 0.045 & 0.192 & 0.077 & 0.0009 + 51 peg & g2 iv & [email protected] & 21.9 & 4.231 & 0.47 & 0.05 & 0.178 & 0.071 & 0.0008 + @xmath1 and & f7 v & [email protected] & 14 & 4.618 & 0.71 & 0.059 & 0.254 & 0.091 & 0.0016 + hd 46375 & k1 iv & @xmath222 & 43 & 3.024 & 0.25 & 0.041 & 0.339 & 0.233 & 0.0011 + hd 73256 & g8/k0 v & [email protected] & 14 & 2.549 & 1.85 & 0.037 & 0.995 & 0.899 & 0.0051 + hd 75289 & g0 v & 4.37 & 15.95 & 3.510 & 0.42 & 0.046 & 0.177 & 0.062 & 0.0003 + hd 76700 & g6v & @xmath222 & & 3.971 & 0.20 & 0.049 & 0.212 & 0.095 & 0.0004 + hd 83443 & k0v & 1.4 & 35 & 2.986 & 0.38 & 0.039 & 0.314 & 0.211 & 0.0010 + @xmath0 cet & g5 v & [email protected] & 9.3 & & & & 0.906 & 0.815 & 0.0059 + @xmath5 cet & g8 v & 1 & 33 & & & & 0.201 & 0.102 & 0.0003 + sun & g2 v & [email protected] & 27 & & & & 0.298 & 0.137 & 0.0005 + the observations were made with the 3.6-m canada - france - hawaii telescope ( cfht ) on 3.5 , 4 , 2 , and 5 nights in 2001 august , 2002 july , 2002 august , and 2003 september , respectively . we used the gecko chellette spectrograph fiber fed by cafe ( cassegrain fiber environment ) from the cassegrain to coud focus ( baudrand & vitry 2000 ) . spectra were centered at 3947 which was isolated by a uv grism ( 300 lines mm@xmath20 ) with @xmath23 60 intercepted by the ccd . the dispersion was 0.0136 pixel@xmath20 and the 2.64-pixel fwhm of the thorium - argon ( th / ar ) lines corresponded to a spectral resolution of r = 110,000 . the detector was a back - illuminated eev ccd ( 13.5 @xmath24m@xmath25 pixels , 200 @xmath26 4500 pixels ) with spectral dispersion along the rows of the device . to remove the baseline from each observation , the appropriate mean darks were subtracted from all the exposures . flat - fields were then normalized to a mean value of unity along each row . all the stellar exposures observed on a given night were combined into a mean spectrum to define a single aperture for the extraction of all stellar and comparison exposures , including subtraction of residual background between spectral orders ( prior to flat - fielding ) . this aperture was ultimately used to extract one - dimensional spectra of the individual stellar and comparison exposures and of the mean , normalized flat - field . this one - dimensional flat used to obtain the most consistent flat - fielded spectra possible . all the data were processed with standard iraf ( image reduction and analysis facility ) routines . wavelength calibration was done using the th / ar arcs taken before and after each spectrum . we required frequent arcs in order to track the ccd drift or any creaking in the system throughout the night . heliocentric and differential radial velocity corrections were applied to each stellar spectrum using iraf s _ rvcorrect _ routine . a specimen , flat - fielded spectrum of @xmath1 and is shown in figure [ upsand_spec ] . the ca ii h & k reversals are weak yet visible at 3968 and 3933 . the final spectra were of high s / n reaching @xmath6 500 per pixel ( or 4300 @xmath20 ) in the continuum and 150 ( or 1290 @xmath20 ) in the h & k cores . spectra with comparable s / n were taken of two stars known not to have close - in giant planets , @xmath5 ceti and the sun ( sky spectra were taken at dusk ) . table 2 of walker et al . ( 2003a ) lists the five cfht program stars plus @xmath5 ceti , including their u magnitudes , exposure times , and typical s / n . differential radial velocities ( @xmath27rvs ) were estimated with the _ fxcor _ routine in iraf ( version pc - iraf v2.12 ) which performs a fourier cross - correlation on dispersion - corrected spectra . we used the first spectrum in the series for each star as the template , hence , all @xmath27rvs are relative to the first spectrum on the first night of the run . both the template and the input spectrum were normalized with a low order polynomial . the correlation used only the @xmath8 20 region between the h & k lines , that part of the spectrum bounded by ( and including ) the two strong al i lines ( 3942 @xmath19 3963 ) as shown in figure [ upsand_spec ] . the @xmath27rvs measured during the 2002 july run for the five ` 51 peg ' stars can be found in walker et al . ( 2003a ) . the @xmath27rvs from 2003 september are plotted in figure [ rv2003 ] . after the orbit is removed , the average @xmath28 is 17 m s@xmath20 . the two stars observed at the highest s / n have @xmath28 of 7 and 9 m s@xmath20 . this precision may be the best achievable for a single spectrum with this spectrograph . the excellent @xmath27rvs yield current orbital ephemerides and hence accurate phases ( @xmath29 ) for each observation . using the ephemerides of the planets discovery orbits , we tabulated the 2003 september times of sub - planetary position ( @xmath30 = 0 ) with revised orbital periods . these are listed in table [ ephemerides ] . ccccccl @xmath5 boo & 33 & 2452892.864 & 0.066 & 3.31250 & 0.00026 + hd 179949 & 19 & 2452894.114 & 0.062 & 3.09246 & 0.00031 + hd 209458 & 17 & 2452893.653 & 0.070 & 3.52490 & 0.00020 + 51 peg & 15 & 2452895.868 & 0.085 & 4.23092 & 0.00014 + @xmath1 and & 9 & 2452892.615 & 0.092 & 4.61750 & 0.00052 + cccccrccl hd 46375 & 9.33 & 8.7 & 300 & 8 & 510 & 150 + hd 73256 & & 8.86 & 300 & 9 @xmath26 2 & 520 & 150 + hd 75289 & 7.04 & 6.94 & 120 & 10 & 790 & 230 + hd 76700 & 9.17 & 8.76 & 300 & 8 & 540 & 160 + hd 83443 & 9.54 & 9.03 & 300 & 8 & 420 & 120 + we obtained high - resolution spectra through visitor mode using the vlt s ultraviolet and visual echelle spectrograph ( uves ) mounted on the 8.2-m kueyen ( ut2 ) over four photometric half - nights ( 2004 april 4 @xmath19 7 ) . the standard blue arm setting was used , centered on 4370 , giving a wavelength range of 3750 to 4990 . we used the cd2 cross - disperser grating ( 660 g mm@xmath20 ) with a ccd of 2048 @xmath26 3000 pixels of 15 @xmath24m@xmath25 . image slicer # 2 with a slit width of 0.44@xmath31 resulted in a resolution of r @xmath6 75,000 . the data was reduced on - site by the uves data reduction pipeline that uses the eso - midas software package within the uves content . the data processing consisted of standard procedures : bias subtraction , interorder background correction , cosmic ray hit removal , flat - fielding , and wavelength calibration . the data were wavelength calibrated with th / ar arcs attached at the beginning and end of each set of 8 to 10 stellar exposures . the exposure times , number of exposures and the s / n for each star are listed in table [ vltstars ] . the very strong ca ii h and k photospheric lines suppress the stellar continuum making it difficult to normalize each 60- spectrum consistently . for this reason , a careful analysis by which to isolate any modulated ca ii emission was devised . figure [ upsand_spec ] shows a flat - fielded spectrum of @xmath1 and with the normalization levels marked with dashed lines . the normalization wavelengths were constant for all spectra of a given star . the 7- spectral ranges , centered on the h and k lines , were chosen to isolate the h and k reversals while minimizing any apparent continuum differences induced by varying illumination of the ccd . this window is , however , wide enough such that a few photospheric absorption features appear which could be tested for variability as well . the mean ca ii k cores for the program stars are shown in figures [ cfht_kcores ] and [ vlt_kcores ] . also , 7- and 2- cuts , centered on the strong photospheric al i line at 3944 were used as internal standards since the line has comparable depth and s / n to the ca ii lines . to normalize each sub - spectrum , the end points were set to 1 and fitted with a straight line . the spectra were grouped by date and a nightly mean was computed for each of the three lines . the rms of the al i residuals for each cfht target star is less than 0.0005 of the normalized mean . this is representative of the al i residuals for all stars in all four runs . the al i line of the vlt data varies with an average rms of 0.0006 . these values demonstrate both the level of stellar photospheric stability as well as the reliability of the data reduction and analysis . for all stars observed , both the h & k emission was non - varying at the 0.001 level on a given night , likely as a result of intra - night ( statistical ) noise . with four cfht observing runs spanning a baseline of over two years , we can compare the long - term variations in the chromospheric levels of the stars . we measure emission strength by integrating across the normalized k cores bounded by the k1 features ( montes et al . each star s average integrated k emission @xmath32k@xmath33 for each observing run is plotted in figure [ yearly ] along with its fractional variation relative to the overall average emission . this is a good start to tracking the intrinsic stellar activity cycles of these stars . in the case of the sun , we see the decrease from 2001 august to 2003 september as it declines from solar maximum . however , we also observed the naked - eye sunspot grouping of 2002 august that appeared in our data as a @xmath6 2% increase in the ca ii emission relative to the other years . since the variability from run - to - run may also be an indication of active regions on the disk of the star , we require more frequent monitoring over several more years to firmly say anything about the activity cycle of any individual program star . when monitoring ca ii emission , intrinsic stellar activity modulated by stellar rotation will appear along with any possible activity stimulated by the planet . the orbital periods of the planets are well known and uniquely established by the prv and transit discovery methods . the rotation periods of the stars are much harder to determine in part due to stellar differential rotation which yields non - unique periods . for our work , it is key to distinguish between the rotational and orbital modulation of chromospheric emission . to isolate the chromospheric activity within the reversals , we took nightly residuals from the average stellar spectrum of all data . each residual spectrum had a broad , low - order curvature removed which was an order of magnitude less than the variations in the h and k lines discussed below . the residuals of the normalized spectra ( smoothed by 21 pixels ) were used to compute the mean absolute deviation ( mad = @xmath34 for @xmath35 spectra ) . as an example , the nightly residuals used to generate the mad plot for @xmath1 and are displayed in figure [ upsand_kcombo ] ( top ) . the mad plot with the corresponding k - core superimposed is in figure [ upsand_kcombo ] ( bottom ) . the identical analysis described above was performed on the ca ii h line of all target stars . the complete ca ii h results can be found in shkolnik ( 2004 ) . as the two resonance lines share the same upper level and connect to the ground state , the same activity is seen in both . as expected , the ca ii h emission and activity levels are @xmath6 2/3 that of ca ii k ( sanford & wilson 1939 ) . the star with the shortest rotation period in our sample is @xmath5 boo . it has the largest @xmath36sin@xmath3 (= 14.8 m s@xmath20 ; gray 1982 ) and is believed to be in synchronous rotation with its tightly - orbiting planet ( @xmath37 = 3.2 @xmath21 0.5 d , henry et al . 2000 ; @xmath10 = 3.31250 @xmath21 0.00026 d , this work ) . if @xmath5 boo is tidally locked to its planet , the planet - star interaction may be minimal ( saar et al . 2003 , 2004 ) due to the fact that there is near zero relative velocity . ( see section [ theory ] for theoretical discussion . ) the integrated residuals from the mean normalized k core are plotted against orbital phase in figure [ intk1 ] ( left ) . looking at the individual observing runs in the plot for @xmath5 boo , the two nights of observation in 2001 august did not show much variation when the star was somewhat less active . the 2002 july data showed an increase in activity near the sub - planetary point relative to the other observations in their respective runs . however , in 2002 august and 2003 september observations show a relative enhancement at @xmath38 0.4 @xmath19 0.5 modulated with a period near @xmath37 and @xmath10 . this chaotic activity does not allow us to draw any conclusions about planet - induced heating . similar to @xmath5 boo , the k emission of hd 209458 showed night - to - night modulation , but with a smaller amplitude during most runs and without any phase coherence , as shown in figure [ intk1 ] ( right ) . in 2001 august , we caught the system immediately after transit at which time we observed a slight enhancement in the ca ii emission relative to all other observations . in the 2002 july run , an increase in emission occurred at @xmath39 0.25 with no apparent rise toward @xmath30 = 0 . due to the relatively low s / n of these data and the large intra - night deviations , we can not form any conclusions . as seen in figures [ intk1 ] and [ intk2 ] , four of the five cfht stars show significant chromospheric variation throughout a single observing run . the standards , @xmath5 ceti and the sun , show no such modulation . one consistent result for the ` active ' stars ( @xmath5 boo , hd 179949 , hd 209458 and @xmath1 and ) is their night - to - night modulation of h & k emission . also , unlike the case for hd 73256 ( see section [ modbyrot ] ) , the night - to - night variations of these stars do not increase or decrease monotonically throughout an observing run , implying the variability can not be explained exclusively with starspots rotating into or out of view . another mechanism is necessary . the night - to - night variations may indicate planet - induced activity or sporadic flaring from hotspots . if coupled to the planet , the localized activity would be travelling on the stellar surface faster than the star is rotating as it tracks the planet in its orbit . other than for @xmath5 boo , the timescale of activity is short compared to the stellar rotation period . unfortunately , due to the large uncertainties in the rotation periods of the other stars , phasing with rotation is uninformative at this stage . however , we do know that @xmath1 and and hd 209458 both have @xmath37 @xmath40 3@xmath10 and rotate only @xmath41 per day . wolf & harmanec ( private communication ) recently made ubv photometric observations of hd 179949 at our request from the saao sutherland observatory . when they combined their v observations with those from hipparcos ( converted to v ) they detected a rotation period of 7.07 d but with an amplitude of only 0.008 mag . given that the rms of the v - observations was 0.006 mag , this periodicity is at the limit of detection . indirect indications of the rotation rate of hd 179949 imply @xmath42 days and are presented in shkolnik et al . ( see also saar et al . these include a high x - ray luminosity for the star , a very long tidal synchronization timescale and a moderate s@xmath43 index . while more photometry is needed to determine a rotation period conclusively , it is highly unlikely that hd 179949 is tidally locked to its planet at 3.092 d. in shkolnik et al . ( 2003 ) we presented the first evidence of planet - induced heating on hd 179949 . the effect lasted for over a year and peaked only once per orbit , suggesting a magnetic interaction . we fitted a truncated , best - fit sine curve with @xmath44 = 3.092 d corresponding to the change in projected area of a bright spot on the stellar surface before being occulted by the stellar limb . figure [ intk2 ] ( left ) updates the integrated k residuals to include the 2003 september data . the spot model is a remarkable fit for the 2001 and 2002 data peaking at @xmath30 = 0.83 with an amplitude of 0.027 . clearly , the average k emission is higher during the latest run ( as shown in figure [ yearly ] ) with a much smaller level of variability . it is interesting to note that the 2003 data still peak between @xmath30 = 0.80 @xmath19 0.95 , consistent with the previous results . the second convincing case of magnetic interaction is between @xmath1 and and its innermost giant planet . and has three known jupiter - mass planets at 0.059 , 0.829 , and 2.53 au ( butler et al . 1999 ) . ] in figure [ intk2 ] ( right ) , the 2002 july , 2002 august and 2003 september runs show good agreement in phase - dependent activity with an enhancement at @xmath30 = 0.53 . the best - fit sine curve has an amplitude of 0.0032 . the 2001 august fluxes are lower than the mean of all four observing runs by almost 3% and still display a significant ( @xmath40 2@xmath45 ) modulation like the quiescent epoch of hd 179949 . again , even the low - amplitude modulation has a rise and fall with a period consistent with @xmath10 and peaks near @xmath46 . for these two cases , the peak of the emission does not directly coincide with the sub - planetary point , @xmath30 = 0 . for hd 179949 , it leads the planet by 60@xmath14 in phase and for @xmath1 and , the ca ii emission is 169@xmath14 out of phase with the sub - planetary point . santos et al . ( 2003 ) also observed a 90@xmath14 lag from the sub - binary point in the periodic activity indicated by the photometric variations for hd 192263 . the phase lead or lag may help identify the nature of the interaction . for example , the phase offset of a starspot or group of starspots can be a characteristic effect of tidal friction , magnetic drag or reconnection with off - center stellar magnetic field lines , including a parker spiral - type scenario ( weber & davis 1967 , saar et al . 2004 ) . in any case , the phasing , amplitude and period of the activity have persisted for over a year between observations . for hd 179949 , this equals 108 orbits or at least 37 stellar rotations and for @xmath1 and , the time spans 88 orbits or approximately 29 rotations . the observations are consistent with a magnetic heating scenario as the chromospheric enhancement occurred only once per orbit . we estimated the excess absolute flux released in the enhanced chromospheric emission of hd 179949 by calibrating the flux with that of the sun . the flux was the same order of magnitude as a typical solar flare , @xmath8 10@xmath47 erg s@xmath20 or 1.5@xmath2610@xmath48 ergs @xmath49 s@xmath20 . this implies that flare - like activity triggered by the interaction of a star with its hot jupiter may be an important energy source in the stellar outer atmosphere . this also offers a mechanism for short - term chromospheric activity on the stars with close - in jupiter - mass planets . of the 10 program stars we monitored for h & k variability , five of them showed no changes down to the 0.001 level : 51 peg , hd 46375 , hd 75289 , hd 76700 , hd 83443 . there are two reasons we offer to explain the relative quiescence of these stars . it was well - known through the many years of the mt . wilson s@xmath43 survey there is a strong correlation between rotation rate ( or inversely with rotation period ) and ca ii emission ( noyes et al . 1984 , pasquini et al . this is a likely contributor though is not obviously clear in our sample set as shown in figure [ madk ] ( left ) where the inverse of the rotation period is plotted against the mean chromospheric emission of the ca ii k line @xmath32k@xmath50 . the photospheric contribution to the emission was removed from @xmath32k@xmath33 using @xmath32k@xmath50 = @xmath32k@xmath33(1 - @xmath51 ) where r@xmath52 = r@xmath43 @xmath19 r@xmath53 and is an empirical function of ( @xmath54 ) and s@xmath43 taken from hartmann et al . the chromospheric contibutions r@xmath53 are from wright et al . ( 2004 ) . ) . ] in table [ starpars ] we list @xmath32k@xmath33 , @xmath32k@xmath50 , and @xmath32madk@xmath33 for all the stars . from figure [ madk ] ( right ) , we deduce that the higher a star s chromospheric emission , the more night - to - night activity it displays . radick et al . ( 1998 ) show the same effect for a much larger sample . this is akin to shot noise since the flaring or stochastic noise associated with the activity will increase with the activity level . secondly , a recent calculation of the magnetic fields in giant extrasolar planets ( snchez - lavega 2004 ) looked at the internal structure and the convective motions of these planets in order to calculate the dynamo - generated surface magnetism . given the same angular frequency ( which is a reasonable approximation for the short - period planets in question ) , the magnetic dipole moment , and hence the magnetospheric strength , increases with planetary mass . this is observed in our own solar system for the magnetized planets where the magnetic moment grows proportionally with the mass of the planet ( arge et al . since only lower limits exist for most of the hot jupiters , we can only plot @xmath2sin@xmath3 against @xmath32madk@xmath33 in figure [ msini_madk ] where we still see an intriguing correlation . the dashed circles for hd 179949 and @xmath1 and are their @xmath32madk@xmath33 values ( 0.0021 , 0.0011 , respectively ) with the orbital modulation removed . of our sample , @xmath5 boo has the most massive planet and yet falls well below the correlation . as we discuss further in section [ theory ] , if the star and planet are tidally locked , as is thought to be the case for @xmath5 boo , then there is little or no free energy left from the orbit and we would expect weak , if any , magnetic coupling . for most of our target stars , rotation periods are not well enough known to accurately phase the ca ii data with @xmath37 . there is even ambiguity in @xmath37 of the often - observed @xmath0 ceti . in late 2003 , @xmath0 ceti was monitored continuously for 30.5 days by most . this best - ever lightcurve obtained for @xmath0 ceti showed a pattern composed of two transiting spots of differing periods , 8.9 and 9.3 days providing a direct and unique measurement of differential rotation ( rucinski et al . the real ambiguity in @xmath37 from the space observations of @xmath0 ceti demonstrates the difficulty in measuring @xmath37 for most stars . nonetheless , two stars do exhibit clear rotational modulation : @xmath0 ceti , a star with no confirmed planet ( halbwachs et al . 2003 ) and hd 73256 , a star with a 1.85-m@xmath55 planet orbiting at 0.037 au ( udry et al . the ca ii emission from @xmath0 ceti has been monitored by baliunas et al . ( 1995 ) through the narrow - band filter of the mt . wilson survey from 1967 to 1991 . these data show long - term stability of a period of 9.4 @xmath21 0.1 d , close to the photometric rotation period of 9.214 d published by messina & guinan ( 2002 ) . the rotation period for hd 73256 is photometrically determined to be 14 d ( udry et al . 2003 ) , consistent with the 13.9 days derived from the r@xmath53 activity index ( donahue 1993 ) . we observed periodic ca ii h & k variability in the chromosphere of @xmath0 ceti during our 2002 and 2003 cfht runs from which we determined a rotation period of 9.332 @xmath21 0.035 d. the results were first published in rucinski et al . ( 2004 ) where we compared the activity seen in ca ii with most s lightcurve showing that the chromospheric activity coincided with the low - latitude spot such that the maxima of the two curves agree . hd 73256 was observed twice per night for four nights at the vlt . the mean k cores of these two stars are shown at the top of figure [ vlt_kcores ] where their similarly strong emission is evident . the high level of chromospheric emission points to a young age for both of these stars : 650 @xmath19 750 myr for @xmath0 ceti ( gdel et al . 1997 ; dorren & guinan 1994 ) and 830 myr for hd 73256 ( donahue 1993 ) . the @xmath32madk@xmath33 given for these two stars in table [ starpars ] are corrected for the rotational modulation resulting in activity levels comparable to hd 179949 . the integrated residual k fluxes for @xmath0 ceti and hd 73256 are plotted against relative rotational phase in figure [ intk_prot ] . for completeness , we plot the residual k fluxes of hd 73256 as a function of orbital phase in figure [ hd73256_intk ] where the flare is apparent at @xmath56 = 0.03 . there is no clear signature of the planet in the activity . in both cases , @xmath0 ceti ( @xmath57 = 4.92 ) and hd 73256 ( @xmath57 = 5.27 ) show sporadic flaring beyond the clear rotational modulation . the periodic best - fit curve for @xmath0 ceti necessitates a non - zero eccentricity while a sine curve is sufficient for hd 73256 . the largest excursion from the rotation curve of @xmath0 ceti has an energy of 2.8@xmath2610@xmath58 erg @xmath49 s@xmath20 ( again , measured by comparing with solar absolute flux ) . we estimate the absolute flux emitted from hd 73256 s flare at @xmath59 = 0.15 to be 4.9@xmath2610@xmath58 erg @xmath49 s@xmath20 ( or @xmath60 erg @xmath49 if the flare lasted for at least the hour for which we observed it . ) the modulation of the k emission due to rotation is @xmath6 6% indicating that the emission is dominated by a large hotspot on the stellar surface . as we have seen on hd 179949 and @xmath1 and , planet - induced variations are at the level of 1 @xmath19 2% suggesting that the reigning hotspot could have diluted any heating caused by the hot jupiter . the @xmath27rvs for @xmath61 ceti are plotted in figure [ rv_k1ceti ] against the 9.332-d phase determined from the k - line residuals where the open symbols correspond to data from 2002 and the solid squares to 2003 . from the 2002 data we derive @xmath6221.8 m s@xmath20 and 23.6 m s@xmath20 when combined with 2003 . these values are very similar to @xmath6224.4 m s@xmath20 , found over 11 years by cumming et al . ( 1999 ) . in 2003 the @xmath27rvs appear significantly different between the two nights , something which seems to be reinforced by the consistency within the pairs of @xmath27rvs . while a planetary perturbation can not be ruled out by the 2002 data and other prv studies , the difference in 2003 might be associated with the velocity field of the star itself . the increase of velocity with increasing k - line strength at @xmath63 0.7 is consistent with the extreme event in 1988 seen by walker et al . however it should be emphasized , the possibility remains of a close giant planet around @xmath61 ceti . for instance , a planet inducing a reflex radial velocity variation @xmath64 m s@xmath20 and tidally synchronized with the star , would have @xmath65 m@xmath66 and @xmath67 = 0.084 au . the enhancements of chromospheric activity on hd 179949 and @xmath1 and appear only once per orbit implying a magnetic , rather than tidal , interaction between the star and its hot jupiter . the two stars are f - type stars with higher x - ray luminosity than the solar value . the rosat catalogue of bright main - sequence stars lists hd 179949 as having at least double the x - ray luminosity ( a measurement independent of @xmath3 ) of most other single f8 @xmath19 9 dwarfs ( hnsch et al . yohkoh solar x - ray observations ( e.g. yokoyama & shibata 2001 ) have shown that the energy release at the site of magnetic reconnection during a solar flare generates a burst of x - ray emitting gas . this hot plasma is funneled along the magnetic field lines down to the surface producing ` footprints ' through anomalous heating of the sun s chromosphere and transition region . it is this same phenomenon that is likely occurring between hot jupiters and their host stars through the reconnection of their magnetic fields . observationally , companion - induced activity is unambiguously observed on rs cvn stars , as discussed in section [ intro ] . the hot jupiter of @xmath1 and is located farther out from the host star than that of hd 179949 , implying that the magnetic interaction between the star and its hot jupiter is diminished . this would result in less ca ii enhancement as is shown in our data . boo is also an f - type star with intense x - ray emission , but its magnetic influence is likely reduced by the small relative motion in the azimuthal direction between the planet and the stellar magnetosphere in an almost final equilibrium state in which the star and the planet are tidally locked to each other . while more data are required to truly verify whether the variability of ca ii emission is correlated to the apparent position of the hot jupiter in these systems , we explore and review a few theoretical aspects of planet - induced heating scenarios in this section . the location of the hot jupiter relative to the alfvn radius ( the distance from the star at which the radial velocity of the wind @xmath68 equals the local alfvn velocity @xmath69 ) plays a significant role in transporting energy toward the star against the stellar winds . since the alfvn radius of the sun is about 10 to 20 r@xmath70 at solar minimum and 30 r@xmath70 at solar maximum ( e.g. lotova et al . 1985 ) , the small distance @xmath71 au of hot jupiters from their host stars suggests that unlike our jupiter , surrounded by a bow shock , some of these hot jupiters are located inside the alfvn radius depending on the magnetic strength of their host stars ( zarka et al . 2001 , ip et al . therefore the direct magnetic interactions between a hot jupiter and its star without a bow shock might resemble the io - jupiter interactions ( zarka et al . 2001 ) or the rs cvn binaries ( rubenstein & schaefer 2000 ) . most of the theoretical models applied to these two cases have focused on the geometry of intertwined magnetic fields as well as the the energy transport through alfvn waves and/or induced currents . alfvn waves can not propagate along the stellar field lines toward the star in the region outside the alfvn radius where the group velocity of alfvn waves is always in the positive radial direction ( e.g. weber & davis 1967 ) . other means of inward energy transport require their energy flux to be at least larger than the energy flux carried by the stellar winds . ip et al . ( 2004 ) estimate the input magnetic power due to the relative motion between the synchronized hot jupiter and the stellar magnetosphere to be @xmath72 ergs s@xmath20 , the same order of magnitude of a typical solar flare . a similar amount of power might be obtained based on the induced current model ( zarka et al . 2001 ) if the radius of the ionosphere of an unmagnetized hot jupiter in the induced current model can be approximated by the radius of the magnetopause of a magnetized hot jupiter . the observed excess energy flux from an unresolved disk of the star is equal to this input magnetic power averaged over the disk of the star . that is , the energy flux is roughly equal to @xmath73 where @xmath74 is the radius of the planet s magnetopause , @xmath75 is the magnetic field at the magnetopause , @xmath76 is the radius of the star , @xmath77 is the radius of the planet , @xmath67 is the distance between the star and the planet , @xmath78 and @xmath10 are the orbital velocity and period respectively , @xmath37 is the rotation period of the star , @xmath79 is the azimuthal component of the stellar wind velocity , and @xmath80 and @xmath81 are the mean magnetic field on the surface of the star and the planet . in deriving the above equation , we have assumed that the stellar magnetic field decays as @xmath82 , the planetary field decays as @xmath83 , @xmath74 was determined by equating the stellar and planetary fields at the magnetopause ( i.e. @xmath84 ) , and the stellar magnetosphere inside the alfvn radius is nearly co - rotating with the stellar rotation ( i.e. @xmath85 ) . the radial component of the stellar wind @xmath68 is left out from the above estimate as long as the condition @xmath86 is valid . the energy flux in eq([eq : energyflux ] ) should be regarded as the maximal input energy from the planet s orbital energy because only some fraction of this amount of energy is transferred to the ca ii emissions . for @xmath87 ( vrsnak et al . 2002 for the case of our sun , and weber & davis 1967 for the case of open fields ) , @xmath88 , @xmath67 = 0.045 au , @xmath89 r@xmath70 , @xmath90 r@xmath66 , @xmath91 d , @xmath92 = 200 g , and @xmath93 = 10 g , the energy flux given by eq([eq : energyflux ] ) is roughly equal to 10@xmath48 erg @xmath49 s@xmath20 , a value comparable to the differential intensity from our data of ca ii k emission from hd 179949 . the same amount of energy flux can be also achieved for the case of a dipole field for the hot jupiter ( @xmath87 , @xmath94 ) where @xmath95 au , @xmath92 = 250 g , and @xmath81 = 10 g. if both the star and the planet have dipole fields ( @xmath96 ) , very strong fields @xmath97 g and @xmath98 g are required to generate the same energy flux . therefore the tight energy budget constrained by the synchronous ca ii emission from hd 179949 strongly suggests that the mean global fields of this f - type star are not likely in a dipole field configuration at the location of the planet but have the radial , open structure ( i.e. @xmath87 ) just like the solar fields have as a result of the outflowing winds . the argument that hot jupiters might have weaker fields than our jupiter due to slower spin rates and weaker convection ( snchez - lavega 2004 ) should be treated with caution since in addition to the uncertainty in the interior structure of the metallic hydrogen region of a hot jupiter , the response of slow convection to various rotation rates in the dynamo process is not well understood . if @xmath87 , and @xmath99 1 g ( snchez - lavega 2004 ) for hd 179949 b , then eq([eq : energyflux ] ) indicates that @xmath92 = 300 g is required to generate the energy flux 10@xmath48 erg @xmath49 s@xmath20 and in this case the stellar field dominates over the planetary field even on the surface of the hot jupiter ( i.e. @xmath100 in eq([eq : energyflux ] ) ) . the observational constraints on the strength of the stellar magnetic fields such as the field versus ca ii relation ( schrijver et al . 1989 , 1992 ) along with the radio cyclotron emissions from the hot jupiter s magnetosphere , where @xmath101 is the speed of light , @xmath102 and @xmath103 are the electron charge and mass respectively . the radiation can reach @xmath104 mhz if @xmath105 g. ] should help to narrow down the field strength of hot jupiters in the magnetic interaction scenario , therefore improving our knowledge of the interior structure and the dynamo processes of gaseous planets . now we turn our attention to the variation of ca ii level , the phase coverage of additional emission , and the phase lead at the different epochs of our observations . the data for hd 179949 seem to suggest that the additional ca ii emission is smaller and the range of phase spanned by the emission is larger when the average ( or minimal ) k emission is larger . presumably this has something to do with the intrinsic stellar activity and therefore the stellar field geometry . the observed phase lead may be caused by spiral stellar fields loaded with stellar winds . when a star is in its quiet phase and therefore the average ca ii emission is very low , no enhancement happens probably because @xmath106 for a hot jupiter located far out from its parent star and therefore being outside the alfvn radius , as may be the case for the 2001 august observations of @xmath1 and . as the star shifts to its more active phase , the average ca ii emission is at the moderate level and @xmath107 is not far smaller than @xmath108 at the hot jupiter . at this time , the configuration of the field lines on the surface of the star may be characterized by the open fields from the coronal holes covering a large area of the star , as well as the closed fields distributed only near the magnetic equator . consequently the foot - points of the open stellar field lines pointing to the planet are located at lower latitudes , as indicated by the 2001 and 2002 data for hd 179949 fitted by a truncated sine curve ( shkolnik et al . 2003 ; also shown in figure [ intk2 ] ) . while a great number of closed fields form at low magnetic latitudes during this time , the coronal holes shrink to the magnetic polar regions . the bright spot at a high latitude implied from the observation for hd 179949 in september 2003 might be caused by the scenario that the hot jupiter perturbs the open field lines emanating from the shrinking coronal holes near the polar regions , leading to a longer phase duration of the additional ca ii flux and perhaps reducing the additional emission due to the smaller projection along the line of sight at higher latitudes . note that at the time close to the stellar - maximum activity , the stellar field lines might be occasionally stretched out like solar streamers emanating from the low latitudes of the star where closed loops of stellar magnetic fields aggregate , possibly giving rise to the planet - induced heating at low latitudes of the stellar surface as well . the picture that we have sketched thus far assumes that most of energy flux released from the vicinity of the hot jupiter is transported along the field lines by alfvn waves and deposited at the foot - points of the magnetic lines . since the field lines inside the alfvn radius are dominated by the poloidal component , detailed calculations for stellar wind models are needed to study how the integration of small pitch angles of the field line can lead to the moderate to large phase angles , @xmath109 for hd 179949 and @xmath110 for @xmath1 and . the phase difference is also determined by the alfvn - speed travel time along the stellar field line . if @xmath111 , the alfvn disturbance roughly takes @xmath112 a few hours to propagate from the hot jupiter to the star with @xmath113 cm s@xmath20 at 0.04 au . this means that the azimuthal angle that the planet has already traveled over the alfvn - speed travel time is not small . since the hot jupiter of @xmath1 and is located farther out ( @xmath114 au ) from its star than hd 179949 b ( @xmath95 au ) , the large phase angle of @xmath110 for @xmath1 and might actually represent a phase lag caused by the small inward group velocity of the alfvn waves @xmath115 that takes considerable amount of time to travel along the field lines right after the waves are generated from the planet . alternatively , a large phase difference between the location of the heating spot in the chromosphere and the position of the hot jupiter might be caused by the entangled or rotationally spiraled magnetic fields connecting directly between the hot jupiter and the star ( rubenstein & schaefer 2000 , saar et al . 2004 , ip et al . 2004 ) . besides considering the energy transported along the field lines and field geometry , understanding how the energy is dissipated is important to construct a complete picture for the ca ii emission in the planet - induced scenario . in our own solar system , a bow shock of the solar wind generally hinders this because alfvn waves can not propagate toward the star from the region outside the alfvn radius . the inward poynting energy of alfvn waves @xmath116 is roughly conserved if the wind velocity @xmath68 is much smaller than the alfvn speed @xmath69 . here @xmath117 is the distance away from the star , @xmath118 is the magnetic disturbance of the wave , and @xmath119 is the mass density of the stellar wind . at first glance , the transition region characterized by a steep decrease of the alfvn speed with depth seems to provide a possible radial stratification for the inward - propagating alfvn waves to pile up the magnetic energy density , leading to the nonlinear dissipation of the growing waves and therefore the heating on the top of the chromosphere . however , the extremely narrow transition region corresponding to sharp gradient in density and alfvn speed should act as a wave barrier to reflect the alfvn waves . in this case , the wavelength is comparable to the size of the magnetopause of the hot jupiter ( wright 1987 ) unless the high - frequency modes are largely excited at the reconnection site . the energy carried by the planet - induced alfvn waves along the open fields might be finally transmitted to coronal loops and therefore might be dissipated via resonance absorption , a damping mechanism however in contradiction to the observations of the solar corona and the coronae of cool stars ( schrijver & aschwanden 2002 ; demoulin et al . the heating due to accelerated particles by alfvn waves ( crary 1997 ) is not important either because unlike the interplanetary space , the dense corona eliminates the kinetic effect of plasmas inside 0.04 au . despite the difficulties , the energy deposit on the surface of the star from the planet - induced alfvn disturbances may be achieved by interacting non - linearly with stellar winds . the stellar wind consists of charges particles , stellar fields , and probably alfvn waves . non - linear interactions between planet - induced incoming and stellar intrinsic outgoing waves is one route of heating the surface of the star . the scenario of magnetic interaction implies the orbital decay of hot jupiters since the ultimate source of energy comes from the orbital energy . theoretically the orbital decay of hot jupiters on the timescale of a few billion years can result from the tidal dissipation in the host star driven by the hot jupiter so long as the tidal dissipation in the solar - type stars is efficient ( rasio et al . 1996 , witte & savonije 2002 , ptzold & rauer 2002 , jiang et al . 2003 ) . in the magnetic interaction scenario for ca ii emission , the time scale of orbital decay is roughly equal to the ratio of the orbital energy of the hot jupiter ( @xmath810@xmath120 ergs ) to @xmath121 erg / s . this gives a timescale as short as several billion years , imposing a non - negligible constraint on modeling the orbital evolution of hot jupiters . in the case of @xmath5 boo , the f - type star might be almost tidally locked to its hot jupiter . therefore there is less free energy available from the planet s orbit . according to eq([eq : energyflux ] ) , the azimuthal relative motion between the stellar and the planetary magnetospheres is smaller than that for the hd 179949 system by a factor of 0.2 if 3.2 d is indeed the rotation period of @xmath5 boo . the mean chromospheric emission @xmath32k@xmath50 from @xmath5 boo is weaker than that from hd 179949 roughly by a factor 0.9 . assuming that the mean stellar field of @xmath5 boo is smaller than that of hd 179949 by 0.9@xmath122 ( @xmath32k@xmath50 @xmath123 @xmath124 ) , one can estimate that the input energy to @xmath5 boo is less than that to hd 179949 roughly by a factor of 0.15 . if the corresponding ca ii emissions are dimmed by this same factor , this effect should have been detected . however unlike the other two f - type stars hd 179949 and @xmath1 and , the variability of ca ii emissions from @xmath5 boo did not show any consistent phase relation with the planet s orbit . note that the radial movement of spiral stellar fields with the stellar winds might be important in providing energy in this case because @xmath125 at 0.04 au in the solar system . however in the case of small pitch angles of the field lines inside the alfvn radius , the alfvn modes driven solely by the radial impact of the stellar winds might not be as efficient as the stellar fields sweeps across the hot jupiter due to the slow relative azimuthal motion . the transiting system hd 209458 did not show synchronized ca ii enhancement , perhaps because g stars have @xmath126 at the distance of the hot jupiter due to weaker stellar fields . the non - synchronized ca ii enhancement of @xmath5 boo and hd 209458 ( as well as the flaring on hd 73256 and @xmath0 ceti , discussed in section [ modbyrot ] ) might still be attributed to direct interactions with their planets but through more chaotic means due to the variability of the stellar winds at the locations of these hot jupiters relative to the nearby alfvn radius . monitoring these stars continuously through several orbital and rotational periods should pin down the cause @xmath19 intrinsic due to fast rotation or induced by a hot jupiter @xmath19 of the strong night - to - night variabilities detected in these systems . of the sample of stars observed from the cfht , those with planets ( with the exception of 51 peg ) show significant night - to - night variations in their ca ii h and k reversals . @xmath5 ceti and the sun , which have no close planets , remained very steady throughout each of the four observing runs . hd 179949 and @xmath1 and exhibited repeated orbital phase - dependent activity with enhanced emission leading the sub - planetary point by 0.17 and 0.47 in orbital phase , respectively . both systems are consistent with a magnetic heating scenario and may be the first glimpse at the magnetospheres of extrasolar planets . the phase - lead or lag of the peak emission relative to the sub - planetary longitude can provide information on the field geometries and the nature of the effect such as tidal friction , magnetic drag or reconnection with off - center magnetic fields , including a parker - spiral type scenario . @xmath5 boo and hd 209458 also exhibited night - to - night variations that could not exclusively be due to stellar rotation . if @xmath5 boo is indeed tidally locked by its hot jupiter , there is no orbital energy available to generate the alfvn modes efficiently as the stellar fields sweep across the planet due to the slow relative azimuthal motion . we measured the excess absolute flux released in the enhanced chromospheric emission of hd 179949 to be the same order of magnitude as a typical solar flare , @xmath8 10@xmath47 erg s@xmath20 or 1.5@xmath2610@xmath48 ergs s@xmath20 @xmath49 . this implies that flare - like activity triggered by the interaction of a star with its hot jupiter may be an important energy source in the stellar outer atmosphere . this offers a mechanism for short - term chromospheric activity . the h & k emission of @xmath0 ceti , an active star with no confirmed planet , was clearly modulated by the star s 9.3-d rotation . similarly , hd 73256 displayed rotational modulation with its 14-d period . in these two cases , the chromospheric emission increases by @xmath6 6% ( relative to the normalization level at 1/3 of the continuum ) . any planet - induced heating at the level of 1 @xmath19 2% could have been diluted by the dominating hotspot on the stellar surface . neither the @xmath27rvs nor the ca ii periodicity exclude the possibility of a sub - stellar companion in a tight orbit around @xmath0 ceti . apart from the cyclical component for four of the stars , short - term chromospheric activity appears weakly dependent on the mean k - line reversal intensities for the sample of 13 stars . also , a suggestive correlation exists with @xmath2sin@xmath3 and thus with the hot jupiter s magnetic field strength . because of their small separation ( @xmath4 0.1 au ) , many of the hot jupiters lie within the alfvn radius of their host stars which allows a direct magnetic interaction with the stellar surface . additional ca ii observations are crucial to confirm the stability of the magnetic interaction as well as to establish better phase coverage . observations on timescales of a few years will begin to characterize the long - term activity of our program stars and allow us to see correlations between intrinsic ca ii emission and night - to - night activity more clearly . this work opens up the possibility of characterizing planet - star interactions with implications for extrasolar planet magnetic fields and the energy contribution to stellar atmospheres . a next step in understanding planet - star interactions is to map the activity as a function of stellar atmospheric height . above the chromosphere lies the thin transition region ( tr ) , where the temperature increases steeply as density and pressure drop , and the corona , which can extend out to several stellar radii . and since the magnetic field drops off as @xmath82 ( where @xmath127 ) , these layers facilitate a stronger interaction with the planet . their fuv and x - ray emissions will be extremely important diagnostics . one indication that the heating is from the outside in is if the increase in emission occurs slightly earlier in phase than in ca ii . moreover , the relative strengths of the different emission lines will tell us where most of the energy is dissipated . the energy sum will point out if there are any discrepancies with the theorized energy budget . orbital phase - dependent variability at these heights will constrain further the nature , form and strength of the interaction as well as specify non - thermal radiative processes in these hot layers of gas . we are grateful to marek wolf and petr harmanec for their photometric observations of hd 179949 made at the south african astronomical observatory ( saao ) . we thank geoff bryden , peng - fei chen , gary glatzmaier , gordon i. ogilvie , and ethan t. vishniac for useful communications regarding section [ theory ] . research funding from the canadian natural sciences and engineering research council ( g.a.h.w . & e.s . ) and the national research council of canada ( d.a.b . ) is gratefully acknowledged . we are indebted to the cfht staff for their care in setting up the cafe fiber feed and the gecko spectrograph , as well as to the staff at eso s vlt for their telescope and instrument support and the real - time data - reduction pipeline . also , we appreciate the helpful comments and suggestions from the referee , steve saar . mazeh , t. , naef , d. , torres , g. , latham , d.w . , mayor , m. , beuzit , j .- l . , brown , t.m . , buchhave , l. , burnet , m. , carney , b.w . , charbonneau , d. , drukier , g. , laird , j.b . , pepe , f. , perrier , c. , queloz , d. , santos , n.c . , sivan , j .- p . , udry , s. , zucker , s. , 2000 , , 532 , l55 saar , s.h . , shkolnik , e. , cuntz , m. , 2003 , in iau symp . 219 , stars as suns : activity , evolution and planets , international astronomical union . symposium no . 219 , held 21 - 25 july , in sydney , australia , meeting abstract walker , g.a.h . , matthews , j.m . , kuschnig , r. , johnson , r. , rucinski , s. , pazder , j. , burley , g. , walker , a. , skaret , k. , zee , r. , grocott , s. , carroll , k. , sinclair , p. , sturgeon , d. , harron , j. , 2003b , , 115 , 1023

the measurements of the higgs boson and top quark masses can be used to extrapolate the standard model higgs potential at energies up to the planck scale . adopting a nnlo renormalization procedure , we : i ) find that electroweak vacuum stability is at present allowed , discuss the associated theoretical and experimental errors and the prospects for its future tests ; ii ) determine the boundary conditions allowing for the existence of a shallow false minimum slightly below the planck scale , which is a stable configuration that might have been relevant for primordial inflation ; iii ) derive a conservative upper bound on type i seesaw right - handed neutrino masses , following from the requirement of electroweak vacuum stability . * higgs boson and top quark masses + as tests of electroweak vacuum stability * isabella masina + + +

1209.0393

the recent discovery of a particle consistent with the standard model ( sm ) higgs boson , announced by the atlas @xcite and cms @xcite collaborations at cern , is a milestone in particle physics ; adding in quadrature statistical and systematic errors , the mass of the particle turns out to be in the range @xmath1 gev at @xmath2 . here we assume that the new particle is actually the sm higgs boson and study the implications that its mass value , together with other relevant parameters such as the top quark mass and the strong gauge coupling , has on the behavior of the higgs potential at very high energy scales and , in particular , for the sake of electroweak vacuum stability . the project of extrapolating the higgs potential up to the planck scale is a long standing one @xcite , and was revamped in the fall of 2011@xcite after the first lhc hints of a higgs boson were reported@xcite . recently , the tools for a next - to - next - to - leading order ( nnlo ) renormalization procedure were derived @xcite . so , there are now all the ingredients necessary to carry out this long standing project . clearly , the extrapolation is based on the assumption that there is a desert up to the planck scale or , better , that possible new physics do not significantly affect the running of the higgs quartic coupling , which dominates the higgs potential at high energy . it is interesting that the recently discovered experimental higgs mass range , combined with the experimental top mass range , indicates a particularly intriguing high energy behavior of the higgs potential , close to the transition between electroweak vacuum stability and metastability . this is due to the fact that , for these higgs and top mass values , the higgs quartic coupling can be very small or even negative . since the dependence on the top mass is strong and quite subtle , it is not surprising that different groups slightly disagree in the interpretation of the results , some of them favoring @xcite and some others disfavoring @xcite electroweak vacuum stability . traditionally the top pole mass was used in the analysis ; however it has been pointed out @xcite that the top pole mass value used in previous analyses and taken to be the one measured at the tevatron , @xmath3 gev @xcite , is not unambiguously derived and that a more careful derivation should be based instead on the running top mass in the @xmath4 scheme , @xmath5 gev . as was shown in @xcite , the top pole mass range consistently derived from the running one , @xmath6 gev , is plagued by a larger error than the tevatron measurement considered in @xcite , rescuing electroweak vacuum stability . in our analysis we keep as a free parameter the running top mass , rather than the pole one . in this way we completely avoid the theoretical uncertainties associated with the top yukawa matching procedure . as we are going to discuss , the theoretical error associated with the higgs quartic coupling matching @xcite turns out to be smaller than the one induced by the experimental uncertainty in the strong gauge coupling , @xmath7 . given the above mentioned range for the running top mass @xcite , we find that electroweak stability is allowed in the whole higgs mass range @xcite . stability could soon be excluded if values of the running top mass @xmath8 gev are excluded by the lhc . otherwise , testing electroweak vacuum stability would become very challenging , since this would require precision measurements of the higgs and top masses , and also of @xmath7 . a stable higgs potential configuration which deserves particular interest is a shallow false minimum close to the planck scale , which could have been relevant for primordial inflation@xcite . we show that such a configuration is realized only if the higgs quartic coupling and its derivative satisfy very specific boundary conditions , possibly having a deep origin in quantum gravity . as is well know , new physics in addition to the sm is required to explain the neutrino masses and mixings , as well as dark matter . the mechanism responsible for the neutrino masses could affect the higgs quartic coupling ; as an example , we consider the impact that the inclusion of neutrino masses via a type i seesaw has on electroweak stability , discussing in some detail the shallow false minimum configuration . the paper is organized as follows . in sec.[sec-1 ] we discuss the input parameters and the nnlo renormalization procedure used to extrapolate the higgs potential up to the planck scale . an analysis of electroweak vacuum stability and the associated constraints on the top and higgs masses , with a detailed discussion of the theoretical errors and the prospects for the future , are presented in sec.[sec-2 ] . in sec.[sec-3 ] we investigate the boundary conditions leading to the particularly interesting configuration of a shallow false minimum below the planck scale . sec.[sec-4 ] is devoted to the upper bound on the seesaw right - handed neutrino masses following from the requirement of electroweak vacuum stability . conclusions are drawn in sec.[sec - concl ] . appendix[app - a ] contains the relevant formulas for the nnlo running procedure in the sm and , and those to incorporate the type i seesaw mechanism are found in appendix[app - b ] . the normalization of the higgs quartic coupling @xmath9 is chosen in this paper so that the potential for the physical higgs @xmath10 contained in the higgs doublet @xmath11 is given , at tree level , by @xmath12 where @xmath13 and @xmath14gev@xmath15 is the fermi constant from muon decay @xcite . the approximation in eq.([eq - v ] ) holds when considering large field values . according to our normalization , the physical higgs mass satisfies the tree level relation @xmath16 . in addition , the mass of the fermion @xmath17 reads , at tree level , @xmath18 , where @xmath19 denotes the associated yukawa coupling . in order to extrapolate the behavior of the higgs potential at very high energies , we adopt the @xmath4 scheme and consider the renormalization group ( rg ) evolution for the relevant couplings which , in addition to the higgs quartic coupling @xmath9 , are the gauge @xmath20 , @xmath21 , @xmath22 , and the top yukawa @xmath23 couplings . we work at nnlo , namely 3-loops for the @xmath24-functions and 2-loops for the matching conditions at some suitable scale . it is customary to introduce the dimensionless parameter @xmath25 , where @xmath26 stands for the renormalization scale and @xmath27 is the @xmath28 boson mass . the rg equations for the relevant couplings are then given by @xmath29 where @xmath30 and the apex on the @xmath24-functions represents the loop order . the 1-loop and 2-loop expressions for the @xmath24-functions can be found _ e.g. _ in ref.@xcite ( see also @xcite ) . recently , the complete 3-loop @xmath24-functions for all the sm gauge couplings have been presented by mihaila , salomon and steinhauser in ref.@xcite , while the leading 3-loop terms in the rg evolution of @xmath9 , @xmath23 and the higgs anomalous dimension have been computed by chetyrkin and zoller in ref.@xcite . for the sake of completeness , the expressions for the @xmath24-functions up to 3-loops are collected in appendix a. the matching of the running gauge couplings is done at the @xmath28 boson pole mass and apply immediately the six flavor running . the correction that would result by running with five flavors up to the top mass is very small and can be neglected , as discussed in @xcite . ] , @xmath27 . the numerical values used for the related @xmath4 observables are taken from the latest particle data group sm fit results @xcite : @xmath31 to match the @xmath32 running quartic coupling @xmath33 with the higgs pole mass @xmath34 is more complicated and requires one to exploit an expansion , @xmath35 which is known at present at nlo : @xmath36 is the 1-loop @xmath37 result of sirlin and zucchini @xcite while @xmath38 is the recently calculated 2-loop result , composed of a qcd contribution of @xmath39 @xcite and a yukawa contribution @xcite . more details can be found in appendix a. as is well known , there is some arbitrariness in the choice of the matching scale @xmath26 in eq.([match - la ] ) , which introduces a `` theoretical '' error in the rg procedure . in this work , we choose to perform the matching of the higgs quartic coupling @xmath9 at the scale @xmath40 . the theoretical uncertainty is estimated by performing the matching also at different scales and by evolving @xmath9 via rg running until @xmath40 . the spread in the numerical values obtained for @xmath41 can then be used to infer the magnitude of the theoretical error . this is illustrated in fig.[fig - match - la ] , assuming for definiteness a top pole mass @xmath42gev . the dashed and solid curves show the value of @xmath41 obtained by including the corrections up to 1-loop and 2-loop respectively , for various choices of the matching scale : from top to bottom @xmath43 , @xmath34 , @xmath44 , @xmath45 . one can see that , working at the 1-loop , the theoretical uncertainty is about @xmath46 . the inclusion of the 2-loop corrections given in ref.@xcite reduces the theoretical uncertainty down to about @xmath47 . notice also that the preferred region shrinks to small @xmath9 values and that @xmath43 and @xmath40 nearly overlap . more generally , one can use the following expression for the 2-loop result , @xmath48- 126 ) + 0.0015 \ , ( m_t[{\rm gev } ] - 172 ) \ , _ { - 0.0060}^ { + 0.0002 } \,\ , , \label{eq - lmh}\ ] ] where the mean value refers to @xmath40 . the reference values of @xmath49 and @xmath44 used in eq.([eq - lmh ] ) are not the central values that will be used in the following analysis ; they are just `` round numbers '' allowing for an easy inspection of the variation of @xmath41 as a function of @xmath34 and @xmath44 . notice that it is not possible to compare directly eq.([eq - lmh ] ) with eq.(63 ) of ref.@xcite , where @xmath50 is rather displayed ( adopting a normalization differing from ours by a factor of @xmath51 and choosing as reference values for @xmath34 and @xmath44 their central ones ) : @xmath52 -125 ) -0.0004 \ , ( m_t [ { \rm gev}]-173.15 ) \pm 0.00140 $ ] , where the error is obtained by varying the matching scale between @xmath27 , @xmath44 , @xmath53 @xcite . we checked that our numerical code gives a result for @xmath50 consistent with the one of eq.(63 ) of ref.@xcite . indeed , choosing @xmath54 gev and @xmath55 gev , our code gives @xmath56 , when the matching scale @xmath57 respectively . this shows that the two results perfectly agree for @xmath58 , that the lower errors ( associated to the difference between @xmath58 and @xmath59 ) are in substantial agreement , while the upper errors ( associated to the difference between @xmath43 and @xmath60 ) are slightly different , ours being smaller . .5 cm obtained by performing the matching at different scales @xmath26 , indicated by the labels , as a function of @xmath34 . the solid ( dashed ) lines are obtained by including corrections up to 2-loop ( 1-loop ) . we fixed @xmath42 gev ( for different values see eq.([eq - lmh ] ) ) . , width=377 ] .1 cm it is common to extrapolate the @xmath61 top yukawa coupling @xmath62 from the matching condition between the running top mass @xmath63 and the top pole mass @xmath44 : @xmath64 where @xmath65 represent the electroweak contribution , which is known at 1-loop @xcite , while @xmath66 is the qcd one . the qcd 1-loop result is known since many years @xcite ; the qcd 2-loop and 3-loop results as a function of the matching scale @xmath26 are given in @xcite ( see also @xcite ) . the matching is usually done at the top pole mass scale , and the theoretical error associated to the arbitrariness of the matching scale can be estimated as before , namely by comparing the values of @xmath67 obtained with different matching scales . this is represented in fig . [ fig - match - h ] , where the curves are obtained by working at 2-loop and using , from bottom to top , @xmath68 . the plot shows that the associated theoretical uncertainty is about @xmath69 . the analytical expression for @xmath67 is : @xmath70- 172)\ , ^{+0.017}_{- 0.013 } \label{eq - fit - h } \,\ , .\ ] ] the variations of @xmath71 due to the experimental range of @xmath72 and @xmath34 have not been explicitly written in eq.([eq - fit - h ] ) because they are negligible ( respectively of order @xmath73 and @xmath74 ) with respect to the variation of @xmath71 due to the experimental range of @xmath44 . the error quoted in eq.([eq - fit - h ] ) then refers only to the theoretical error coming from varying the matching scale @xmath26 from @xmath27 , @xmath44 ( mean value ) , @xmath53 . notice that our result perfectly agrees with the analogous expression derived in ref . @xcite , where however the error due to the variation of the matching was not estimated . .5 cm and @xmath75 as a function of @xmath44 . the curves are obtained by matching at different scales , which are indicated by the labels . we fixed @xmath76 gev for definiteness but the results do not significantly dependent on @xmath34 , provided it is chosen in its experimental range.,title="fig:",width=377 ] .1 cm the procedure adopted in previous analyses of the stability of the electroweak vacuum , including the latest ones @xcite , was to use the experimental value of @xmath44 , identified with the one measured at the tevatron by the cdf and d0 collaborations , @xmath77gev@xcite , to extrapolate the running yukawa @xmath67 via eq.([eq - fit - h ] ) . however , as discussed in ref.@xcite , it is not meaningful to use the mass parameter provided by the tevatron as the pole top mass to be inserted in eq.([eq - fit - h ] ) : the running top mass in the @xmath4 scheme is instead a well defined parameter that can be directly extracted at nnlo from tevatron measurements of the inclusive top pair production cross - section , giving @xmath78 gev@xcite . so , it is conceptually more robust and practically more convenient to extract the top yukawa coupling directly from @xmath79 , as will be done in the following is translated into a value of @xmath44 , to be inserted in the expression of the lower bound on @xmath34 ensuring electroweak vacuum stability as derived in ref.@xcite . ] . our results will thus be presented as a function of @xmath79 . notice that , according to eq.([eq - fit - h ] ) , the value of the top pole mass can be easily recovered via the relation @xmath80 gev , which however is plagued by a large uncertainty . in ref.@xcite it was found that , by doing a scheme transformation to nnlo accuracy from the running to the pole top mass , the range @xmath78 gev is equivalent to @xmath6 gev . hence , while displaying our results as a function of @xmath79 as already stated , motivated by the results of ref.@xcite , in some plots ( as the one in fig.[fig - line ] ) we will link the value of the top pole mass to the running mass via the simple relation @xmath81 gev . before presenting the results of our analysis in the following sections , we recall that , in order to carefully study the shape of the higgs potential at high energy , one should consider the renormalization improved effective potential . this can be done by introducing an effective coupling , @xmath82 , so that @xmath83 the expression for @xmath84 is known up to 2-loop @xcite ( and given , for instance , in@xcite ) . since the scalar contribution is not well defined when @xmath9 is negative ( a logarithm of a negative quantity appears ) , in the following we consider the renormalization improved potential at the tree level , and identify @xmath26 with @xmath10 . it is well known that this simplification has a negligible impact in the determination of the vacuum stability bound ( for a detailed discussion see _ e.g. _ ref.@xcite ) to be discussed in the next section . the experimental region of the values of the higgs and top masses is very intriguing from the theoretical point of view , since the higgs quartic coupling could be rather small , vanish or even turn negative at a scale slightly smaller than the planck scale . accordingly , the behavior of the higgs potential at high energy changes drastically : if @xmath33 is always positive , the electroweak vacuum is a global minimum , possibly accompanied by another local minimum just below the planck scale , which could have played a role in primordial inflation@xcite ; if @xmath33 turns negative below @xmath85 , the electroweak vacuum correspondingly becomes metastable @xcite . .8 cm , for @xmath76 gev and different values of @xmath75 , increasing from top to bottom by the amount indicated by the labels . the dashed curve in the right plot shows the associated value of @xmath86 . the other input parameters are fixed at the central values discussed in the previous section . , title="fig:",width=268 ] , for @xmath76 gev and different values of @xmath75 , increasing from top to bottom by the amount indicated by the labels . the dashed curve in the right plot shows the associated value of @xmath86 . the other input parameters are fixed at the central values discussed in the previous section . , title="fig:",width=283 ] .1 cm these drastically different possibilities for the behavior of the renormalization improved higgs potential at high energy are illustrated in the left plot fig.[fig - shapes ] , where @xmath76gev and some specific values for @xmath75 have been selected , increasing from top to bottom . the right plot shows the associated values of @xmath33 . let start considering the value @xmath87 gev . increasing the latter by just @xmath88 mev , the potential develops an inflection point ; notice that the associated @xmath33 becomes as small as @xmath89 . increasing again @xmath75 by about @xmath90 kev , the minimum of @xmath33 is equal to zero : a second vacuum degenerate with the electroweak one is obtained . further increasing @xmath75 makes @xmath33 turn negative : the electroweak vacuum becomes metastable . the dashed curve in the right plot in fig.[fig - shapes ] shows the evolution of @xmath91 for the same parameter values ; there is only a single dashed curve because @xmath86 mildly depends on @xmath75 if the latter is in the range @xmath92 gev . let call @xmath93 the renormalization scale such that @xmath94 . clearly , only in the case of two degenerate vacua the conditions @xmath94 and @xmath95 are simultaneously met . for a shallow false minimum we instead have @xmath94 and @xmath96 , as already mentioned . in fig . [ fig - mubmt ] we show how @xmath93 depends on @xmath75 , for various values of @xmath34 . it is interesting that @xmath93 is maximized and nearly constant for the values of @xmath75 for which @xmath33 is very small . 0.5 cm as a function of @xmath75 and for different values of @xmath34 , as indicated by the labels.,title="fig:",width=359 ] .5 cm we now turn to the determination of the points in the plane @xmath97 $ ] allowing for the existence of a second minimum degenerate with the electroweak one . these points belong to a line separating the stability from the metastability region , see fig.[fig - line ] : in the lower part of the plot @xmath33 is always positive , while in the upper part it becomes negative before reaching the planck scale . the configuration of a shallow false minimum belongs to the stability region , but the associated points are so close to the transition line that they could not be distinguished visually . $ ] where a second vacuum , degenerate with the electroweak one , is obtained just below the planck scale . the ( red ) diagonal arrow shows the effect of varying @xmath98 @xcite ; the ( blue ) horizontal one shows the effect of varying @xmath99 ( the matching scale of @xmath9 ) from @xmath27 up to @xmath45 . the shaded ( yellow ) vertical region is the @xmath2 atlas@xcite and cms@xcite combined range , @xmath100 gev ; the shaded ( green ) horizontal region is the range @xmath78gev , equivalent to @xmath101 gev@xcite . , width=434 ] .5 cm the transition line of fig.[fig - line ] was obtained with the input parameter values discussed in the previous section and by matching the running higgs quartic coupling at @xmath34 . clearly , it is also important to estimate the theoretical error associated to experimental ranges of the input parameters and the one associated to the matching procedure . to illustrate this , we consider in particular the point on the transition line associated to the value @xmath76 gev ; for such point , @xmath9 and @xmath102 both vanish at a certain scale @xmath93 ( see fig.[fig - mubmt ] ) . the arrows show how , if some inputs or the matching scale are changed , the position of this point have to change in order to keep having , at the same scale @xmath93 , a vacuum degenerate with the electroweak one . the diagonal arrow is obtained by varying the strong coupling in its allowed range , @xmath103 @xcite ; the short ( long ) dashed line shows how the solid line would move if @xmath7 were equal to its minimum ( maximum ) presently allowed value . notice that the error on @xmath7 induces an uncertainty in both the higgs and top masses of about @xmath104 gev . in ref.@xcite the impact of the variation of @xmath7 on @xmath34 was estimated to be @xmath105 gev ( see their table 1 ) . the two results are in substantial agreement , considering that in our analysis @xmath103 at @xmath106 @xcite , while ref.@xcite considers a smaller error , @xmath107 at @xmath106 . since the variation of the other input parameters in eq.([eq-3 ] ) induces much smaller effects then the one due to @xmath7 , they have not been reported in the fig.[fig - line ] . the horizontal arrow represents instead the theoretical error obtained by varying @xmath99 , the matching scale of the higgs quartic coupling , from @xmath43 to @xmath108 ; notice that the associated error is very asymmetric ( see fig.[fig - match - la ] ) : essentially it can only enhance @xmath34 , by at most @xmath109 gev . clearly , similar considerations apply to each point of the transition line . we note that in ref.@xcite the impact of the variation of the matching scale of @xmath9 on @xmath34 was estimated to be @xmath104 gev ( see their table 1 ) , hence close to our estimate but with a symmetric error . fig.[fig - line ] shows that stability can be achieved in the whole experimental range for @xmath34 ( shaded vertical region ) , but this is not the case for @xmath75 ( shaded horizontal region ) . so , it is convenient to write down the condition of electroweak vacuum stability under the form of an upper bound on the running top mass : @xmath110 \le 162.0 + 0.47 \ , ( m_h[{\rm gev } ] - 126 ) + 0.7 \ , \left ( \frac{\alpha_3(m_z ) - 0.1196}{0.0017 } \right ) -0.2^{(\mu_\lambda)}_{th } \,\,\ , , \label{eq - boundmt}\ ] ] where the last term accounts for the ( very asymmetric ) theoretical error induced by the matching of @xmath9 . the latter turns out to be smaller than the variation induced by varying @xmath7 in its presently allowed experimental range . we recall that the relation between the running and pole top mass is simply @xmath111 gev . fig.[fig - resi ] summarizes our results for the determination of the transition line between stability and metastability in the @xmath112 $ ] plane . the three lines correspond to the central and @xmath113 values of @xmath7 @xcite and their thickness represents the theoretical error due to the matching of @xmath9 . the shaded rectangle emphasizes the present allowed region for @xmath44 @xcite and @xmath34 @xcite . according to our analysis it is not possible , given the present experimental situation , to understand whether we live in a stable or metastable vacuum configuration happens as in the sm up to energies close to the planck scale , without significant modifications . ] . $ ] and for fixed values of @xmath98 @xcite . the thickness of the lines represents the theoretical error due to the variation of @xmath99 ( the matching scale of @xmath9 ) from @xmath27 up to @xmath45 . the shaded region is obtained by intersecting the @xmath2 atlas@xcite and cms@xcite combined range ( @xmath100 gev ) with the running top mass range given by ref.@xcite , @xmath78gev , equivalent to @xmath101 gev . , width=434 ] .5 cm in order to discriminate between the two possibilities , it would be crucial to better determine @xmath75 . as discussed in @xcite , after lhc the higgs mass will presumably be known with an accuracy of @xmath114 mev @xcite , but the precision on the top mass would improve only by a factor of two . for instance , if the whole range of @xmath8 gev ( or , equivalently @xmath115 gev ) would be excluded , we would conclude that our vacuum is metastable ; otherwise the investigations should continue . a self - consistent and precise determination of the top quark mass can best be performed at a high - energy electron - positron collider , with a planned accuracy of @xmath114 mev . moreover , at an electron - positron collider @xmath7 could be determined with an accuracy close to or better than @xmath116 ( this precision is sometimes currently adopted@xcite but can not be considered to be conservative according to ref.@xcite ) . at this stage , if the stability region will still have an overlap with the allowed ranges of the top and higgs masses , we will be mostly limited by the theoretical uncertainty associated to @xmath99 . notice also that it is not realistic to hope to distinguish the case of two degenerate minima with the one of a shallow false minimum , since the difference in the top mass is just about @xmath90 kev ( see fig . [ fig - shapes ] ) . we now discuss how to compare eq.([eq - boundmt ] ) and fig.[fig - resi ] with previous literature results , in particular those of ref.@xcite , since the authors claimed that `` absolute stability of the higgs potential is excluded at @xmath117 c.l . for @xmath118 gev '' ( see also their fig.5 ) . in the latter work the stability condition is indeed expressed under the form of a lower bound on the higgs mass : @xmath119 , where all masses are in gev and the last term represents the overall theoretical error . combining in quadrature their theoretical uncertainty and their experimental errors on @xmath44 and @xmath7 , the authors derive @xmath120 gev , which motivates the quoted claim of ref.@xcite . in order to carry out the comparison , one must rewrite the inequality of @xcite under a form directly comparable with eq.([eq - boundmt ] ) , namely : @xmath121 } [ { \rm gev } ] \le 171.8 + 0.5\ , ( m_h[{\rm gev}]-126)+ 0.61\ , \left ( \frac{\alpha_3(m_z ) - 0.1196}{0.0017 } \right ) \pm 0.5_{th } \ , . \label{eq-14}\ ] ] so , our results eq.([eq - boundmt ] ) and the one obtained in ref.@xcite are perfectly compatible , as the central value of eq.([eq-14 ] ) essentially overlaps with the lower value of eq.([eq - boundmt ] ) . the theoretical error in eq.([eq - boundmt ] ) is however smaller than the theoretical error of ref.@xcite : this is mainly due to the fact that in the present analysis we used directly the running top mass , thus avoiding to introduce the theoretical error due to the matching scale of the top yukawa coupling ( see table 1 of ref.@xcite ) . the left plot of fig.[fig - compis ] shows the comparison between eq.([eq - boundmt ] ) and eq.([eq-14 ] ) in the determination of the transition line between stability and metastability in the @xmath112 $ ] plane ; we choose @xmath122 for definiteness , so that the thickness of the lines represents just the theoretical error . according to eq.([eq - boundmt ] ) the thickness of the line is @xmath123 gev , while according to eq.([eq-14 ] ) it is @xmath88 gev ( as can also be checked by inspecting table 1 of @xcite ) . ) ( thinner ) and eq.([eq-14 ] ) of ref.@xcite ( thicker ) . for definiteness we choose @xmath122 . right : transition line between stability and metastability according to eq.([eq-14 ] ) of ref.@xcite ; the thickness of the band accounts for both the @xmath88 gev theoretical error and the experimental error due to the variation of @xmath7 in the range @xmath124 , as done in @xcite . the ( brown ) shaded disks represent the @xmath106 and @xmath2 combined ranges for @xmath44 and @xmath34 used in @xcite ( see their fig.5 ) . the ( green ) rectangle allows for the comparison with the ranges of @xmath44 and @xmath34 used here ( see fig.[fig - resi]).,title="fig:",width=294 ] ) ( thinner ) and eq.([eq-14 ] ) of ref.@xcite ( thicker ) . for definiteness we choose @xmath122 . right : transition line between stability and metastability according to eq.([eq-14 ] ) of ref.@xcite ; the thickness of the band accounts for both the @xmath88 gev theoretical error and the experimental error due to the variation of @xmath7 in the range @xmath124 , as done in @xcite . the ( brown ) shaded disks represent the @xmath106 and @xmath2 combined ranges for @xmath44 and @xmath34 used in @xcite ( see their fig.5 ) . the ( green ) rectangle allows for the comparison with the ranges of @xmath44 and @xmath34 used here ( see fig.[fig - resi]).,title="fig:",width=294 ] .5 cm clearly all these considerations do not justify the different conclusions of the two papers and rather show that the different conclusions have to come from the different ranges used for the three most relevant parameters : @xmath44 , @xmath34 and @xmath7 . in ref.@xcite it is assumed that @xmath125 gev and @xmath126 gev ; these errors are further combined in quadrature and the @xmath106 and @xmath2 ( brown ) disks in the right plot of fig.[fig - compis ] are obtained . these disks have to be confronted with our ( green ) rectangular region , obtained by using @xmath127 gev , as suggested in ref.@xcite , and @xmath128 gev , as suggested combining the atlas@xcite and cms@xcite ranges at @xmath2 . we have a rectangular region since we think that in this kind of analysis it is not really justified to combine in quadrature the errors on @xmath44 and @xmath34 , thus enhancing the exclusion of the interesting low-@xmath44 and high-@xmath34 values . a small value of @xmath7 also goes in such direction , since it lowers the transition line towards smaller values of @xmath44 . in ref.@xcite it is assumed that @xmath129 : the corresponding transition line is displayed in the right plot of fig.[fig - compis ] , using for consistency eq.([eq-14 ] ) . the line perfectly reproduces the results of fig.5 of @xcite ; its thickness accounts for both the theoretical error and the experimental error due to the variation of @xmath7 , as derived in @xcite . ( we can not display three separate lines as done in fig.[fig - resi ] since in eq.([eq-14 ] ) the theoretical error and the one associated to the variation of @xmath7 are comparable ) . as the transition line marginally overlaps with the @xmath2 disk , the authors of ref.@xcite concluded that stability is disfavored . a very different conclusion would be derived by considering instead the broad overlap with the rectangle . this is the main reason of the different conclusions . a small effect is also played by the different values used for @xmath7 . the range of @xmath7 used in ref.@xcite has a very small error and has already been questioned in ref.@xcite ( see bottom of pag 8) . in the present analysis we rather use @xmath130 @xcite , whose central value and experimental error are bigger than those used in ref.@xcite . as an effect , the ensemble of the three ( red ) lines in fig.[fig - resi ] forms a band slightly wider and higher than the ( blue ) band in fig.[fig - compis ] depicting the results of @xcite . summarizing , upon comparison of our results in fig.[fig - resi ] with the results of ref.@xcite reproduced in the right plot of fig.[fig - compis ] , one can conclude that the difference in the physical interpretation of the results is mainly due to the fact that ref.@xcite adopts a too small experimental error for @xmath44 , as already pointed out in ref.@xcite . it is interesting to study in some detail the boundary conditions which must be satisfied in order to have a very shallow false minimum just below the planck scale , since it could be relevant for inflation @xcite . to study this particular configuration , we denote with @xmath131 the renormalization scale where the higgs potential has an inflection point ; we also recall that @xmath93 has been defined to be the scale where @xmath95 and @xmath94 are simultaneously fulfilled . both @xmath131 and @xmath93 increase is slightly smaller than @xmath93 . this can be easily understood , since the condition for having an inflection point at @xmath131 reads @xmath132 , which implies @xmath133 . ] with @xmath34 , as shown in fig.[fig - mu ] , where the shaded region accounts for the experimental range of @xmath7 . .3 cm ( left ) and @xmath93 ( right ) as a function of @xmath34 . for the solid lines , the input parameters are fixed at their central values and the matching of @xmath9 is done at @xmath40 . the shaded regions shows the uncertainty induced by the experimental error of @xmath103 : the short and long dashed curves refer to the lower and upper value at @xmath106 , respectively . , title="fig:",width=283 ] ( left ) and @xmath93 ( right ) as a function of @xmath34 . for the solid lines , the input parameters are fixed at their central values and the matching of @xmath9 is done at @xmath40 . the shaded regions shows the uncertainty induced by the experimental error of @xmath103 : the short and long dashed curves refer to the lower and upper value at @xmath106 , respectively . , title="fig:",width=283 ] 1 cm .3 cm as a function of @xmath34 . right : the higgs potential at @xmath131 and the associated prediction for @xmath134 as a function of @xmath34 . the short and long dashed curves refer to the @xmath106 lower and upper values of @xmath7 , respectively . the shaded ( yellow ) vertical region marks the preferred range of @xmath34 at @xmath2 @xcite . the upper region in the right plot is excluded because @xmath135 @xcite.,title="fig:",width=283 ] as a function of @xmath34 . right : the higgs potential at @xmath131 and the associated prediction for @xmath134 as a function of @xmath34 . the short and long dashed curves refer to the @xmath106 lower and upper values of @xmath7 , respectively . the shaded ( yellow ) vertical region marks the preferred range of @xmath34 at @xmath2 @xcite . the upper region in the right plot is excluded because @xmath135 @xcite.,title="fig:",width=298 ] .5 cm it is interesting that , for the whole experimental range of @xmath34 , a shallow false minimum is obtained only if the following boundary condition holds : ( _ ) ( 8.750.15 ) 10 ^ -6 . [ eq - lamub ] one could speculate that such value could be originated by some quantum gravity effect @xcite . in the left plot of fig.[fig - lamub ] we show that @xmath136 has a mild dependence on @xmath34 ; in the right plot we show instead the value of the higgs potential at the inflection point , which turns out to be of @xmath137 gev . as before , the shaded regions account for the experimental range of @xmath7 . as pointed out in @xcite , a way of testing the hypothesis that inflation occurred when the higgs field was trapped into a shallow false vacuum below the planck scale is to look at the tensor - to - scalar ratio @xmath134 of cosmological perturbations . the amplitude of density fluctuations in the observed universe as seen by the cmb and large - scale structure data is parametrized by the power spectrum in @xmath138-space , @xmath139 , where @xmath140 is the amplitude at some pivot point @xmath141 , whose best - fit value is @xmath142 at @xmath143 @xcite . in models where inflation happened while the higgs was trapped in the shallow minimum @xcite , the higgs potential at the inflection point and the amount of gravity waves that can be produced - parametrized via the tensor - to - scalar ratio @xmath134 - are linked via a simple relation : _ r^2 = , [ eq - r ] where @xmath144 is the reduced planck scale . such prediction for @xmath134 is reported in the right plot of fig . [ fig - lamub ] . notice that , for these models , only if @xmath34 is in its upper allowed range and @xmath7 is quite low , there are chances for the planck satellite mission @xcite to measure @xmath134 . however , the forthcoming experiment epic @xcite should be able to test @xmath134 down to @xmath145 , while core @xcite down to about @xmath146 . we now consider the effect of including neutrino masses via a type i seesaw . this issue has been already considered in a series of papers @xcite . although the precise amount of the effect is quite model dependent , here we obtain a conservative estimate of the effect by considering only one right handed neutrino with mass @xmath147 , associated to a light majorana neutrino with mass @xmath148 ev , the scale of the atmospheric oscillations . this is supported by the following argument . it is well known that the @xmath24-function of the higgs quartic coupling is affected only if @xmath149 , the yukawa coupling of the dirac mass term ( defined only for @xmath150 ) , is large enough . as the top yukawa coupling , also the neutrino yukawa coupling induces a suppression of the higgs quartic coupling at high energy . by increasing @xmath147 and @xmath151 , the neutrino yukawa coupling at the threshold scale @xmath147 also increases : h_(m _ ) = 2 . this justifies that the fact that we equate @xmath151 to the the atmospheric mass scale , about @xmath152 ev , which is the lowest possible value for the heaviest among the three light neutrinos . in addition , two other majorana neutrinos with masses lighter than @xmath151 can be accommodated via the seesaw but , if their right - handed neutrinos are lighter than @xmath147 , the associated dirac yukawa couplings are naturally expected to be smaller , and their effect on @xmath33 negligible . in appendix b we provide the additional terms ( with respect to the pure sm ) for the relevant @xmath24-functions , above and below the scale @xmath147 . .3 cm as a function of the running top mass , following from the requirement that the electroweak vacuum is not destabilized because of the inclusion of the seesaw , for @xmath76 gev . the shaded region is obtained by varying @xmath7 in its @xmath106 range . , title="fig:",width=359 ] .5 cm since the effect of @xmath153 is a suppression of @xmath9 , a sm configuration with a stable electroweak vacuum could be rendered metastable because of the addition of the seesaw interactions . for a fixed value of @xmath34 , and in the range of the top mass values allowing the electroweak vacuum to be the global one , one can find the upper bound on @xmath147 following from the requirement that the electroweak vacuum remains the global one even after the inclusion of the seesaw interactions . clearly such upper limit can not be derived in the range of the top mass values for which the electroweak vacuum is already metastable . as shown in fig.[fig - m ] for @xmath76 gev ( but similar upper bounds are obtained in the whole experimental range of @xmath34 ) , such upper bound strongly depends on the top mass and is affected by an uncertainty which is mainly due to @xmath7 ( shaded region ) . the smaller the top mass is , the more the configuration of the higgs potential is stable and the less stringent is the @xmath147 upper bound ensuring that the electroweak vacuum remains the global one and does not become metastable , @xmath154 gev . but increasing the top mass , the electroweak vacuum becomes less stable and the upper bound on @xmath147 becomes accordingly more and more stringent . increasing further the top mass the electroweak vacuum becomes metastable even without seesaw interactions , so that no meaningful bound can be derived . the upper bound on @xmath147 following from the requirement of electroweak vacuum stability has to be taken _ cum grano salis _ , in the sense that it is not a physically robust bound , but just a bound that should be respected in the case one has a model in which the higgs potential has to remain stable for reason . let consider in particular the upper bound on @xmath147 needed to avoid destabilization of an inflection point configuration , as the one depicted via the dashed line in fig.[fig - nu ] . notice that an inflection point becomes a not so shallow local second minimum if @xmath155 gev and that electroweak vacuum destabilization is avoided only if the condition @xmath156 gev is satisfied . the latter bound might be relevant for models of inflation based on the sm shallow false minimum @xcite ; note however that it is well compatible with the thermal leptogenesis mechanism to explain matter - antimatter asymmetry , for which the lower bound on the lightest majorana neutrino is about @xmath157 gev @xcite . .3 cm gev , @xmath158 and @xmath159 gev ( the value of the top mass leading to an inflection point configuration in the sm case once fixed the former two parameters ) . the solid lower curves display the effect of adding the seesaw , with three increasing values of @xmath147 from top to bottom.,title="fig:",width=415 ] .5 cm clearly , the neutrino yukawa coupling @xmath160 is not the only additional term beyond the sm capable of modifying the running of @xmath9 at high energy . always in the context of type i seesaw , in the case that the vacuum expectation value of a singlet scalar field @xmath161 ( violating the lepton number by two units ) is actually at the origin of the right - handed majorana neutrino mass , the @xmath161 couplings induce an enhancement of @xmath9 , thus helping the stability of the electroweak vacuum @xcite . such effect is indeed generically expected when adding to the sm a singlet field @xmath161 @xcite . the recent discovery of a particle consistent with the sm higgs boson @xcite provides a strong motivation to pursue @xcite the old project @xcite of investigating the behavior of the sm higgs potential at very high energies . in particular , one would understand whether the sm electroweak vacuum is a global minimum up to the planck scale , namely whether we live in a stable vacuum assuming a desert ( or assuming that new interactions do not modify the running of @xmath9 with respect to the sm case ) . in particular , a stable configuration which deserves a special interest is a shallow false minimum below the planck scale : the higgs field could have been primordially trapped there , leading to a stage of inflation@xcite . stability below the planck scale is required also in higgs inflation models with non minimal gravitational couplings @xcite . in our analysis , we adopted the recently derived tools for a nnlo renormalization procedure@xcite . at difference of previous analyses , we considered as free parameter the running top mass rather than the ( tevatron ) top pole mass , as suggested in ref.@xcite . given the present range of the running top mass and of the higgs mass , we found that electroweak vacuum stability is at present allowed , as shown in figs.[fig - line ] and [ fig - resi ] . to further test stability , a more precise measurement of the top mass would be crucial . as apparent from the stability condition of eq.([eq - boundmt ] ) , in case that lhc will not exclude values of the running top mass below @xmath162 gev ( or equivalently values of the pole top mass below @xmath163 gev ) , an electron - positron collider would probably be needed to discriminate between stability and metastability . we also determined the high scale boundary conditions allowing for a shallow false minimum slightly below the planck scale , @xmath164 ( @xmath93 is the renormalization scale were the @xmath24-function of the higgs quartic coupling vanishes ) , and discussed the prospects for the cosmological tests of such configuration . finally , a conservative upper bound on type i seesaw right - handed neutrino masses , following from the requirement of electroweak vacuum stability , was derived , analyzing in particular its dependence on the top mass . the present analysis does not consider the effect of the gravitational couplings because it is far from clear how the quantum effects of the latter would impact the study at very high energies . other sources of uncertainty could also come from the treatment of the effective potential itself , such as the fine - tuning required for both the cosmological constant and the higgs mass . we would like to thank g. isidori , g. moore , a. notari and a. strumia for useful discussions . here we provide the expressions for the @xmath24-functions up to 3-loops , see eq . ( [ eq - rge ] ) . at 1-loop they are given by : @xmath165 at 2-loop they are : @xmath166 @xmath167 \ , \\ \nonumber \end{aligned}\ ] ] @xmath168 the leading terms in the 3-loop @xmath24-functions of @xmath9 and @xmath23 are @xcite : @xmath169 \ , , \nonumber \\ \nonumber \beta_{h_t}^{(3)}&= & 2 \left [ ( -\frac{2083}{3 } + 320 \zeta_3 ) g_3(t)^6 + ( \frac{3827}{12 } - 114 \zeta_3 ) g_3(t)^4 h_t(t)^2 - \frac{157}{2 } g_3(t)^2 h_t(t)^4 \right . \\ & & + ( \frac{339}{16 } + \frac{27}{4 } \zeta_3 ) h(t)^6 + \frac{4}{3 } g_3(t)^2 h_t(t)^2 \lambda(t ) + \frac{33}{2 } h_t(t)^4 \lambda(t ) + \left . \frac{5}{96 } h_t(t)^2 \lambda(t)^2 -\frac{1}{12 } \lambda(t)^3 \right ] \ , , \nonumber\end{aligned}\ ] ] where @xmath170 is the riemann zeta function . the complete 3-loop @xmath24-functions for the gauge couplings are @xcite : @xmath171 according to sirlin and zucchini @xcite , the 1-loop matching is given by @xmath172 where @xmath173 and , introducing @xmath174 , @xmath175-z\left[\frac{c^2}{\xi } \right]+\frac{9}{2 } \left(\frac{25}{9}-\frac{\pi } { \sqrt{3}}\right)\ , , \nonumber\\ f_0(\mu)&=&\frac{3 c^2}{s^2}\log \left(c^2\right)+12 \log c^2\left(c^2\right ) + \frac{3 \xi c^2}{\xi -c^2}\log \left(\frac{\xi } { c^2}\right)+4c^2\,z\left[\frac{c^2}{\xi } \right ] -\frac{15}{2 } \left(2 c^2 + 1\right ) \nonumber\\ \nonumber & & -6 \left(2 c^2-\frac{2m_t^2}{m_z^2}+1\right ) \log \left(\frac{\mu^2}{m_z^2}\right ) -\frac{3m_t^2}{m_z^2 } \left(4 \log \left(\frac{m_t^2}{m_z^2}\right ) + 2\,z\left[\frac{m_t^2}{m_z^2 \xi}\right]-5\right ) \nonumber\\ & & + 2\ , z\left[\frac{1}{\xi } \right]\,,\nonumber \\ \nonumber f_{-1}(\mu)&=&8 \left(2 c^4 + 1\right)-12c^4 \log \left(c^2\right ) -12c^4\ , z\left[\frac{c^2}{\xi } \right ] + 6 \left(2 c^4-\frac{4 m_t^4}{m_z^4}+1\right ) \log \left(\frac{\mu^2}{m_z^2}\right)\nonumber\\ & & -6\,z\left[\frac{1}{\xi } \right]+\frac{24 m_t^4 } { m_z^4}\left(\log \left(\frac{m_t^2}{m_z^2}\right)+z\,\left[\frac{m_t^2}{m_z^2 \xi } \right]-2\right)\,,\nonumber\\ z[z]&=&\left\ { \begin{array}{lll } 2 a(z ) \,{\rm arctan}\left(\frac{1}{a(z)}\right ) & \textrm{if } & z>\frac{1}{4}\nonumber \\ a(z ) \log \left(\frac{a(z)+1}{1-a(z)}\right ) & \textrm{if } & z<\frac{1}{4}\nonumber \end{array } \right . \ , , \\ a(z)&=&\sqrt{|1 - 4 z| } \,.\nonumber\end{aligned}\ ] ] we compute the qcd and the yukawa contribution to @xmath176 following the expressions of @xcite ( multiplied them by a factor of @xmath51 to compensate for the different definition of the quartic coupling ) . below the right handed neutrino mass scale , the running of the effective light majorana neutrino mass is given by @xcite = ( - 3 g_2(t)^2 + 6 h_t(t)^2 + ) m_(t ) . g. aad _ et al . _ [ atlas collaboration ] , phys . b [ arxiv:1207.7214 [ hep - ex ] ] . s. chatrchyan _ et al . _ [ cms collaboration ] , phys . b [ arxiv:1207.7235 [ hep - ex ] ] . p. q. hung , phys . * 42 * ( 1979 ) 873 . n. cabibbo , l. maiani , g. parisi and r. petronzio , nucl . phys . b * 158 * ( 1979 ) 295 . m. lindner , z. phys . c * 31 * ( 1986 ) 295 . m. lindner , m. sher and h. w. zaglauer , phys . lett . b * 228 * ( 1989 ) 139 . m. sher , phys . * 179 * ( 1989 ) 273 . m. sher , phys . b * 317 * , 159 ( 1993 ) [ addendum - ibid . b * 331 * , 448 ( 1994 ) ] [ arxiv : hep - ph/9307342 ] . g. altarelli and g. isidori , phys . lett . b * 337 * , 141 ( 1994 ) . c. d. froggatt and h. b. nielsen , phys . b * 368 * ( 1996 ) 96 [ hep - ph/9511371 ] . b. schrempp and m. wimmer , prog . phys . * 37 * ( 1996 ) 1 [ hep - ph/9606386 ] . j. a. casas , j. r. espinosa and m. quiros , phys . lett . b * 342 * , 171 ( 1995 ) [ arxiv : hep - ph/9409458 ] . j. r. espinosa and m. quiros , phys . b * 353 * , 257 ( 1995 ) [ arxiv : hep - ph/9504241 ] . j. a. casas , j. r. espinosa and m. quiros , phys . b * 382 * ( 1996 ) 374 [ hep - ph/9603227 ] . g. isidori , g. ridolfi and a. strumia , nucl . b * 609 * ( 2001 ) 387 [ hep - ph/0104016 ] . j. r. espinosa , g. f. giudice and a. riotto , jcap * 0805 * ( 2008 ) 002 [ arxiv:0710.2484 [ hep - ph ] ] . g. isidori , v. s. rychkov , a. strumia and n. tetradis , phys . d * 77 * ( 2008 ) 025034 [ arxiv:0712.0242 [ hep - ph ] ] . n. arkani - hamed , s. dubovsky , l. senatore and g. villadoro , jhep * 0803 * ( 2008 ) 075 [ arxiv:0801.2399 [ hep - ph ] ] . j. ellis , j. r. espinosa , g. f. giudice , a. hoecker and a. riotto , phys . lett . b * 679 * ( 2009 ) 369 [ arxiv:0906.0954 [ hep - ph ] ] . m. holthausen , k. s. lim and m. lindner , jhep * 1202 * ( 2012 ) 037 [ arxiv:1112.2415 [ hep - ph ] ] . i. masina and a. notari , phys . d * 85 * ( 2012 ) 123506 [ arxiv:1112.2659 [ hep - ph ] ] . j. elias - miro , j. r. espinosa , g. f. giudice , g. isidori , a. riotto and a. strumia , phys . b * 709 * ( 2012 ) 222 [ arxiv:1112.3022 [ hep - ph ] ] . z. -z . xing , h. zhang and s. zhou , phys . d * 86 * ( 2012 ) 013013 [ arxiv:1112.3112 [ hep - ph ] ] . atlas collaboration , phys . b * 710 * , 49 ( 2012 ) ; phys . lett . * 108 * , 111803 ( 2012 ) ; arxiv:1202.1415 [ hep - ex ] . cms collaboration , arxiv:1202.1488 [ hep - ex ] . l. n. mihaila , j. salomon and m. steinhauser , phys . * 108 * ( 2012 ) 151602 [ arxiv:1201.5868 [ hep - ph ] ] . k. g. chetyrkin and m. f. zoller , jhep * 1206 * ( 2012 ) 033 [ arxiv:1205.2892 [ hep - ph ] ] . f. bezrukov , m. y. kalmykov , b. a. kniehl and m. shaposhnikov , arxiv:1205.2893 [ hep - ph ] . s. alekhin , a. djouadi and s. moch , arxiv:1207.0980 [ hep - ph ] . [ tevatron electroweak working group and cdf and d0 collaborations ] , arxiv:1107.5255 [ hep - ex ] . i. masina and a. notari , phys . * 108 * ( 2012 ) 191302 [ arxiv:1112.5430 [ hep - ph ] ] . i. masina and a. notari , arxiv:1204.4155 [ hep - ph ] . j. beringer et al . ( particle data group ) , review of particle physics " , phys . d86 , 010001 ( 2012 ) c. ford , d. r. t. jones , p. w. stephenson and m. b. einhorn , nucl . b * 395 * , 17 ( 1993 ) [ arxiv : hep - lat/9210033 ] ; c. ford , i. jack and d. r. t. jones , nucl . phys . b 387 ( 1992 ) 373 [ erratum - ibid . b 504 ( 1997 ) 551 ] [ arxiv : hep - ph/0111190 ] . d. j. gross and f. wilczek , phys . lett . * 30 * ( 1973 ) 1343 . d. j. gross and f. wilczek , phys . d * 8 * ( 1973 ) 3633 . h. d. politzer , phys . lett . * 30 * ( 1973 ) 1346 . d. r. t. jones , nucl . b * 75 * ( 1974 ) 531 . w. e. caswell , phys . lett . * 33 * ( 1974 ) 244 . o. v. tarasov and a. a. vladimirov , sov . j. nucl . * 25 * ( 1977 ) 585 [ yad . fiz . * 25 * ( 1977 ) 1104 ] . o. v. tarasov , a. a. vladimirov and a. y. zharkov , phys . b * 93 * , 429 ( 1980 ) ; d. r. t. jones , m. fischler and j. oliensis , phys . b * 119 * ( 1982 ) 385 . m. e. machacek and m. t. vaughn , nucl . b * 222 * ( 1983 ) 83 . m. e. machacek and m. t. vaughn , nucl . b * 236 * ( 1984 ) 221 . m. e. machacek and m. t. vaughn , nucl . b * 249 * ( 1985 ) 70 . n. gray , d. j. broadhurst , w. grafe and k. schilcher , z. phys . c * 48 * , 673 ( 1990 ) . d. j. broadhurst , n. gray and k. schilcher , z. phys . c * 52 * , 111 ( 1991 ) . j. fleischer , f. jegerlehner , o. v. tarasov and o. l. veretin , nucl . b * 539 * , 671 ( 1999 ) [ erratum - ibid . b * 571 * , 511 ( 2000 ) ] [ arxiv : hep - ph/9803493 ] . k. melnikov and t. v. ritbergen , phys . b * 482 * ( 2000 ) 99 [ hep - ph/9912391 ] . f. jegerlehner and m. y. kalmykov , nucl . b * 676 * , 365 ( 2004 ) [ arxiv : hep - ph/0308216 ] . g. aad _ et al . _ [ atlas collaboration ] , arxiv:0901.0512 [ hep - ex ] . g. l. bayatian _ et al . _ [ cms collaboration ] , j. phys . g g * 34 * ( 2007 ) 995 . m. shaposhnikov and c. wetterich , phys . b * 683 * ( 2010 ) 196 [ arxiv:0912.0208 [ hep - th ] ] . e. komatsu _ et al . _ [ wmap collaboration ] , astrophys . j. suppl . * 192 * ( 2011 ) 18 [ arxiv:1001.4538 [ astro-ph.co ] ] . [ planck collaboration ] , astro - ph/0604069 . see also : c. burigana , c. destri , h. j. de vega , a. gruppuso , n. mandolesi , p. natoli and n. g. sanchez , astrophys . j. * 724 * ( 2010 ) 588 [ arxiv:1003.6108 [ astro-ph.co ] ] . j. bock _ et al . _ [ epic collaboration ] , arxiv:0906.1188 [ astro-ph.co ] . f. r. bouchet _ et al . _ [ core collaboration ] , arxiv:1102.2181 [ astro-ph.co ] . w. rodejohann and h. zhang , jhep * 1206 * ( 2012 ) 022 [ arxiv:1203.3825 [ hep - ph ] ] . j. chakrabortty , m. das and s. mohanty , arxiv:1207.2027 [ hep - ph ] . s. davidson and a. ibarra , phys . b * 535 * ( 2002 ) 25 [ hep - ph/0202239 ] . j. elias - miro , j. r. espinosa , g. f. giudice , h. m. lee and a. strumia , jhep * 1206 * ( 2012 ) 031 [ arxiv:1203.0237 [ hep - ph ] ] . see also : l. basso , s. moretti and g. m. pruna , phys . d * 82 * ( 2010 ) 055018 [ arxiv:1004.3039 [ hep - ph ] ] . o. lebedev , eur . j. c * 72 * ( 2012 ) 2058 [ arxiv:1203.0156 [ hep - ph ] ] . f. l. bezrukov and m. shaposhnikov , phys . lett . b * 659 * ( 2008 ) 703 [ arxiv:0710.3755 [ hep - th ] ] . f. bezrukov and m. shaposhnikov , jhep * 0907 * ( 2009 ) 089 [ arxiv:0904.1537 [ hep - ph ] ] . f. bezrukov , d. gorbunov and m. shaposhnikov , jcap * 1110 * ( 2011 ) 001 [ arxiv:1106.5019 [ hep - ph ] ] . for a classification of models with non minimal couplings see : k. kamada , t. kobayashi , t. takahashi , m. yamaguchi and j. i . yokoyama , phys . d * 86 * ( 2012 ) 023504 [ arxiv:1203.4059 [ hep - ph ] ] . k. s. babu , c. n. leung and j. t. pantaleone , phys . b * 319 * ( 1993 ) 191 [ hep - ph/9309223 ] . y. f. pirogov and o. v. zenin , eur . j. c * 10 * ( 1999 ) 629 [ hep - ph/9808396 ] .

the orion nebula cluster toward the hii region m42 is the most outstanding young cluster at the smallest distance 410pc among the rich high - mass stellar clusters . by newly analyzing the archival molecular data of theco(=10 ) emission at 21 resolution , we identified at least three pairs of complementary distributions between two velocity components at 8kms and 13kms . we present a hypothesis that the two clouds collided with each other and triggered formation of the high - mass stars , mainly toward two regions including the nearly ten o stars ,ori andori , in m42 and the b star , nuori , in m43 . the timescale of the collision is estimated to be.1myr by a ratio of the cloud size and velocity corrected for projection , which is consistent with the age of the youngest cluster members less than 0.1myr . the majority of the low - mass cluster members were formed prior to the collision in the last one myr . we discuss implications of the present hypothesis and the scenario of high - mass star formation by comparing with the other eight cases of triggered o star formation via cloud - cloud collision .

1701.04669

high - mass stars are influential in the evolution of galaxies by injecting a large amount of energy of ultraviolet photons , stellar winds and supernova explosions . it is of fundamental importance to understand the formation mechanism of high - mass stars in order to elucidate structure and evolution of galaxies , and considerable efforts have been devoted to understand the formation mechanism of high - mass stars . it has been discussed that self - gravitational massive aggregation of molecular gas or protostars are probable sites of high - mass star formation . the theories of high - mass star formation include the monolithic collapse and the competitive accretion ( for recent reviews see * ? ? ? * ; * ? ? ? * ) , and massive and dense clouds including the hot cores and the infrared dark clouds are studied as possible sites of such high - mass star formation ( e.g. , * ? ? ? it is however puzzling that these massive objects are not commonly associated with hii regions , the most obvious signature of o / early b star formation , and we do not yet have convincing confrontation between the theories and observations @xcite . recently , supersonic collision between the interstellar molecular clouds is discussed as a possible triggering mechanism of high - mass star formation both observationally and theoretically . up to now we have 8 regions of cloud - cloud collision toward the super star clusters and the hii regions in the milky way disk , and toward the young o stars in the lmc @xcite . these collisions take place supersonically at a velocity of 1030kms@xmath3 , and the high velocity indicates that the cloud motion is not governed by the cloud - cluster self - gravity . it is shown by theoretical works that cloud - cloud collision increases the cloud density and the effective sound speed in the shocked interface layer @xcite . according to the recent magnetohydrodynamical numerical simulations for a head - on collision at 20kms@xmath3 by @xcite , the mass accretion rate is enhanced by two orders of magnitude to @xmath7@xmath8@xmath9yr@xmath3 due to the collision as compared with the pre - collision state . the mass accretion rate @xmath7@xmath8@xmath9yr@xmath3 satisfies the theoretical requirement to overcome the radiation pressure feedback of the forming o star @xcite . the typical collision time scale is estimated to be 0.1myr from a ratio of the o star cluster size and velocity , 1pc/10kms@xmath3 , and stellar mass of 10100@xmath9 can be accumulated at the enhanced mass accretion rate in the timescale . currently , a key question is if cloud - cloud collision is one of the major mechanisms of high - mass star formation . it is premature to conclude that cloud - cloud collision is common because the number of samples is yet too small . one of the difficulties in studying high - mass star formation is that the number of high - mass star forming regions is small as compared with low - mass star formation sites . the nearest young o star cluster is the orion nebula cluster ( onc ) including m42 and m43 located at 410pc from the sun ( see for reviews * ? ? ? * ; * ? ? ? * ; * ? ? ? the onc has been best studied among the o star clusters at various wavelengths from x rays to radio emissions and is the primary object where theories of high - mass star formation are tested . the other known regions of high - mass star formation are located at a distance around 2kpc or larger . the orion a molecular cloud which harbors the onc is an active star forming region , and includes omc(orion molecular cloud)-1 , omc-2 , and omc-3 in the m42m43 region . a number of molecular line observations were made toward the orion a cloud in the last few decades mainly in the millimeter co emission lines @xcite . @xmath6100 ob stars and 2000 low mass young stars are distributed in and around the molecular cloud having @xmath10@xmath9 . in particular , the hii region m42 is the most active site of o star formation which includes about ten o / early b stars . on a large scale including the orion b cloud , four ob associations named ori ob1a , ori ob1b , ori ob1c , and ori ob1d are identified @xcite and they may be distributed spatially in age sequence . a model of the sequential formation of the ob associations was therefore proposed , where ob stars are formed due to triggering by the ionization - shock fronts @xcite . the orion a cloud appears to be converging toward m42 showing a v shape pointing to the north and a model consisting of self - gravitationally - bound converging motion was presented to explain the shape and formation of the onc @xcite , whereas details of the star formation remain elusive ( see also for low - mass star formation , * ? ? ? in this paper we present a new analysis of the archival data of @xmath0co(@xmath1=10 ) transition obtained with large single - dish telescopes aiming to test the scenario of cloud - cloud collision as a possible mechanism of high - mass star formation . the present paper focuses on the distribution and kinematics of the cloud . in section 2 we give a summary of observational and theoretical papers on cloud - cloud collision and present common signatures characteristic to the collision scenario . section 3 explains an analysis of the co data , and section 4 presents a model of the colliding clouds along with discussion . section 5 gives discussion from a broader perspective and section 6 concludes the paper . evidence for triggered formation of o / early b star(s ) by cloud - cloud collision is found in the four super star clusters , westerlund2 , ngc3603 , rcw38 and [ dbs2003]179 ( * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * kuwahara et al . 2016 , in prep . ) , in the two hii regions with single o stars , rcw120 and m20 ( @xcite , 2016 ) , and in the single high - mass stars in the lmc , n159w - south and the papillon - nebula yso of n159e @xcite . the physical parameters of the 8 objects along with the onc are summarized in table 1 . the four super star clusters include 1020 o stars with age of 0.14.0myrs , while the others with a single o star have age less than 1.0myr . theoretical studies of cloud - cloud collision are made by using hydrodynamical numerical simulations @xcite . these studies allow us to gain an insight into the observational / theoretical signatures of colliding clouds . it is also notable that synthetic observations present observable characteristics of cloud - cloud collision including the complementary distribution at the two velocities and the bridging feature between the two colliding clouds @xcite . we shall summarize the major properties obtained by these previous studies in the followings . the observed cloud - cloud collision indicates the relative velocity between the two clouds is supersonic in a range of 1030kms@xmath3 , except for a few cases where the relative velocity is apparently small probably due to projection ( n159w - south and n159e - papillon ) . such high velocity is gravitationally unbound by the typical cloud mass within a 10pc radius of the formed o star(s ) , @xmath11@xmath12@xmath9 . observational verification of heating by o stars / hii regions is used to establish the physical association of the clouds with the formed o stars in spite of their large velocity difference ( e.g. , * ? ? ? * ) which is otherwise interpreted as due to different kinematic distance . the number density of the molecular clouds in the galactic spiral arms is high enough to cause frequent cloud - cloud collision every 8myr for a giant molecular cloud as demonstrated by numerical simulations @xcite and co observations by nanten ( fukui et al . 2016b , in prep . the magnetohydrodynamical numerical simulations demonstrate that in the shock - compressed interface layer between the two clouds the density increases by a factor of 10 and the turbulent velocity and alfvn velocity increase by a factor of 5 @xcite . this is because the shock front is deformed into disordered turbulence by density inhomogeneities of the initial clouds and the turbulence amplifies the field strength . the mass accretion rate is proportional to the third power of the effective sound speed including the turbulent velocity and alfvn velocity , and the supersonic motion produces a high - mass accretion rate in the order of @xmath7@xmath8@xmath9yr@xmath3 , which is large enough to overcome the stellar radiation pressure and allow mass growth of the star(s ) . we note that the turbulent velocity field in the interface layer is isotropic , providing a signature of the collision even when the velocity between the colliding clouds becomes small by the projection effect as shown in the interface layer of n159w - south and the papillon nebula in n159e @xcite . we see in table 1 that the number of forming o stars has a threshold value of molecular column density @xcite . 1020 o star formation takes place for high column density arround @xmath13@xmath14 derived from @xmath0co(@xmath1=10 ) line intensity by using the x@xmath15 factor , whereas formation of a single o star happens for lower column density around @xmath16@xmath14 . o star(s ) is formed if one of the two clouds , not always both of them , satisfies these conditions . we suggest that in the high column density case the o star formation takes place in a small path length of the collision , e.g. , @xmath60.1pc ( see rcw38 , * ? ? ? * ) , and that in the low column density case the path length prior to the o star formation is longer , @xmath61pc ( see rcw120 , * ? ? ? * ) molecular clouds within 110pc of o star formation will be rapidly dispersed by the ionization / winds and may not be observable at a timescale larger than @xmath62myr @xcite , while the details will depend on the number of o stars , i.e. , the ultraviolet radiation field , and the gas density distribution . collisional interaction also converts the natal gas into the shocked interface layer and stars , and dissipates the natal clouds as discussed further in section 2.3 . figure 1 shows the radial distribution of the associated molecular emission in the nine regions of o star formation in table 1 . the super star clusters including 1020 o stars in a volume smaller than 1pc show often a central cavity of 10pc radius as seen in westerlund2 , ngc3603 , and [ dbs2003]179 , whose ages are 24myr . similar cavities are observed in the other super star clusters including arches , quintuplet , and r136 ( e.g. , * ? ? ? * ; * ? ? ? the only exception is rcw38 which shows rich molecular gas at apc scale . a close look at rcw38 indicates a cavity of 0.5pc radius in the molecular gas created by the ionization , and the small ionization cavity likely reflects the small age 0.1myr @xcite . these super star clusters suggest that the ionization propagates at a typical velocity of @xmath65kms@xmath3 ( a ratio 0.5pc/0.1myr , or 10pc/2myr ) . this velocity is consistent with theoretical values ( e.g. , * ? ? ? we expect that the clouds within 1pc will soon be ionized and dispersed in another 0.1myr , leading to termination of star formation in a few times 0.1myr . the age spread of the cluster members formed by triggering is therefore short in the order of 0.1myr in super star clusters . such a small age spread is consistent with that of 0.10.3myr of the young cluster members estimated by careful vlt and hst observations of the two super star clusters ngc3603 and westerlund1 @xcite . among the regions of single o star formation , the three regions m20 , n159w - south and n159e - papillon show rich molecular gas within 1pc and rcw120 shows a molecular cavity of 2pc radius the size of the molecular cavity in single o star formation seems to be smaller than in the superstar clusters perhaps due to the less ionizing photons . m20 whose age is 0.3myr is extended over 3pc , whereas it also shows a clear cavity of neutral matter with a 0.1pc radius centered on the o star in the hst image ( esa / hubble & nasa , bruno conti ) . the papillon - nebula yso in n159e shows a molecular cavity of 0.5pc radius , whose inside is filled by the hii region and the age of the papillon - nebula yso is estimated to be 0.2myr @xcite . n159w - south has no hii region and no molecular cavity , and is driving protostellar outflow with the associated accretion disk ( fukui et al . n159w - south has a small age of 0.1myr , and is younger than the papillon - nebula yso . so , the molecular cavities in m20 and the papillon - nebula are likely created by the ionization . it is suggested that the molecular cavity in rcw120 was created by the cloud - cloud collision as first modeled by @xcite , and the effect of the ionization is not dominant in forming the cavity ( @xcite ; see also for a different scenario @xcite ) . the single o star forming regions in the current sample are fairly young and the effect of the ionization does not seem to be prevailing , whereas the cloud dispersal by ionization will become dominant later . it is often the case that colliding clouds are not of the same size as simulated by @xcite , @xcite , and @xcite . if one of the colliding clouds is smaller than the other , the small cloud can create a cavity in the large cloud . this produces complementary distribution of the two clouds at different velocities , which is used to identify candidates for colliding clouds unless the cloud dispersal by the ionization is significant . in order to visualize the collision , we adopt the numerical simulations by @xcite , and table 2 lists the simulation parameters . a schematic is shown in figure 2(a ) . in the model where the two clouds are spherical , the cross section of the cavity is determined by the size of the small cloud and the length of the cavity by the travel distance since the initiation of the collision . the cavity is identified observationally as an intensity depression in the molecular distribution at the velocity of the large cloud . the projected distributions of the small cloud and the intensity depression are complementary , and they are in general displaced with each other because the angle of the relative motion to the line of sight , @xmath17 , is not 0deg . ( figure 5 ) . while the real cloud shape can be more complicated , it is still worthwhile to use the simple two - sphere model to gain an insight into the cloud distribution created by the collision . the relative motion of the colliding clouds generally has an inclination angle to the line of sight , and three cases of the inclination angle @xmath17 , 0deg . , 45deg . , and 90deg . , are shown in the velocity channel distributions in figures 35 . for @xmath17=0deg . ( figure 3 ) we see two velocity features whose distribution varies with velocity . we identify the initial two clouds at @xmath18@xmath19kms@xmath3 ( panels ( e)(g ) in figure 3 ) and @xmath20@xmath21kms@xmath3 ( panels ( i)(k ) in figure 3 ) and the intermediate velocity feature mainly at @xmath19@xmath20kms@xmath3 ( panels ( g)-(i ) in figure 3 ) . the small cloud fits well with the cavity created in the large cloud by collision , showing the complementary distribution between them ( figure 3(l ) ) . for @xmath17=45deg . ( figure 4 ) we see the two clouds at @xmath22@xmath23kms@xmath3 ( panels ( f)(g ) in figure 4 ) and @xmath20@xmath21kms@xmath3 ( panels ( i)(j ) in figure 4 ) and the intermediate velocity feature mainly at @xmath23@xmath20kms@xmath3 ( panels ( h ) in figure 4 ) . in figure 4 we see a displacement of the complementary distribution between the cavity and the small cloud , which is a natural result of projection for @xmath17=45deg . if the relative motion is vertical to the line of sight @xmath17=90deg . ( figure 4 ) , we do not see the two velocity components . all these simulations show only single - peaked line profiles with skew in velocity ( figures 3(l ) and 4(l ) ) , where the two clouds seem to be continuous due to the @xmath24bridging@xmath25 feature ( section 2.4 ) . figures 35 illustrate that , even if we see only a singly - peaked velocity component in observations , a possibility of cloud - cloud collision is not excluded . in regions of single o star formation , the most significant complementary distribution toward an o star at a 0.1pc scale is found in m20 ( torii et al . 2016 , in prep . ) . in rcw120 the dispersal of the small cloud is significant and only the large cloud with the intensity depression , one of the complementary clouds , is seen along with the remnant of the small cloud outside the `` bubble '' @xcite . in n159 it is possible that the angular resolution of alma , [email protected] , is not high enough to resolve the complementary distribution which is likely to be in a scale of 0.1pc . in the super star clusters where the ionization is more significant , it is probable that the complementary distribution is dispersed more quickly . even in the youngest cluster rcw38 we see the ionization is dominant in shaping the clouds @xcite . we note that the oldest cluster [ dbs2003]179 among the four shows a large - scale complementary distribution at a scale more than 10pc , which is not affected much by the ionization ( kuwahara et al . 2016 , in prep . ) . to summarize , the colliding clouds with small age and relatively low ionization are the most probable site to find the complementary distribution , whereas it is possible that the complementary distribution is often dispersed by the stellar feedback and the collisional interaction , depending on the cloud distribution . another possible signature of cloud - cloud collision is seen in the optical / near - infrared extinction toward the hii region . in cloud - cloud collision the blue - shifted cloud is always on the far - side of the red - shifted cloud prior to the collision but moves to the near - side after the collision . as a result , one may expect that only the blue - shifted cloud , not the red - shifted cloud , is located in front of the hii region and can create a dark heavily obscured feature . such dark lanes are in fact observed in m20 where the collision took place @xmath60.3myr ago to form a single o star ; the dark lanes toward m20 show a good correspondence with the blue - shifted cloud @xcite . a caveat is that this may not always be the case if the collision happened very recently where the line - of - sight separation between the two clouds is still small . the super star cluster rcw38 shows a narrow dark lane in the near infrared as irs1 , whereas irs1 is identified as part of the red - shifted cloud @xcite . the blue - shifted cloud in rcw38 also corresponds to the near - infrared dark features surrounding the cluster . these authors interpret that irs1 remains neutral without full ionization due to the small age of the system ; the o star formation occurred only 0.1myr ago and part of the red - shifted cloud within 1pc of the cluster center is still surviving in front of or inside the hii region . so , the dark lanes can arise from the both clouds depending on the age and shape of the clouds , and we need to be cautious in examining the extinction features when the collision is a very recent event . numerical simulations show that the two colliding clouds become connected in velocity by a bridging feature in synthetic observations if the projected velocity separation is larger than the linewidths of the clouds @xcite . the shocked interface layer appears as the bridging feature connecting the clouds in the position - velocity plots as shown in figures 3 and 4 . the bridging features are observable in the spots of collision , although all of them are not necessarily related to o star formation as shown later . in table 1 we see that only the youngest two cases , rcw38 and m20 , show the bridging feature toward the o star formation . rcw38 shows the bridging feature within 0.5pc of the @xmath620 o star candidates , where the small colliding cloud is localized @xcite . the connection among the three features [ the bridging feature , @xmath620 o stars , and the small colliding cloud ] is tight ( see figure 3 in * ? ? ? m20 shows three bridging features and one of them is found toward the o star ( torii et al . 2016 , in prep . ) . the three super star clusters , ngc3603 , westerlund2 and [ dbs2003]179 , show the bridging features in the outer regions of more than 10pc radius of the cluster , where molecular gas still remains . rcw120 also shows the bridging features outside the hii region in co and hi @xcite . n159w - south and n159e - papillon have a small projected velocity separation of a fewkms@xmath3 and the bridging features are not seen . to summarize , the main observational signatures of o star formation by cloud - cloud collision are listed as follows ; 1 . the supersonic velocity separation of the two clouds , 2 . the complementary distribution of the two velocity components , and 3 . the bridging feature connecting the two clouds . we note that the cloud dispersal by the ionization and the collisional merging of the clouds , in addition to the projection effect , can weaken or extinguish these signatures , and one is required to carefully examine these effects which can smear out the collision signatures . even if a cloud - cloud collision happens , it is possible that one sees none of the signatures . in order to catch the signatures observationally it is required that the system is young , without significant dispersal of the natal gas , and that the projection effect is small . in more than a fewmyrs after the o star formation by collision , it probably becomes impossible to find any of the observable signatures , for instance , as in case of the relatively young cluster ngc6231 with an age of 35myr ( baume et al . 1999 ) which has no molecular gas within 30pc of the center ( fukui et al . 2016c , in prep . the @xmath0co(@xmath1=10 ) distributions in the m42m43 region obtained with the nro 45-m telescope are presented and discussed by @xcite . the observational parameters are listed in table 3 . the paper shows that the orion a molecular cloud is singly peaked in velocity , which is consistent with the previous co results summarized in section 1 . the cloud properties created in cloud - cloud collision may however show no obvious double component ( section 2 ) , and it is worthwhile to make a thorough investigation of the kinematic properties in the m42m43 region from a view point of the cloud - cloud collision scenario . the numerical simulations of the cloud - cloud collision in section 2 assumed physical parameters which do not exactly coincide with the observed properties of the present cloud . nonetheless , it is conceivable that the basic characteristics of the colliding clouds hold for different physical parameters , and we here consider the qualitative properties presented in section 2 . in the followings , we apply an x@xmath15 factor of 2@xmath2610@xmath27 ( kkms@xmath3)@xmath3@xmath14 @xcite to convert the @xmath0co(@xmath1=10 ) intensity into the molecular column density . figure 6 shows the dec .- velocity diagram covering @xmath61deg in dec . integrated in a r.a . range from @xmath28 to @xmath29 , and figure 7 shows velocity - channel distributions in a region of @xmath60.5deg ( r.a . ) by @xmath61deg ( dec . ) centered on @xmath4ori . although most ( 75% ) of the co emission is apparently seen in a velocity range of 613kms@xmath3 as a single component , in figure 7 we find the distribution of the @xmath0co emission is significantly different between the two velocity ranges 5.39.8kms@xmath3 ( panels ( b ) , ( c ) , and ( d ) ) and 11.414.4kms@xmath3 ( panels ( f ) and ( g ) ) . the strongest peak of the @xmath0co emission is seen at 9kms@xmath3 toward @xmath4ori where the @xmath0co linewidth is enhanced from 5kms@xmath3 to 14kms@xmath3 at a 1 kkms@xmath3 level toward the onc . this main peak having a size of @xmath61pc@xmath261pc is called the omc-1 clump in the followings . we see that the blue - shifted emission at @xmath68kms@xmath3 , which constitutes the integral shape of the cloud ( mainly in panel ( d ) of figure 7 , * ? ? ? * ) , is extended on the northern and southern sides of the omc-1 clump , whereas the red - shifted component at @xmath612kms@xmath3 ( mainly in panel ( f ) of figure 7 ) has a sharp intensity decrease toward the south at dec.=@xmath30 . we see the trend also in figure 6 . these different spatial distributions at the two velocity ranges suggest a possibility that the two clouds are overlapping as modeled later in section 4 . the intermediate velocity range 1011kms@xmath3 ( panel ( e ) of figure 7 ) shows mixed morphology of the two velocity components . based on the empirical signatures of cloud - cloud collision ( section 2 ) , we searched the @xmath0co data for complementary distribution and found three possible pairs . figure 8 shows the three complementary distributions at the two velocity ranges in panels ( a ) , ( b ) and ( c ) . figure 8(a ) shows the distribution toward m42 , and figure 8(b ) toward m43 and omc-2 . figure 8(c ) is the distribution toward omc-3 , where no hii region is associated ( e.g. , * ? ? ? * ) . in figure 8(a ) , we define the omc-1 clump in the blue - shifted cloud at the contour level of 15 kkms@xmath3 . we find the u - shaped cloud in the red - shifted cloud appears to surround the omc-1 clump , and define the cloud at the contour level of 2 kkms@xmath3 . the defined levels are used for calculating the cloud mass . in figure 8(b ) and 8(c ) we find that the red - shifted component shows complementary distributions to the blue - shifted component as detailed below . in order to show the complementary distributions clearly figure 9 overlays the two velocity components in the three regions with appropriate displacements as indicated by arrows in figures 9(b ) and ( c ) , whereas no displacement is applied in figure 9(a ) . in figure 9(a ) , the complementary distribution is clear in the two components . the omc-1 clump is extended by more than 1pc only toward the north - west at an intensity level of @xmath615 kkms@xmath3 ( figure 8) . it is possible that this extension forms a complementary pair with the intensity depression in the red - shifted component toward ( r.a . , dec.)=(@xmath31 , @xmath32 ) ( panel ( f ) of figure 7 ) , although the diffuse nature of the extension makes the distribution less evident than toward omc-1 . we note that the intensity depression in the south of the omc-1 clump in the blue - shifted component , which is extended from r.a.=@xmath33 to @xmath29 at dec.=@xmath34 , shows also a possible complementary distribution to the red - shifted component in the same direction . further , in the south of the omc-1 clump we see an intensity depression elongated in the east to west ( r.a.=@xmath35@xmath36 , dec.=@xmath37@xmath38 ) in the blue - shifted component ( figure 10(a ) ) . this depression seems similar in shape to the southern part of the complementary distribution of the red - shifted component , and possibly represents another sign of complementary distribution . the depression appears to be shifted to the south as compared with the red - shifted component . a displacement similar to figure 9(b ) may be applicable , while the large size of the omc-1 clump @xmath61pc tends to complicate the complementary correspondence more than in figure 9(b ) . in order to explain the estimate of the displacement in figure 9(b ) , we show in figure 10 , the upper panels , the inner part of the two @xmath0co components toward the central part of m42 and m43 in the red- and blue - shifted velocity ranges . figure 10 also shows enlarged @xmath0co images toward m43 in the lower two panels . the blue - shifted cloud shows an intensity depression toward the exciting star nuori , which is elongated in the north - south by @xmath60.5pc with a width of @xmath60.2pc as enclosed by a box in figure 8(c ) . the eastern side of the depression is bound by a filamentary feature that corresponds to the dark lane in m43 ( see also figure 12 ) . a corresponding emission feature to the intensity depression is seen in the red - shifted component which has a similar shape to the depression . we call the two features as the @xmath24orion keyhole@xmath25 and the @xmath24orion key@xmath25 in order to avoid confusion with the keyhole nebula in carina . we applied the method to estimate the displacement in the complementary distribution which is presented in the appendix . in fitting the displacement , we restricted the area of the analysis to the region of the orion keyhole and the orion key as shown in figure 11 ; in the north of the orion keyhole , the possible intensity depression is not closed , making it difficult to apply the method . by inspection we find that the orion key has a discontinuous @xmath24bent@xmath25 in the middle at @xmath60.25pc from the southern edge , which seems to morphologically correspond to the edge of the eastern filament of the orion keyhole ( figure 8) . we then chose the direction of the displacement at a position angle of 108deg . so that the two features coincide spatially after the displacement . figure 11 shows the overlapping function @xmath39 in pc@xmath40 defined in the appendix and indicates that @xmath41=0.3pc gives the optimum fit . we confirmed that the present position angle adopted gives the best fit , i.e. , the maximum value of @xmath39 , by changing the angle in a range 103113deg . a displacement of 0.3pc shown by the arrow in figure 9(b ) gives the fit between the orion keyhole and key . by the displacement , we also see that the western edge of the red - shifted component shows good correspondence with the eastern edge of the blue - shifted component over @xmath61pc in the north - south , which lends support for the displacement . in figure 9(c ) , while not so clear as in the other two cases , the two peaks of the blue - shifted component at ( r.a . , dec)=(@xmath42 , @xmath43 ) and ( @xmath44 , @xmath45 ) are located toward the intensity depression of the red - shifted component , forming a possible pair of complementary distribution , after a displacement of 0.1pc shown by an arrow . this displacement was chosen based on the eye inspection because the distribution is relatively simple . figures 12 and 13 show overlays of the two velocity components on the optical image . the most notable correspondence is seen toward m43 where the thin dark lane in the center of the nebula coincides with the co filament in the east of the orion keyhole ( figure 12(a ) ) . the other blue - shifted features of @xmath60.5deg in length show coincidence with the dark areas in the northeast of m42 . this correspondence is consistent with that the blue - shifted component lies on the nearside of the optical nebula . the red - shifted component in figure 13 shows no correspondence with the optical features , suggesting that the red - shifted component lies on the far side or the inside of the nebula . we remark that the orion key has no corresponding dark lane in m43 . we present a model of the two molecular clouds which are colliding to trigger the formation of the o / b stars in m42 and m43 . table 5 lists the main physical parameters of the model clouds and figures 14 gives a schematic of the clouds before and after the collision . the model consists of two clouds of projected velocities peaked at @xmath68kms@xmath3 and @xmath613kms@xmath3 , while the actual relative velocity between the two clouds is in the order of 10kms@xmath3 . both of the clouds are elongated along the galactic plane . the mass of the blue - shifted component is dominant ( @xmath680% ) toward omc-1 , explaining the apparent single peak , while the border velocity 11kms@xmath3 between the two model clouds may have some uncertainty ; i.e. , a border velocity between 10kms@xmath3 and 11kms@xmath3 may not be excluded . the total mass of the two clouds @xmath610@xmath46@xmath9 is too small by an order of magnitude to gravitationally bind the estimated relative velocity @xmath67kms@xmath3 . the model clouds have an overlap toward the omc-1 clump at dec.=@xmath47@xmath48 which corresponds to the region of the enhanced velocity span in figure 6 . the enhancement is interpreted as due to the turbulence by the collisional interaction @xcite , while the effect of the protostellar outflow may also be in part responsible for the enhancement . the model offers an explanation of the complementary distribution which is found the present cloud . the collisional interaction is seen at least in the three regions as indicated by the complementary spatial distribution ; one is toward m42 and triggered formation of @xmath4ori and @xmath5ori ( figure 9(c ) ) , and another toward m43 , which triggered formation of nuori ( figure 9(b ) ) . the other is toward the north of m43 without an hii region ( figure 9(a ) ) . two more regions associated with possible complementary distribution are noted in section 3 , while no clear star formation is seen except for m42 and m43 . in the present case , the half - power linewidths of the two clouds are 23kms@xmath3 and the projected velocity separation @xmath65kms@xmath3 is too small for observing the bridging feature . in the model , @xmath4ori and @xmath5ori , including nearly ten o stars , were formed on the near side of the blue - shifted cloud which was collided and shock - compressed by the more extended red - shifted cloud . this collision created the cavity in the red - shifted cloud following the cloud - cloud collision model @xcite . the interface layer between the two clouds became strongly - compressed and highly turbulent to form o stars as shown by @xcite . the o stars formed on the nearside of the dense blue - shifted cloud ionized the nearside of the cloud to form the m42 nebula as exposed to us . nuori was formed in the collision between the orion key and the red - shifted component having the orion keyhole . nuori is a b3 star and ionized m43 ( see section 4.3 for more details ) . the ionization possibly has widened the size of the orion keyhole , whereas the smaller uv photon flux of nuori than @xmath4ori and @xmath5ori is not powerful enough to dissipate the orion keyhole at present . the timescale of the collision is approximately estimated from a ratio of the cloud size and the relative velocity between the two clouds . if we assume tentatively the relative motion between the two clouds has an angle of 45deg to the line of sight , the relative velocity corrected for the projection is estimated to be @xmath67kms@xmath3 , which is consistent with the enhanced @xmath0co broadening toward the onc ( figure 6 ) . the apparent spatial extent of the colliding clouds forming o / b stars ranges from @xmath60.3pc to @xmath60.8pc ( figure 8) . the ratio between the timescale and the cloud size yields roughly 0.050.11myr as the collision timescale and poses a constraint on the upper limit for the age of the stars formed by triggering . the complementary distribution is characterized by the projected displacement between the colliding clouds ( figure 3 , see also the appendix ) . the displacement is small for a small collision timescale . also , the displacement @xmath49 to the size of the small cloud becomes smaller , if the size of the small cloud in the collision model is large . figure 10 shows different displacements between the two velocity components among the three regions . we apply a detailed analysis of the timescale to the orion key and keyhole , where information is obtained on their relative location by the analysis in section 3 thanks to the small size , @xmath60.2pc , of the small cloud . the separation between the two features are estimated to be 0.3pc at a relative velocity is 4kms@xmath3 . if we assume 45deg as the angle of the relative motion to the line of sight , the time scale of the collision is estimated to be 0.4pc/5.7kms@xmath3=0.07myr if no deceleration of the relative motion is assumed . the northern region in figure 9(c ) shows a displacement of @xmath60.1pc , while it may be less accurate than in figure 9(b ) . the omc-1 clump in figure 9(a ) shows no appreciable displacement , whereas the larger clump size , @xmath61pc , may tend to smear out a possible displacement . the separation between the two velocity components may vary from region to region reflecting the initial distribution of the two components , leading to different collision parameters . in summary , we conclude that the typical collision timescale is @xmath60.1myr with some regional minor variation . at the present epoch , the blue - shifted component has high peak column density toward the peak of the omc-1 clump , @xmath62@xmath2610@xmath50@xmath14 , while the red - shifted cloud significantly smaller column density , @xmath610@xmath51@xmath14 . in m43 , the typical column density in the orion key and keyhole is @xmath610@xmath51@xmath14 . we assume that the initial conditions of the two clouds are similar to those of the present clouds , whereas it is probable that the present column densities have become somewhat lower than the initial values due to the collisional interaction and ionization by the formed stars . by comparing o star formation in several regions , @xcite argue that the number of o stars formed by collisional triggering depends on the initial column density ; more than ten o stars are formed for a threshold column density @xmath610@xmath50@xmath14 and a single o star for a threshold column density @xmath610@xmath51@xmath14 . the difference in the o / b star formation in m42 and m43 is consistent with the suggestion . the mhd numerical simulations by @xcite have shown that the interface layer between the colliding clouds become denser and more turbulent with amplified magnetic field , realizing the high mass accretion rate @xmath610@xmath5210@xmath53@xmath9yr@xmath3 which satisfies the formation of the high - mass stars greater than 20@xmath9 . in the present typical collision timescale 0.1myr , the stellar mass which is attained by mass accretion is 10100@xmath9 . so , the present short timescale is consistent with the o star formation . more details of the star formation may be explored . in m42 , the ionized gas has higher density in @xmath4ori than in @xmath5ori . there are even younger heavily - embedded high - mass protostars showing outflows , the bn / kl object and the orion - s region , in the west of @xmath4ori . we see an age sequence from the east to the west , where @xmath5ori is the oldest and the bn / kl object and the orion - s region the youngest . a possible scenario is that the age sequence is caused by the time sequence of the collision ; i.e. , the colliding two components has a spatial gradient in the sense that the western part collided first and then the collision is propagating to the west , and the most recent triggering formed the bn / kl object and the orion - s region . we speculate that this apparent time sequence from the east to the west is due to a gradient in the separation between the colliding surfaces of the two clouds in the sense that the western part collided later than the eastern side . a similar age gradient is seen in rcw38 ; the eastern part is already ionized , whereas the western part is more heavily embedded @xcite . in m43 , the position of nuori is separated to the west by @xmath60.15pc from the eastern filament of the orion keyhole . the velocity of the interface layer generally becomes smaller by the momentum conservation ( e.g. , * ? ? ? * ) , which places the formed star behind the small cloud . a possible scenario is that the b star formation took place by the collision in the eastern edge of the orion key . this explains the location of the nuori and the situation is similar to m20 where a single o star is formed by triggering in the eastern edges of the clouds ( torii et al . 2016 , in prep . ) . the age of the onc members ranges from less than 0.1myr to more than 1myr while the majority is in the range 0.11myr as estimated in the hr diagram by comparing with the theoretical stellar evolutionary tracks @xcite . the collision timescale less than 0.1myr corresponds to the youngest end of the age distribution . we infer that the blue - shifted cloud which dominates the molecular column density in the orion a cloud was forming low - mass stars of the onc prior to the collision in the last 1myr . the distribution of low - mass young stars of the onc , 2mass variables stars in figures 15 and 16 @xcite , indicates that these stars toward the omc-1 clump are correlated well with the column density distribution of the blue - shifted cloud except for the @xmath0co peak including the bn / kl object where the extinction may be large at k band @xcite . these low - mass members are spatially more extended than the o stars @xcite , beyond the region of the collisional interaction . we see little molecular gas in the two regions in the north of m42 , ( r.a . , dec.)=(@xmath54@xmath55 , @xmath32@xmath43 ) , and the south of the m42 clump ( r.a . , dec.)=(@xmath56@xmath57 , @xmath58@xmath34 ) , whereas 2mass stars are distributed there . the two correspond to the regions of the suggested collisional interaction and the blue - shifted gas in the regions has possibly been removed by the collision after the low - mass star formation . the red - shifted cloud had lower column density , not actively forming stars before the collision , while the northern part of the red - shifted component may have been forming part of the stars as indicated by the spatial correspondence with the stars ( figure 15 ) . it is possible that not only o / b stars but also low mass - stars were formed by the collisional triggering ( cf . * ? ? ? the young low - mass stars in the hr diagram by @xcite suggest very young low - mass stars of an age less than 0.1myr . part of them might represent protostars whose accretion disk was destructed by the uv photons of @xmath4ori and @xmath5ori before they reach @xmath5910@xmath9 . the monolithic collapse and the competitive accretion are the two scenarios for massive cluster formation and have been discussed for more than a decade ( e.g. , * ? ? ? * ) . in either case the massive gas / star system is self - gravitationally bound and multiple sources evolve into stars of various masses by mass accretion . it is also discussed that the bound system may evolve in time and becomes unbound by gas dispersal ( e.g. , * ? ? ? these two gravitationally - bound scenarios have no explicit physical mechanism of realizing the high - mass accretion rate like the shock - induced turbulence in the cloud - cloud collision scenario , and it is assumed that the gas achieves the high - mass accretion rate by some gravitationally driven mechanism . the colliding clouds in the present hypothesis are not gravitationally bound by the supersonic velocity and the collision increases the mass accretion rate to be @xmath7@xmath8@xmath9yr@xmath3 , which is higher than in the bound scenarios , favoring the high - mass star formation @xcite . the evolution of the monolithic mass aggregation has been tested theoretically . in the simulations of proto - clusters , a single gravitationally bound n - body system is adopted as the initial condition of a cluster whose number of stars are assumed to be 800 2000 in the same order with that of the onc ( e.g. , * ? ? ? * ; * ? ? ? the effect of the stellar feedback is also incorporated in recent simulations ( e.g. , * ? ? ? the typical timescale of the evolution of the n - body system is myrs , which is significantly larger than that of the cloud - cloud collisionless than 0.1myr . it is probable that these simulations are applicable to the pre - collision cluster of low - mass stars in the orion blue - shifted cloud of the present hypothesis . some n - body simulations assume @xmath24rapid@xmath25 gas dispersal by ionization due to the high - mass stars , and it is argued that such rapid dispersal is consistent with the observations @xcite . the rapid dispersal is however yet of a large time scale , close to myrs . there is increasing observational evidence for a short time scale less than 0.1myr in o star formation ; observations of the onc shows that the youngest stars have age less than 0.1myr , and in the two super star clusters ngc3603 and westerlund1 the age spread of young clusters is short in the order of 0.1myr @xcite . it is also becoming established that the age and mass segregation is outstanding in young massive clusters ; the o stars are concentrated in the inner most part of ngc3603 , the onc , and r136 . while the stellar mass function depends on the area for averaging the stellar properties , the innermost part tends to have a top - heavy mass function in these clusters . we need to be cautious that the onc may consist of at least two different populations . the onc has a usual mass function similar to the field initial mass function imf having a salpeter - like slope @xcite . in the present model the stars toward m42 is a superposition of two different populations ; the older low - mass members formed prior to the collision are likely dominant in the blue - shifted cloud and the other youngest stars including o / early b stars were formed in triggering by the collision , which are not dominant in mass . the pre - existent low - mass stars dominate the stars of the onc , making the steep slope of the mass function , while about a half of them may still be deeply embedded in the blue - shifted cloud @xcite . one may expect that the imf has discontinuous slopes if the star formation by the two different modes are working . we suggest that the total mass of the o stars in the center of the onc is @xmath6150@xmath9 @xcite . on the other hand , the molecular mass of the omc-1 clump is 3.3@xmath2610@xmath60@xmath9 . if we assume tentatively that the depth where of the shock front propagated in the omc-1 clump is 0.3pc , one third of the clump size , the molecular mass already shocked is estimated to be @xmath61000@xmath9 , large enough to form the ten o stars of 2030@xmath9 . the present collision model suggests that the forming stars by triggering are located on the front side of the m42 clump , producing a non - spherically symmetric distribution of the cluster . this is a natural consequence of the collision that has a directivity at least in the early phase where part of the molecular gas still remains without ionization . we note that the n - body simulations result in spherically symmetric system after myrs do not explain such asymmetry . the total mass of the orion key and keyhole region is estimated to be @xmath6150@xmath9 for a size of @xmath60.5pc . this is large enough to form nuori with 17@xmath9 @xcite . the initial conditions of the present cloud - cloud collision assume a molecular cloud having high column density 10@xmath50@xmath14 , which has been continuously forming low - mass stars . these low - mass stars are not able to ionize the cloud and the blue - shifted cloud continues to form low - mass stars until it forms high - mass stars which are capable of ionizing the cloud . the o star formation is suddenly triggered by the impact of the supersonic collision with the red - shifted cloud which has lower column density . the epoch of o star formation may range in a certain time span , whereas the cloud - cloud collision poses a limit on the age spread of the youngest members including o stars by the ratio of the cluster size divided by the collision velocity to be @xmath60.1myr . it is unlikely that the two gravitationally bound scenarios can explain the very short discontinuous history of star formation less than 0.1myr . we do not see any gap in the age distribution of the cluster member . this is a natural consequence of the cloud - cloud collision following myr low - mass star formation . we have made a new analysis of the molecular gas toward m42 and m43 , and test a hypothesis that two clouds collided about 0.1myr ago to form the northern part of the orion a cloud . we have shown the collision model offers a reasonable explanation of the observed cloud and cluster properties in the m42 and m 43 region . although , currently , we do not have an extensive observational basis that the cloud - cloud collision is a dominant mechanism of o star formation in the galaxy . the main conclusions of the present study are summarized as follows . 1 . the present analysis of the high resolution co data shows a possibility that the orion a cloud toward m42 and m43 consists of two velocity components , one at 4.011.1kms@xmath3 and the other at 11.114.9kms@xmath3 . the mass of the blue - shifted component is 1.5@xmath2610@xmath46@xmath9 with its velocity range 4.011.1kms@xmath3 and the red - shifted component 3.4@xmath2610@xmath60@xmath9 with its velocity range 11.114.9kms@xmath3 having a size of 4pc by 7pc in r.a . and we find at least three pairs of complementary distributions between the two clouds having scale sizes of 0.11pc . these complementary distributions of molecular gas are an observational signature characteristic to cloud - cloud collision according to the hydrodynamical numerical simulations at the early phase of o star formation . the major complementary distribution is found toward @xmath4ori and @xmath5ori where the blue - shifted component of @xmath60.5pc radius , the omc-1 clump , is surrounded by the red - shifted component , the u - shaped cloud , which is extended with a radius of @xmath61pc . the secondary complementary distribution is seen toward m43 as named the orion key and keyhole , where a spatial displacement of 0.3pc is seen between the two complementary distributions . we frame a hypothesis that the two velocity components collided with each other to trigger formation of the nearly ten o stars in m42 and the b star in m43 . the displacement between the two complementary features in m43 suggests that the relative motion makes a large angle to the line - of - sight . by assuming the angle to be 45deg , the relative velocity and separation between the clouds is estimated to be @xmath66kms@xmath3 and @xmath60.5pc , respectively , after correction for the projection , and the typical collision time scale is calculated to be @xmath60.1myr . although the collision is extended over an area of 4pc@xmath267pc in r.a . and the triggering of the o / b star formation is seen only toward the localized regions with ob stars . we suggest that this trend is consistent with the threshold for o star formation in cloud - cloud collision suggested by @xcite ; multiple o star formation is possible for column density @xmath610@xmath50@xmath14 while the single o / b star formation for @xmath610@xmath51@xmath14 . the high segregation of the high - mass cluster members to the cluster center is explained as a reflection of the initial column density distribution prior to the collision . the stellar contents of the onc has been best studied among the rich clusters thanks to its unrivaled small distance to the sun although the cluster is not overwhelmingly rich in high - mass stars . the ob stars in the onc are minor members in terms of stellar mass , and the initial mass function is not significantly influenced by the formation of the ob stars . the youngest members including the o / early b stars are formed by the collision within 0.1myr , and the large age spread of the cluster , from a few 0.01myr to a few times 1.0myr , is understood by a combination of low - mass star formation in the blue - shifted cloud prior to the collision and the recent o / b star formation by the collisional triggering . formation of o stars may still be continuing toward the western edge of the m42 clump where the bn object and omc-1-s are located along the north - south direction . in the present study we have shown that cloud - cloud collision offers a possible consistent interpretation of the formation of the o / b stars in m42 and m43 . since the immediate vicinity of @xmath4ori and @xmath5ori is significantly ionized , we are not allowed to directly witness the triggering as in case of rcw38 . we note that the bn / kl object can be such a candidate as discussed in section 5 , and suggest that a future more careful scrutiny of the region and its surroundings at higher resolutions with alma etc . will offer a further insight into the possible role of the collisional interaction . we are grateful to thomas haworth for valuable discussion on this paper . we also thanks to robert gendler for optical image of the onc , and kazuki tokuda for alma archive data . this publication 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 . nobeyama radio observatory is a branch of the national astronomical observatory of japan , national institutes of natural sciences . the aste telescope is operated by national astronomical observatory of japan ( naoj ) . this paper makes use of the following alma data : ads / jao.alma#2012.1.00554.s . alma is a partnership of eso ( representing its member states ) , nsf ( usa ) and nins ( japan ) , together with nrc ( canada ) , nsc and asiaa ( taiwan ) , and kasi ( republic of korea ) , in cooperation with the republic of chile . the joint alma observatory is operated by eso , aui / nrao and naoj . this work was supported by jsps kakenhi grant numbers 15h05694 , 15k17607 , 15k05014 , 16k17664 , and 26247026 ; by jsps and by the mitsubishi foundation . natexlab#1#1 , s. v. 2010 , , 405 , 1431 , j. , langer , w. d. , stark , a. a. , & wilson , r. w. 1987 , , 312 , l45 , s. , & kroupa , p. 2015 , arxiv e - prints , arxiv:1512.03074 , a. 1991 , in nato advanced science institutes ( asi ) series c , vol . 342 , nato advanced science institutes ( asi ) series c , ed . c. j. lada & n. d. kylafis , 125 , j. m. , hillenbrand , l. a. , & skrutskie , m. f. 2001 , , 121 , 3160 , a. , duvert , g. , dutrey , a. , et al . 1990 , , 234 , 469 , j. e. , ercolano , b. , & bonnell , i. a. 2013 , , 431 , 1062 , g. , paresce , f. , panagia , n. , et al . 2011 , , 739 , 27 , c. l. , pringle , j. e. , & duarte - cabral , a. 2015 , , 446 , 3608 , m. p. , shipman , r. f. , price , s. d. , et al . 1998 , , 494 , l199 , b. g. , & lada , c. j. 1977 , , 214 , 725 , e. d. , getman , k. , townsley , l. , et al . 2005 , , 160 , 379 , y. , ohama , a. , hanaoka , n. , et al . 2014 , , 780 , 36 , y. , harada , r. , tokuda , k. , et al . 2015 , , 807 , l4 , y. , torii , k. , ohama , a. , et al . 2016 , , 820 , 26 , n. , dawson , j. r. , ohama , a. , et al . 2009 , , 696 , l115 , a. , & ohta , k. 1992 , , 44 , 203 , l. , & burkert , a. 2007 , , 654 , 988 , t. j. , shima , k. , tasker , e. j. , et al . 2015 , arxiv e - prints , arxiv:1509.00859 , t. j. , tasker , e. j. , fukui , y. , et al . 2015 , , 450 , 10 , m. h. , morgan , j. , schloerb , f. p. , snell , r. l. , & goldsmith , p. f. 1992 , , 395 , l99 , l. a. 1997 , , 113 , 1733 , l. a. , & carpenter , j. m. 2000 , , 540 , 236 , t. , & fukui , y. 2013 , , 774 , l31 , p. 2001 , , 322 , 231 , p. , aarseth , s. , & hurley , j. 2001 , , 321 , 699 , n. , brandner , w. , gennaro , m. , et al . 2012 , , 750 , l44 , m. l. , tucker , k. d. , chin , g. , & thaddeus , p. 1977 , , 215 , 521 , r. j. , morris , m. , moscowitz , j. , & thaddeus , p. 1986 , , 303 , 375 , f. , pomars , m. , deharveng , l. , zavagno , a. , & bouret , j. c. 2010 , , 510 , a32 , c. f. , & tan , j. c. 2003 , , 585 , 850 , a. , getman , k. , hillenbrand , l. , & preibisch , t. 2008 , star formation in the orion nebula i : stellar content , ed . b. reipurth , 483 , t. , mizuno , a. , ogawa , h. , & fukui , y. 1998 , , 116 , 336 , f. , miura , t. , kitamura , y. , et al . 2012 , , 746 , 25 , a. , tokuda , k. , kimura , k. , et al . 2015 , , 216 , 18 , c. r. , muench , a. , smith , n. , & zapata , l. 2008 , star formation in the orion nebula ii : gas , dust , proplyds and outflows , ed . b. reipurth , 544 , a. , dawson , j. r. , furukawa , n. , et al . 2010 , , 709 , 975 , d. e. , & megeath , s. t. 2008 , the orion molecular cloud 2/3 and ngc 1977 regions , ed . b. reipurth , 590 , r. , bensch , f. , howe , j. e. , et al . 2000 , , 539 , l133 , f. , heyer , m. h. , gutermuth , r. , snell , r. l. , & brunt , c. m. 2013 , , 431 , 1296 , k. , onishi , t. , nayak , o. , et al . 2016 , arxiv e - prints , arxiv:1604.06010 , s. , hayashi , m. , hasegawa , t. , handa , t. , & oka , t. 1994 , , 425 , 641 , y. , kawabe , r. , takakuwa , s. , et al . 2011 , , 63 , 105 , a. , humann , b. , olczak , c. , et al . 2015 , , 578 , a4 , a. w. , bloemen , j. b. g. m. , dame , t. m. , et al . 1988 , , 207 , 1 , a. m. , & gould , a. 2016 , , 590 , a2 , k. , fukui , y. , ogawa , h. , & kawabata , k. 1986 , , 303 , 667 , k. , yamaguchi , c. , nakai , n. , et al . 2000 , in , vol . 4015 , radio telescopes , ed . h. r. butcher , 237246 , h. , fukui , y. , fujimoto , y. , et al . 1986 , , 166 , 276 , k. , tasker , e. j. , & habe , a. 2014 , , 792 , 63 , j. c. , beltrn , m. t. , caselli , p. , et al . 2014 , protostars and planets vi , 149 , k. , enokiya , r. , sano , h. , et al . 2011 , , 738 , 46 , k. , hasegawa , k. , hattori , y. , et al . 2015 , , 806 , 7 , g. j. , & sandell , g. 1995 , , 299 , 179 , a. p. , & zinnecker , h. 2004 , , 427 , 299 , b. a. , dame , t. m. , masheder , m. r. w. , & thaddeus , p. 2005 , , 430 , 523 , m. g. , & cassinelli , j. p. 1986 , , 310 , 207 , c. , sunada , k. , iizuka , y. , iwashita , h. , & noguchi , t. 2000 , in , vol . 4015 , radio telescopes , ed . h. r. butcher , 614623 , a. , russeil , d. , motte , f. , et al . 2010 , , 518 , l81 , h. , & yorke , h. w. 2007 , , 45 , 481 [ table:1 ] [ table:2 ] .velocity range of the molecular emission in the 8 cases ( table 1 ) . [ cols="^,^,^,^,^,^,^,^,^,^,^,^ " , ] we describe the method for a simple case of the complementary distribution in the x - y plane ( figure 17 ) where the small cloud and the intensity depression are assumed to be circles of the same radius a with an initial displacement of @xmath61 along the x - axis in the sky . by shifting the small cloud by @xmath41 in the direction of the intensity depression along the x - axis , we calculate the overlapping product between the two circles represented as @xmath62 and @xmath63 by assuming that they have uniform intensity 1.0 in an arbitrary unit inside the circle and 0.0 outside the circle . by taking a product of the circle and the intensity depression at @xmath41 , the overlapping between the two circles is given by @xmath64 , the product of @xmath65 and @xmath66 . @xmath67 is integrated in the x - y plane and is expressed by @xmath39 . the displacement of the two circles is then estimated to be @xmath68 at the peak of @xmath39 ( figure 2 ) . the method is applied to the synthetic observations in figures 2(b ) , ( c ) , and ( d ) . in figures 2(b ) and ( c ) , the distributions of the small cloud and the intensity depression are taken from figure 4 . grid spacing is given as [email protected] in the x - y plane according to the numerical simulations @xcite , and we assign a numerical value to be 1.0 , for each pixel inside the both components , and to be zero outside of them as above ( figures 17(f ) and ( g ) ) . @xmath39 is calculated in the same manner as above with an incremental step of 0.1pc in @xmath41 and is shown in figure 2(d ) . the peak position of @xmath39 gives the displacement to be @xmath62pc . by correcting for the projection effect , the displacement corresponds to the length of the cavity @xmath63pc , where the size of the small cloud is taken into account . we find the value is roughly consistent with the length of the cavity in the large cloud , @xmath64pc , at 1.6myr in the numerical simulations . in section 3 , we applied the method to estimate the displacement between the orion key and keyhole .

recently , dessins denfants has been proven to be a powerful tool in certain types of supersymmetry gauge theories , bringing a relationship to combinatorics and number theory . dessins denfants are-cell embeddings of a finite and connected bipartite graph in a closed orientable surface . in general , a finite and connected bipartite graph may appear as the underlying bipartite graph of many non - isomorphic dessins denfants . in this article we provide a constructive procedure to compute all those non - isomorphic dessins denfants having the same underlying bipartite graph , together their automorphims group , monodromy group and orientability type .

1611.02901

a dessin denfant , introduced by grothendieck in its _ esquisse de un programme _ @xcite , is a @xmath0-cell embedding of a finite and connected bipartite graph @xmath1 on a closed orientable surface @xmath2 ; see @xcite for some generalities . the dessin denfant is called orientable if its faces can be labelled by signs @xmath3 and @xmath4 , so that adjacent faces have different label . as a consequence of belyi s theorem @xcite , there is a natural action of the absolute galois group @xmath5 on dessins denfants , which is known to be faithful @xcite . grothendieck observed that using that action may permits to obtain information on the internal structure of @xmath5 by codified by such an action in terms of simple combinatorial objects . in the recent years there has been an interest on dessins denfants in the field of supersymmetric gauge theories @xcite and conformal field theories @xcite to mention some of the applications in theoretical physics . in this way , it seems interesting searching for algorithms to provide examples of dessins denfants , together some of their invariants , up to isomorphisms . by the definition , a dessin denfant has associated a bipartite graph , but the dessin is not uniquely determined up to isomorphisms by it . in this paper we provide an algorithm which permits to construct all non - isomorphic dessins denfants , together their automorphims group , monodromy groups and orientability type , having the same underlaying bipartite graph . in section [ sec : appendix ] we describe , for a particular bipartite graph , the implementation of the above algorithm into gap @xcite which we have used for all our examples in section [ sec : ejemplos ] for certain classical bipartite graphs . as an example of our algorithm , for the double - prism shown in figure [ doubleprism ] ( see example [ sec : doble - prisma ] ) , we are able to find two non - isomorphic dessins denfants @xmath6 and @xmath7 with the same bipartite graph , both of genus one , the same passport @xmath8 and isomorphic monodromy groups . the dessin denfant @xmath6 is chiral and @xmath7 is reflexive ; so they can not be in the same galois orbit . the dessin @xmath9 is defined by the following permutations @xmath10 @xmath11 @xmath12 @xmath13 : 1 . a finite connected bipartite graph ( vertices are either black or white ) @xmath1 whose @xmath14 black vertices are @xmath15 , its @xmath16 white vertices are @xmath17 , and it has @xmath18 edges . 2 . the group @xmath19 of bipartite graph automorphisms of @xmath1 ( graph automorphisms preserving vertices of a fixed colour ) . : 1 . a maximal collection of non - isomorphic dessins denfants together their monodromy groups , automorphisms group and orientability type , whose subjacent bipartite graph is @xmath1 . : 1 . fix an enumeration of the @xmath20 edges of @xmath1 with numbers in the set @xmath21 without repeating . the above enumeration provides a natural injective homomophism from @xmath22 . 3 . for each black vertex @xmath23 ( respectively , white vertex @xmath24 ) we consider the collection @xmath25 ( respectively , @xmath26 ) all possible cycles @xmath27 ( respectively , @xmath28 ) of length equal to the degree of such vertex in @xmath1 using the numbers at all the edges at such a vertex . set @xmath29 , where @xmath30 is the collection of all the permutations @xmath31 , where @xmath32 , and @xmath33 is the collection of all the permutations @xmath34 , where @xmath35 . the connectivity of @xmath1 asserts that for each pair @xmath36 , the group @xmath37 is a transitive subgroup of @xmath38 , so it defines a dessin denfant @xmath39 whose bipartite subjacent graph is @xmath1 . 5 . consider the natural action of @xmath19 on the set @xmath40 given as : @xmath41 in the above , multiplication of permutations are from the left to the right as is done in gap @xcite . 6 . as a consequence of theorem [ teomain ] ( see section [ sec : mainpart ] ) we obtain the following facts . 1 . if @xmath42 is a dessin denfant whose underlying bipartite graph is isomorphic to @xmath1 , then it is isomorphic to @xmath39 for a suitable @xmath36 . 2 . two pairs in @xmath40 define isomorphic dessins denfant if and only if they belong to the same @xmath19-orbit . the group of automorphism of the dessin denfant @xmath43 , where @xmath36 is naturally isomorphic to the @xmath19-stabilizer of @xmath44 . a ( finite ) _ graph _ is a triple @xmath45 , where @xmath46 and @xmath47 are finite sets and @xmath48 is the incident map . the elements of @xmath46 are called the vertices and those of @xmath47 the edges . if @xmath49 , then we say that @xmath20 and the vertices @xmath50 and @xmath51 are _ incident _ ( in case that @xmath52 , we say that @xmath20 is a loop ) . if @xmath53 , then its _ degree _ @xmath54 is the number of incident edges with @xmath55 ( loops are counted twice ) the graph is called _ connected _ if for every pair of vertices @xmath56 there is ( finite ) collection of edges @xmath57 so that @xmath58 and @xmath50 are incident , @xmath59 and @xmath51 are incident and @xmath60 with @xmath61 are both incident to a common vertex . in this paper all graphs will be assumed to be connected and finite . an _ isomorphism _ of the graphs @xmath62 and @xmath63 is a pair @xmath64 , where @xmath65 and @xmath66 are bijective functions respecting the incident maps , i.e. , @xmath67 \circ \varepsilon_{1}=\varepsilon_{2 } \circ \phi_{2},\ ] ] where @xmath68 $ ] is the induced map by @xmath69 . a self - isomorphism of a graph is called a _ graph - automorphism _ of it and we denote by @xmath70 the _ group of graph automorphisms _ of the graph @xmath1 . if we enumerate the @xmath20 edges of the graph @xmath1 with numbers in @xmath21 ( without repeating ) , then each automorphism @xmath71 defines a permutation @xmath72 ; so providing in this way a natural homomorphism @xmath73 . observe at this point that a graph automorphism may be the identity at the level of the edges , but not at the level of the vertices . the only cases of connected graphs with such a pathology are those having exactly two vertices and without loops , called _ special graphs _ ( see figure [ fig : mulase ] ) @xcite . so , if the graph is not a special graph , the homomorphism @xmath74 is injective . edges*,width=75 ] a _ bipartite graph _ is a graph together a coloring of its vertices using two colors , black and white , so that adjacent vertices have different colors . the passport of a bipartite graph @xmath1 is a tuple of the form @xmath75 where @xmath76 @xmath77 let us note that , if @xmath20 is the number of edges of the bipartite graph , then @xmath78 in the case that all white vertices have degree two , we say that the bipartite graph is _ clean_. an element of @xmath70 either keeps invariant the vertices of a fixed color or it interchanges the vertices of different colours . we will denote by @xmath19 the subgroup of automorphisms of the bipartite graph @xmath1 , that is , those automorphisms which sends black vertices to black vertices . in most of the cases , @xmath79 ; otherwise @xmath80=2 $ ] ( in this case , the elements in @xmath81 permutes the white vertices with the black ones ) . we have that the restriction @xmath82 is then an injective homomorphism . if @xmath1 is a finite and connected graph , then a clean bipartite graph @xmath83 is obtained by colouring all vertices of @xmath1 in black and then taking a white vertex in the interior of each of its edges . in this case , every graph automorphism of @xmath1 induces a bipartite graph automorphism of @xmath83 ( observe that @xmath84 if at least one of the vertices of @xmath1 has degree at least three ) , so providing an embedding @xmath85 ; which is surjective if and only if the graph @xmath1 has no loops . in the case when @xmath1 has loops , each loop @xmath20 provides an extra involution @xmath86 that permutes both edges of @xmath83 contained in @xmath20 and acts as the identity on all other edges ; and the group generated by @xmath87 and all the involutions @xmath88 ( where @xmath20 runs over all lops of @xmath1 ) generates the group @xmath89 . in particular , if @xmath1 has no loops , @xmath89 is naturally isomorphic to @xmath70 . next , we will recall some definitions and facts about dessins denfants ( the reader may look , for instance , at @xcite ) . an _ embedding _ ( or a drawing ) of a graph @xmath45 on a closed orientable surface @xmath2 , denoted this by the symbol @xmath90 , is a pair of injective functions @xmath91 and @xmath92 , so that @xmath93 is a an arc homeomorphic to the unit open interval , @xmath94 for @xmath95 , @xmath96 ( for every @xmath53 and every @xmath97 ) , if @xmath53 is incident to @xmath20 , then @xmath98 belong to one extreme of @xmath93 and every extreme of @xmath93 belongs to @xmath99 . the embedding is called a _ @xmath0-cell embedding _ if each connected component of @xmath100 is simply - connected , called the _ faces _ of the embedding . this last condition ensures that @xmath1 must be connected . a @xmath0-cell embedding @xmath90 induces a natural @xmath0-cell embedding @xmath101 . in this way , the study of @xmath0-cell embedding of finite and connected graphs on closed orientable surfaces is , in some way , equivalent to the study of @xmath0-cell embeddings of clean bipartite graphs . dessin denfant _ , as defined by grothendieck in @xcite , is a triple @xmath102 , where @xmath2 is a closed orientable surface , @xmath1 is a finite bipartite graph ( vertices are coloured in black and white ) and @xmath90 is a @xmath0-cell embedding . each face of the embedding , i.e. a connected component of @xmath103 , is called a _ face _ of the dessin . the genus of @xmath2 is also called the genus of the dessin denfant . a face of the dessin denfant is topologically a polygon of @xmath104 sides , where @xmath105 ; we say that that face has degree @xmath106 . the passport of the dessin denfant is the tuple @xmath107 where @xmath76 @xmath108 @xmath109 note that the tuple @xmath110 is the passport of the underlying bipartite graph @xmath1 . if @xmath20 is the number of edges of the dessin denfant , then @xmath111 and , as a consequence of euler s formula , the genus of @xmath2 is @xmath112 let us denote the black ( respectively , white ) vertices of @xmath1 as @xmath15 ( respectively , @xmath17 ) . next , we label the edges of @xmath1 with numbers in @xmath21 without repeating . for each black ( respectively , white ) vertex we chose a cyclic permutation of the edges at the @xmath113-image of that vertex in counterclockwise order ( here we are using the orientation of @xmath2 ) . we then consider , in @xmath38 , the permutation @xmath114 ( respectively , @xmath115 ) obtained as the product of all these cyclic permutations at black ( respectively , white ) vertices . the subgroup @xmath37 of @xmath38 , which is a transitive subgroup by the connectivity of the underlying graph , is called the _ monodromy group _ of the dessin denfant . by the construction , the number of disjoint cycles of @xmath114 ( respectively , @xmath115 ) is @xmath14 ( respectively , @xmath16 ) and the number @xmath116 of faces is equal to the number of disjoint cycles of @xmath117 . the integers @xmath118 are the lengths of the cycles of @xmath114 , the integers @xmath119 are the lengths of the cycles of @xmath115 and the integers @xmath120 are the lengths of the cycles of @xmath117 . any two - generator transitive subgroup @xmath121 of @xmath38 is the monodromy of some dessin denfant with @xmath20 edges . if @xmath114 and @xmath115 are two generators of @xmath121 , then the dessin denfant is constructed so that the disjoint cycles of @xmath114 correspond to the black vertices , the disjoint cycles of @xmath115 correspond to the white vertices and the disjoint cycles of @xmath117 corresponds to the faces . two dessins denfants @xmath122 and @xmath123 are called _ isomorphic _ ( respectively , _ non - orientable - isomorphic _ ) if there is an orientation - preserving ( respectively , orientation - reversing ) homeomorphism @xmath124 and an isomorphism of bipartite graphs @xmath125 ( i.e. , @xmath126 sends black ( respectively , white ) vertices to black ( respectively , white ) vertices , so that @xmath127 and @xmath128 . we say that such a homeomorphism @xmath129 ( or the pair @xmath130 ) is an _ isomorphism _ ( respectively , _ non - orientable isomorphism _ ) of the above two dessins denfants . the two dessins are called _ chirals _ if they are non - orientable isomorphic but they are not isomorphic . in terms of the cartography groups , the isomorphism of the dessins denfants @xmath6 and @xmath7 ( with the same number @xmath20 of edges ) can be stated as follows . let us enumerate the edges of each dessin and consider the corresponding monodromy groups @xmath131 and @xmath132 , respectively . then the dessins denfants are isomorphic if and only there is a permutation @xmath133 so that @xmath134 and @xmath135 . let us observe that the passport of ( orientable or non - orientable ) isomorphic dessins denfants is the same , but there known examples of non - isomorphic dessins denfant with the same passport . there are non - isomorphic dessins denfants ( with same number of edges @xmath20 ) with the same monodromy group ( this happens since we may have @xmath136 ) . an _ automorphism _ of a dessin denfant @xmath102 is given by any self - isomorphism @xmath130 of it . the _ group of automorphisms _ of @xmath42 is denoted by the symbol @xmath137 . there is a subgroup ( of index at most two ) @xmath138 , called the _ group of orientation - preserving automorphisms _ of the dessin denfant , which corresponds to those self - isomorphisms @xmath130 where @xmath129 is orientation - preserving . in most of the cases we have that @xmath139 ( this happen when the dessin has no orientation - reversing automorphisms ) . a dessin denfant admitting an orientation - reversing automorphism is called _ reflexive_. there is a natural homomorphism @xmath140 ( which is injective if the graph can not be embedded in a circle ) and its restriction @xmath141 ( which is always injective ) . if we fix a labelling of the @xmath20 edges of @xmath1 using numbers in @xmath21 without repeating , then the obtain a monodromy group @xmath142 of @xmath42 and also an injective homomorphism @xmath22 . the isomorphic image @xmath143 is the centralizer of the monodromy group , that is , the group formed by those permutation commuting with @xmath114 and with @xmath115 . the group @xmath144 is the monodromy group of some dessin denfant @xmath145 ; called the _ conjugated dessin _ of @xmath42 . the riemann surface structure defined by @xmath145 is the conjugated of that defined by @xmath42 . these two conjugated dessins are isomorphic if and only if @xmath42 admits an anticonformal automorphism ( i.e. , if the dessin is reflexive ) ; otherwise , @xmath42 and @xmath145 is a chiral pair . belyi pair _ is a pair @xmath146 , where @xmath147 is a closed riemann surface and @xmath148 is a non - constant meromorphic map whose branch values are contained in the set @xmath149 ; in this case we say that @xmath147 is a _ belyi curve _ and that @xmath150 is a _ belyi map _ for @xmath147 . as consequence of weil s descent theorem @xcite , every belyi pair can be defined over the field @xmath151 of algebraic numbers . on the other direction , belyi s theorem @xcite asserts that if @xmath147 is a closed riemann surface which can be defined by an algebraic curve over @xmath151 , then @xmath147 is a belyi curve ( and its has a belyi map over @xmath152 ) . in particular , there is a natural action of the absolute galois group @xmath5 on belyi pairs which is known to be faithfull ; in genus one observed by grothendieck , in genus zero by schneps @xcite , in hyperelliptic belyi pairs by gonzlez - diez and girondo @xcite and in the non - hyperelliptic case in @xcite . it is well known that there is bijective correspondence between the equivalence classes of the following objects ( see e.g @xcite ) 1 . dessins denfants @xmath153 with @xmath20 edges ; 2 . belyi pairs @xmath154 of degree @xmath20 ; 3 . subgroups @xmath155 of index @xmath20 of triangle groups @xmath156 ; 4 . pairs of permutations @xmath157 generating transitive subgroups @xmath158 of @xmath38 . the link between these four classes of objects is made as follows . given a belyi pair @xmath154 one gets a dessin denfant by setting @xmath159 , @xmath160)$ ] and black ( respectively , white ) vertices are provided by @xmath161 ( respectively , @xmath162 . a fuchsian group @xmath155 as in ( 3 ) defines a belyi function @xmath150 by simply considering the natural projection @xmath163 . finally , two permutations @xmath164 , of orders @xmath165 and @xmath166 as in ( 4 ) , with @xmath167 , give rise to a fuchsian group @xmath155 as in ( 3 ) by considering the epimorphism @xmath168 obtained by sending the generators @xmath169 of @xmath170 to @xmath114 and @xmath115 , respectively , and setting @xmath171 . we use freely this correspondence and we will speak of the belyi pair , the fuchsian group and the permutation representation ( or monodromy group ) defining ( or associated to ) a dessin @xmath42 . as a consequence of the previous correspondence , there is a faithful action of @xmath5 on the collection of dessins denfants . if @xmath42 is a dessin d enfant and @xmath172 , then @xmath173 will denote the image of the dessin denfant @xmath42 by the action of @xmath114 . the following properties of the dessin denfant are invariant under the action of the absolute galois group @xcite : 1 . the number of edges . 2 . the passport . 3 . the genus . 4 . the monodromy group . the group of automorphisms . a consequence of the above is that the passport of the underlying bipartite graph is also kept invariant under the action of the absolute galois group ; but not necessarily the isomorphism type of the graph . unfortunately , the above galois - invariants are not in general enough to decide if two dessins denfants belong to the same galois orbit ( examples can be found in @xcite ) . a dessin denfant @xmath102 is called _ orientable _ if we may paint its faces in two different colours so that adjacent faces have different colours . in @xcite it has been shown that orientability is another galois invariant of a dessin denfant and the following fact . let @xmath102 be a dessin denfant with associated permutations @xmath114 and @xmath115 . then @xmath42 is orientable if and only if there exists a homomorphism @xmath174 so that @xmath175 and @xmath176 . the _ field of moduli _ of a dessin denfant @xmath42 is the fixed field of the subgroup of @xmath5 formed of those field automorphisms @xmath114 so that @xmath173 is isomorphic to @xmath42 . a _ field of definition _ of a dessin denfant is a subfield @xmath177 of @xmath178 so that there is a belyi pair defined over @xmath177 defining the dessin . it is well known that every field of definition of a dessin contains its field of moduli and that the intersection of all of the fields of definitions is exactly the field of moduli @xcite ( at this point observe that this last fact will be in general false if we only consider fields of definitions inside @xmath151 ) . in this section we proceed to describe the theoretical bases for our algorithm which permits to obtain , up to isomorphisms , those dessins denfants having the same underlying bipartite graph . let @xmath1 be a ( connected and finite ) bipartite graph , whose black ( respectively , white ) vertices are @xmath15 ( respectively , @xmath17 ) and let @xmath19 be its bipartite graph automorphisms . let us fix a labelling of the @xmath20 edges of @xmath1 with numbers in @xmath21 without repetitions . as seen in section [ sec : grafo ] , this enumeration provides an injective homomorphism @xmath22 . for each black vertex @xmath23 ( respectively , white vertex @xmath24 ) we consider the collection @xmath25 ( respectively , @xmath26 ) of all possible cycles @xmath27 ( respectively , @xmath28 ) of length equal to the degree @xmath179 ( respectively , @xmath180 ) using all the labels at the edges at such a vertex . clearly , @xmath181 and @xmath182 . let @xmath30 ( respectively , @xmath33 ) be the collection of all the permutations @xmath31 , where @xmath32 ( respectively , @xmath34 , where @xmath35 ) . set @xmath29 , whose cardinality is @xmath183 the connectivity of @xmath1 asserts that , for each @xmath36 , the subgroup @xmath37 of @xmath38 is transitive ; so it is the monodromy group of a dessin denfant @xmath39 , whose underlying bipartite graph is @xmath1 . let us observe that every dessin denfant whose underlying graph is isomorphic to @xmath1 must be isomorphic to one of the dessins denfants defined by an element of @xmath40 . if @xmath83 is the clean bipartite graph associated to a finite and connected graph @xmath1 , then in the definition of the collection @xmath184 we may only consider the first coordinate @xmath114 s as the second one @xmath115 is uniquely determined . if @xmath36 and @xmath185 , then ( as @xmath186 preserves the colours and incidences of edges ) @xmath187 . this provides a natural action of @xmath19 over the collection @xmath40 . [ teomain ] 1 . two pairs in @xmath40 define isomorphic dessins denfants if and only if they belong to the same @xmath188-orbit . in particular , the cardinality of the quotient set @xmath189 is equal to the number of different isomorphic dessins denfant having @xmath1 as underlying bipartite graph . the @xmath188-stabilizer of the dessin denfant @xmath42 whose monodromy group is @xmath37 , for @xmath36 , is equal to @xmath190 and it is the group of automorphisms of it . let @xmath191 and let the corresponding dessins denfants be @xmath192 and @xmath193 . if these are isomorphic dessins denfant , then there is an orientation - preserving homeomorphism @xmath194 and an automorphism @xmath64 of @xmath1 ( as a bipartite graph ) so that @xmath195 . conversely , let us assume there is an automorphism @xmath186 of the bipartite graph @xmath1 so that @xmath196 conjugates @xmath197 to @xmath198 and @xmath199 to @xmath200 . then it also conjugates @xmath201 to @xmath202 . this permits to construct a an orientation - preserving homeomorphism @xmath203 so that @xmath130 is an automorphism of @xmath1 . this provides part ( 1 ) . part ( 2 ) is consequence of part ( 1 ) . let @xmath191 and let us assume that @xmath204 belongs to the @xmath19-orbit of @xmath205 . then the following holds . 1 . if @xmath205 and @xmath206 belong to different orbits ( i.e. they are non - isomorphic dessins ) , then @xmath207 and @xmath208 form a chiral pair . 2 . if @xmath205 and @xmath206 belong to the same orbit ( i.e. they are isomorphic dessins ) , then @xmath207 is reflexive . if @xmath1 is a bipartite graph with trivial group of automorphisms ( as bipartite graph ) , then the number of different isomorphic dessins denfant admitting @xmath1 as its bipartite graph is @xmath209 . in @xcite there were defined the wilson s operations on dessins denfants . these operations are defined as follows . let fix @xmath210 be positive integers so that @xmath106 ( respectively , @xmath211 ) is co - prime to all degrees of black ( respectively , white ) vertices of the bipartite graph @xmath1 . the wilson s operation @xmath212 is defined by sending @xmath36 to the new pair @xmath213 where @xmath214 and @xmath215 . a graph theoretic characterization of certain quasiplatonic curves defined over cyclotomic fields , based on wilson s operations on maps , is developed in @xcite given a finite and connected graph @xmath1 , there are many embeddings into a closed orientable surface ( not necessarily being @xmath0-cell embeddings ) . the minimal genus of such an embedding is called the graph genus of @xmath1 ; say @xmath216 . the characterization of the graph genus has been given in @xcite ; it states that it occurs for a @xmath0-cell embedding with a maximal number of faces . the determination of @xmath216 seems to be a difficult task , but in the same paper an algorithm to determine it was obtained ( claimed that such an algorithm is lengthy ) . it is known that the problem of finding the graph genus is np - hard and the problem of determining whether an n - vertex graph has genus @xmath217 is np - complete @xcite . for some types of graphs ( vertex transitive ones ) some information is known ; for instance , @xmath218 , @xmath219 ( see , @xcite ) , @xmath220 ( see , @xcite ) . our algorithm permits to determine the graph genus of @xmath1 as follows . assume the number of edges of the graph is @xmath20 and the number of its vertices is @xmath14 . we consider its associated clean bipartite graph @xmath83 , an enumeration of its @xmath221 edges and the corresponding collection @xmath184 . as there is only one possible permutation @xmath115 ( this being a product of transpositions ) , we may just consider only the permutations @xmath114 . now , for each @xmath114 , we consider the product permutation @xmath117 and we let @xmath116 be the number of its disjoint cycles . if @xmath222 is the maximal possible value of @xmath116 , then the minimal genus of @xmath1 is @xmath223 similarly , if we let @xmath224 be the minimal value for @xmath116 , then the maximal embedding genus of @xmath1 is @xmath225 in this section we consider some well known bipartite graphs in order to compute the number of non - isomorphic dessins denfants with that subjacent bipartite graph . let @xmath226 be the clean bipartite graph obtained from the one skeleton of the tetrahedra ( its vertices being the black vertices and as white vertices we use a middle point of each side ; see figure [ fig2 ] ) . in this case @xmath227 , @xmath228 and @xmath229 . in @xcite two non - isomorphic dessins denfants ( one of genus zero and the other of genus one ) admitting @xmath226 as bipartite graph were provided . our algorithm permits to see that there are exactly three non - isomorphic dessins denfants with such property ; the missing one has also genus one . in this case , @xmath230 and @xmath231 , where @xmath232 @xmath233 * , width=113 ] the set @xmath234 has @xmath235 elements and @xmath236 has three elements , these are represented by the pairs @xmath237 , where @xmath238 @xmath239 @xmath240 that is , there are exactly three non - isomorphic dessins denfant with @xmath241 as bipartite graph . the @xmath242-orbit of @xmath243 has length @xmath244 , the one of @xmath245 has length @xmath246 and the one of @xmath247 has length @xmath0 . the dessin denfant defined by the pair @xmath248 has genus @xmath249 , its monodromy group is isomorphic to @xmath250 and its group of automorphisms is isomorphic to @xmath251 ( see the left in figure [ figd1 ] ) . the dessin denfant defined by the pair @xmath252 has genus @xmath249 , its monodromy group is isomorphic to @xmath253 and its group of automorphisms is isomorphic to @xmath254 ( the dihedral group of order @xmath244 ) ( see the right in figure [ figd1 ] ) . the dessin denfant defined by the pair @xmath247 has genus @xmath255 , is regular and its monodromy group ( isomorphic to the group of automorphisms ) is isomorphic to @xmath256 . all these three dessins are reflexive ones . with @xmath226 as bipartite graph*,title="fig:",width=113 ] with @xmath226 as bipartite graph*,title="fig:",width=151 ] let @xmath257 be the clean bipartite graph associated to @xmath258 ( its black vertices are the vertices of @xmath258 and the white vertices are given at middle points of all of the edges ) , so @xmath259 , @xmath260 and @xmath261 . in this case , @xmath262 and @xmath231 , where @xmath263 @xmath264 @xmath265 the set @xmath266 has @xmath267 elements and @xmath268 has three elements , these are represented by the pairs @xmath237 , where @xmath269 @xmath270 @xmath271 @xmath272 so , there are exactly three non - isomorphic dessins denfant with @xmath257 as bipartite graph ( i.e. , there exactly three topologically types of embedding the graph @xmath258 in an orientable closed surface as a map ) . the @xmath273-orbit of @xmath243 has length @xmath274 , the one of @xmath245 has length @xmath275 and the one of @xmath247 has length @xmath276 . the dessin denfant defined by the pair @xmath248 has genus @xmath249 , is regular and its monodromy group ( isomorphic to the group of automorphisms ) is isomorphic to @xmath277 . this regular dessin denfant is defined over the fermat curve @xmath278 ; the factor @xmath251 is generated by the automorphism @xmath279:=[x : y:\omega_{3}z]$ ] and the factor @xmath280 is generated by the involution @xmath281=[y : x : z]$ ] and the order three automorphism @xmath282=[\omega_{3}x:\omega_{3}^{2}y : z]$ ] , where @xmath283 . the dessin denfant defined by the pair @xmath252 has genus @xmath284 , its monodromy group is isomorphic to @xmath285 and its group of automorphisms is isomorphic to @xmath251 . this corresponds to the genus two riemann surface defined by @xmath286 . the dessin denfant defined by the pair @xmath247 has genus @xmath249 , its monodromy group is isomorphic to @xmath287 and its group of automorphisms is isomorphic to @xmath288 . all the above dessins are reflexive . the first example of a finite and connected graph @xmath289 with trivial group of automorphisms was provided by r. frucht in @xcite ( see figure [ figfrucht1 ] ) . the associated clean bipartite graph @xmath290 is shown in figure [ figfrucht2 ] . in this case , @xmath291 , @xmath292 and @xmath293 , the set @xmath294 has cardinality @xmath295 and its elements represent the non - isomorphic dessins denfants admitting the bipartite graph @xmath290 . the elements of @xmath294 are given by the pairs @xmath296 , where @xmath297 @xmath298 and the @xmath295 @xmath299 are of the form @xmath300 @xmath301 * , width=113 ] associated to frucht s graph*,width=151 ] the genus formula @xmath302 , where @xmath116 denotes the number of faces of the dessin denfant , ensures that the possible genus are @xmath303 . it is easy to see that @xmath255 is possible ( the figure [ figfrucht2 ] shows a dessin denfant of genus zero ) ; this is provided with @xmath304 @xmath305 and the eight faces correspond to cycles of @xmath306 @xmath307 a dessin of genus one is provided by @xmath308 @xmath309 in which case we have six faces and @xmath310 @xmath311 a dessin of genus two is provided by @xmath312 @xmath313 in which case we have four faces and @xmath314 @xmath315 a dessin of genus three is provided by @xmath316 @xmath317 in which case we have two faces and @xmath318 @xmath319 all the above dessins are reflexive . let @xmath321 be the clean bipartite graph associated to the complete graph @xmath320 ( see figure [ figk5 ] ) . with the given enumeration of the edges we have @xmath322 and there are @xmath323 choices for @xmath114 . in this case , @xmath324 and @xmath325 , where @xmath326 @xmath327 * , width=188 ] the set @xmath328 has @xmath329 elements , that is , there are exactly @xmath329 non - isomorphic dessins denfant whose bipartite graph is @xmath321 . between these @xmath329 dessins denfants , there are exactly nine of genus @xmath249 ; these are given by the following choices for @xmath114 : @xmath330 @xmath331 @xmath332 @xmath333 @xmath334 @xmath335 @xmath336 @xmath337 @xmath338 let us denote by @xmath339 the dessin denfant defined by @xmath340 , for @xmath341 . the dessins denfant @xmath339 , for @xmath342 , are non - uniform and those defined by @xmath343 and @xmath344 are regular . the two regular ones have passport @xmath345 and they form a chiral pair . the automorphism group is isomorphic to @xmath346 and they are defined over the same elliptic curve @xmath347 . the dessin @xmath348 is the only one whose monodromy group is isomorphic to @xmath349 ( whose group of automorphisms is @xmath350 ) ; this again is over the elliptic curve @xmath347 . this dessin is reflexive and its passport is @xmath351 . the genus one dessin denfant @xmath352 is the only with monodromy group of order @xmath353 and it has trivial group of automorphisms . this dessin is reflexive and its passport is @xmath354 . the other five dessins have monodromy group of order @xmath355 and group of automorphisms isomorphic to @xmath288 . the dessins @xmath356 and @xmath357 ( respectively , by @xmath358 and by @xmath359 ) form a chiral pair . the passport of @xmath356 and @xmath357 is @xmath360 and the passport of @xmath358 and @xmath359 is @xmath351 . the dessin @xmath254 is reflexive and its passport is @xmath361 . let us now consider the bipartite graph @xmath1 as shown in figure [ doubleprism ] with the given enumeration of the edges . with the given enumeration , @xmath362 and there are @xmath363 choices for @xmath114 . the dessins denfants admitting the above bipartite graph are of genus @xmath303 and there are exactly @xmath364 non - isomorphic ones : two non - isomorphic dessins of genus zero , @xmath365 of genus one , @xmath366 of genus two and @xmath367 of genus three . of these genus one dessins denfants , there are exactly @xmath368 with passport @xmath8 , all of them have monodromy group of order @xmath369 , exactly @xmath370 of them which are reflexive and the others are chirals pairs . two of these dessins , say @xmath6 and @xmath7 are provided by the permutations @xmath11 @xmath12 @xmath371 using gap @xcite one can check that both of them have isomorphic monodromy group ; but , the dessin denfant @xmath6 is chiral ( with group of automorphisms isomorphic to @xmath288 ) and @xmath7 is reflexive ( with trivial group of automorphisms ) ; so they can not be in the same galois orbit . there are , up to isomorphisms , exactly three bipartite graphs with passport @xmath372 ( @xmath373 , @xmath374 ) ; these being the bipartite graphs @xmath258 , @xmath375 and @xmath376 shown in figure [ fig1 ] . in each case we fix a labelling of the black vertices as @xmath377 , white vertices as @xmath378 and the nine edges with numbers in @xmath379 without repeating . for each @xmath380 , the cardinality of @xmath40 is @xmath267 . we proceed , in each case , to describe all non - isomorphic dessins denfants whose bipartite graph is isomorphic to @xmath1 . as the black vertices have odd degree , none of the dessins admitting these bipartite graphs can be orientable . * , width=264 ] in this case , @xmath381 and @xmath382 , where @xmath383 the set @xmath384 has four elements , these are represented by the pairs @xmath205 , @xmath385 , @xmath386 and @xmath206 , where @xmath387 @xmath388 that is , there are exactly four non - isomorphic dessins denfant with @xmath258 as bipartite graph . the @xmath19-orbit of @xmath205 has length @xmath274 , the orbits of @xmath385 and also of @xmath386 have lengths @xmath389 and that of @xmath206 has length @xmath276 . these four non - isomorphic dessins have the following properties : 1 . the dessin denfant with monodromy group @xmath390 has genus @xmath249 ( since @xmath391 ) , its passport is @xmath392 , its @xmath393-stabilizer is given by the group @xmath394 ( its group of automorphisms ) and the underlying riemann surface of genus one is defined by the fermat curve of degree three : @xmath395 . in this case , the group of automorphisms @xmath396 is generated by the order three automorphisms @xmath397:=[\omega_{3}x : y : z]$ ] and @xmath398:=[x:\omega_{3}y : z]$ ] . this dessin is reflexive . 2 . the dessin denfant with monodromy group @xmath399 has genus @xmath284 ( since @xmath400 ) , its passport is @xmath401 , its @xmath393-stabilizer is @xmath402 ( its group of automorphisms ) and the underlying riemann surface of genus two is defined by @xmath286 . 3 . the dessin denfant with monodromy group @xmath403 has genus @xmath284 ( since @xmath404 ) , its passport is @xmath401 , its @xmath393-stabilizer is @xmath405 ( its group of automorphisms ) and the same underlying riemann surface of genus two as above ( @xmath406 and @xmath407 are chirals ) . 4 . the dessin denfant with monodromy group @xmath408 has genus @xmath249 ( since @xmath409 ) , its passport is @xmath410 , it is non - uniform , and it has trivial @xmath393-stabilizer ( its group of automorphisms ) . this dessin is reflexive . as the dessin denfant has trivial group of automorphisms , it can be defined over its field of moduli . as the dessin denfant is unique , its field of moduli is @xmath152 . in particular , the genus one riemann surface it defines can be defined over @xmath152 . in this example there are four wilson s operations : @xmath411 ( identity ) , @xmath412 and @xmath413 ( each one of order two ) ; they form a copy of @xmath414 inside @xmath287 . this group keeps invariant @xmath415 and , moreover , keeps invariant each of the @xmath274 @xmath19-orbits . the action is transitive on the @xmath19-orbit of @xmath205 , produces @xmath370 orbits in each of the @xmath19-orbits of @xmath385 and of @xmath386 and produces @xmath416 orbits in the @xmath19-orbit of @xmath206 . in particular , there are isomorphic dessins denfants which are not equivalent by the wilson s operations . in this case , @xmath417 and @xmath418 , where @xmath419 the set @xmath420 has four elements , these are represented by the pairs @xmath421 , @xmath422 , @xmath423 and @xmath424 , where @xmath425 @xmath426 @xmath427 that is , there are exactly four non - isomorphic dessins denfant with @xmath375 as bipartite graph . the @xmath19-orbit of @xmath421 has length @xmath275 , the orbit of @xmath422 has length @xmath244 , the orbit of @xmath423 have lengths @xmath275 and that of @xmath424 has length @xmath244 . these four non - isomorphic dessins have the following properties : 1 . the dessin denfant with monodromy group @xmath428 has genus @xmath249 ( since @xmath429 ) , its passport is @xmath430 and its @xmath431-stabilizer is trivial ( its group of automorphisms ) . 2 . the dessin denfant with monodromy group @xmath432 has genus @xmath249 ( since @xmath433 ) , its passport is @xmath392 and its @xmath431-stabilizer @xmath434 ( its group of automorphisms ) . 3 . the dessin denfant with monodromy group @xmath435 has genus @xmath249 ( since @xmath436 ) , its passport is @xmath437 and its @xmath431-stabilizer is trivial ( its group of automorphisms ) . 4 . the dessin denfant with monodromy group @xmath438 has genus @xmath255 ( since @xmath439 ) , its passport is @xmath440 and its @xmath431-stabilizer @xmath441 ( its group of automorphisms ) . all these dessins are reflexive . in this case , @xmath442 and @xmath443 , where @xmath444 the set @xmath445 has eight elements , these are represented by the pairs @xmath421 , @xmath422 , @xmath423 , @xmath424 , @xmath446 , @xmath447 , @xmath448 and @xmath449 , where @xmath425 @xmath450 @xmath451 @xmath452 @xmath453 that is , there are exactly eight non - isomorphic dessins denfant with @xmath376 as bipartite graph . the @xmath19-orbit of each @xmath454 has length @xmath244 . these eight non - isomorphic dessins have the following properties : 1 . the dessin denfant with monodromy group @xmath455 has genus @xmath284 ( since @xmath456 ) , its passport is @xmath401 and its @xmath457-stabilizer is trivial ( its group of automorphisms ) . this dessin is reflexive . 2 . the dessin denfant with monodromy group @xmath458 has genus @xmath249 ( since @xmath459 ) , its passport is @xmath460 and its @xmath457-stabilizer is trivial ( its group of automorphisms ) . this dessin is chiral to the one defined by @xmath461 . 3 . the dessin denfant with monodromy group @xmath462 has genus @xmath284 ( since @xmath463 ) , its passport is @xmath401 and its @xmath457-stabilizer is trivial ( its group of automorphisms ) . this dessin is reflexive . 4 . the dessin denfant with monodromy group @xmath464 has genus @xmath249 ( since @xmath465 ) , its passport is @xmath466 and its @xmath457-stabilizer is trivial ( its group of automorphisms ) . this dessin is reflexive . the dessin denfant with monodromy group @xmath467 has genus @xmath249 ( since @xmath468 ) , its passport is @xmath460 and its @xmath457-stabilizer is trivial ( its group of automorphisms ) . this dessin is chiral to the one defined by @xmath469 . the dessin denfant with monodromy group @xmath470 has genus @xmath249 ( since @xmath471 ) , its passport is @xmath437 and its @xmath457-stabilizer is trivial ( its group of automorphisms ) . this dessin is reflexive . the dessin denfant with monodromy group @xmath472 has genus @xmath255 ( since @xmath473 ) , its passport is @xmath474 and its @xmath457-stabilizer is trivial ( its group of automorphisms ) . this dessin is reflexive . the dessin denfant with monodromy group @xmath475 has genus @xmath249 ( since @xmath476 ) , its passport is @xmath466 and its @xmath457-stabilizer is trivial ( its group of automorphisms ) . this dessin is reflexive . [ [ section ] ] table [ tabla1 ] summarizes all the above . we may see the following facts . 1 . there are exactly two non - isomorphic genus zero dessins denfants whose bipartite graphs have passport @xmath372 ; so each one has field of moduli @xmath152 . for instance , the one corresponding to @xmath375 has belyi map @xmath477 where @xmath283 ( so it is defined over the extension of degree two @xmath478 ) . if @xmath479 , then @xmath480 + in this way , noticing that @xmath481 is a weil s co - cycle with respect to the galois extension @xmath482 , we may see that this dessin denfant is in fact definable over @xmath152 . there are nine non - isomorphic genus one dessins denfants whose bipartite graphs have passport @xmath372 . with the exceptions of those with monodromy group @xmath483 and @xmath484 , each of them has field of moduli @xmath152 . the two dessins @xmath483 and @xmath484 form a galois orbit and they are chirals . 3 . there are three genus two dessins denfants , up to strong - isomorphisms , whose bipartite graphs have passport @xmath372 , all of them with field of moduli @xmath152 and reflexive . all those dessins in table [ tabla1 ] with trivial group of automorphisms together the regular one are definable over their field of moduli . [ tabla1 ] 99 s. ashok , f. cachazo and e. dellaquila . strebel differentials with integral lengths and argyres - douglas singularities . preprint ( 2006 ) . arxiv : hep - th/0610080 . s. ashok , f. cachazo and e. dellaquila . children?s drawings from seiberg - witten curves . _ commun number theory phys . _ * 1 * ( 2007 ) , 237305 . doi : 10.4310/cntp.2007.v1.n2.a1 y - h . he and j. read . dessins denfants in @xmath485 generalised quiver theories . _ j. high energy phys . _ * 85 * ( 2015 ) , 129 . doi : 10.1007/jhep08(2015)085 g. v. belyi . on galois extensions of a maximal cyclotomic field . _ mathematics of the ussr - izvestiya _ * 14 * no.2 ( 1980 ) , 247256 . r. frucht . herstellung von graphen mit vorgegebener abstrakter gruppe . _ compositio mathematica _ ( in german ) * 6 * ( 1939 ) , : 239250 . the gap group . groups , algorithms , and programming , version 4.8.3 ; 2016 . ( http://www.gap-system.org ) . e. girondo and g. gonzlez - diez . _ introduction to compact riemann surfaces and dessins denfants_. london mathematical society student texts * 79*. cambridge university press , cambridge , 2012 . e. girondo and g. gonzlez - diez . a note on the action of the absolute galois group on dessins . london math . soc . _ * 39 * no . 5 ( 2007 ) , 721723 . e. girondo , g. gonzez - diez and r. a. hidalgo . orientable dessins denfant . in preparation . a. grothendieck . esquisse dun programme ( 1984 ) . in geometric galois actions . l. schneps and p. lochak eds . , _ london math . notes ser . _ * 242*. cambridge university press , cambridge , 1997 , 547 . a. hanany , y - h . he , v. jejjala , j. pasukonis , s. ramgoolam and d. rodriguez - gomez . the beta ansatz : a tale of two complex structures . _ j. high energy phys . _ * 1106 * ( 2011 ) , 056 . doi : 10.1007/jhep06(2011)056 v. jejjala , s. ramgoolam and d. rodriguez - gomez . toric cfts , permutation triples and belyi pairs . _ j. high energy phys . _ * 1103 * ( 2011 ) 065 . doi:10.1007/jhep03(2011)065 r.a . the fiber product of riemann surfaces : a kleinian group point of view . _ analysis and mathematical physics _ * 1 * ( 2011 ) , 37 - 45 . r. a. hidalgo and p. johnson . field of moduli of generalized fermat curves of type @xmath486 with an application to non - hyperelliptic dessins denfants . _ journal of symbolic computation _ * 77 * ( 2015 ) 6072 . r. jagadeesan . a new @xmath5-invariant of dessins denfants . _ proc . of the london math . ( 2015 ) doi : 10.1112/plms / pdv039 v. jejjala , s. ramgoolam and d. rodriguez - gomez . toric cfts , permutation triples and belyi pairs . _ jhep _ * 1103 * ( 2011 ) , 65 . g. a. jones and d. singerman . belyi functions , hypermaps and galois groups . london math . _ * 28 * ( 6 ) ( 1996 ) , 561590 . g. a. jones and j. wolfart . _ dessins denfants on riemann surfaces_. springer monographs in mathematics . g. a. jones , m. streit and j. wolfart . wilson s map operations on regular dessins and cyclotomic fields of definition . _ proc . of the london math . * 100 * ( 2 ) ( 2010 ) , 510532 . s. koizumi . fields of moduli for polarized abelian varieties and for curves . _ nagoya math . j. _ * 48 * ( 1972 ) , 3755 . s. k. lando and a. k. zvokin . _ graphs on surfaces and their applications_. encyclopeadia of mathematical sciencies * 141 * , springer - verlag , berlin 2004 . with an appendix by don b. zagier , low dimensional tolpology ii . m. mulase and m. penkava . ribbon graphs , quadratic differentials on riemann surfaces , and algebraic curves defined over @xmath151 . _ the asian journal of mathematics _ * 2 * ( 4 ) ( 1998 ) , 875920 . g. ringel . der vollstndige paare graph auf nichtorientierbaren flchen . ( german ) _ j. reine angew . _ * 220 * ( 1965 ) , 8893 . g. ringel and j. w. t. youngs . solution of the heawood map - coloring problem . _ * 60 * ( 1968 ) , 438445 . l. schneps . dessins denfant on the riemann sphere . in _ the grothendieck theory of dessins denfants_. edited by leila schneps . _ london math . notes ser . _ * 200*. cambridge university press , cambridge , 1994 , 4777 . c. thomassen . the graph genus problem is np - complete . _ journal of algorithms _ * 10 * ( 4 ) ( 1989 ) , 568576 . the field of definition of a variety . _ * 78 * ( 1956 ) , 509524 . s. e. wilson . operators over regular maps . _ pacific j. math . _ * 81 * ( 1979 ) 559568 . j. wolfart . the obvious part of belyi s theorem and riemann surfaces with many automorphisms . in geometric galois actions , vol . 1 , _ london math . lecture note ser . _ , * 242 * , cambridge univ . press , cambridge , 97112 ( 1997 ) . j. wolfart . @xmath487 for polynomials , dessins denfants and uniformization a survey . elementare und analytische zahlentheorie , 313345 , schr . wiss . ges . johann wolfgang goethe univ . frankfurt am main , 20 , franz steiner verlag stuttgart , stuttgart , 2006 . j. w. t. youngs . minimal imbeddings and the genus of a graph . _ journal of mathematics and mechanics _ * 12 * no . 2 ( 1963 ) , 303315 . l. zapponi . fleurs , arbres et cellules : un invariant galoisien pour une famille darbres . _ compositio mathematica _ * 122 * ( 2000 ) , 113133 . in this section we describe the algorithm ( written for gap ) to obtain all non - isomorphic dessins denfants , together their automorphisms groups and orientability for the complete bipartite graph @xmath488 , which is obtained from the complete bipartite graph @xmath489 by replacing each edge by two edges . in this case , we have two black vertices @xmath490 and two white vertices @xmath491 and there are exactly two edges connecting each black vertex to each white vertex . the group @xmath492 is isomorphic to @xmath493 . we enumerate both edges connecting @xmath50 and @xmath494 as @xmath495 and @xmath0 ; both edges connecting @xmath50 and @xmath496 as @xmath370 and @xmath274 ; both edges connecting @xmath51 and @xmath494 as @xmath497 and @xmath246 ; and both edges connecting @xmath51 and @xmath496 as @xmath498 and @xmath244 . with the above enumeration of edges , we first introduce the group @xmath499 of automorphisms of the bipartite graph ( with the enumeration provided ) 1 . a1:=[(1,2,3,4),(1,2,4,3),(1,3,2,4),(1,4,2,3),(1,3,4,2),(1,4,3,2 ) ] ; 2 . a2:=[(5,6,7,8),(5,6,8,7),(5,7,6,8),(5,7,8,6),(5,8,6,7),(5,8,7,6 ) ] ; 3 . b1:=[(1,2,5,6),(1,2,6,5),(1,5,2,6),(1,5,6,2),(1,6,2,5),(1,6,5,2 ) ] ; 4 . b2:=[(3,4,7,8),(3,4,8,7),(3,7,4,8),(3,7,8,4),(3,8,4,7),(3,8,7,4 ) ] ; 5 . j1:=0 ; ; j2:=0 ; ; x:= [ ] ; ; y:= [ ] ; ; 6 . for j1 in [ 1 .. 6 ] do + for j2 in [ 1 .. 6 ] do + add(x , a1[j1]*a2[j2 ] ) ; add(y , b1[j1]*b2[j2 ] ) ; + od ; od ; 7 . xy:=cartesian(x , y ) ; the cardinality of @xmath501 is @xmath504 . now , we proceed to list all non - isomorphic dessins denfants of a fixed genus ( this is specified by the variable ot " which is the number of disjoint cycles of the permutation @xmath117 , i.e. , the number of faces ) , together their monodromy group , automorphism group and its orientability type ( a dessin denfant is orientable if we may paint its faces in two different colours so that adjacent faces have different colours ) . 1 . e:=8 ; ; 2 . alpha:=size(orbits(group([oo[1][1][1]]),[1 .. e],onpoints ) ) ; ; 3 . beta:=size(orbits(group([oo[1][1][2]]),[1 .. e],onpoints ) ) ; ; 4 . j:=0;;count:=0;;u:=group([(1,2 ) ] ) ; ; 5 . for j in [ 1 .. size(oo ) ] do + t:=group([oo[j][1][2]*oo[j][1][1 ] ] ) ; ot:=orbits(t,[1 .. e],onpoints ) ; + if size(ot)=2 then count:=count+1 ; + g1:=1+(e - alpha - beta - size(ot))/2 ; + m:=group([oo[j][1][1],oo[j][1][2 ] ] ) ; + hom:=grouphomomorphismbyimages(m , u,[oo[j][1][1],oo[j][1][2]],[(1,2),(1,2 ) ] ) ; + print(``j = ' ' ) ; print ( j ) ; print(``@xmath505n'');print(``@xmath505n '' ) ; + print(``genus = ' ' ) ; print(g1);print(``@xmath505n'');print(``@xmath505n '' ) ; + print(``sigma= '' ) ; print(oo[j][1][1 ] ) ; print(``@xmath505n'');print(``@xmath505n '' ) ; + print(``tau= '' ) ; print(oo[j][1][2 ] ) ; print(``@xmath505n'');print(``@xmath505n '' ) ; + print(``tau*sigma='');print(oo[j][1][2]*oo[j][1][1 ] ) ; print(``@xmath505n'');print(``@xmath505n '' ) ; + print(``monodromy group= ' ' ) ; print(m ) ; print(``@xmath505n'');print(``@xmath505n '' ) ; + print(``order of the monodromy group = ' ' ) ; print(size(m ) ) ; print(``@xmath505n'');print(``@xmath505n '' ) ; + print(``automorphism group = ' ' ) ; + print(intersection(stabilizer(g , oo[j][1][1],ac),stabilizer(g , oo[j][1][2],ac ) ) ) ; print(``@xmath505n'');print(``@xmath505n '' ) ; + print(``the stabilizer of 1 in the monodromy group is '' ) ; print(``@xmath505n'');print(``@xmath505n '' ) ; + s:=stabilizer(m,1 ) ; print(s ) ; print(``@xmath505n'');print(``@xmath505n '' ) ; + print(hom ) ; print(``@xmath505n'');print(``@xmath505n '' ) ; + if hom = fail then print(``the dessin is non - orientable '' ) ; else ; + k:=kernel(hom ) ; ks:=issubgroup ( k , s ) ; + if ks = true then print(``the dessin is orientable '' ) ; else ; + print(``the dessin is non - orientable '' ) ; + fi ; fi ; print(``@xmath505n'');print(``@xmath505n '' ) ; print(`` '' ) ; print(``@xmath505n'');print(``@xmath505n '' ) ; + else ; fi ; od ; + print(count ) ; +

this paper provides a new construction of-coalescents called `` measure division construction '' . this construction is pathwise and consists of dividing the characteristic measure into several parts and adding them one by one to have a whole process . using this construction , a `` universal '' normalization factor for the randomly chosen external branch length has been discovered for a class of coalescents . this class of coalescents covers processes similar to bolthausen - sznitman coalescent , the coalescents without proper frequencies , and also others .

1302.1083

let @xmath3 , @xmath4 be a subset of @xmath5 and @xmath6 a partition of @xmath4 such that @xmath7 ( @xmath8 denotes the number of blocks in @xmath6 ) . the @xmath0-coalescent process starting from @xmath6 , introduced independently by pitman@xcite and sagitov@xcite , is denoted by @xmath9 , where @xmath10 and @xmath0 is a finite measure on @xmath11 $ ] . here we specify that a finite measure on @xmath11 $ ] can be a null measure and hence its total mass is a non - negative real value . if @xmath12 , i.e. , the set of first @xmath13 singletons , then the process is simply denoted by @xmath14 . in this paper , we will frequently use two other notations @xmath15 for finite measures . we define then @xmath16 as the @xmath17-coalescent and @xmath18 the @xmath19-coalescent , both taking @xmath20 as initial value . this process @xmath21 is a continuous time markov process with cdlg trajectories taking values in the set of partitions of @xmath4 . more precisely : assume that at time @xmath22 , @xmath23 has @xmath24 blocks , then after a random exponential time with parameter @xmath25 @xmath26 @xmath21 encounters a collision and the probability for a group of @xmath27 blocks to be merged into a bigger block with the other @xmath28 blocks unchanged is @xmath29 then @xmath30 is the probability to have @xmath31 blocks right after the collision . this definition gives the exchangeability of blocks . in particular , for any permutation @xmath32 on @xmath33 , @xmath34 remark that if @xmath35 , then we get the following well known formula : @xmath36 the definition shows that the law of @xmath21 is determined by the initial value @xmath6 and the measure @xmath0 which is hence called characteristic measure . notice that @xmath4 can be an abstract set and the coalescing mechanism works all the same . the reason why one takes @xmath4 as a subset of @xmath37 relies on its applications in the genealogies of populations . we take @xmath14 as an example where @xmath38 . at time @xmath39 , we have @xmath40 which is interpreted as a sample of @xmath13 individuals labelled from @xmath41 to @xmath13 . if at time @xmath22 , @xmath14 has its first coalescence where @xmath42 and @xmath43 are merged together with the others unchanged , then @xmath44 which is interpreted as getting the mrca ( most recent common ancestor ) @xmath45 of individuals @xmath41 and @xmath46 with the others unchanged at that time . hence @xmath47 is an absorption state of @xmath14 and is the mrca of all individuals . for more details , we refer to @xcite or @xcite . let @xmath48 and @xmath49 the restriction from @xmath33 to @xmath50 . we have the consistency property : @xmath49 @xmath51 @xmath52 ( see @xcite ) . according to this property and exchangeability of blocks , if @xmath53 is a subset of @xmath6 , then the restriction of @xmath21 from @xmath6 to @xmath53 has the same distribution as that of @xmath54 . we can also define @xmath21 when @xmath55 by using the consistency property and the definition in finite cases ( see @xcite ) . let @xmath56 be the block counting process associated to @xmath14 such that @xmath57 is the number of blocks of @xmath58 for any @xmath59 . then it decreases from @xmath13 at time @xmath39 . we denote by @xmath60 the decrease of number of blocks at the first coalescence . for @xmath61 , we define @xmath62 the length of the @xmath63th external branch and @xmath2 the length of a randomly chosen external branch . by exchangeability , we denote by @xmath65 the total external branch length of @xmath14 , and by @xmath66 the total branch length . there are four classes of @xmath0-coalescents having been largely studied . we give the results concerning @xmath2 , which show a common regularity that we will discuss later . * @xmath67 : kingman coalescent ( see @xcite , @xcite ) . then @xmath68 is asymptotically distributed with density function @xmath69 ( see @xcite , @xcite , @xcite ) . * @xmath70 : bolthausen - sznitman coalescent ( see @xcite ) . here @xmath71 denotes the lebesgue measure on @xmath11 $ ] . then @xmath72 converges in distribution to @xmath73 ( we denote by @xmath74 , the exponential variable with parameter @xmath75)@xcite . * @xmath76 @xmath77-coalescent . here @xmath78 denotes euler s beta function . then @xmath79 converges in distribution to a random variable @xmath80 which has density function @xmath81 ( see @xcite ) . * @xmath82 : these processes are called coalescents without proper frequencies ( @xcite ) . this category contains @xmath77-coalescents with @xmath83 ( see @xcite , @xcite ) . then @xmath84 converges in distribution to @xmath73 ( see @xcite , @xcite ) . we see a common property for the last three cases concerning one external branch length which is that the normalization factor for @xmath2 is @xmath85 . more precisely , * bolthausen - sznitman coalescent : notice that @xmath86 . hence directly we have @xmath87 . * @xmath77-coalescent with @xmath88 : @xmath89 hence @xmath90 converges in distribution to @xmath91 . * if @xmath82 , then @xmath92 . hence @xmath90 converges in distribution to @xmath73 . kingman coalescent can be viewed as the formal limit of @xmath77-coalescent with @xmath88 when @xmath93 tends to @xmath39 , since the measure @xmath94 tends weakly to the dirac measure on @xmath39 . the normalization factor in the case of @xmath77-coalescent is @xmath95 , and of kingman coalescent is @xmath13 . then we see that these two factors show also some kind of continuity as @xmath93 tends to @xmath39 . we can formally take @xmath13 as @xmath1 in the case of kingman coalescent . therefore @xmath1 is characteristic for the randomly chosen external branch length in those processes considered . notice that @xmath1 concerns only the measure @xmath96}$ ] , so it is natural to think about the influences of measures @xmath96}$ ] and @xmath97 on the external branch lengths . more generally , if @xmath98 , how can we evaluate each influence on the construction of the whole @xmath0-coalescent ? if @xmath17 is small enough , we can imagine that @xmath14 looks like @xmath99 ( recall that @xmath99 is the @xmath19-coalescent ) . in this case , we call @xmath17 the noise measure and @xmath19 the main measure . to separate @xmath17 and @xmath19 , we introduce in the next section the `` measure division construction '' of a @xmath0-coalescent . the idea of this construction can be at least tracked back to @xcite where the authors consider also a coupling of two finite measures on @xmath11 $ ] but in a slightly different manner . the main results are as follows : [ gnmu0 ] if @xmath0 satisfies : @xmath100 then @xmath101 . [ nonzero ] * condition @xmath102 implies that @xmath103 indeed , if @xmath104 , then @xmath105 and @xmath106)$ ] . then @xmath102 is invalid . * the class of coalescents satisfying condition @xmath102 does not contain the beta@xmath107-coalescents with @xmath108 and @xmath109 . the following conjecture uses a description similar to condition @xmath102 to include them : + * conjecture : * let @xmath110 . if @xmath111 then @xmath112 , where @xmath113 is a random variable with density @xmath114 . here @xmath115 is the unique solution of the equation @xmath116 + this conjecture is true for beta@xmath107-coalescents with @xmath88 . in this case , we have @xmath117 . the coalescents , which are more general than but similar to beta@xmath107-coalescents with @xmath88 , studied in @xcite also satisfy this conjecture . * examples : * we give a short list of typical examples satisfying condition @xmath102 which are processes without proper frequencies or similar to bolthausen - szitman coalescent . define @xmath118 . * ex 1 : * @xmath82 : it suffices to prove that @xmath119 recalling the expression ( [ gn1 ] ) of @xmath120 , we have , for @xmath121 @xmath122 the second term @xmath123 for the first term , let @xmath124 and @xmath125 , then @xmath126 notice that @xmath127 can be arbitrarily small and @xmath128 tends to @xmath39 as @xmath13 tends to @xmath129 . then we get that @xmath130 tends to @xmath39 . hence if @xmath82 , condition @xmath102 is satisfied . * ex 2 : * bolthausen - sznitman coalescent : in this case , it is straightforward to prove that @xmath131 and @xmath86 , then @xmath132 . * ex 3 : * @xmath0 has a density function @xmath133 on @xmath134 where @xmath135 and there exists a positive number @xmath136 such that @xmath137 on @xmath134 : this kind of processes can be considered as being _ dominated _ by the bolthausen - sznitman coalescent . if @xmath82 , we turn back to the first example . if @xmath138 , then we have @xmath139 for @xmath13 large enough , hence @xmath140 . it turns out that this kind of coalescent also satisfies condition @xmath102 . * ex 4 : * @xmath0 has a density function @xmath141 on @xmath134 where @xmath135 and @xmath142 are positive numbers : using ( [ gnn ] ) , we have @xmath143 for two real sequences @xmath144 , we write @xmath145 , if there exist two positive constants @xmath146 such that @xmath147 for @xmath13 large enough . then it is not difficult to find out that @xmath148 , @xmath149 , @xmath150 . hence we get @xmath151 . [ kd ] if @xmath0 satisfies condition @xmath102 and @xmath138 , then we have : @xmath152 where @xmath153 are independently distributed as @xmath73 . the same result has been proved for bolthausen - sznitman coalescent in @xcite . the authors have used a moment method . we can apply this theorem to example 4 and example 3 when @xmath154 if @xmath155 then ( [ mainkd ] ) is not true and there is no more asymptotic independence ( see @xcite ) . the following three corollaries have also been proved for bolthausen - sznitman coalescent ( see @xcite , @xcite , @xcite ) . [ momk ] if @xmath0 satisfies condition @xmath102 , then for any @xmath156 @xmath157=\mathbb{e}[e_1^r],\ ] ] where @xmath158 is distributed as @xmath73 . moreover , if @xmath159 then for any @xmath160 and any @xmath161 we have : @xmath162=\mathbb{e}[\prod_{i=1}^{k}e_i^{r_i}],\]]where @xmath163 are independently distributed as @xmath164 [ ext ] if @xmath0 satisfies condition @xmath102 and @xmath138 , then the total external branch length @xmath165 satisfies : @xmath166 converges in @xmath167 to @xmath168 [ total ] if @xmath0 satisfies condition @xmath102 and @xmath138 , then the total branch length @xmath66 satisfies : @xmath169 converges in probability to @xmath168 * in fact , we will prove that @xmath170=1 $ ] . notice that corollary [ momk ] gives @xmath171=1 $ ] . hence we deduce this corollary using corollary [ ext ] . * if @xmath82 , then ( [ multiple ] ) and corollaries ( [ ext ] ) and ( [ total ] ) are not true ( see again @xcite ) . in section @xmath46 , we introduce the main object of this paper : the measure division construction . at first , one needs to define the restriction by the smallest element which serves as a preliminary step of measure division construction . in the same section , we then introduce the two - type @xmath0-coalescent which is defined using the measure division construction . this process gives a label _ primary _ or _ secondary _ to every block and its every element of a normal @xmath0-coalescent . using this process , we can see more clearly the coalescent times of some singletons . for a technical use , we then give a tripling to estimate the number of blocks at small times of @xmath16 which is related to the noise measure @xmath17 . in section @xmath172 , we at first give a characterization for the condition @xmath102 . then we apply the general results obtained in section 2 to those processes satisfying @xmath102 . finally , we give all the proofs for the results presented in the section @xmath41 . let @xmath173 , @xmath174 be two partitions of @xmath33 . we define @xmath175 ( resp . @xmath176 ) as the smallest number in the block @xmath177 ( resp . @xmath178 ) . we define also the notation @xmath179 if @xmath180 and for any @xmath181 , where @xmath182 is a partition of @xmath183 . roughly speaking , @xmath184 is finer than @xmath185 . if @xmath179 we define the stochastic process @xmath186 which is the restriction by the smallest element of @xmath187 from @xmath184 to @xmath185 : * @xmath188 ; * for any @xmath189 , if @xmath190 , where @xmath191 denotes a block , then @xmath192 where the empty sets in @xmath193 are removed . notice that the restriction by the smallest element is defined from path to path ( see figure [ fig : restriction small ] ) . [ restby ] @xmath186 has the same distribution as @xmath194 . every block in @xmath185 is identified by its smallest element which belongs to a unique block in @xmath184 . hence for any @xmath178 in @xmath185 , there exists a unique @xmath195 such that @xmath196 , @xmath197 and @xmath198 with @xmath199 let @xmath200 and define a new process @xmath201 as follows : * @xmath202 * for any @xmath189 , if @xmath190 , then @xmath203 where the empty sets in @xmath204 are removed . it is easy to see that @xmath201 is a natural restriction of @xmath187 from @xmath184 to @xmath205 . by the consistency property , we get @xmath206 . in the construction of @xmath201 and @xmath186 , what is determinant is the smallest element in each block . hence to obtain @xmath186 from @xmath201 , at time @xmath39 , one needs to complete every @xmath195 by some other numbers larger than @xmath207 to get @xmath178 and then follow the evolution of @xmath201 . it turns out that @xmath186 is a coalescent process with initial value @xmath185 . hence we can conclude . let @xmath208 be three finite measures such that @xmath98 . we denote by @xmath209 the stochastic process constructed by the measure division construction using @xmath17 and @xmath19 . here the index @xmath210 is for @xmath98 with @xmath17 called noise measure and @xmath19 main measure . recall that @xmath16 is the @xmath17-coalescent with @xmath211 . * step 0 : given a realization or a path @xmath212 of @xmath213 , we set @xmath214 , for any @xmath59 . we set also @xmath215 . * step 1 : let @xmath216 be the coalescent times after @xmath217 of @xmath218 ( if there is no collision after @xmath217 , we set @xmath219 ) . within @xmath220 , @xmath218 is constant . then we run an independent @xmath19-coalescent with initial value @xmath221 from time @xmath217 . * * if the @xmath19-coalescent has no collision on @xmath220 , we pass to @xmath222 . similarly , we construct another independent @xmath19-coalescent with initial value @xmath223 from time @xmath224 , and so on . * * otherwise , we go to the next step . * step 2 : if finally within @xmath225 , the related independent @xmath19-coalescent has its first collision at time @xmath226 and its value at @xmath227 is @xmath228 . we then modify @xmath229 in the following way : * * we change nothing for @xmath230 . * * let @xmath231 be the restriction by the smallest element of @xmath232 from @xmath233 to @xmath228 . then let @xmath234 and go to the step 1 by taking @xmath227 as a new starting point . notice that , due to lemma [ restby ] , @xmath232 has the same distribution as a @xmath17-coalescent from time @xmath227 with initial value @xmath228 , . [ defmdc ] * the measure division construction works path by path . * if we take @xmath235 as noise measure and @xmath236 as main measure , then @xmath237 for any @xmath59 and @xmath238 . [ pi12pi ] let @xmath0 , @xmath17 and @xmath19 be three finite measures and @xmath98 . then we have @xmath238 . let @xmath22 be a coalescent time of @xmath218 . we consider the time of the next coalescence and the value at that moment . in the measure division construction of @xmath218 , we can see appearing two independent processes with one being a @xmath17-coalescent with initial value @xmath239 and the other one being a @xmath19-coalescent with initial value @xmath239 from time @xmath240 the process @xmath218 gets the next coalescence whenever one of them first encounters a coalescence and picks up the value of the process at that moment . then we follow the same procedure from the new coalescent time of @xmath218 . it is easy to see that @xmath218 behaves in the same way as @xmath14 . hence we can conclude . the theorem shows that if we exchange the noise measure and the main measure , the distribution of the process is not changed and is uniquely determined by their sum . the measure division construction also works for more than two measures . if there are @xmath241 finite measures @xmath242 and @xmath243 one can get a stochastic process by first giving a realization of @xmath213 which will be modified by @xmath19 in the way described in the measure division construction , and then we apply @xmath244 on the modified process , etc . the equivalence in distribution can be obtained in a recursive way . we give a corollary to show an immediate application of the measure division construction . the following corollary is essentially the same as lemma 3.2 in @xcite . but we prove it again in our way . [ measto ] let @xmath17 , @xmath19 be two finite measures such that @xmath245 , then on can construct @xmath16 and @xmath18 such that @xmath246 for all @xmath59 . @xmath99 can be regarded as the measure constructed process by imposing the measure @xmath247 on the paths of @xmath213 . then we can deduce this corollary . let @xmath248 , @xmath19 be three finite measures and @xmath98 and @xmath19 satisfies @xmath249 . a two - type @xmath0-coalescent , denoted by @xmath250 is to give a label _ primary _ or _ secondary _ to every block and also to its every element at any time @xmath22 of a normal @xmath0-coalescent . a block is _ secondary _ if and only if every element in this block is _ secondary_. the construction is via the measure division construction . let @xmath251 be independent random variables following the distribution of @xmath252 , @xmath253 i.i.d copies of @xmath254 and @xmath255 . * construction of a two - type @xmath0-coalescent : * * step 0 : we pick a realization or a path @xmath212 of @xmath213 . every element and every block of @xmath212 at any time is labeled _ primary_. we also fix independent realizations of @xmath251 and @xmath256 . let @xmath257 be the path @xmath212 with labels . * step 1 : at time @xmath258 , every block of @xmath259 is independently marked `` head '' with probability @xmath260 and `` tail '' with probability @xmath261 . every element in a `` head '' block is then labelled _ secondary_. all those blocks marked `` head '' are merged into a bigger block , provided that there are at least two `` head ' 's . in this case , we use the restriction by the smallest element to modify @xmath257 at time @xmath262 in the same way as in the measure division construction in section 2.2 . we still call the modified path @xmath257 and then forward to the time @xmath263 and do the same operations . this procedure can be continued until mrca . it is easy to verify that without labels , @xmath257 has the same distribution as @xmath14 . we call @xmath256 the _ _ marking time__s . we define @xmath264 as the _ first marking time _ of @xmath265 when @xmath265 is marked `` head '' for the first time . let @xmath266 , if @xmath265 is never marked as `` head '' . if @xmath267 , then we get a coupling between @xmath19-coalescent and its related annihilator process ( see @xcite ) . more precisely , the whole process without labels is the @xmath19-coalescent and the restriction to _ primary _ elements and blocks is the annihilator process . the above construction of two - type coalescents shows that coalescences happen only at the _ marking times_. this property will help us to understand the coalescent times of singletons in terms of their _ first marking times_. [ rela2 ] let @xmath212 be the path of @xmath16 chosen at the step 0 of the construction of two - type @xmath0 coalescent . assume that at some time @xmath268 , @xmath269 , @xmath270 with @xmath271 . let @xmath272 be the probability for @xmath42 to be coalesced at its _ first marking time _ within @xmath273 . then we have @xmath274,\ ] ] where @xmath275 ; @xmath276 for @xmath277 . notice that the parameter @xmath13 is hidden in @xmath278 . let @xmath279 be the @xmath280 smallest elements respectively in each block at time @xmath22 with @xmath281 . conditional on @xmath282 , @xmath283 is the probability for @xmath42 to have its _ first marking time _ at @xmath284 ( assume that @xmath285 ) . to let @xmath42 be coalesced at @xmath284 , one needs also at least one other block marked `` head '' at that time . to get a lower bound of @xmath286 , one can consider the propability to have at least one _ primary _ block containing one element of @xmath287 to be marked `` head '' at that time and this probability is @xmath288 [ rela1 ] in addition to the assumptions in the previous lemma , we assume further that all @xmath289 for @xmath290 and @xmath291 . define the probability @xmath292 for every @xmath265 to be coalesced at its _ first marking time _ within @xmath293 . then we have @xmath294 let @xmath295 , which denotes the assumptions of @xmath296 in this lemma . then @xmath297 the last inequality is due to the fact that @xmath298 which is true due to the same arguments used in the proof of the last lemma . if @xmath299 are large enough such that under some assumptions , we could prove that @xmath278 is very close to @xmath41 . then the coalescent times are almost the _ first marking times _ which are easier to deal with . in section 3.3 , we will see such a situation for @xmath0 satisfying condition @xmath102 and @xmath300}.$ ] the following corollary studies the _ first marking times _ in this particular case . [ lll ] let @xmath268 and @xmath301 . assume that @xmath0 satisfies condition @xmath102 and @xmath302 , @xmath303}$ ] . let @xmath212 be a path of @xmath213 . recall that @xmath264 is the _ first marking time _ of @xmath265 for @xmath304 * if @xmath305 , then for any @xmath306 , @xmath307 * assume moreover @xmath138 and @xmath308 for any @xmath290 with @xmath301 and fixed . let @xmath309 , we then have @xmath310 the first case is easy to see , due to the definition of @xmath311 . for the second case , we only consider @xmath312 for @xmath313 , the proof is similar . assume that within @xmath314 $ ] , there are @xmath315 _ _ marking time__s and for @xmath316 $ ] , there are @xmath317 _ _ marking time__s . @xmath315 and @xmath317 are independently poisson distributed with parameters respectively @xmath318 and @xmath319 ( here we have @xmath320 ) . then we get @xmath321\\ & = \mathbb{e}[(1 - 2\mathbb{e}[\eta^{(2)}_1]+\mathbb{e}[(\eta^{(2)}_1)^2])^{n_1}]\mathbb{e}[(1-\mathbb{e}[\eta^{(2)}_1])^{n_2}]\\ & = exp\left(\frac{\bar{\mu}^{(n)}t_1}{\mu^{(n)}}(-2\mathbb{e}[\eta^{(2)}_1]+\mathbb{e}[(\eta^{(2)}_1)^2])\right)exp\left(\frac{\bar{\mu}^{(n)}(t_2-t_1)}{\mu^{(n)}}(-\mathbb{e}[\eta^{(2)}_1])\right),\end{aligned}\ ] ] where the last equality is due to the probability generating function of poisson distribution . recall that @xmath322=\frac{\mu^{(n)}}{\bar{\mu}^{(n)}}$ ] and @xmath323=\frac{\int_{1/n}^1\lambda(dx)}{\bar{\mu}^{(n)}}$ ] . therefore , @xmath324=\frac{\int_{1/n}^{1}x^{-1}\lambda(dx)}{\mu^{(n)}}=1 , \text{and}\quad \frac{\bar{\mu}^{(n)}}{\mu^{(n)}}\mathbb{e}[(\eta^{(2)}_1)^2]=\frac{\int_{1/n}^{1}\lambda(dx)}{\mu^{(n)}}\stackrel{}{\rightarrow 0}.\ ] ] then we can conclude ( [ mis2 ] ) . we often have some results on the coalescent related to a special measure , for example , the @xmath325-coalescent . when the process is perturbed by a noise measure , we would wonder whether this damage is negligible . one example is to estimate the number of blocks of the coalescent related to the noise measure . to this aim , we use the tool of tripling . * tripling : * notice that @xmath14 encounters its first collision after time @xmath326 , which is a random variable . at this collision , the number of blocks is reduced to @xmath327 , where @xmath328 is a positive integer valued random variable . then we add @xmath328 new blocks ( these blocks can contain any number belonging to @xmath329 ) and consider the whole new @xmath13 ones . by the consistency property , the evolution of the original @xmath327 blocks can be embedded into that of the new @xmath13 blocks , i.e. after time @xmath330 , we have the collision in the new @xmath13 blocks whose total number is reduced to @xmath331 and we can calculate the distribution of the number of blocks coalesced among the original @xmath327 blocks ( we call any block containing at least one of @xmath33 as `` original block '' and it is very possible that nothing happens for the @xmath327 blocks ) . then we add again new blocks containing different elements to have another @xmath13 ones . this procedure is stopped when every element of @xmath33 is contained in one block . by the definition of @xmath0-coalescent , @xmath332 are independent exponential random variables with parameter @xmath120 and @xmath333 are i.i.d copies of @xmath60 . the above procedure gives a tripling of @xmath332 , @xmath333 and @xmath14 . we define @xmath334 then we have the following proposition : [ tripleprop ] suppose that @xmath332 , @xmath333 and @xmath14 are tripled , then at any time @xmath59 , we have @xmath335 where @xmath336 , which is poisson distributed with parameter @xmath337 and independent of @xmath333 . meanwhile , @xmath338=\frac{n\int_0 ^ 1(1-(1-x)^{n-1})x^{-1}\lambda(dx)}{g_n}-1 , \text{and } \quad\mathbb{e}[(w_i^{(n)})^2]=\frac{n(n-1)\int_0 ^ 1\lambda(dx)}{g_n}-\mathbb{e}[w_i^{(n)}].\ ] ] the number of @xmath63s within @xmath339 $ ] follows the poisson distribution with parameter @xmath337 . due to the tripling , at any time @xmath340 with @xmath341 , the decrease of number of blocks ( i.e. @xmath342 ) of original blocks is less than or equal to @xmath343 . hence we get ( [ triple ] ) . notice that @xmath344 , then ( [ w ] ) is a consequence of two equalities in @xcite with eq ( 17 ) for the first one and p.1007 for the second one . some notations for this section : let @xmath0 be a finite measure on @xmath11 $ ] and @xmath302 , @xmath303}$ ] ; @xmath345 , @xmath346 with @xmath347 notice that the definitions of @xmath348 and @xmath349 are consistent with that of @xmath1 and @xmath120 when @xmath35 . these notations help to examine carefully different measures . here we are going to prove theorem [ gnmu0 ] , theorem [ kd ] , corollary [ momk ] , corollary [ ext ] and corollary [ total ] . under condition @xmath102 , we decompose @xmath0 into @xmath19 and @xmath17 . the idea is to construct @xmath14 using measure division construction with noise measure @xmath17 and main measure @xmath19 . at first , we need to show more details implied by condition @xmath102 . for any real number @xmath350 , let @xmath351 and @xmath352 where @xmath361 , @xmath362 . notice that for @xmath13 large , using monotone property , we have @xmath363 and @xmath364 hence condition @xmath102 is equivalent to @xmath365 then we deduce that @xmath366 indeed , for @xmath367 and @xmath368 , we have @xmath369 one thing to notice is that @xmath370 is true for any finite @xmath371 in fact , for any positive number @xmath136 and @xmath372 , we have @xmath373 where both terms can be made as small as we want by taking @xmath136 large enough and @xmath374 close enough to @xmath39 . looking into details of @xmath375 when @xmath376 , we have the following equality , using integration by parts and @xmath370 , @xmath377 where @xmath378 . due to ( [ secon ] ) , we get that @xmath379 . therefore , ( [ secon ] ) and ( [ parts ] ) give @xmath380 notice that @xmath381 and @xmath348 is a cgld function . hence there exists a cgld function @xmath355\rightarrow [ 0,1]$ ] , continuous at @xmath39 with @xmath356 such that @xmath382 @xmath384 using the fundamental theorem of newton and leibniz which also works for cgld functions whose primitive functions take left derivatives . then for @xmath385 , @xmath386 therefore , @xmath387 by taking the left derivatives on the both sides and noticing that @xmath388 , we can conclude . * part 2 * : we now assume that @xmath354 is true . in the first part , we proved implicitly that ( [ secon ] ) is equivalent to the @xmath354 . hence we will use ( [ secon ] ) to prove ( [ 2con ] ) which is equivalent to condition @xmath102 and only the first convergence in ( [ 2con ] ) is needed to be proved . let @xmath136 be a positive number and @xmath389 , @xmath359 , then @xmath390 the first term can be made as small as we want by taking @xmath136 large , and the third term @xmath391 let @xmath124 and @xmath13 large enough such that @xmath392 on @xmath393 $ ] . then @xmath394 which can be made as close as possible to @xmath41 with @xmath395 small enough . hence we can conclude . let @xmath402 . we write @xmath403 where @xmath404 and @xmath405 it is easy to see that for @xmath406 , @xmath407 for the second term , @xmath408 notice that for @xmath13 large , there exists a positive constant @xmath409 such that @xmath410 hence @xmath411 it suffices to take @xmath412 to conclude . if @xmath417 with some @xmath418 , then for any @xmath419 , @xmath17 is the null measure and hence @xmath420 for any @xmath421 which proves this lemma . in consequence , one needs only to consider the case where @xmath422 for any @xmath423 we recall @xmath424 defined in lemma [ lemgnlambda ] . let @xmath425 be the decrease of the number of blocks at the first coalescence of @xmath16 . thanks to proposition [ tripleprop ] where we pick up the notations , @xmath426 where @xmath427 is poisson distributed with parameter @xmath428 independent of @xmath333 which are i.i.d copies of @xmath429 . then we have , for @xmath13 large , @xmath430\geq n\epsilon-\frac{g_n^{(1)}m}{\mu^{(n)}}\mathbb{e}[w_1^{(n)}]\right)\nonumber\\ & \leq \frac{{{\rm var}}(\sum_{i=1}^{n{(\lambda_{1 } , n , m/\mu^{(n)})}}w_i^{(n)})}{(n\epsilon-\frac{g_n^{(1)}m}{\mu^{(n)}}\mathbb{e}[w_1^{(n)}])^2}=\frac{\frac{mg_n^{(1)}}{\mu^{(n)}}\mathbb{e}[(w_1^{(n)})^2]}{(n\epsilon-\frac{g_n^{(1)}m}{\mu^{(n)}}\mathbb{e}[w_1^{(n)}])^2},\end{aligned}\ ] ] where the second inequality needs @xmath431>0 $ ] which is justified by the following calculations : notice that due to proposition [ tripleprop ] and lemma [ lemgnlambda ] , for @xmath13 large enough , @xmath432 + 1\leq \frac{n(n-1)\int_0^{1/n}\lambda_{1}(dx)}{g_n^{(1)}}\leq \frac{1}{c_1 } ; \mathbb{e}[(w_1^{(n)})^2]\leq \frac{n(n-1)\int_0^{1/n}\lambda_{1}(dx)}{g_n^{(1)}}\leq\frac{1}{c_1},\ ] ] where @xmath400 is the positive constant in lemma [ lemgnlambda ] . these terms are probabilities defined in section 2.3.1 , which measure the possibility to make one or several singletons coalesced in their _ first marking times _ within @xmath293 . in fact , we will study @xmath437 , since we want to prove that the normalization factor of the external branch length is @xmath1 . we denote by `` @xmath438 '' the stochastic domination between two real random variables . the following corollary together with the remark at the end play an important role in getting the asymptotics of the three probabilities . recall @xmath251 , @xmath253 , @xmath441 which are associated to @xmath19 and defined in section 2.3 . at first , we remark that @xmath442=1.$ ] one only needs to prove that @xmath443=0 $ ] . it is easy to see that @xmath444=\mathbb{e}[\bar{\delta}^{(2)}_i(1-\bar{\delta}^{(2)}_i)^{n-1}]$ ] , where @xmath445 . it is obvious that @xmath446 is a markov chain . for @xmath447 , we define a stopping time @xmath448 then we get @xmath449&=\mathbb{e}[\sum_{i=1}^{+\infty}\bar{\delta}^{(2)}_i(1-\bar{\delta}^{(2)}_i)^{n-1}]\nonumber\\ & = \mathbb{e}[\sum_{i=1}^{\tau_n-1}\bar{\delta}^{(2)}_i(1-\bar{\delta}^{(2)}_i)^{n-1}+\sum_{i=\tau_n}^{+\infty}\bar{\delta}^{(2)}_i(1-\bar{\delta}^{(2)}_i)^{n-1}].\\\nonumber\end{aligned}\ ] ] notice that @xmath450 , if @xmath451 and @xmath452 , if @xmath453 . then ( [ 2delta ] ) gives @xmath454\leq \mathbb{e}[\frac{\tau_n-1}{n}+\sum_{i=\tau_n}^{+\infty}\bar{\delta}^{(2)}_i]\leq \mathbb{e}[\frac{\tau_n-1}{n}]+\frac{1}{\mathbb{e}[n\eta^{(2)}_1]}.\ ] ] to calculate @xmath455 $ ] , we use renewal theory . let @xmath456 $ ] . depending on whether @xmath457 is finite or not , we separate the discussion into two parts . * part 1 * : assume that @xmath458 we denote by @xmath459 the distribution function and @xmath460 the density function of @xmath461 and @xmath462 an independent random variable with density function @xmath463 let @xmath124 , then using integration by parts , @xmath464 now together with ( [ 0x1 ] ) , one gets @xmath471 we fix @xmath472 and define a new markov chain @xmath473 and a stopping time @xmath474 for @xmath447 . it is clear from the definitions of @xmath475 and @xmath476 that @xmath477=\mathbb{e}[\tau_{se^{{-\epsilon}}}-1].\ ] ] then @xmath478&=\mathbb{e}[\tau'_{n\eta^{(2)}_1}{{\bf 1}}_{0\leq x\leq \epsilon}]+\mathbb{e}[\tau'_{n\eta^{(2)}_1}{{\bf 1}}_{x > \epsilon}]\\ & \geq \mathbb{p}(0\leq x\leq \epsilon)\mathbb{e}[\tau_{nexp(-\epsilon)}-1]+\mathbb{e}[\tau'_{n\eta^{(2)}_1}{{\bf 1}}_{x > \epsilon}],\\\end{aligned}\ ] ] notice that @xmath481 , hence @xmath482 therefore , ( [ tautau ] ) gives @xmath483\leq \frac{\mathbb{e}[\tau'_{ne^{\epsilon}\eta^{(2)}_1}]}{\mathbb{p}(0\leq x\leq \epsilon)}+1=\frac{\mathbb{e}[\ln(ne^{\epsilon}\eta^{(2)})]}{\mu\mathbb{p}(0\leq x\leq \epsilon)}+1.\ ] ] it is easy to see that , using ( [ gn1 ] ) , there exists a positive constant @xmath489 such that @xmath490=\frac{n\int_{1/n}^{1}x^{-1}\lambda(dx)}{\bar{\mu}^{(n)}}\geq c_2\frac{n\mu^{(n)}}{g_n}$ ] , for any @xmath491 hence @xmath490 $ ] tends to @xmath129 since @xmath0 satisfies @xmath102 . for the second convergence , we fix @xmath492 . then , @xmath493}{\mathbb{e}[n\eta^{(2)}_1]}&=\frac{\mathbb{e}[\ln(n\eta^{(2)}_1){{\bf 1}}_{n\eta^{(2)}_1\geq m}]+\mathbb{e}[\ln(n\eta^{(2)}_1){{\bf 1}}_{n\eta_1^{(2)}<m}]}{\mathbb{e}[n\eta^{(2)}_1]}\\ & \leq \frac{\mathbb{e}[\ln(n\eta^{(2)}_1){{\bf 1}}_{n\eta^{(2)}_1\geq m}]}{\mathbb{e}[n\eta^{(2)}_1]}+\frac{\ln m}{\mathbb{e}[n\eta^{(2)}_1]}\\ & \leq \frac{\mathbb{e}[\ln(n\eta^{(2)}_1){{\bf 1}}_{n\eta^{(2)}_1\geq m}]}{\mathbb{e}[n\eta^{(2)}_1{{\bf 1}}_{n\eta^{(2)}_1\geq m}]}+\frac{\ln m}{\mathbb{e}[n\eta^{(2)}_1]}\\ & \leq \frac{\ln m}{m}+\frac{\ln m}{\mathbb{e}[n\eta^{(2)}_1]}.\\ \end{aligned}\ ] ] the last inequality is due to the fact that for any @xmath494 , we have @xmath495 . since @xmath136 can be chosen as large as we want , then @xmath496}{\mathbb{e}[n\eta^{(2)}_1]}=0.$ ] hence we can conclude . * part 2 * : if @xmath497 we define @xmath498 and for @xmath499 @xmath500 . notice that @xmath501<+\infty$ ] , then we return to the first case and get ( [ taunmain1 ] ) by replacing @xmath502 by @xmath503 and keeping the same @xmath260 but with different @xmath462 ( depending on @xmath504 ) . in this setting , @xmath505 we see that the closer @xmath506 is to @xmath41 , larger the @xmath507 and hence @xmath508 then we can conclude . for @xmath509 , we also have @xmath510=0.\ ] ] the proof is all the same . the only thing different is that in place of ( [ 1taun ] ) , we have @xmath511\leq c\mathbb{e}[\frac{\tau_n-1}{n}+\sum_{i=\tau_n}^{+\infty}\bar{\delta}^{(2)}_i],$ ] with @xmath512 larger than @xmath41 and depends on @xmath513 due to proposition [ plambda ] , we have @xmath520\leq \lim_{n\rightarrow + \infty}\mathbb{e}[\sum_{i=1}^{+\infty}\delta_i^{(2)}(1-\delta_i^{(2)})^{n-1}]=0.\ ] ] then it suffices to prove that @xmath521=1.\ ] ] let @xmath522 , which is the sum of @xmath523 i.i.d unit exponential variables . let @xmath524 . then @xmath525 for any fixed @xmath526 where the last equality is a large deviation result ( for example , see theorem 1.4 of @xcite ) . ) . notice that @xmath528 is independent of @xmath441 , then @xmath529&=\mathbb{e}[1-(1 - 1/i_n)^{\tau_n(t)+1}]\\ & = \mathbb{e}[1-(1 - 1/i_n)^{\tau_n(t)+1}{{\bf 1}}_{ti_n\beta \leq \tau_n(t)\leq ti_n/\beta}]+o((ti_n)^{-1})\\ & \geq \mathbb{e}[1-(1 - 1/i_n)^{ti_n\beta}{{\bf 1}}_{ti_n\beta \leq \tau_n(t)\leq ti_n/\beta}]+o((ti_n)^{-1}).\\\end{aligned}\ ] ] notice that @xmath530 and the term at the right of the above inequality satisfies @xmath531+o((ti_n)^{-1})=1\ ] ] then we can conclude ( [ taun1 ] ) . fix @xmath268 and @xmath509 . considering the measure division construction for two - type @xmath0-coalescents , let @xmath212 be the path of @xmath213 chosen at the step @xmath39 and define the event @xmath536 recall that @xmath537 implies that there are at least @xmath538 singletons at time @xmath539 . for @xmath13 large enough , using the exchangeability property , we have @xmath540 , where @xmath541 and @xmath542 due to the inequality ( [ lambda1 < ] ) . for @xmath395 small enough and @xmath13 large enough , we have @xmath543 as close as we want to @xmath41 . we define another event @xmath544 corollary [ lll ] tells that @xmath549 and it has been proved that @xmath550 can be made as close as possible to @xmath41 by taking @xmath395 small enough and @xmath22 large enough and @xmath13 tending to @xmath129 . hence the first term of ( [ thm1 ] ) can be made as close as we want to @xmath551 and the second term is close to @xmath39 . then we can conclude . we prove instead for @xmath160 : @xmath552 which is equivalent to ( [ mainkd ] ) ( see billingsley [ @xcite , p.19 ] ) . we will give the proof for @xmath553 and leave the easy extension to readers . the proof is similar to that of theorem [ gnmu0 ] . let @xmath212 be the path of @xmath213 chosen at step 0 . let @xmath554 and define the event @xmath555 using the same arguments , we get @xmath556 . we then define the event @xmath557 then due to ( [ 1k ] ) and @xmath545 is increasing on @xmath280 , we get @xmath558 which is as close as possible to @xmath41 for @xmath22 large and @xmath13 tending to @xmath559 as shown that @xmath562 can be made as close as possible to @xmath41 by taking @xmath22 large enough and @xmath395 small enough , tending @xmath13 to @xmath129 . then the second term in ( [ thm ] ) is close to @xmath563 using corollary [ lll ] , the first term is as close as possible to @xmath564 by tending @xmath13 to @xmath129 with @xmath22 large enough . then we can conclude . we prove at first the case of one external branch length . we seek to prove the uniform integrability of @xmath565 for any @xmath566 . one needs only to show that for any @xmath160 , @xmath567| n\geq 2\}<+\infty$ ] ( see lemma 4.11 of @xcite and problem 14 in section 8.3 of @xcite ) . let @xmath568 , @xmath569 and @xmath570 . to avoid invalid calculations , we set @xmath571 if @xmath572 . using the markov property , we have @xmath573 where @xmath574 and conditional on @xmath575 @xmath576 is independent of @xmath577 . then for @xmath578 @xmath579&\leq \mathbb{e}[(m+\mu^{(n)}\bar{t}_1^{(\beta_n)}{{\bf 1}}_{\mu^{(n)}t_1^{(n ) } > m})^{k}]\leq ( 2m)^k+ \mathbb{e}[(2\mu^{(n)}\bar{t}_1^{(\beta_n)}{{\bf 1}}_{\mu^{(n)}t_1^{(n ) } > m})^{k}]\nonumber\\ & \leq ( 2m)^{k}+(\mathbb{e}[2\mu^{(n)}\bar{t}_1^{(n)}{{\bf 1}}_{\beta_n = n}])^k]+\mathbb{e}[(2\mu^{(n)}\bar{t}_1^{(\beta_n)}{{\bf 1}}_{\mu^{(n)}t_1^{(n ) } > m , n\epsilon\leq \beta_n\leq n-1})^{k}]\nonumber\\ & \quad+\mathbb{e}[(2\mu^{(n)}\bar{t}_1^{(\beta_n)}{{\bf 1}}_{\mu^{(n)}t_1^{(n ) } > m , \beta_n < n\epsilon})^{k}]\nonumber\\ & \leq ( 2m)^{k}+exp(-\frac{mg_n}{\mu^{(n)}})\mathbb{e}[(2\mu^{(n)}\bar{t}_1^{(n)})^k]\nonumber\\ & \quad+\mathbb{p}(\mu^{(n)}t_1^{(n ) } > m)(2\frac{\mu^{(n)}}{\mu^{(n\epsilon)}})^{k}\max\{\mathbb{e}[(\mu^{(j)}\bar{t}_1^{(j)})^{k}]|j\in[n\epsilon , n-1]\}\nonumber\\ & \quad+\mathbb{p}(\beta_n < n\epsilon)\mathbb{e}[\frac{\beta_n}{n}(2\frac{\mu^{(n)}}{\mu^{(\beta_n)}})^k(\mu^{(\beta_n)}\bar{t}_1^{(\beta_n)})^{k}|\beta_n < n\epsilon],\end{aligned}\ ] ] where @xmath580 in the second term at right of the last inequality is the probability for no coalescence within @xmath581 $ ] . the third term is due to the fact that @xmath1 is an increasing function of @xmath13 when @xmath582 . the fourth term is due to exchangeability which says that the probability for @xmath42 not to have coalesced at @xmath583 when there exist only @xmath584 blocks is less than @xmath585 one needs the following three estimates to prove the boundedness of @xmath586)_{n\geq 2}$ ] . * estimation of @xmath587 notice that for @xmath582 , @xmath588 and if @xmath589 , we have @xmath590 hence if @xmath136 is large enough , we have , for any @xmath121 @xmath591 * estimation of @xmath592 due to corollary [ maincor ] , we get @xmath593 and theorem [ gnmu0 ] gives @xmath594 hence by taking @xmath136 large enough , we have for any @xmath121 + @xmath595 * estimation of @xmath596 using the notations in proposition [ muequi ] , for @xmath597 we have + @xmath598 + let @xmath599 such that for any @xmath600 , we have @xmath601 . hence for any @xmath602 , @xmath603 . this @xmath604 can be found since @xmath605 tends to @xmath39 as @xmath13 tends to @xmath559 then ( [ betan ] ) implies , for @xmath606 + @xmath607 + hence if @xmath608 and @xmath609 , @xmath610 if @xmath611 due to corollary [ maincor ] , one could find a large number @xmath612 such that @xmath613 and for any @xmath614 @xmath615 + in total , if @xmath614 and @xmath616 then @xmath617 @xmath618&\leq \frac{4}{3}(2m)^{k}+\frac{1}{3}\max\{\mathbb{e}[(\mu^{(j)}\bar{t}_1^{(j)})^{k}]|j\in[n\epsilon , n-1]\}+\frac{1}{3}\mathbb{e}[(\mu^{(\beta_n)}\bar{t}_1^{(\beta_n)})^{k}|\beta_n < n\epsilon]\nonumber\\ & \leq \frac{4}{3}(2m)^{k}+\frac{2}{3}\max\{\mathbb{e}[(\mu^{(j)}\bar{t}_1^{(j)})^{k}]|j\leq n-1\}.\end{aligned}\ ] ] notice that @xmath622 are exchangeable . hence corollary [ momk ] shows that @xmath623=\lim_{n\rightarrow + \infty}\mathbb{e}[\mu^{(n)}(t_1^{(n)}+t_2^{(n)}+\cdots+t_n^{(n)})/n]=\lim_{n\rightarrow + \infty}\mathbb{e}[\mu^{(n)}t_1^{(n)}]=1,\ ] ] @xmath624+n(n-1)\mathbb{e}[(\mu^{(n)})^2t_1^{(n)}t_2^{(n)}]-n^2(\mathbb{e}[\mu^{(n)}t_1^{(n)}])^2}{n^2}\\ & = \displaystyle \lim_{n\rightarrow + \infty}\frac{{{\rm var}}(\mu^{(n)}t_1^{(n)})+(n-1)cov(\mu^{(n)}t_1^{(n ) } , \mu^{(n)}t_2^{(n)})}{n}=0.\end{aligned}\ ] ] before proving corollary [ total ] , we study at first a problem of sensibility of a recurrence satisfied by @xmath625 . more precisely , if @xmath626 $ ] , then @xmath627 satisfies a recurrence ( see @xcite ) : @xmath628 , and for @xmath406 , we have where @xmath630 and @xmath631 is defined in ( [ pbk ] ) . due to corollary [ momk ] , we have @xmath632 the question is as follows : what is the limit behavior of @xmath627 if we set initially the values of @xmath633 with @xmath634 without using ( [ rec1 ] ) and replace @xmath635 by @xmath636 ? it is answered in the next lemma . let us at first look at ( [ rec1 ] ) which has the following interpretation using random walk : a walker stands initially at point @xmath13 , then after time @xmath635 , he jumps to point @xmath646 with probability @xmath647 , then after time @xmath648 , he jumps to @xmath649 with probability @xmath650 , and then after time @xmath651 , he jumps to the next point , etc . if he falls at point @xmath41 , then this walk is finished . it is easy to see that @xmath627 is the expectation of the total walking time . one notices that there is a scaling effect on the walking time . more precisely , let @xmath652 and @xmath653 such that the walker jumps from @xmath654 to @xmath655 for @xmath656 . then conditional on this walking history , the remaining walking time is @xmath657 the recurrence ( [ rec2 ] ) has the same interpretation . the difference is that one should stop the walker when he arrives at a point @xmath63 within @xmath658 $ ] and one adds a scaled value of @xmath659 to the walking time ( notice that @xmath659 can be non - positive ) . to estimate @xmath660 , we use a markov chain @xmath661 to couple the jumping structures of ( [ rec1 ] ) and ( [ rec2 ] ) : @xmath662 , notice that the jumping dynamics of both recurrences is characterized by @xmath661 until arriving at a point within @xmath671 $ ] . and we also see that @xmath661 is the discrete time markov chain related to the block counting process @xmath56 stopped at the first time arriving within @xmath671 $ ] . due to the definitions of @xmath680 , we obtain @xmath681\leq \mathbb{e}[t_{\varsigma_n}]\leq a_n ; \quad a'_n - m\mathbb{e}[c_{\varsigma_n}]\leq \mathbb{e}[t'_{\varsigma_n}]\leq a'_n+m\mathbb{e}[c_{\varsigma_n } ] ; \quad \mathbb{e}[t'_{\varsigma_n}]\leq ( 1+\epsilon)\mathbb{e}[t_{\varsigma_n}].\ ] ] notice that @xmath682\leq \frac{n_{\epsilon}}{n}$ ] and due to corollary [ maincor ] , we have @xmath683 hence @xmath684\mu^{(n)}=0.$ ] then we can conclude that for @xmath13 large , @xmath685 in the same way , we can prove also @xmath686 for another small positive number @xmath687 with @xmath13 large enough . hence we deduce the lemma . using corollary [ maincor ] , we have @xmath697 then for @xmath13 large enough , @xmath698+\frac{\mu^{(n)}}{\mu^{(n - ma)}n}\\ & \leq \frac{\mu^{(n)}}{mg_n}\max\{\frac{1}{\mu^{(k)}}| 1\leq k\leq n\}+\frac{\mu^{(n)}}{\mu^{(n - ma)}n},\end{aligned}\ ] ] where the first term at the right of the the last inequality is due to ( [ xxx ] ) and can be made as small as we want w.r.t @xmath699 when @xmath136 is large enough . notice that @xmath700 due to @xmath102 . then the second term @xmath701 using also @xmath702 then @xmath694 . where @xmath512 is a positive number . if @xmath706 , this is exactly a transformation of the recurrence ( [ rec1 ] ) let @xmath707 . then it is easy to see that @xmath708 . let @xmath634 , such that for @xmath582 , we have @xmath709 then for @xmath710 the author wants to thank his supervisor , prof jean - stphane dhersin , for fruitful discussions and also for his careful reading of a draft version resulting in many helpful comments . the author also wants to thank prof martin mhle for a discussion on theorem [ gnmu0 ] .

we investigate the stability of the extra dimensions in a warped , codimension two braneworld that is based upon an einstein - maxwell - dilaton theory with a non - vanishing scalar field potential . the braneworld solution has two 3-branes , which are located at the positions of the conical singularities . for this type of brane solution the relative positions of the branes ( the shape modulus ) are determined via the tension - deficit relations , if the brane tensions are fixed . however , the volume of the extra dimensions ( the volume modulus ) is not fixed in the context of the classical theory , implying we should take quantum corrections into account . hence , we discuss the one - loop effective potential of the volume modulus for a massless , minimally coupled scalar field . given the scale invariance of the background solution , the form of the modulus effective potential can _ only _ be determined from the sign of the logarithmic term in the effective potential that depends on the renormalization scale . this term can be evaluated via heat kernel analysis and we show that in most cases the volume modulus is stabilized . in the actual evaluation , due to a lack of knowledge of the uv contributions from the conical branes in a six - dimensional spacetime , we consider its four - dimensional counterpart . we then go on to discuss the mass scale of the modulus itself and find that it becomes comparable to the gravitational scale for mild degrees of warping , when the renormalization scale is set nominally to the gravitational scale . then , we make some suggestions on the original six - dimensional model . finally , we close this article , after discussing some phenomenological implications relating to the hierarchy problem of the fundamental energy scales and the smallness of the effective vacuum energy on the brane . we find that one - loop corrections in this model appear to alleviate some of these problems .

hep-th0606238

string theory suggests that our universe is not actually four - dimensional , but in fact a submanifold ( brane ) embedded into a higher - dimensional spacetime ( bulk ) . braneworld gravity and cosmology , especially based on the proposal by randall and sundrum @xcite , has developed a lot in the literature in the codimension one context @xcite . string theory ; however , suggests that there are as many as six or seven extra dimensions and thus , one may consider braneworld models with higher codimensions . recently , codimension two braneworld has been investigated eagerly , because it may give a more _ natural _ resolution of the cosmological constant problem than the codimension one case does , namely the vacuum energy of the brane may affect only the extra dimensions , not the geometry on the brane , see e.g. , @xcite . the basic nature of gravity and cosmology in codimension two braneworld has been reviewed , e.g. , in ref . @xcite ( see also references therein ) . the stability of the extra dimensions is one of the most significant issues in braneworld models with two branes . the interbrane distance between the branes appears as a scalar degree of freedom in the four - dimensional effective theory and affects the geometry and cosmology on the brane . for instance , in rs - type codimension one brane models , this modulus , called the radion , behaves as a scalar degree of freedom in an effective brans - dicke gravity @xcite . several stabilization mechanisms have been discussed in the rs model , both by classical dynamics @xcite and quantum corrections to the bulk vacuum state @xcite . stability of the extra dimensions is also a significant issue in higher codimensions . in this paper , we focus on the issue of quantum mechanical stabilization of the modulus in codimension two braneworld . various codimension two braneworld solutions with two branes have been found , where the compact extra dimensions are supported by the dynamics of flux fields @xcite . solutions with warped compact extra dimensions have also been found @xcite , especially in the context of six - dimensional supergravity theory @xcite . in this paper , we focus on the solution discussed by aghababaie et al . in @xcite . similar warped solutions have been recently discussed in @xcite , in the context of six - dimensional einstein - maxwell theory . in these two - brane models , the magnetic flux plays an essential role in order to obtain regular warped solutions . by regular " we mean that singularities which are stronger than conical ones are not permitted . the branes are situated at the conical defects , whose tensions are determined by the conical deficit angles . in warped brane models the conical defects correspond to the horizons of the bulk geometry and thus , the deficit angles depend on the parameters of the bulk geometry , e.g. , the mass and charge . by the tension - deficit relations , if the brane tensions are fixed , then part of bulk geometry is also fixed . in the models discussed in @xcite , the bulk geometry is also completely determined . however , in the model we shall consider here @xcite , which is based on six - dimensional supergravity , only the warp factor is fixed by the classical analysis . thus , to fix the size of the extra dimensions , quantum corrections of the bulk field should be taken into account . in this paper , we consider such a supersymmetric warped flux compactification model . we shall consider the perturbations of a massless scalar field on this brane background and calculate the effective potential of the volume modulus , focusing on whether or not one - loop quantum corrections can stabilize the absolute size of the extra dimensions . due to the scale invariance of the background field theory , the form of the modulus effective potential is almost completely fixed and therefore we only need to evaluate the sign of the coefficient of the logarithmic term in the effective potential to investigate the possibility of volume stabilization . this is directly related to the appropriate heat kernel coefficient , which is composed of contributions from both the bulk and the conical branes . the paper is organized as follows : in sec . ii , we present a warped flux compactification model that is based on an einstein - maxwell - dilaton theory with a non - vanishing scalar field potential . in sec . iii , we consider the perturbation of a massless , minimally coupled bulk scalar field in order to derive the modulus effective potential and then investigate whether or not the effective potential has a minimum . as a result , the stability of the modulus is determined by the appropriate heat kernel coefficient . then , we present results for our analysis of the heat kernel coefficients and we also discuss the _ rigidity _ of the stabilization , i.e. , is the modulus mass stable against kk perturbations . in the actual investigation of stability , rather than the original six - dimensional model , we consider a four - dimensional counterpart , because of the difficulties associated with evaluating the uv contributions from the conical branes . then , we make some suggestions for the original six - dimensional model . in sec v , we discuss phenomenological implications relating to the hierarchy of the fundamental energy scales and the effective vacuum energy realized on the brane . in sec . vi , we summarize this paper and discuss possible extensions of our present analyses . in appendix a , we introduce the four - dimensional version of the warped flux compactification solution and consider the nature of the massless scalar field perturbations on such a background , including quantum effects . in appendix b , we show the conformal invariance of the @xmath0 heat kernel coefficient . in appendix c we use the conformal invariance of the @xmath1 heat kernel coefficient on the cone to find the smeared coefficient @xmath2 , which is required for the cocycle function . in appendix d we apply the wkb method to give an estimate for the zeta function in four dimensions , which contributes to the mass scale of the modulus . we consider a six - dimensional einstein - maxwell - dilaton theory with a non - vanishing scalar potential @xcite as @xmath3 where @xmath4 is a dilaton field , @xmath5 represents a @xmath6 gauge field strength and @xmath7 represents the dilaton potential . this theory corresponds to the bosonic part of the salam - sezgin , six - dimensional gauged supergravity theory @xcite with vanishing kalb - ramond 2-form and chiral scalar fields . hereafter we set @xmath8 for simplicity and if needed , we put it back in explicitly . the theory given in eq . ( [ theory6d ] ) has a dilatonic , charged black hole solution @xcite with metric @xmath9 where @xmath10 and @xmath11 are integration constants . at this stage it will be convenient to perform a _ double _ wick rotation , @xmath12 , @xmath13 with the reparameterizations @xmath14 , @xmath15 @xcite . this leads to @xmath16 in this scenario , the branes are located at positions which are determined by the _ horizon _ condition @xmath17 ; @xmath18 we can then rewrite @xmath19 and whence we also obtain the following useful relations @xmath20 these give a direct relation between the mass and charge of the magnetic flux . the global period of @xmath21 is called @xmath22 , which is determined by the brane tensions through the tension - deficit relations as will be discussed later . note that for the @xmath23 limit , we obtain a single brane solution with a naked curvature singularity in the bulk . by changing coordinates to @xmath24 eq . ( [ bulkmetricori ] ) becomes @xmath25\ , . \label{bulkmetric}\end{aligned}\ ] ] thus , there are two cones at @xmath26 , whose deficit angles are given by @xmath27 respectively . once the brane tensions are fixed , then @xmath28 is fixed . but the absolute size of the extra dimensions , i.e. , @xmath29 , is not fixed due to scale invariance of the bulk solution , unless quantum corrections are taken into account . in the following , instead of fixing @xmath30 , we shall fix @xmath31 and @xmath32 . the branes are codimension two boundaries , embedded onto conical deficits at @xmath33 , whose actions are given by @xmath34 respectively , where @xmath30 denotes the brane tensions . note that we do not assume any coupling of the dilaton with the brane tensions and such vanishing dilaton couplings are required in order that effective cosmological constant on the branes vanishes completely , at least at the classical level @xcite . we also assume that there is no ordinary matter on the branes other than their tensions . the brane tensions are related to the conical deficit angles by @xmath35 we stress that these relations are only valid for sufficiently small brane tensions in comparison with the bulk scale @xmath36 . then , the angle period @xmath37 is determined by @xmath38 where @xmath39 denotes the surface gravities on each brane : @xmath40 thus , we obtain @xmath41 after eliminating @xmath32 , we obtain @xmath42 we emphasize that , in general , @xmath43 implying that the bulk angular period can be completely determined by the brane tensions . for later convenience , we regard it as a function of @xmath32 and @xmath31 : @xmath44 once the brane tensions , @xmath45 and @xmath46 are fixed , then @xmath32 is also fixed and thus , we now regard the free parameters as @xmath32 and @xmath31 , along with the dilaton bulk coupling @xmath7 . the remaining degree of freedom used to determine the bulk geometry is the absolute size of the bulk , i.e. , @xmath29 . due to the scale invariance of the background brane solution , @xmath29 can only be fixed by quantum corrections of bulk fields . note that there is also magnetic flux constraint given by @xmath47 but the magnetic flux only depends on @xmath32 and @xmath31 and does not contribute to the volume of the bulk . we consider the modulus dynamics in a four - dimensional model in the context of the _ moduli approximation _ , assuming @xmath29 is a function of @xmath48 . this approximation may be valid for sufficiently low energy scales compared with the bulk gravitational one @xmath49 . like for the codimension one case , we expect that this approximation is valid for low energies , namely when the energy scale on the brane is below that of the bulk . in this subsection , we shall reintroduce @xmath49 in the action in order to clarify the dimensionality . we regard @xmath29 as a function of the coordinate on the brane @xmath50 in eq . ( [ theory6dmet ] ) and thus , as the volume modulus . then , we expand the components of the lagrangian density and take the modulus parts out . it is rather useful to move to the four - dimensional physical frame : @xmath51 \nonumber\\ & = & 2\rho_+(x^{\mu } ) \tilde \rho(z ) \big [ \frac{dz^2}{g_1 ^ 2 ( 1+z^2)^2 } + \frac{g_1 ^ 2 ( 1-r^2)^2}{4 ( 1+r^2 z^2)^2 } z^2 d\theta^2 + \eta_{\mu\nu}dx^{\mu}dx^{\nu } \big ] \nonumber\\ & = & 2 \tilde \rho(z ) \big [ \rho_+(\tilde x^{\mu})\big ( \frac{dz^2}{g_1 ^ 2 ( 1+z^2)^2 } + \frac{g_1 ^ 2 ( 1-r^2)^2 } { 4 ( 1+r^2 z^2)^2 } z^2 d\theta^2 \big ) + \eta_{\mu\nu}d\tilde x^{\mu}d\tilde x^{\nu } \big ] \,,\end{aligned}\ ] ] where @xmath52 has the dimension of physical length . the metric form is just like a kaluza - klein theory with two compact dimensions , apart from the overall conformal ( warp ) factor @xmath53 . then , we evaluate the effective action of the volume modulus @xmath54 note that the flux and dilaton potential terms do not give any kinetic term contributions to the moduli kinetic term because the vector potential @xmath55 has no scale dependence . the second term is a total derivative term and does not contribute to the effective action . integrating over the extra dimensions , we obtain @xmath56 after redefining the modulus as @xmath57 we obtain the canonical form of the modulus kinetic term as @xmath58 next , we introduce a massless , minimally coupled scalar field ( not to be confused with the dilaton field that supports the background configuration ) . from this we can investigate the one - loop effective action for such a scalar field and thus , the effective potential of the volume modulus on the co - dimension two warped brane background . from now on , we shall work mainly in euclidean space , rather than in the original lorentzian frame . the action for the massless scalar field perturbations is given by @xmath59 the one - loop effective action for a massless minimally coupled scalar field is defined as @xmath60 where @xmath61 is the six - dimensional laplacian , which is divergent and needs to be regularized and renormalized . for this purpose , we define @xmath62 where @xmath63 is the ( integrated ) heat kernel defined by @xmath64 the ( integrated ) zeta function is related to the heat kernel by a mellin transform : @xmath65 and after analytically continuing to @xmath66 we obtain the renormalized one - loop effective action . the explicit expression for the renormalized effective action is @xmath67 where @xmath68 is euler s constant @xmath69 and the pole term at @xmath70 is removed by renormalization . thus , after a redefinition of the renormalization scale we obtain the renormalized effective action and the renormalized scalar field effective action can be written as @xmath71 by integrating over the internal dimensions , the four - dimensional effective potential is @xmath72 where @xmath73 has the dimensions @xmath74 . in the next section , we shall derive the effective potential in a codimension two warped brane model . for brevity , from now on we shall omit the subscript ren " . in the case of codimension two , the zeta function is given by the summation @xmath75 where the eigenvalues are defined by @xmath76 it is straightforward to show that @xmath77 where @xmath78 is a heat kernel coefficient , defined by the asymptotic expansion of the heat kernel @xcite : @xmath79 in the warped codimension two case , the eigenvalues are unknown and there are not even any analytic solutions for the eigenfunctions so we must resort to using an approximate wkb method to estimate the effective action . one strategy to evaluate the one - loop effective action and the effective potential is to define a continuous conformal transformation ( parameterized by @xmath80 ) @xmath81 and thus @xmath82 where for @xmath83 we have the original metric , which we shall denote as @xmath84 . the classical action of this scalar field is changed under a conformal transformation , see eq . ( [ conformal ] ) , by ( here we are considering a massless , minimally coupled bulk scalar field ) @xmath85 where @xmath86 due to technical reasons , which will be explained later , we shall evaluate the zeta function in the non - warped frame @xmath87 . the correction associated with such a conformal transformation is commonly known as the cocycle function ( obtained from integration along the paths of the conformal transformation ) : @xmath88 furthermore , thanks to the conformal invariance of @xmath89 we arrive at @xmath90 the term @xmath91 is given by the volume integration of linear combinations of cubic order curvature tensors @xcite : @xmath92 \nonumber\\ & & \qquad + ( { \rm contribution\,\ , of\,\ , conical\,\ , branes } ) \big\ } \,.\label{a6}\end{aligned}\ ] ] in order to avoid a volume divergence we shall employ the one - loop effective potential , as given by eq . ( [ effact6d ] ) , to try and stabilize the extra dimensions . from our discussion in the previous subsection , the effective potential takes exclusively the following form @xmath93 where we have defined @xmath94 clearly , if @xmath95 , then the modulus effective potential has a minimum at @xmath96 after a redefinition of the modulus , as given by eq . ( [ effrad6d ] ) , the effective potential can be rewritten as @xmath97 the field value at the minimum is then given by @xmath98 and hence , the effective modulus mass becomes @xmath99 thus , the ratio between the stabilized mass and the effective bulk cosmological constant is @xmath100 the expression above allows us to quantify the _ rigidity _ of stabilization ; namely , is the modulus mass stable against kk perturbations . however , the problem is that the @xmath91 contribution for conical branes has never been formulated as far as the authors are aware . in this section , in order investigate the one - loop effective potential of the volume modulus , we consider the four - dimensional counterpart of the six - dimensional case in the theory @xcite , @xmath101 where , again , @xmath4 is a dilaton , @xmath5 represents a @xmath6 gauge field strength and @xmath102 represents the dilaton potential ( corresponding to @xmath7 in the original model ) . hereafter , we set the bulk gravitational scale to @xmath103 for simplicity ; we shall reinsert it if and when it is needed . the essential properties of this solution is summarized in appendix a. the reason we use this model is that we can evaluate the uv contributions from the conical branes . we shall extrapolate our results to the original model in six dimensions . the zeta function and one - loop effective action in four dimensions can be defined similarly to the case of six dimensions , as in eq . ( [ zeta6d ] ) and eq . ( [ effact6d ] ) , just by replacing @xmath104 ( and @xmath105 ) . thus , the one - loop effective potential of the volume modulus takes the same form as eq . ( [ effact6d ] ) , after the replacement of @xmath106 with @xmath107 . the effective action also takes the same form as eq . ( [ dec ] ) , given the scale invariant nature of the flux compactification solution , implying the effective potential can be written similarly as @xmath108 where @xmath109 note that the stabilized volume of the internal space is given by @xmath110 again , if @xmath111 , then the modulus effective potential has a minimum at @xmath112 , which is determined by @xmath113 and @xmath114 . importantly , given the similarity of the spacetime structure in this four dimensional model with the original six dimensional model , the form of the effective modulus field can be estimated by dimensional arguments , in comparison to eq . ( [ effmodu ] ) . namely , @xmath115 acts as the modulus field in two dimensions . the field value at the minimum is then given by @xmath116 and hence , the effective modulus mass is @xmath117 the _ rigidity _ of the stability is determined by the quantity @xmath118 given the conformal invariance of the heat kernel coefficients , @xmath119 is straightforward to evaluate in the @xmath83 frame . in the original four - dimensional spacetime , including the contribution of the cones @xcite we have @xmath120 \big\},\label{a4}\end{aligned}\ ] ] where @xmath121 runs from @xmath122 to @xmath123 which denotes the two orthonormal directions to the conical branes @xmath124 and @xmath125 is the extrinsic curvature defined by @xmath126 . here @xmath127 denotes two mutually orthogonal ( inward pointing ) unit normals and @xmath128 and @xmath129 , where @xmath130 is the induced metric on the brane . the terms @xmath131 are undetermined parameters associated with conformal invariance of the conical heat kernels . but , as shown in appendix c , the extrinsic curvatures vanish on the cones and we need not specify them . note that this formula is valid for any deficit angle @xcite . now we present the results of our analyses on the one - loop effective potential of the modulus . the function @xmath114 that corresponds to @xmath132 in the original six - dimensional model is composed of two parts ; a bulk and conical brane contribution and thus , the total is @xmath133 from the properties of the heat kernel coefficients @xmath114 is invariant under conformal transformations , as is explicitly demonstrated in appendix c. the bulk part of @xmath114 is obtained from eq . ( [ bulk ] ) for @xmath83 and as shown in appendix b , it is in fact independent of @xmath80 . similarly , the brane contribution to @xmath114 is obtained from eq . ( [ inflationary ] ) . in fig . 1 , the bulk contribution to @xmath114 is shown as a function of @xmath134 , for @xmath135 and @xmath136 . note that in the following plots , which are described as functions of @xmath32 , we have set @xmath136 . there is no significant dependence on @xmath31 and thus we mainly focus on @xmath32 . for other values of @xmath102 , the behavior of @xmath114 is also almost the same . the bulk contribution is always positive and as @xmath137 it suffers from a divergence because of the presence of a naked curvature singularity . is shown as a function of @xmath32 , for @xmath135 . ] is shown as a function of @xmath32 , for @xmath135 . ] in fig . 2 , the contribution from the branes , to @xmath114 , is shown as a function of @xmath32 . comparing this with fig . 1 , one can readily see that the cone contribution becomes comparable to that of the bulk one but the sign is opposite . thus , the positivity of the total , @xmath114 , is a rather non - trivial issue . in fig . 3 , the total , @xmath114 , is shown as a function of @xmath32 and we find that the sign of @xmath114 is always positive . thus , for all values of @xmath32 , the modulus can be stabilized ; this is one of the main results in this article . though here we discuss the four - dimensional counterpart , we will make suggestions on the more realistic case of six dimensions later . note that the behavior as shown in fig . 3 is almost the same for other values of the conical deficit angle @xmath31 . ) is shown as a function of @xmath32 for @xmath135 and @xmath136 . in the following all logarithmic plots are to the base @xmath138 . ] ) is shown as a function of @xmath32 for @xmath135 and @xmath136 . in the following all logarithmic plots are to the base @xmath138 . ] next , we wish to discuss the minimum potential energy and and thus , the effective mass of the modulus . these can only be considered through the combination @xmath139 , see eq . ( [ exponent ] ) , where @xmath113 corresponds to @xmath140 in the original model . this ratio should then be compared with the renormalization scale , @xmath141 . we give an explicit derivation of the cocycle function and the derivative of the zeta function in appendix c and d , respectively . the term @xmath113 is obtained from eq . ( [ 4dcoef ] ) . the first term is the derivative of the zeta function in the non - warped conformal frame , @xmath87 , in the context of wkb approximation . the second term is known as the cocycle function , which corresponds to corrections associated with a conformal transformation . by combining these two contributions we obtain @xmath142 the bulk and brane contribution to the cocycle function , @xmath143 , is given in eq . ( [ inflationary2 ] ) , while the wkb contribution , @xmath144 , in the non - warped frame , is given in eq . ( [ inflationary3 ] ) . actually , the value of @xmath113 itself is not as important as the ratio @xmath145 , because the stabilized size of the extra dimensions is given by eq . ( [ exponent ] ) . in fig . 4 , we show a logarithmic plot of the exponential factor as a function of @xmath32 for @xmath135 and @xmath136 . for all values of @xmath32 , the stabilized size of the extra dimensions is larger than that of the renormalization scale . especially in the limits @xmath137 and @xmath146 , much larger extra dimensions are permitted . furthermore , in order to show how the stability of the extra dimensions are set _ rigid _ , we can consider the ratio between the effective modulus mass and the bulk curvature scales @xmath147 . in fig . 5 , we show logarithmic plots of eq . ( [ massratio ] ) as a function of @xmath32 for various dilaton couplings @xmath102 . ) for @xmath148 is shown as a function of @xmath31 . the red ( solid ) , green ( dashed , with wider intervals ) and blue ( dashed , with shorter intervals ) curves correspond to @xmath149 , respectively and @xmath150 . ] ) for @xmath148 is shown as a function of @xmath31 . the red ( solid ) , green ( dashed , with wider intervals ) and blue ( dashed , with shorter intervals ) curves correspond to @xmath149 , respectively and @xmath150 . ] in fig . 6 , we also show them as a function of @xmath31 . if the ratio is greater than unity , then , the stability of the modulus is generic . from fig . 5 and 6 , the value of eq . ( [ massratio ] ) becomes larger for larger couplings @xmath102 ( in fact the relative mass is almost proportional to @xmath151 ) . thus , for large dilaton couplings the stability of the modulus is rigid against kk perturbations . in fig . 5 , we also see that for smaller values of @xmath32 , the modulus mass becomes relatively light , whereas for larger @xmath32 it is insensitive to @xmath32 . in the limit @xmath137 the modulus mass diverges , because a bulk naked singularity is formed . as a final comment , fig . 6 is not particularly sensitive to changes in @xmath31 , but it does exhibit a divergent behavior for @xmath152 . this is because in such a limit the bulk essentially disappears . now , we shall make some suggestions on the stability of the volume modulus in six dimensions . because of our poor knowledge about the contribution from the conical branes we can evaluate only the bulk contribution . however , we can compare the result of the bulk contribution in six - dimensions to the four - dimensional one discussed previously . evaluating the bulk part of @xmath132 from eqs . ( [ a6 ] ) and ( [ 6dcoef ] ) leads to @xmath153 where @xmath154 in fig . 7 , we plot eq . ( [ b6conc ] ) as a function of @xmath32 for @xmath155 and @xmath136 ( this behavior is essentially the same for other values of the coupling @xmath7 ) . the behavior of the bulk part of @xmath132 is similar to that of @xmath114 as shown in fig . 1 . as we mentioned previously , we have no analytic formulae to evaluate the contribution from the conical branes for @xmath78 . however , by analogy we expect similar behavior for the conical part of the heat kernel coefficient . it may be worth mentioning that for larger @xmath32 and appropriate values of @xmath7 , @xmath132 is very likely to be positive and thus the modulus can probably be stabilized . we could also evaluate @xmath156 , via the wkb method , and due to symmetry would follow almost identical steps to the four dimensional case , see appendix c. however , we have no way to include the cocycle correction , at the level of our semi - analytic approach . is shown as a function of @xmath32 , for @xmath136 . the red ( solid ) , green ( dashed , with wider intervals ) and blue ( dashed , with shorter intervals ) curves correspond to @xmath149 , respectively . ] is shown as a function of @xmath32 , for @xmath136 . the red ( solid ) , green ( dashed , with wider intervals ) and blue ( dashed , with shorter intervals ) curves correspond to @xmath149 , respectively . ] thus , any further investigation of the conical contribution and the effective mass of the stabilized modulus ( in six dimensions ) is somewhat out of the scope of this article and is left for future work . finally , before closing this article , we shall discuss some phenomenological implications , i.e. , the hierarchy structure between the fundamental energy scales and the effective vacuum energy on the brane . even though we are investigating , primarily , the four - dimensional case , we expect similar results to hold in six dimensions . one of the most longstanding issues in theoretical physics to date , is the problem of the hierarchy structure of the fundamental energy scales , i.e. , the huge gap between the gravitational scale , @xmath157 , and the electroweak scale , @xmath158 . in the original six - dimensional model , the _ effective _ four - dimensional planck scale in the four - dimensional model . ] is @xmath159 if we assume a brane localized field whose bare mass is given by @xmath160 on either brane at @xmath161 then the observed mass scales are @xmath162 thus , the mass ratio between the field and the effective planck mass is given by @xmath163 assuming that the factor of @xmath164 takes the optimal value of @xmath165 at the unification of the fundamental scales , the mass ratio becomes @xmath166 where we have used the value of @xmath167 , given by eq . ( [ sta - san ] ) , at stabilization . at present , the best we can do is use the results obtained from the four - dimensional model : @xmath168 the results are shown in fig . 8 and indicate that smaller values of @xmath102 and @xmath32 lead to a large hierarchy on both branes . however , the mass of the stabilized modulus is perhaps too light to support the realization of the hierarchy at the fundamental scales . this is somewhat similar to the case of rs codimension one branes @xcite . another interesting issue is the origin of the observed vacuum energy density , which is much smaller than the field theoretical prediction ; the so - called the cosmological constant problem . given the background brane solution that we are considering , the brane vacuum energy ( the brane tensions ) only contributes to the bulk geometry and therefore does not act as an energy source for today s accelerated cosmic expansion ( at the background level the branes are to be kept flat ) . however , in this model , the observed vacuum energy density can arise from one - loop quantum corrections as follows . at the stabilized volume , from eqs . ( [ originaleffpot ] ) and ( [ sta - san ] ) , the effective potential takes the value @xmath169 and hence , the brane realized vacuum energy is almost completely determined by the renormalization scale . it is then quite natural that the renormalization scale is taken to the unification scale , i.e. , of order @xmath49 . however , in such a case the exponential factor is also very small and therefore we may be able to obtain a sufficiently small vacuum energy , which acts as a dark energy source . although the vacuum energy is negative definite one might expect some kind of uplifting mechanism , for instance , due to the presence of test branes @xcite . as in the case of the hierarchy problem , we shall consider the potential energy at the time of the stabilization . whence , the vacuum energy is given by @xmath170 the coefficient acting on the renormalization scale is very similar to the quantity plotted in fig . , for smaller @xmath102 and @xmath32 , we can expect smaller vacuum energy densities of the volume modulus . the mass of the modulus is also very light , but this might be a sign of a dynamically varying vacuum energy . we have discussed the stability of the volume of a warped codimension two brane model with flux compactification , by taking the one - loop quantum corrections of a massless scalar field into account . we have also examined the _ rigidity _ of stability by comparing the ratio between comparison the effective modulus mass and the bulk curvature scales . first , we introduced a warped flux compactification model in six dimensions , where we started from the solution with a spacetime structure similar to the reisner - nortstr@xmath171 m black hole . after a _ double _ wick rotation , we were able to derive the desired braneworld solution . the branes are located at the horizon positions and correspond to conical singularities . for relatively low energies , namely when the brane tensions are less than the bulk gravitational energy scale , the brane tensions are related to the deficit angles directly . once the brane tensions are fixed , via eq . ( [ tendef6d ] ) , the ratio of the brane positions , @xmath28 is also fixed . but , the volume of the extra dimensions is not completely determined , because of the scale invariance of the solution . to this end , we investigated whether or not this dynamical degree of freedom can be fixed by considering the one - loop quantum effective action for a massless , minimally coupled bulk scalar field . we investigated the stability of the modulus for given values of the two brane tensions , or equivalently of @xmath32 and @xmath31 ( the deficit angle of the brane located at @xmath29 ) . the form of the effective potential can be fixed up to overall coefficients , due to scale invariance and dimensional arguments , which depend on @xmath32 and @xmath31 as in eq . ( [ originaleffpot ] ) . from this the condition for the volume modulus to have a minimum ( and hence be stabilized ) is determined only by the sign of the coefficient in the logarithmic term , i.e. , @xmath132 . this coefficient is directly related to the zeta function @xmath172 and thus , the heat kernel coefficient @xmath173 . in the actual investigation of stability , we have used an alternative model in four dimensions that has a very similar spacetime structure to the original six - dimensional model . this is because the contribution for the cone to the @xmath91 heat kernel coefficient has _ never _ been formulated as far as the authors are aware . we then showed that the heat kernel coefficient is positive definite for most dilaton couplings , @xmath102 , independently of the choice of @xmath32 and @xmath31 . the contribution from the bulk is positive , whereas that from the conical branes is negative for smaller @xmath32 ; though they are of the same order . for all of the cases with @xmath174 , the bulk dominates over the brane parts and thus , the volume is _ stabilized_. however , one might ask whether or not the stability is _ rigid _ against the various kk perturbations . this rigidity can be determined from the ratio between the effective modulus mass and the stabilized curvature radius of the bulk , as given by eq . ( [ massratio ] ) , implying a knowledge of the ratio of @xmath175 . if this ratio is less than unity , then the mass of the volume modulus becomes too light and may be destabilized by other kk contributions , whereas if it is greater than unity , the modulus field is not easily perturbed from the stabilized " minimum . we showed that the mass ratio becomes larger for larger coupling @xmath102 , which is proportional to @xmath151 and therefore , stability is _ rigid_. for larger degrees of warping , i.e. , for smaller value of @xmath32 , other than @xmath176 , we found that the stabilized modulus mass becomes relatively light and that this ratio is not so sensitive to different values of deficit angle @xmath31 . note that it only exhibits a divergence at @xmath177 , where the bulk essentially vanishes . then , we made some suggestions for the six - dimensional case and by comparing the bulk part with that of the four - dimensional @xmath178 bulk part we found similar behavior indicating that our conclusion in four dimensions will carry through to the six - dimensional case : the modulus can be stabilized and furthermore our approximate wkb approach suggests that the mass of the modulus leads to rigidity for certain couplings and parameters . of course , once formal expressions for the six - dimensional conical heat kernel coefficients are found , then we can extend the same analysis as that employed in this paper . however , a more accurate ( semi - analytical / numerical ) method to evaluate @xmath179 should be found , rather than the approximate wkb eigenvalue method that we used in this article . we also discussed some phenomenological issues , i.e. , the hierarchy problem of the fundamental energy scales and the extremely tiny value of the observed vacuum energy density . the six - dimensional results were extrapolated from the corresponding results in four dimensions . we found that in the hierarchy problem , the ratio between the energy scale of the brane localized field and the gravitational scale becomes much smaller than unity , especially for larger degrees of warping @xmath180 . furthermore , the effective vacuum energy density can also be much smaller than the gravitational scale . the vacuum energy is negative by definition and we therefore need some kind of uplifting mechanism , to obtain a more realistic cosmology . in these cases , as was mentioned previously , the mass of the volume modulus may also be small and thus , might be destabilized . it is also interesting to compare our results with the codimension one case . an important difference is that in the codimension one case all the moduli are not dynamically fixed and should be stabilized by quantum corrections of bulk fields ; however in the codimension two case at least one of the moduli fields is fixed dynamically , because the brane tensions are directly related to the bulk geometry . in terms of heat kernel coefficients another obvious difference is that we are considering even dimensions . in odd dimensions only boundary terms occur and thus , the renormalization only affects the brane tensions , whereas in even dimensions there is a bulk contribution and hence logarithmic terms appear . probably the most important difference is that the stabilization of the volume modulus can be realized by the contribution from the bulk , not from the branes , for small warpings . for @xmath180 ( large warpings ) the stabilized modulus effectively behaves as a very light scalar field on the brane , which is somewhat similar to the codimension one case , e.g. , see @xcite . before closing this article we shall refer to some extensions of this work : one issue is to do with the quantum corrections of other fields , such as the magnetic flux and graviton . the tensor part of the graviton perturbations should be equivalent to a minimally coupled scalar field , which we investigated in this article and such analyses might give us important insights into warped flux compactifications with self - gravitating branes . besides quantum effects , there are many interesting issues that still remain in codimension two braneworld , e.g. , gravitational waves in the bulk @xcite , non - linear dynamics of the bulk and branes , thick branes and so on . such work might be key to understanding the self - tuning mechanism of the effective cosmological constant as a dynamical process . we hope to report on these topics in ongoing work @xcite . we are grateful to a. flachi and k. uzawa for valuable and interesting discussions . this work was supported in part by monbukagakusho grant - in - aid for scientific research(s ) no . 14102004 and ( b ) no . 17340075 . m. m. and m. s. were also supported by the japan - france research cooperative program . in this appendix , we consider an alternative model of a warped flux compactification in a four - dimensional einstein - maxwell - dilaton theory given by eq . ( [ theory ] ) . this is a toy - model analogue of the original six - dimensional co - dimension two braneworld model with a warped flux compactification and a non - vanishing scalar potential @xcite : hereafter , we set the bulk gravitational scale to @xmath103 for simplicity ; we shall reinsert it if and when it is needed . as for six dimensions , we first consider a black hole solution of the theory @xcite . after a _ double _ wick rotation , we obtain @xmath181 in this scenario the branes are located at positions that are determined by the _ horizon _ condition @xmath17 ; @xmath182 we can then rewrite @xmath183 and whence we also obtain the following useful relations @xmath184 this gives a direct relation between the mass and charge to the magnetic flux . the global period of @xmath21 , denoted @xmath22 , is determined by the brane tensions through the tension - deficit relations as is discussed later . after a coordinate transformation @xmath185 the bulk metric becomes @xmath186 where our choice of factorization is for later convenience . in this frame we can now easily investigate the asymptotic behavior of the spacetime near each of the conical branes . for @xmath187 , there are cones whose deficit angles are obtained from @xmath188 the modulus corresponds to the proper length between two branes ( strings ) , which is characterized by @xmath161 , or equivalently @xmath29 and @xmath32 . the string tensions are related to the conical deficits by @xmath189 we stress that these relations are only valid for sufficiently small brane tensions in comparison with the bulk scale @xmath190 . again , @xmath37 is given by a relation identical to eq . ( [ angle period ] ) and thus , @xmath191 after eliminating @xmath32 , we obtain @xmath192 once the brane tensions , @xmath45 and @xmath46 are fixed , then @xmath32 is also fixed and so we now regard the free parameters as @xmath32 and @xmath45 , along with the dilaton bulk coupling @xmath102 . the remaining degree of freedom used to determine the bulk geometry is the absolute size of the bulk , i.e. , @xmath29 . however , because of the scale invariance , @xmath29 can only be fixed by quantum corrections of bulk fields . the magnetic flux is given by @xmath193 thus , the gauge field flux does not contribute to the size modulus . we discuss the basic properties of a massless , minimally coupled scalar field on this background , which is introduced in order to investigate the one - loop quantum effects in the counterpart model discussed in sec iv and v , whose action is given by @xmath194 as before , we consider a continuous conformal transformation of the metric , parameterized by @xmath80 ; @xmath195 and thus @xmath196 where for @xmath83 we have the original metric , which we shall define as @xmath197 . the classical action changes as @xmath198 where @xmath199 the classical action of a massless , minimally coupled scalar field is changed under this conformal transformation , see eq . ( [ class4d ] ) . in four dimensions , including conical branes , the @xmath178 heat kernel coefficient is given by @xmath200 \bigg\}\ , . \label{a4conf}\end{aligned}\ ] ] the integrand of the bulk part of eq . ( [ a4conf ] ) is @xmath201 where @xmath202 at a glance , eq . ( [ bulk ] ) appears to depend on the parameter @xmath80 ; however , we can show that it is independent of @xmath80 , i.e. , it is conformally invariant . we can also evaluate the part coming from the cones . starting with the normal derivatives @xmath203 in the bulk we obtain @xmath204 these values are invariant for the limits , @xmath205 . the extrinsic curvature in each direction is given by @xmath206 and we find @xmath207 thus , we see that these terms vanish at the locations of the conical branes : at @xmath208 and @xmath209 . given this fact it is easy to show that @xmath210 is also conformally invariant . the explicit form of @xmath114 for conical branes is finally obtained as @xmath211 \nonumber\\ & & + \frac{g^2}{288\pi } \frac{\frac{\delta_+}{2\pi}-(1-r)}{r } \frac{r+(1-\frac{\delta_+}{2\pi } ) } { 1-\frac{\delta_+}{2\pi } } \big[4 -\frac{1}{5 } \frac{r^2+(1-\frac{\delta_+}{2\pi})^2 } { ( 1-\frac{\delta_+}{2\pi})^2 } ( 1-\frac{3}{r } ) \big]\,,\label{inflationary } \ ] ] where we have used @xmath212 we have also confirmed that the bulk part of @xmath173 is conformally invariant and we expect a similar behavior for the cone part , as in four dimensions . in this appendix , we shall give a derivation of the cocycle function for conical boundaries in four dimensions . in deriving such a term , we shall use the conformal properties of the heat kernel coefficients , see e.g. , @xcite . in general , for a conformal transformation of the metric , given by @xmath213 , the heat kernel coefficient in @xmath214-dimensions satisfies the following relation @xmath215 = \frac{1}{d-2k } \delta a^{(d)}_{k } [ e^{2\omega } g_{ij},1]\big|_{\omega=0}\ , . \label{heatconf}\end{aligned}\ ] ] we shall use this formal relation to derive the _ smeared _ heat kernel coefficients . at a glance , there seems to be a pole at @xmath216 in eq . ( [ heatconf ] ) . however , because @xmath217 is conformally invariant in four - dimensions ( see appendix b ) , then the leading order pole term vanishes . if @xmath218 is _ not _ conformally invariant , then we should add the term , @xmath219|_{\omega=0}$ ] , which includes mass and curvature couplings @xcite . this is in order to keep conformal invariance on the right - hand - side . in our case @xmath220 is already conformally invariant , because of the @xmath221 term that appears as a result of the conformal transformation of the classical scalar action . the conical contribution to @xmath210 is given by @xmath222\end{aligned}\ ] ] and in our case the conformal transformation is defined by eq . ( [ confort4d ] ) . for a general @xmath214-dimensional spacetime , we obtain @xmath223 \,,\end{aligned}\ ] ] where in the second step we have used the conformal scaling property of the laplacian , e.g. , see @xcite . note that @xmath83 corresponds to the original metric for the codimension two brane . given the above conformal transformation the ricci scalar , ricci and riemann curvature terms scale as @xcite @xmath224 \nonumber\\ \big(\sum_a r_{aa } -2\sum_{a , b } r_{abab } + \frac{1}{2}\sum_a \kappa^{(a ) } \kappa^{(a)}\big)(\epsilon ) & = & e^{-2(\epsilon-1)\omega } \big [ \sum_a r_{aa } -2\sum_{a , b } r_{abab } + \frac{1}{2 } \sum_{a , b } \kappa^{(a)}\kappa^{(a ) } \nonumber\\ & & -(\epsilon-1 ) \big ( ( d-4 ) \sum_a \omega_{;ij}n^{(a)i}n^{(a)j } + 2 \delta_2 \omega + ( d-2)\sum_a \kappa^{(a)}\omega_{;i } n^{(a)i } \big ) \nonumber\\ & & + ( \epsilon-1)^2 \big ( \frac{1}{2}(d-4)(d+2 ) \sum_a \omega_{;i}\omega_{;j}n^{(a)i } n^{(a)j } -2(d-4)\omega_{;k}\omega^{;k } \big ) \big ] \nonumber\\ \sum_a \big ( \kappa^{(a)}{}^2 -2 \kappa^{(a)}_{ij}\kappa_{(a)}^{ij } \big ) \big)(\epsilon ) & = & e^{-2(\epsilon-1)\omega } \sum_a \big [ \kappa^{(a)}{}^2 -2 \kappa^{(a)}_{ij}\kappa_{(a)}^{ij } -2(d-4)(\epsilon - 1)\omega_{;i}n^{(a ) i } \kappa^{(a ) } \nonumber\\ & & + ( d-2)(d-4)(\epsilon-1)^2 \omega_{;i } \omega_{;j } n^{(a)i } n^{(a)j}\big]\ , , \label{coldspot}\end{aligned}\ ] ] where @xmath225 denotes the covariant derivative with respect to the metric @xmath83 and curvature tensors which appear on the right - hand - side are also defined for the same metric . here , @xmath226 represents the laplace operator defined by @xmath227 note that terms including @xmath228 and @xmath229 vanish on the cone in the model which we are considering and thus , we neglect these terms . all the remaining terms proportional to @xmath230 are also proportional to @xmath231 . the terms which include @xmath232 ( or @xmath233 ) are also vanishing on the cones . however , in order to see the conformal properties of these terms , in this subsection , we keep them . by multiplying the conformal transformation of the trace of the brane induced metric , @xmath234 , with these combinations of intrinsic and extrinsic curvature terms , we can expand the terms as a series in powers of ( @xmath231 ) . for instance , expanding the first term in eq . ( [ a4f1 ] ) we obtain @xmath235 \nonumber\\ & = & \sqrt{h } e^{(d-4)(\epsilon-1)\omega } \big [ \frac{1}{6 } r + \lambda_{1}\sum_a \big(\kappa^{(a)}{}^2 -2 \kappa^{(a)}_{ij}\kappa_{(a)}^{ij}\big ) \nonumber \\ & & + ( d-4 ) \big\ { \frac{1}{6}(\epsilon-1 ) \delta \omega + \frac{1}{12}(\epsilon-1)^2(d-2 ) \omega_{;k}\omega^{;k } + \lambda_1 \big ( -2(\epsilon-1)\sum_a \omega_{;i}n^{(a)i}\kappa^{(a ) } \nonumber\\ & & + ( d-2)(\epsilon-1)^2\sum_a \omega_{;i}\omega_{;j}n^{(a)i}n^{(a)j } \big ) \big\ } \big ] \nonumber\\ & = & \sqrt{h}\big [ \frac{1}{6 } r + \lambda_{1}\sum_a \big ( \kappa^{(a)}{}^2 -2 \kappa^{(a)}_{ij}\kappa_{(a)}^{ij}\big)\big ] \nonumber\\ & + & ( d-4)\sqrt{h } \big [ ( \epsilon-1)\omega \big(\frac{1}{6 } r + \lambda_{1}\sum_a \big ( \kappa^{(a)}{}^2 -2 \kappa^{(a)}_{ij}\kappa_{(a)}^{ij } \big ) \big ) + \frac{1}{6}(\epsilon-1 ) \omega_{;k}{}^{;k } + \frac{1}{12}(d-2)(\epsilon-1)^2 \omega_{;k}\omega^{;k } \nonumber\\ & & + \lambda_1 \big ( -2(\epsilon - 1)\sum_a \omega_{;i}n^{(a ) i } \kappa^{(a ) } + ( d-2)(\epsilon-1)^2 \sum_a \omega_{;i } \omega_{;j } n^{(a)i } n^{(a)j } \big ) \big ] + o((d-4)^2)\,.\end{aligned}\ ] ] similarly , expanding the second piece in eq . ( [ a4f1 ] ) leads to : @xmath236 + o((d-4)^2 ) \,.\end{aligned}\ ] ] now we can construct the cocycle function for the conical boundaries . the @xmath237 terms can be reduced to @xmath238 and because @xmath239 vanishes on the cone , at @xmath240 , we can discard these terms . whence @xmath241 \big\}\ , . \nonumber\end{aligned}\ ] ] finally , performing the integration over @xmath80 we obtain the cocycle contribution for the conical branes : @xmath242 \big\ } \nonumber\\ & = & \int d^2 x \big\ { -\frac{g^2(2\pi-\delta_+)}{1440\pi^2 } \int_0 ^ 1 d\epsilon \int^{\infty}_0 dz \big(\frac{1}{2}\ln \big(\frac{2\rho}{\rho_+}\big ) \big ) \frac { z } { ( 1+z^2)(1+r z^2 ) } \frac{\psi_4(1,r,\epsilon , z)}{(1 + r z^2)^4 } \nonumber\\ & -&\frac{g^2}{144\pi } \frac{\delta_+}{2\pi } \frac{2-\frac{\delta_+}{2\pi } } { 1-\frac{\delta_+}{2\pi } } \big[2 \ln(2\rho_+ ) -1+r -\frac{1}{5 } \frac{1+(1-\frac{\delta_+}{2\pi})^2 } { ( 1-\frac{\delta_+}{2\pi})^2 } \big ( \frac{1}{2}\ln(2\rho_+ ) ( 1 - 3r ) + \frac{1-r}{2 } \big ) \big ] \nonumber\\ & - & \frac{g^2}{144\pi } \frac{\frac{\delta_+}{2\pi}-(1-r)}{r } \frac{r+(1-\frac{\delta_+}{2\pi } ) } { 1-\frac{\delta_+}{2\pi } } \nonumber\\ & \times & \big[2\ln(2\rho_- ) -1+\frac{1}{r } -\frac{1}{5 } \frac{r^2+(1-\frac{\delta_+}{2\pi})^2 } { ( 1-\frac{\delta_+}{2\pi})^2 } \big ( \frac{1}{2}\ln(2\rho_- ) ( 1-\frac{3}{r } ) + \frac{1-\frac{1}{r}}{2 } \big ) \big ] \big\ } \,.\end{aligned}\ ] ] combining the above result with the bulk piece , we obtain the total cocycle part of @xmath113 , defined in eq . ( [ 4dcoef ] ) , as @xmath243 \nonumber\\ & - & \frac{g^2}{144\pi } \frac{\frac{\delta_+}{2\pi}-(1-r)}{r } \frac{r+(1-\frac{\delta_+}{2\pi } ) } { 1-\frac{\delta_+}{2\pi } } \nonumber\\ & \times & \big[2\ln(2r ) -1+\frac{1}{r } -\frac{1}{5 } \frac{r^2+(1-\frac{\delta_+}{2\pi})^2 } { ( 1-\frac{\delta_+}{2\pi})^2 } \big ( \frac{1}{2}\ln(2r ) ( 1-\frac{3}{r } ) + \frac{1-\frac{1}{r}}{2 } \big ) \big ] \ , , \label{inflationary2}\end{aligned}\ ] ] where @xmath244 is given in eq . ( [ psiconf ] ) . in this appendix , we derive the mass spectrum in the conformal frame ( @xmath87 ) : @xmath245 and we consider the corresponding eigenvalue problem @xmath246 where @xmath247 is obtained from eq . ( [ good ] ) and we shall decompose the mass eigenfunction as @xmath248 from now on , we concentrate on the radial mode function and derive the mass spectrum in the context of the wkb approximation . changing variables to @xmath249 leads to the diagonalized equation of motion in terms of @xmath250 : @xmath251 where @xmath252 to apply the wkb method we require a langer transformation @xcite @xmath253 and a further redefinition of the eigenfunction @xmath254 . thus , eq . ( [ bulkquintom ] ) becomes @xmath255 where @xmath256 the coordinate @xmath257 runs from @xmath258 to @xmath259 , while the turning points , defined by @xmath260 , are given by @xmath261 where @xmath262 satisfies @xmath263 . ( in the original frame , @xmath83 , the turning point problem is even more difficult to solve . ) furthermore , for the @xmath264 modes the turning points extend to @xmath265 and in such a case the mass spectrum becomes exact in the context of the wkb method . for @xmath266 , turning points appear if and only if @xmath267 the mass spectrum , which will be determined later via the wkb method , is certainly valid for the modes that satisfy the above condition . however , we will use the wkb result for deriving the full zeta function , because this is the best we can do at present . nevertheless , it should give a useful order of magnitude approximation to @xmath268 . in the wkb approximation , the mass spectrum is given by the quantization condition @xmath269 where the integration can be performed as follows : first , we change the integration variable @xmath270 where @xmath271 . then , we use the fact that @xmath272 in the case that we are now considering , @xmath273\,.\end{aligned}\ ] ] then , we evaluate each term ; @xmath274_{u_+}^{u_- } = -\pi \alpha ( 1+r ) |n|\,,\end{aligned}\ ] ] and @xmath275\,.\label{integral}\end{aligned}\ ] ] unfortunately , in order to find the approximated mass spectrum , we must make one further approximation : @xmath276 . this limit corresponds to the case where @xmath277 and @xmath278 respectively . hence , @xmath279 = ( m+\frac{1}{2})\pi\ , . \label{wkbmass}\end{aligned}\ ] ] the modes with @xmath264 correspond to the limits @xmath280 and @xmath281 . thus , from eq . ( [ integral ] ) the mass spectrum in the wkb approximation becomes _ exact _ : @xmath282 for these modes we use a single summation in terms of @xmath283 . to investigate the accuracy of wkb method we can compare the result to the non - warped kk limit , where @xmath284 this has an exact analytic solution for the mode functions with @xmath285 where we have made a coordinate transformation from @xmath286 and @xmath287 the exact mass spectrum is then determined by the regularities of the mode functions on both boundaries , at @xmath288 and thus , we find @xmath289 where @xmath290 as before . in order to have positivity of the eigenvalue , @xmath291 , we shall take the latter condition . from this the _ exact _ mass spectrum is found to be @xmath292 the result of our wkb method , eq . ( [ wkbmass ] ) , agrees with the one above ( when @xmath284 ) except for a constant term in the eigenvalue spectrum . however , given that our wkb approximation is valid for large quantum numbers this constant term is negligible in such a limit . given the wkb spectrum derived in the previous subsection , we derive an expression for the zeta function and the associated effective action . the zeta function is given by the mode sum @xmath293 hereafter , we omit `` wkb '' in the subscript of the zeta function . here , we shall use the density of states method @xcite , which from eq . ( [ wkbmass ] ) the density of states is easily derived : @xmath294 we can therefore replace the sum over @xmath283 with an integration in terms of @xmath291 @xcite @xmath295 it the following it will be convenient to decompose the zeta function into two pieces , i.e. , the @xmath296 and @xmath264 modes . the total of the derivative of zeta function is finally given by @xmath297 similar considerations can easily be applied to the six - dimensional case , working in the conformal frame . for the @xmath298 mode , we obtain the following expression @xmath299 where because of the condition that there be two turning points , see eq . ( [ conde ] ) , we must restrict the range of integration . the @xmath300 denotes that the @xmath264 mode is to be dropped , because the potential for this mode is not oscillator - like . performing the @xmath301-integration first , we obtain @xmath302 where @xmath303 is defined by @xmath304 for sufficiently large @xmath305 , we obtain the following integration formulae @xmath306 \big\ } \ , , \label{yamada}\end{aligned}\ ] ] where again @xmath307 is a hypergeometric function . note that the first term in eq . ( [ yamada ] ) has been analytically continued by employing the standard duplication formula for the gamma function ; otherwise the result will be ill defined when taking the derivative at @xmath70 . finally , expressing the summations over @xmath308 in terms riemann zeta functions we obtain ( substituting back the expression for @xmath309 , see eq . ( c6 ) ) @xmath310 -1 \big ) \big\}\big]\,,\end{aligned}\ ] ] where @xmath311 is riemann s zeta function . then , taking the derivative in terms of @xmath312 and an analytic continuation @xmath313 , we find @xmath314 \big\ } \,.\end{aligned}\ ] ] the zeta function for @xmath264 mode is given by @xmath315^{-s } = \frac{\pi ( 2g)^{2 - 2s}}{(s-1 ) } \zeta_{h } ( 2s-2,\frac{1}{2 } ) \,,\end{aligned}\ ] ] where @xmath316 is hurwitz s zeta function . taking the derivative with respect to @xmath312 and continuing to @xmath66 gives @xmath317 s. m. carroll and m. m. guica , arxiv : hep - th/0302067 . i. navarro , jcap * 0309 * , 004 ( 2003 ) [ arxiv : hep - th/0302129 ] . y. aghababaie , c. p. burgess , s. l. parameswaran and f. quevedo , nucl . b * 680 * , 389 ( 2004 ) [ arxiv : hep - th/0304256 ] . t. tanaka and x. montes , nucl . b * 582 * , 259 ( 2000 ) [ arxiv : hep - th/0001092 ] . c. csaki , m. graesser , l. randall and j. terning , phys . d * 62 * , 045015 ( 2000 ) [ arxiv : hep - ph/9911406 ] . j. garriga , o. pujolas and t. tanaka , nucl . b * 605 * , 192 ( 2001 ) [ arxiv : hep - th/0004109 ] . j. garriga , o. pujolas and t. tanaka , nucl . b * 655 * , 127 ( 2003 ) [ arxiv : hep - th/0111277 ] . g. w. gibbons , r. guven and c. n. pope , phys . b * 595 * , 498 ( 2004 ) [ arxiv : hep - th/0307238 ] . y. aghababaie _ et al . _ , jhep * 0309 * , 037 ( 2003 ) [ arxiv : hep - th/0308064 ] . h. m. lee and c. ludeling , jhep * 0601 * , 062 ( 2006 ) [ arxiv : hep - th/0510026 ] . b. m. n. carter , a. b. nielsen and d. l. wiltshire , arxiv : hep - th/0602086 . h. nishino and e. sezgin , phys . b * 144 * , 187 ( 1984 ) . h. nishino and e. sezgin , nucl . b * 278 * , 353 ( 1986 ) . s. mukohyama , y. sendouda , h. yoshiguchi and s. kinoshita , jcap * 0507 * , 013 ( 2005 ) [ arxiv : hep - th/0506050 ] . h. yoshiguchi , s. mukohyama , y. sendouda and s. kinoshita , jcap * 0603 * , 018 ( 2006 ) [ arxiv : hep - th/0512212 ] . d. v. vassilevich , phys . rept . * 388 * , 279 ( 2003 ) [ arxiv : hep - th/0306138 ] . c. p. burgess and d. hoover , arxiv : hep - th/0504004 . d. hoover and c. p. burgess , jhep * 0601 * , 058 ( 2006 ) [ arxiv : hep - th/0507293 ] .

we investigate the symanzik improvement of the wilson quark action on anisotropic lattices . taking first a general action with nearest - neighbor and clover interactions , we study the mass dependence of the ratio of the hopping parameters , the clover coefficients , and an improvement coefficient for heavy - light vector and axial vector currents . we show how tree - level improvement can be achieved . for a particular choice of the spatial wilson coupling , the results simplify , and improvement is possible . ( here is the bare quark mass and the temporal lattice spacing . ) with this choice we calculate the renormalization factors of heavy - light bilinear operators at one - loop order of perturbation theory employing the standard plaquette gauge action .

hep-lat0103026

the anisotropic lattice has become an important tool in lattice qcd simulations . with a small temporal lattice spacing @xmath2 one can more easily follow the time evolution of correlators , while keeping the spatial lattice spacing @xmath3 comparatively modest @xcite . this approach is especially effective when the signal - to - noise ratio deteriorates quickly , as , for example , in the case of glueballs @xcite . the better signal - to - noise ratio is beneficial also for heavy quark systems @xcite . in addition , it is hoped that the anisotropy can be exploited to reduce lattice artifacts @xcite , which are a special concern with heavy quarks . in current work on heavy quarks , lattice artifacts are controlled with non - relativistic qcd ( nrqcd ) and heavy - quark effective theory ( hqet ) . this is done either _ a priori _ , by discretizing the nrqcd action @xcite , or _ a posteriori _ , by using the effective theories to describe lattice gauge theory with wilson fermions @xcite . these strategies are possible because the typical _ spatial _ momenta in heavy quark systems are much smaller than the heavy quark mass . heavy quarkonia have momenta @xmath4 and @xmath5 , where @xmath60.3 is the heavy - quark velocity ; heavy - light hadrons have momenta only of order @xmath7 . in these approaches one is left with discretization effects of order @xmath8 from the light quarks and gluons and of order @xmath9 from the heavy quark . the method of ref . @xcite smoothly connects to the usual continuum limit , so one can , in principle , reduce discretization effects to scale as a power of the lattice spacing @xmath10 , but only by making @xmath10 too small to be practical . klassen proposed using anisotropic lattices with the anisotropy @xmath11 chosen so that @xmath12 and @xmath13 are both small @xcite . clearly , this proposal works only if @xmath14 is smaller than @xmath15 , as in the approaches based explicitly on heavy - quark theory . it also works only if renormalization constants have a smooth limit as @xmath16 , where @xmath1 is the bare quark mass . in particular , one would like to be able to expand the renormalization constants in powers of @xmath17 even when @xmath18 . then it may be possible to adjust the improvement parameters of the lattice action ( and currents ) in a non - perturbative , mass - independent scheme @xcite . if , on the other hand , @xmath19 dependence appears in an essential way , then one would be forced back to a non - relativistic interpretation , as explained for isotropic lattices in refs . @xcite . to our knowledge there is no proof that cutoff effects always appear as powers of @xmath20 . in this paper we try to gain some experience by calculating the full mass dependence of several ( re)normalization constants , first at tree level and then at one - loop in perturbation theory . we focus on the fermilab action @xcite , which is the most general action without doubler states , having different nearest - neighbor and clover couplings in the temporal and spatial directions . this action has been applied on anisotropic lattices to the charmonium system @xcite , as have some actions with next - to - nearest neighbor interactions . the self - energy has been calculated at the one - loop level in perturbation theory @xcite . in the numerical work on charmonium , two different choices for tuning the spatial wilson term have been made . one choice is that of refs . @xcite , where @xmath21 . another choice is that of refs . @xcite , where @xmath22 . in the first part of this paper , we study improvement conditions for these two choices , as a function of the heavy quark mass . ( in a perturbative calculation more generally improved actions with @xmath23 also have been considered @xcite . ) by studying the full functional dependence on @xmath24 and @xmath17 , we can test whether @xmath25 appears in an essential way . we find that the limit of small @xmath17 is benign at the tree - level only for the first choice , @xmath21 . for the other choice , @xmath22 , the continuum limit is reached only for @xmath26 . it turns out that with the first choice ( @xmath21 ) two of the improvement parameters vanish at the tree - level as @xmath16 . this simplifies the one - loop analysis , so in the second part of the paper we concentrate on this choice . this calculation has two purposes . the first is to study cutoff effects of the renormalization coefficients and to test at the one - loop level whether they still appear only as powers of @xmath27 . the second is for phenomenological applications to heavy - light matrix elements . even if a non - relativistic interpretation is necessary , anisotropic lattices are a good method for reducing the signal - to - noise ratio @xcite . this paper is organized as follows . [ sec : formulation ] describes the quark field action and discusses its parametrization in detail . in sec . [ sec : one - loop ] , the expression for the one - loop perturbative calculation is given . the numerical result for these perturbative constants are presented in sec . [ sec : result ] . the last section is devoted to our conclusions . we give the feynman rules in appendix [ sec : feynman ] and explicit expressions for the one - loop diagrams in appendix [ sec : expressions ] . this section describes the actions with wilson fermions @xcite on anisotropic lattices . we denote the renormalized anisotropy with @xmath24 , and the spatial and temporal lattice spacings with @xmath28 and @xmath29 respectively : @xmath30 these lattice spacings would be defined through the gauge field with quantities such as the wilson loops or the static quark potential . we therefore consider @xmath24 to be independent of the quark mass . following ref . @xcite , let us introduce an action with two hopping parameters @xcite and two clover @xcite coefficients , @xmath31 \nonumber \\ & & - \kappa_s \sum_i [ ( r_s - \gamma_i ) u_{n , i } \psi_{n+\hat{\imath } } + ( r_s + \gamma_i ) u^\dagger_{n-\hat{\imath},i}\psi_{n-\hat{\imath } } ] \label{eq : kappa_action } \\ & & + \ihalf r_s c_b \kappa_s \sum_{i , j , k } \epsilon_{ijk } \sigma_{ij } b_{n , k } \psi_n + i c_e \kappa_s \sum_{i } \sigma_{4i } e_{n , i } \psi_n \big ] . \nonumber\end{aligned}\ ] ] this is the most general nearest - neighbor clover action . note that the notation is slightly different than in ref . @xcite ; @xmath32 of ref . @xcite corresponds to @xmath33 in ( [ eq : kappa_action ] ) . it is helpful to change to a notation with a quark mass . we rescale field by @xmath34 and introduce the bare mass @xmath35,\ ] ] with @xmath36 . then one can rewrite the action as @xmath37 \psi(x ) . \label{eq : mass_action}\end{aligned}\ ] ] the covariant difference operators @xmath38 , @xmath39 and @xmath40 , and the fields @xmath41 and @xmath42 are defined as in ref . @xcite , except that the lattice spacing @xmath10 is replaced by @xmath29 or @xmath28 in the obvious way . the action @xmath43 has six parameters @xmath1 , @xmath44 , @xmath45 , @xmath46 , @xmath32 , and @xmath47 . two are redundant and can be chosen to solve the doubling problem @xcite . in particular , we choose @xmath48 to eliminate doubler states . we then rename @xmath49 , but discuss how to adjust it below . the other four parameters are dictated by physics . the bare mass is adjusted to give the desired physical quark mass , and @xmath46 , @xmath32 , and @xmath47 are chosen to improve the action . following ref . @xcite we also consider a rotated field @xmath50\psi(x ) , \label{eq : rot}\ ] ] where @xmath51 is the rest mass , defined and given below , and @xmath52 is an improvement parameter . this field is convenient for constructing heavy - light bilinears @xmath53 which , at the tree - level , are correctly normalized currents for all @xmath17 . beyond the tree - level one may add dimension - four terms to these currents , and one must multiply with suitable renormalization factors . the renormalization factors and the improvement parameters @xmath46 , @xmath32 , @xmath47 , and @xmath52 must , in general , be chosen to be functions of @xmath17 and the anisotropy @xmath24 . below we shall give the full mass dependence to check whether , for small @xmath17 , power series such as @xmath54 can be admitted . if @xmath55 enters into the full mass - dependent expression , this series would not be accurate when @xmath18 . in the past @xcite the behavior in ( [ m0at - exp ] ) was implicitly assumed . if the expansions of the form ( [ m0at - exp ] ) do work , then for full @xmath56 improvement one must adjust @xmath57 , @xmath58 , @xmath59 , and @xmath60 in the action , and @xmath61 , @xmath62 , and @xmath63 of the currents @xmath64 , @xmath65 . from ( [ eq : mass_action ] ) one can see that conditions for the improvement coefficients can be obtained by simply replacing @xmath66 in formulae in ref . for example , the energy of a quark with momentum @xmath14 is given by @xmath67 where @xmath68 and @xmath69 . for small momentum @xmath70 , where the rest mass @xmath51 and kinetic mass @xmath71 are @xmath72 to obtain a relativistic quark one sets the rest mass and kinetic mass equal to each other . this yields the condition @xmath73 which can be read off from ref . matching of on - shell three - point functions yields the conditions @xmath74 on the clover coefficients , and @xmath75 } { 2m_0a_\tau(2+m_0a_\tau)(1+m_0a_\tau ) } \label{d1}\end{aligned}\ ] ] on the rotation parameter . these tree - level formulae ( [ coshe])([d1 ] ) have been obtained independently by m. okamoto @xcite . we see that essential dependence on @xmath55 indeed may arise , depending on how @xmath76 is tuned . from the energy ( [ coshe ] ) one can also find the energy of states at the edge of the brillouin zone . the energy of a state with @xmath77 components of @xmath14 equal to @xmath78 is @xmath79 although there is some freedom to choose @xmath76 , discussed below , one still wants to keep @xmath80 and @xmath51 well separated . for small @xmath17 the interesting taylor expansions are @xmath81 with the mass dependent factor in ( [ eq : rot ] ) there is no mass dependence at the tree - level in the currents normalization factors . let us now discuss the choice of the redundant coefficient of the spatial wilson term @xmath76 . two choices have been used in numerical calculations : 1 . @xmath82 @xcite . this is a natural choice because then the mass form of the action takes a symmetric - looking form , without @xmath24 . in the small @xmath17 limit , the tree - level improvement parameters become @xmath83 a key advantage is that @xmath25 does not appear ; the continuum limit is reached for small @xmath17 . furthermore , both @xmath58 and @xmath63 vanish at the tree - level , which is especially helpful in one - loop calculations . a disadvantage is that with @xmath84 the energy splitting between physical states and states at the edge of the brillouin zone is not large . one can circumvent this problem to some extent by choosing appropriate cutoffs and anisotropy @xcite . 2 . @xmath85 @xcite . now all hopping terms in ( [ eq : kappa_action ] ) have projection matrices @xmath86 , and the anisotropic nature appears only in @xmath46 and @xmath87 . but now , if one considers what happens to the conditions when @xmath88 while @xmath89 , then @xmath90 \\ c_e(\xi , m_0a_\tau ) & = & \half\left[1 + \xi\left(1 + \half m_0a_\sigma + { \textstyle\frac{1}{8}}(m_0a_\sigma)^2\right ) \right ] \\ d_1(\xi , m_0a_\tau ) & = & \frac{(m_0a_\sigma)^2}{16m_0a_\tau } + \quarter(1-\xi ) \end{aligned}\ ] ] keeping terms of order @xmath91 and @xmath92 . clearly , the continuum limit sets in only when @xmath26 . even then @xmath93 and @xmath52 are non - zero already at the tree - level . an advantage is that the splitting between the physical states and the edge of the brillouin zone is larger than in case ( i ) . + in passing we mention that refs . @xcite take @xmath94 which agrees with the taylor expansion ( [ ii - zeta ] ) , but does not agree with the full mass dependence ( [ zeta ] ) . the denominator of this expression also is of the form that reaches the continuum limit only when @xmath95 . it is instructive to examine the difference between the two conditions on @xmath93 by considering the full mass dependence of @xmath46 . figure [ fig : zeta_tree ] plots the right - hand side of ( [ zeta ] ) against @xmath96 , for several values of @xmath24 and the two choices @xmath82 and @xmath85 . the mass in spatial lattice units , @xmath97 is chosen not because it is a natural variable , but because one usually would first fix the spatial lattice spacing @xmath3 so that @xmath13 is small enough , while @xmath18 . one would then choose the anisotropy @xmath24 to make @xmath17 small . for example , let us consider the charmed quark on a lattice with @xmath98 gev . the quark mass in spatial lattice units is @xmath99 , so if @xmath100 , then @xmath101 , which seems small . for @xmath82 one finds @xmath102 , which is only 4 percent larger than @xmath103 . in this sense , @xmath101 _ is _ small . on the other hand , for the choice @xmath85 , @xmath104 , which is 20 percent smaller than @xmath105 . even worse , the taylor expansion ( [ ii - zeta ] ) estimates only 0.14 . thus , only with the choice @xmath82 does it seem possible to approximate the improvement coefficients by the small @xmath17 limit . with this choice it seems possible to treat charmed quarks , without appealing to the heavy - quark expansion , at accessible spatial lattice spacings and anisotropy around 34 . lattice artifacts appear under control and there is probably enough room between @xmath51 and the energies at the edge of the brillouin zone , @xmath80 , to accommodate the lowest excitations of the @xmath106 meson . on the other hand , it seems that reasonable choices of @xmath3 and @xmath24 do not exist for treating the @xmath107 quark : @xmath108 remains big , requiring a non - relativistic interpretation along the lines of refs . @xcite . the choice @xmath82 requires no tree - level rotation for the quark field . this is a great simplification for one - loop renormalization . then the quark and anti - quark field operators are multiplied by the factor @xmath109 . with the choice @xmath85 one would have to include the rotation term for a consistent @xmath110 calculation . in the rest of this paper , we therefore focus on @xmath82 . to carry out one - loop perturbative calculations , we must specify the gauge field action as well as the quark action . we begin this section with the gauge field action and remark on the gauge couplings , which , in general , differ for the spatial and temporal components of the gauge field . feynman rules required at one - loop level are summarized in appendix [ sec : feynman ] . the self - energy at the one - loop level is represented by the diagrams in fig . [ fig : diagram](a)(b ) . + ( a ) ( b ) ( c ) [ fig : diagram ] we calculate , as a function of @xmath17 , the one - loop contribution to the quark rest mass and wave function renormalization factor . these quantities require the self - energy and its first derivative with respect to the external momentum @xmath111 , evaluated on the mass shell @xmath112 @xcite . by obtaining the full mass dependence , we can check how the one - loop corrections behave for @xmath18 and @xmath17 small . we also discuss mean field improvement of the self - energy . in the past , the full mass dependence of the one - loop quark self - energy has been obtained for the wilson action on isotropic lattices @xcite for the fermilab action on isotropic lattices @xcite , and for several improved actions with @xmath23 on anisotropic lattices @xcite . we also discuss vertex corrections at the one - loop level , shown in fig . [ fig : diagram](c ) , and present matching factors for the vector and axial vector currents . we again obtain the full mass dependence first , and use it to study the practical situation with @xmath18 and @xmath17 small . in the past , the full mass dependence of the one - loop quark vertex functions has been obtained on isotropic lattices for the wilson action @xcite , the clover action @xcite , and the fermilab action @xcite . we use the standard wilson gauge action on the anisotropic lattice @xcite : @xmath113 , \label{eq : sgauge}\ ] ] where @xmath114 denotes parallel transport around a plaquette in the @xmath115 plane . the bare anisotropy @xmath116 coincides with the renormalized isotropy @xmath24 at the tree - level . in gauge field theory with @xmath117 colors , the coupling @xmath118 is related to the usual bare gauge coupling @xmath119 by @xmath120 . there is a subtlety in the gauge coupling , because the temporal and spatial gluons have different couplings @xcite . one can rewrite @xmath118 and @xmath116 as @xmath121 where @xmath122 and @xmath123 are couplings for spatial and temporal gluons , respectively . although at the one - loop level @xmath124 and @xmath125 need not be distinguished , it is convenient to separate the results for spatial and temporal parts . to improve perturbative series , it is crucial to use renormalized couplings , defined at the momentum scale typical for the process under consideration @xcite . these couplings , and therefore the scales , could be defined separately for spatial and temporal gluons . with this end in mind , we shall present the coefficients of @xmath124 and @xmath125 separately . the relation between the rest mass to the self - energy is @xcite @xmath126a_\tau , \label{rest mass}\ ] ] where @xmath127 and the self - energy @xmath128 is the sum of all one - particle irreducible two - point diagrams . the formula ( [ rest mass ] ) is valid for all masses and at every order in perturbation theory . since it is obtained from the pole position , the rest mass is infrared finite and gauge independent at every order in perturbation theory @xcite . we write the perturbation series as @xmath129}(p_4,\vec{p } ; m_0),\ ] ] where we explicitly specify the bare quark mass @xmath1 . the quark is massless ( @xmath130 ) when the bare mass is tuned to @xmath131.\ ] ] it is more convenient to introduce a subtracted bare mass @xmath132 , which vanishes for a massless quark . then the formula for the rest mass becomes @xmath133a_\tau , \label{m1renorm}\ ] ] where @xmath134 = \sigma(p_4,\vec{p};m_0 ) - m_{0c}.\ ] ] in developing the perturbation series , now @xmath135 is treated independently from @xmath136 . the perturbative series for @xmath51 is @xmath137 } + \sum_{l=1}^\infty g^{2l } m_1^{[l ] } .\ ] ] from ( [ m1renorm ] ) @xmath138 } & = & a_\tau^{-1 } \ln ( 1+m_0a_\tau ) , \\ m_1^{[1 ] } & = & - \frac{\tr[p_+\bar{\sigma}^{[1]}(im_1,\veg{0};m_0)]}% { 1+m_0a_\tau}. \label{eq : mass_renorm2}\end{aligned}\ ] ] in evaluating @xmath139}$ ] one may disregard the distinction between @xmath135 and @xmath1 , because @xmath140 starts at one - loop order . to show the mass dependence it is convenient @xcite to introduce the multiplicative renormalization factor @xmath141 defined by , @xmath142}a_\tau.\ ] ] from @xmath143}=m_1^{[1]}a_\tau/\tanh m_1^{[0]}a_\tau$ ] one can then remove the anomalous dimension by writing @xmath144 } = c_f \left[c^{[1 ] } - \frac{3}{16\pi^2}\ln(m_1 ^ 2 a_{\tau}^2)\right].\ ] ] numerical results for @xmath145}a_\tau$ ] and @xmath146}$ ] are given in sec . [ sec : result ] . the all orders formula for the wave function renormalization factor is @xcite @xmath147a_\tau \label{z2}\ ] ] where @xmath148 in view of the mass dependence , we write @xmath149},\ ] ] so that the @xmath150}$ ] are only mildly mass dependent . this definition of @xmath150}$ ] is slightly different from that of ref . @xcite for @xmath151 . the wave function renormalization factor is infrared divergent and gauge dependent . therefore we express the one - loop term as @xmath152 } & = & ( 1+m_0a_{\tau})^{-1 } \tr[p_+(\bar{\sigma}^{[1]}+\dot{\sigma}^{[1 ] } ) ] , \\ & = : & w^{[1 ] } + l^{[1 ] } , \label{eq : wl_latt}\end{aligned}\ ] ] where @xmath153}$ ] and @xmath154}$ ] are the infrared finite and singular parts , respectively . the infrared divergence does not depend on the ultraviolet regulator , so it is the same as in the continuum theory . we define @xmath154}$ ] by the continuum expression . for a massive quark ( @xmath155 ) in feynman gauge , @xmath156 } = \frac{c_f}{16\pi^2 } \left[- \frac{9}{2 } + 3 \ln(m^2 a_{\tau}^2 ) - 2 \ln ( \lambda^2a_{\tau}^2)\right ] , \label{eq : z2h_latt}\ ] ] which we use for the heavy quark . in particular , @xmath157}$ ] is defined by combining ( [ eq : wl_latt ] ) and ( [ eq : z2h_latt ] ) with @xmath158}$ ] . for a massless quark ( @xmath159 , @xmath160 ) , the mass singularity seen in ( [ eq : z2h_latt ] ) can still be regulated by the gluon mass . in feynman gauge , @xmath161 } = \frac{c_f}{16\pi^2 } \ln ( a_{\tau}^2\lambda^2 ) , \label{eq : z2l_latt}\ ] ] which we use for the light quark . here [ @xmath163 for su(3 ) ] . because the infrared and mass singularities have been subtracted consistently , we should ( and do ) find @xmath164}=w_l^{[1]}$ ] . numerical results for @xmath157}$ ] and @xmath165}$ ] are in sec . [ sec : result ] . mean field improvement @xcite has been employed extensively in monte carlo work to improve tree - level estimates of couplings . the approximation works , because most of the one - loop coefficients can be traced , via tadpole diagrams , to a mean field . on the anisotropic lattice , the mean field values of the link variables can be defined individually for the temporal and the spatial links . we denote them by @xmath166 and @xmath167 respectively . then mean field improvement is achieved by replacing the link variables with @xcite @xmath168 with mean field improvement , the one - loop counter - terms of @xmath169 and @xmath170 must be removed from perturbative coefficients . here we consider generically the @xmath171 contributions from the mean field to the self - energy . from the feynman rules in appendix [ sec : feynman ] , the self - energy and its first derivative with respect to @xmath111 , on the mass shell @xmath172 , are @xmath173}(im_1,\veg{0 } ) & = & - g^2 u^{[1]}_{\sigma } 3 r\zeta - g^2 u^{[1]}_{\tau } e^{m_1^{[1]}a_\tau } \\ \bar{\sigma}_{\rm mf}^{[1]}(im_1,\veg{0 } ) & = & - g^2 u^{[1]}_{\tau } m_0a_\tau\\ \dot{\sigma}_{\rm mf}^{[1]}(im_1,\veg{0 } ) & = & + g^2 u^{[1]}_{\tau } e^{m_1^{[1]}a_\tau } \end{aligned}\ ] ] then , the mean field contribution to the rest mass is @xmath174}a_\tau = \frac{m_0a_\tau}{1+m_0a_\tau } u^{[1]}_{\tau},\ ] ] and to the wave function renormalization factor @xmath175 } = \frac{1}{1+m_0a_\tau } u^{[1]}_{\tau},\ ] ] which holds for massive and massless ( @xmath176 ) quarks . the explicit values of @xmath177}$ ] and @xmath178}$ ] depend on the definition of the mean field . since one can easily incorporate the contributions from the mean field improvement to the one - loop coefficients , we do not employ a specific scheme and quote only the contributions from the loop integrations . to obtain improved matrix elements , operators also must be improved @xcite . as discussed in sec . [ sec : formulation ] , with the choice @xmath82 only the multiplicative factor @xmath179 is required at the tree - level . in particular , with @xmath82 no new dimension - four operator is needed to achieve tree - level improvement . at higher loop order the counterpart of the mass - dependent factor comes from the quark self - energy through the wave function factor , as seen in ( [ z2 ] ) , and dimension - four terms are needed . because the tree - level rotation coefficient @xmath52 vanishes as @xmath180 , we consider here currents of the form @xmath181 where @xmath182 and @xmath183 are the light and heavy quark fields respectively . we consider the vector and axial vector currents , so the the @xmath184 matrix @xmath185 is one of @xmath186 ( @xmath187 ) , @xmath188 ( @xmath189 ) , @xmath190 ( @xmath191 ) , and @xmath192 ( @xmath193 ) . we seek the matching factors @xmath194 such that @xmath195 has the same matrix elements ( for @xmath196 ) as the continuum bilinear . these matching factors are composed of two parts : the wave function of each quark field and the correction to the vertex . since the former is already obtained in previous subsection , here we discuss the vertex corrections . the vertex function @xmath197 , which is the sum of one - particle irreducible three - point diagrams , can be expanded @xmath198}.\ ] ] as with the wave function renormalization , @xmath197 is gauge dependent and suffers from infrared and mass singularities . for the one - loop term we again subtract of the divergent part , @xmath199 } = v_{\gamma}^{[1]}+l_{\gamma}^{[1]},\ ] ] where , in feynman gauge , @xmath200 } = \frac{c_f}{16\pi^2 } \left(-\half ( hg-1)-\ln(\lambda^2 a_{\tau}^2)\right).\ ] ] the constants are again taken from the continuum expression , so @xmath201 for temporal components of the currents and @xmath202 for spatial components . the sought - after matching factor @xmath194 is simply the ratio of the lattice and continuum radiative corrections : @xmath203^{\rm cont}}{% \left[z_{2h}^{1/2}\lambda_\gamma z_{2l}^{1/2}\right]^{\rm lat}}.\ ] ] in view of the mass dependence , we write @xmath204},\ ] ] so that the @xmath205}$ ] are only mildly mass dependent . at the one - loop level we have consistently defined the finite lattice parts so that @xmath206 } = - \left ( \half w_h^{[1 ] } + v_{\gamma}^{[1 ] } + \half w_l^{[1 ] } \right ) \label{eq : z}\ ] ] is the desired one - loop coefficient of the matching factor . it is gauge invariant and independent of the scheme for regulating the infrared and ( light - quark ) mass singularities . numerical results for @xmath207}$ ] and @xmath208}$ ] are in sec . [ sec : result ] . in this section we present our results for the one - loop coefficients . they are obtained numerically using the monte carlo integration program bases @xcite . we give one - loop terms for the rest mass , _ i.e. _ , @xmath145}$ ] and @xmath146}$ ] ; for the infrared - finite parts of the wave function renormalization factors and the vertex functions , _ i.e. _ , @xmath165}$ ] , @xmath157}$ ] and @xmath207}$ ] ; and for the currents matching factors @xmath208}$ ] . in this section we are concerned with zero three - momentum , so for brevity we set the temporal lattice spacing @xmath209 . when the spatial lattice spacing is needed , we use the anisotropy @xmath24 . the spatial and the temporal parts of @xmath145}$ ] are listed separately , namely @xmath210 } \to g^2_{\tau}m^{[1]}_{1\tau } + g^2_{\sigma}m^{[1]}_{1\sigma},\ ] ] so that one could use different ( improved ) couplings in a practical evaluation of the perturbative rest mass . on the other hand , for the other quantities we show the combined values of the spatial and the temporal parts , because we are interested mostly in seeing how they behave when @xmath211 while @xmath212 small . the spatial and the temporal parts of @xmath145}$ ] , @xmath213}$ ] and @xmath214}$ ] respectively , are listed in table [ tab : mass ] for a range of @xmath215 at four values of @xmath24 : 1 , 2 , 3 and 4 . .the one - loop correction to the rest mass , @xmath216}$ ] , for various values of @xmath135 at four values of @xmath24 . [ cols="^,<,<,<,<,<",options="header " , ] the magnitude of the one - loop correction decreases as @xmath24 increases , in all cases . figures [ fig : za ] and [ fig : zv ] are the most important results of this section . they are relevant to phenomenological applications to charm physics . moreover , these results test , at the one - loop level , whether the matching factors are well - behaved in the interesting region with small @xmath135 but @xmath211 . for this reason we have plotted them , as in fig . [ fig : zeta_tree ] , not against @xmath135 but @xmath217 . by inspection of figs . [ fig : za ] and [ fig : zv ] , one can see that the small @xmath135 taylor series continues to be a good approximation to the full mass dependence for all @xmath218 , for @xmath219 . had we found a stronger mass dependence ( like that of @xmath220}$ ] for @xmath85 ) , one would begin to doubt the feasibility of the ideas laid out in ref . @xcite also for our choice @xmath82 . for @xmath221 , our results should reproduce previous calculations on isotropic lattices . in the case of the mass and the wave function renormalization , we ( independently ) reproduced the full mass dependence of ref . @xcite . for the matching factors of the currents , only the result for the massless quark is ( independently ) available @xcite , and we find agreement . in conclusion , in the whole region of @xmath135 and @xmath24 we surveyed , we found good behavior connecting the continuum limit with the region of practical interest . for charmed hadrons , the target region of the heavy quark mass @xmath135 is around 0.10.3 on lattices with anisotropy @xmath24=34 . the required one - loop coefficients of the renormalization factors are easily obtained by interpolating the values in the tables using , for example , spline interpolation . in this paper , we have studied the @xmath56 improvement of wilson quarks on anisotropic lattices . at the tree - level we find that a certain choice of the parameters , @xmath82 @xcite , is well - behaved in the region of practical interest for charmed hadrons , namely @xmath222 , while @xmath223 is small . on the other hand , with a different choice , @xmath85 @xcite , continuum behavior is reached only for @xmath224 . with this latter choice a non - relativistic interpretation @xcite is still possible , but a mass - independent renormalization , which was proposed in ref . @xcite , is obstructed . the choice @xmath82 also simplifies tree - level @xmath56 improvement . the action does not require separate temporal and spatial hopping parameters . the currents require mass - dependent matching factors , but no intrinsically dimension - four terms . we therefore have started to examine the behavior of this choice at the one - loop level . we have computed the one - loop contributions to the rest mass and to the matching factors of the vector and axial vector currents . the matching factors depend significantly on @xmath24 . a more critical observation is that they are well approximated by taylor expansions @xmath225 for @xmath226 and @xmath227@xmath228 . this region encompasses the one suitable for the charmed quark with currently available computer resources . there are several issues that remain to be studied . the first is to compute the one - loop corrections to the ratio of hopping parameters @xmath46 , the clover coefficients @xmath32 and @xmath47 , and dimension - four terms in the currents . the calculation of @xmath46 is especially difficult , because it requires the one - loop kinetic mass @xmath229}$ ] . as at the tree - level , it is crucial to compute the full mass dependence , so one can check whether low - order taylor expansions work well for @xmath222 . only with the full mass dependence can one check whether @xmath24 , which comes with the couplings in the action , and @xmath223 , which also comes from the on - shell condition , come together to form @xmath230 . if not , then one could proceed with a non - perturbative calculation of @xmath57 , @xmath58 , @xmath59 , @xmath60 , etc . a more practical problem is to define renormalized couplings . the scale - setting scheme of brodsky , lepage , and mackenzie ( blm ) is usually a good way to absorb the dominant part of two- and higher - order contributions @xcite . on an anisotropic lattice , it may make sense to define separate scales for temporal and spatial gluons . these results are of interest in any case : even if anisotropic lattice calculations require a non - relativistic interpretation for heavy quarks , anisotropy remains a useful tool for improving the signal - to - noise ratio . finally , after these problems are resolved , it will be important to combine the results with numerical simulation data to obtain the matrix elements relevant to experimental measurements of charmed hadrons . the authors would like to thank shoji hashimoto and masataka okamoto for useful discussions . a.s.k . would like to thank akira ukawa and the center for computational physics for hospitality while this work was being completed . is supported by the center - of - excellence ( coe ) program at research center for nuclear physics , osaka university . is supported by the grant - in - aid of the ministry of education ( no . 12640279 ) . fermilab is operated by universities research association inc . , under contract with the u.s . department of energy . + ( a ) @xmath231 ( b ) @xmath232 + the feynman rules for perturbative calculation are shown the same as in ref . @xcite except for two points . one is that the @xmath32 is replaced by @xmath233 as was mentioned in sec . [ sec : qaction ] and the other is the gluon propagator . with a gauge fixing term that is symmetric under exchange of temporal and spatial axis , the free propagator of the gauge field , in feynman gauge , is @xmath234 @xmath235 where we replaced the anisotropy parameter @xmath116 with the tree - level value @xmath24 . a fictitious gluon mass @xmath236 is introduced to regulate infrared divergences . the feynman diagram for the counter - term from the mean field is obtained by expanding @xmath237 } + o(g^4 ) , \\ u_{\sigma } & = & 1 + g^2 u_{\sigma}^{[1 ] } + o(g^4).\end{aligned}\ ] ] with the replacement of the link variables as eq . ( [ eq : mfreplace ] ) in the action ( [ eq : mass_action ] ) , the feynman rules required for the the one - loop calculation are @xmath238 } [ \gamma_4 i \sin p_4 - \cos p_4 ] \\ { \cal v}^{\rm ( mf)}_{\sigma}(p ) & = & g^2 u_{\sigma}^{[1 ] } \sum_{j } [ \zeta \gamma_j i \sin p_j - r \zeta \cos p_j ] \end{aligned}\ ] ] with the diagrams ( a ) and ( b ) in fig . [ fig : feynman ] , respectively . in the following , we show the explicit representations of the self - energy and the vertex correction . in order to simplify the expressions , we introduce the following abbreviations : @xmath239 @xmath240 = \sum_{i=1}^{3 } s_{i}^2 \check{c}_{i}^2.\ ] ] to reduce the volume of notation , we also define @xmath241 which always appear in these combinations . to reduce the dirac matrix structure , it is convenient to introduce @xmath242 , with the upper ( lower ) sign for massive quarks ( anti - quarks ) on the external leg , and @xmath243 and @xmath244 , defined by @xmath245 with an implied sum on @xmath246 . the following quantities are convenient for representing the one - loop expressions below and in our integration programs : @xmath247 and @xmath248 the symbols with superscript @xmath249 are essentially massless versions of those with superscript @xmath250 . from the feynman rules in appendix [ sec : feynman ] , the contributions to the self - energy from the rainbow diagram , fig . [ fig : diagram](a ) , are @xmath251}(im_1,{\bf 0 } ) + g^2_{\tau } \sigma_{a\tau}^{[1]}(im_1,{\bf 0 } ) + o(g^4 ) , \\ \sigma_{a\sigma}^{[1]}(im_1,{\bf 0 } ) & = & c_f \int \frac{d^4 k}{(2\pi)^4 } s_q(im_1+k_4,\vec{k } ) g_{\sigma}(k ) \left[\check{\vec{c}}^2 \bar{b}^q(e^q)^2 + r^2\zeta^2\check{\vec{s}}^2 b^q \right . \nonumber \\ & & \hspace{1.5 cm } \left . + i r \zeta^2\vec{s}^2 e^q + \left ( i \zeta c'_b e^q + \quarter c_b^{\prime 2 } b^q \right ) \left(\vec{s}^2\check{\vec{c}}^2-[\vec{s}\cdot\check{\vec{c } } ] \right ) \right ] , \\ \sigma_{a\tau}^{[1]}(im_1,{\bf 0 } ) & = & c_f \int \frac{d^4k}{(2\pi)^4 } s_q(im_1+k_4,\vec{k } ) g_{\tau}(k ) \nonumber \\ & & \hspace{2 cm } \times \left [ ( a^{q})^2b^q+\vec{s}^2\left(i\zeta c'_e \check{c}_4 a^q g^q + \quarter c_e^{\prime 2 } \check{c}_4 ^ 2 \bar{b}^q \right ) \right ] . \end{aligned}\ ] ] where @xmath252^{-1}.\ ] ] similarly , the contributions from the tadpole diagram , fig . [ fig : diagram](b ) , are @xmath253}(im_1,{\bf 0 } ) + g^2_{\tau } \sigma_{b\tau}^{[1]}(im_1,{\bf 0 } ) + o(g^4 ) , \\ \sigma_{b\sigma}^{[1]}(im_1,{\bf 0 } ) & = & - \half c_f \int \frac{d^4 k}{(2\pi)^4 } g_{\sigma}(k ) 3r\zeta , \\ \sigma_{b\tau}^{[1 ] } ( im_1,\veg{0 } ) & = & - \half c_f \int \frac{d^4 k}{(2\pi)^4 } g_{\tau}(k ) \left ( g^q\sinh m_1+\cosh m_1 \right).\end{aligned}\ ] ] the derivative of the self - energy with respective to @xmath111 is separated into temporal and spatial contributions @xmath254}(im_1,\veg{0 } ) + g^2_{\tau } \dot{\sigma}_{\tau}^{[1]}(im_1,\veg{0 } ) + o(g^4)\ ] ] the contributions from fig . [ fig : diagram](a ) are @xmath255}(im_1,\veg{0 } ) & = & -ic_f \int \frac{d^4 k}{(2\pi)^4 } g_{\sigma}(k ) \left [ -2 s_q(im_1+k_4,\vec{k})^2 s_4 \left(1 + 2 r \zeta \check{\vec{s}}^2 + m_{0}\right ) \right.\\ & & \hspace{-1 cm } \times \left\ { \check{\vec{c}}^2 ( e^q)^2 \bar{b}^q + r^2\zeta^2\check{\vec{s}}^2 b^q + i r \zeta^2 \vec{s}^2 e^q + \left ( i \zeta c'_b e^q + \quarter c_b^{\prime 2 } b^q \right ) \left ( \vec{s}^2 \check{\vec{c}}^2 - [ \vec{s}\cdot\check{\vec{c } } ] \right ) \right\ } \nonumber \\ & & \hspace{-1 cm } \left . + s_q(im_1+k_4,\vec{k } ) \left\ { \check{\vec{c}}^2 ( e^q)^2 \bar{j}^q + r^2 \zeta^2 \check{\vec{s}}^2 j^q + \quarter c_b^{\prime 2 } ( \vec{s}^2\check{\vec{c}}^2 - [ \vec{s}\cdot\check{\vec{c}}])j^q \right\ } \right ] \nonumber \\ \dot{\sigma}_{a\tau}^{[1]}(im_1,\veg{0 } ) & = & -ic_f \int \frac{d^4 k}{(2\pi)^4 } g_{\tau}(k ) \left [ -2 s_q(im_1+k_4,\vec{k})^2 s_4 \left(1 + 2 r \zeta \check{\vec{s}}^2 + m_{0}\right ) \right.\\ & & \hspace{-1 cm } \times \left\ { ( a^q)^2b^q+\vec{s}^2\left(i\zeta c'_e \check{c}_4a^qg^q + \quarter c_e^{\prime 2 } \check{c}_4 ^ 2 \bar{b}^q \right ) \right\ } \nonumber \\ & & \hspace{-1 cm } + s_q(im_1+k_4,\vec{k } ) \left . \left\ { -2i(a^q)^2 b^q g^q + \zeta c'_e\check{c}_4\vec{s}^2 a^q + \quarter c_e^{\prime 2}\check{c}_4 ^ 2\vec{s}^2\bar{j}^q + ( a^q)^2 j^q \right\ } \right ] . \nonumber \end{aligned}\ ] ] there is only one contribution from fig . [ fig : diagram](b ) @xmath256}(im_1,{\bf 0 } ) = -\ihalf c_f \int \frac{d^4 k}{(2\pi)^4 } g_{\tau}(k)[g^q \cosh m_1 + \sinh m_1 ] , \ ] ] with a temporal gluon . the vertex function is split as follows : @xmath257 } = g^2_{\sigma } \lambda^{[1]}_{\gamma \sigma } + g^2_{\tau } \lambda^{[1]}_{\gamma \tau } , \ ] ] and the contributions are @xmath258}_{\gamma \sigma } & = & c_f \int \frac{d^4 k}{(2\pi)^4 } s_q(im_1+k_4,\vec{k } ) s_q(k ) g_{\sigma}(k ) \left [ \frac{1}{3}\check{\vec{c}}^2(h - g)e^q\overline{b}^qe^q\overline{b}^q \right . \nonumber \\ & & \hspace{0.8 cm } + \,\frac{i}{6}\zeta(h - g ) \left\ { c'_b(\vec{s}^2\check{\vec{c}}^2-[\vec{s}\cdot\check{\vec{c } } ] ) + r\zeta\vec{s}^2\right\ } ( e^q\overline{b}^q+\overline{b}^qe^q ) \nonumber\\ & & \hspace{0.8 cm } + \,\ihalf \zeta \left\ { \frac{1}{6}c'_b ( \vec{s}^2\check{\vec{c}}^2-[\vec{s}\cdot\check{\vec{c } } ] ) \left ( ( h - g)^2 - 3 \right)+r\zeta\vec{s}^2\right\ } ( e^q b^q+b^q e^q ) \nonumber \\ & & \hspace{0.8 cm } + \,\left\{\frac{1}{24}c_b^{\prime 2 } ( \vec{s}^2\check{\vec{c}}^2-[\vec{s}\cdot\check{\vec{c } } ] ) \big((h - g)^2 - 3\big ) + r^2\zeta^2\check{\vec{s}}^2\right\ } b^q b^q \nonumber\\ & & \hspace{0.8 cm } -\,\zeta^2 \left\ { \frac{1}{6}(\vec{s}^2\check{\vec{c}}^2 - [ \vec{s}\cdot\check{\vec{c}}])\left ( ( h - g)^2 - 3 \right ) + [ \vec{s}\cdot\check{\vec{c } } ] \right\ } e^q e^q \nonumber \\ & & \hspace{0.8 cm } \left . -\,\frac{1}{12}c_b^{\prime 2}\zeta^2\vec{s}^2 ( \vec{s}^2\check{\vec{c}}^2-[\vec{s}\cdot\check{\vec{c}}])(h - g ) -\frac{1}{3}r^2\zeta^4\vec{s}^2\check{\vec{s}}^2(h - g ) \right ] \gamma , \\ \lambda^{[1]}_{\gamma \tau } & = & c_f\int \frac{d^4 k}{(2\pi)^4 } s_q(im_1+k_4,\vec{k } ) s_q(k ) g_{\tau}(k ) \times \nonumber \\ & & \hspace{0.6 cm } \left [ \big(a^q b^q+\ihalf \zeta c'_e \check{c}_4\vec{s}^2 g\big ) \big(a^q b^q+\ihalf \zeta c'_e \check{c}_4\vec{s}^2 g^q\big ) \right . \nonumber\\ & & \hspace{0.8cm}\left . + \frac{1}{3}(h - g)\vec{s}^2 \big(i\zeta a^q+\half c'_e \check{c}_4\overline{b}^qg\big ) \big(i\zeta a^q+\half c'_e \check{c}_4\overline{b}^qg^q\big ) \right ] \gamma.\end{aligned}\ ] ] 99 f. karsch , nucl . b * 205 * ( 1982 ) 285 ; + g. burgers , f. karsch , a. nakamura , and i. o. stamatescu , _ ibid . _ * 304 * ( 1988 ) 587 . c. j. morningstar and m. j. peardon , phys . d * 56 * ( 1997 ) 4043 ; * 60 * ( 1999 ) 034509 . j. fingberg , phys . b * 424 * ( 1998 ) 343 ; + k. j. juge , j. kuti , and c. j. morningstar , nucl . * 63 * ( 1998 ) 543 ; phys . rev . * 82 * ( 1999 ) 4400 ; nucl . * 83 * ( 2000 ) 304 ; + i. t. drummond , r. r. horgan , t. manke , and h. p. shanahan , nucl . * 73 * ( 1999 ) 336 ; + t. manke _ et al . _ [ cp - pacs collaboration ] , phys . * 82 * ( 1999 ) 4396 ; nucl . suppl . * 86 * ( 2000 ) 397 ; + i. t. drummond _ et al . _ , phys . b * 478 * ( 2000 ) 151 ; + s. collins _ et al . _ , hep - lat/0101019 . t. r. klassen , nucl . b * 509 * ( 1998 ) 391 ; nucl . * 73 * ( 1999 ) 918 . g. p. lepage and b. a. thacker , nucl . . suppl . * 4 * , 199 ( 1987 ) ; + b. a. thacker and g. p. lepage , phys . d * 43 * , 196 ( 1991 ) ; + g. p. lepage , l. magnea , c. nakhleh , u. magnea , and k. hornbostel , _ ibid . _ * 46 * , 4052 ( 1992 ) . a. x. el - khadra , a. s. kronfeld , and p. b. mackenzie , phys . d * 55 * ( 1997 ) 3933 . a. s. kronfeld , phys . d * 62 * ( 2000 ) 014505 . k. jansen _ et al . _ , phys . b * 372 * ( 1996 ) 275 ; + m. lscher , s. sint , r. sommer , p. weisz , and u. wolff , nucl . b * 491 * ( 1997 ) 323 . de forcrand _ et al . _ [ qcd - taro collaboration ] , nucl . * 83 * ( 2000 ) 411 . t. umeda , r. katayama , o. miyamura , and h. matsufuru , hep - lat/0011085 ( to appear in int . j. mod . p. chen , hep - lat/0006019 ; + p. chen , x. liao , and t. manke , nucl . * 94 * ( 2001 ) 342 . a. ali khan _ et al . _ [ cp - pacs collaboration ] , nucl . * 94 * ( 2001 ) 325 . s. groote and j. shigemitsu , phys . d * 62 * ( 2000 ) 014508 . k. g. wilson , in _ new phenomena in subnuclear physics _ , edited by a. zichichi ( plenum , new york , 1977 ) . b. sheikholeslami and r. wohlert , nucl . b * 259 * ( 1985 ) 572 . m. okamoto , unpublished . g. mertens , a. s. kronfeld , and a. x. el - khadra , phys . d * 58 * ( 1998 ) 034505 . y. kuramashi , phys . d * 58 * ( 1998 ) 034507 . ishikawa , s. aoki , s. hashimoto , h. matsufuru , t. onogi , and n. yamada , nucl . . suppl . * 63 * , ( 1998 ) 344 ; + k .- i . ishikawa , t. onogi , and n. yamada , nucl . * 83 * , ( 2000 ) 301 . j. harada _ _ , to appear . a. s. kronfeld , phys . d * 58 * ( 1998 ) 051501 . g. p. lepage and p. b. mackenzie , phys . d * 48 * ( 1993 ) 2250 . g. heatlie _ et al . _ , nucl . b * 352 * ( 1991 ) 266 ; + m. lscher , s. sint , r. sommer , and p. weisz , nucl . b * 478 * ( 1996 ) 365 . s. kawabata , comput . * 88 * ( 1995 ) 309 . s. aoki , k. nagai , y. taniguchi , and a. ukawa , phys . d * 58 * ( 1998 ) 074505 . s. j. brodsky , g. p. lepage and p. b. mackenzie , phys . d * 28 * ( 1983 ) 228 .

we investigate the non - perturbative renormalization group behavior of the gauge coupling constant using a truncated form of the functional flow equation for the effective average action of the yang - mills gravity system . we find a non - zero quantum gravity correction to the standard yang - mills beta function which has the same sign as the gauge boson contribution . our results fit into the picture according to which quantum einstein gravity ( qeg ) is asymptotically safe , with a vanishing gauge coupling constant at the non - trivial fixed point .

0910.4938

recently a lot of efforts went into the computation of gravitational corrections to the beta function of the running yang - mills coupling constant . robinson and wilczek @xcite , in an effective field theory setting , obtained a non - zero correction at the one - loop level . it has the same negative sign as the familiar term already present in absence of gravity , and so it would render even pure abelian theories asymptotically free . after pietrykowski @xcite had realized that this result is gauge fixing dependent , toms @xcite reanalyzed the problem using the vilkovisky - dewitt method . in a manifestly gauge invariant as well as gauge fixing independent formulation of the effective action he finds that the quantum gravity contributions to the running charge vanish . all of these computations employ the dimensional regularization scheme . in @xcite , ebert , plefka and rodigast pointed out that its use might be problematic since it is insensitive to quadratic divergences , and it is precisely such quadratic divergences that are responsible for the non - zero result obtained in @xcite . using a cutoff regularization instead they found that all gravitational quadratic divergences cancel so that there is again no correction to the beta function . thereafter tang and wu @xcite argued that the use of a cutoff regularization is not permissible here since it does not respect gauge invariance . performing a calculation in a scheme which both retains quadratic divergences and preserves gauge invariance ( `` loop regularization '' ) they obtained a non - zero gravitational correction to the one - loop beta function . furthermore , toms @xcite demonstrated that , with a cosmological constant included , also dimensional regularization yields a non - vanishing gravitational correction , albeit of a different type . in the present paper we shall analyze the running of the gauge coupling constant in the framework of the asymptotic safety approach to quantum gravity @xcite . contrary to the calculations mentioned above it considers metric gravity not merely an effective but rather a fundamental quantum field theory , with the continuum limit taken at a non - trivial renormalization group ( rg ) fixed point @xcite . instead of perturbation theory , the main tool will be the gravitational average action @xcite and a suitably truncated form of the associated functional rg equation ( frge ) . originally developed for matter field theories @xcite the effective average action turned out an ideal tool for investigating the rg flow of quantum einstein gravity ( qeg)@xcite and exploring its potential physics implications @xcite . the purpose of this paper is twofold : first , we develop a general framework for both exact and approximate ( `` truncated '' ) investigations of the yang - mills gravity system by means of a _ gauge invariant _ running effective action . in particular we shall see that because of the semi - direct product structure of the pertinent gauge group there arises a subtlety as for the appropriate construction of the ghost action . second , we use the resulting framework in order to find the rg flow in a simple truncation of the space of actions which , however , is general enough to allow for an approximate determination of the beta function of the yang - mills coupling constant in presence of quantized gravity . the remaining sections of this paper are organized as follows . in section 2 we describe the general setup of the effective average action for a yang - mills field coupled to gravity , the corresponding frge in particular . then , in section 3 , we discuss the problem of background covariant ghost actions and explain its solution . the explicit computation of the running gauge coupling is performed in section 4 , and section 5 contains a summary as well as a brief discussion of various immediate applications of our results . schematically the dynamics of the yang - mills gravity system is governed by the path integral @xmath0}\end{aligned}\ ] ] here @xmath1 and @xmath2 are the quantum metric and the quantum gauge field , respectively , and @xmath3 denotes the bare action . as usual , both of these fields are supposed to transform tensorially with respect to diffeomorphisms , @xmath4 . in addition , @xmath2 defines a connection with respect to yang - mills gauge transformations , @xmath5 . denoting the vector field that generates the diffeomorphism by @xmath6 and the parameter of the yang - mills transformation by @xmath7 , we have ( @xmath8 denotes the lie derivative along @xmath6 ) : @xmath9 from now on we will assume the yang - mills gauge group to be @xmath10 , so @xmath11 runs from @xmath12 to @xmath13 , @xmath14 are the associated structure constants . we demand @xmath3 to be invariant under both @xmath4 and @xmath5 . employing the background formalism , the dynamical fields are decomposed according to @xmath15 with fixed , but arbitrary background configurations @xmath16 and @xmath17 and fluctuations @xmath18 and @xmath19 @xcite . assuming a translational invariant measure , the fluctuation fields will replace the full quantum fields as the variables of integration in ( [ pi - sketchy ] ) . there are now two possibilities to realize the gauge transformations @xmath4 and @xmath5 at the level of the background decomposition : * the _ background gauge transformations _ @xmath20 and @xmath21 are defined such that under diffeomorphisms all the fields transform tensorially , i.e. @xmath22 with respect to yang - mills transformations , the background gauge field transforms as a connection and the fluctuation transforms homogeneously : @xmath23 * on the other hand , we can define _ true gauge transformations _ @xmath24 and @xmath25 by requiring that these shall only affect the fluctuations , not the background fields : @xmath26 the crucial idea is to choose a gauge fixing term @xmath27 that breaks only the true gauge invariance but retains background gauge invariance . it gives rise to an associated ghost action @xmath28 in the usual way . introducing ghost fields @xmath29 and @xmath30 for the diffeomorphisms and @xmath31 and @xmath32 for the yang - mills transformations , respectively , we arrive at the following path integral which depends parametrically on the background fields : @xmath33 - s^{\rm gf } - s_{\rm gh}}\end{aligned}\ ] ] in order to set up a functional rg equation , we follow the well - known construction of the effective average action @xcite and add a higher derivative ir cutoff term @xmath34 that is quadratic in the fluctuations : @xmath35}^{\mu\nu\rho\sigma}h_{\rho\sigma}\nonumber\\ & & + \frac{1}{2}\int\:{\rm d}^d x\:\sqrt{\bar{g}}\:a^a_\mu{{\cal r}^{\rm ym}_k [ \bar{g } , \bar{a}]}^{a\mu b\nu}a^b_\nu\nonumber\\ & & + \sqrt{2 } \int\:{\rm d}^d x\:\sqrt{\bar{g}}\:\big(\bar{\cal{c } } , \bar{\sigma}\big){{\cal r}^{\rm gh}_k [ \bar{g } , \bar{a}]}\bigg(\!\!\begin{array}{c } { \cal c } \\[-0.2 cm ] \sigma \end{array}\!\ ! \bigg)\end{aligned}\ ] ] with @xmath36 and @xmath37 denoting newton s constant ( see below ) . furthermore , adding appropriate source terms enables us to easily compute expectation values of the quantum fields . with the notation @xmath38 the background decomposition ( [ decomp ] ) now reads @xmath39 the legendre transformation of the now @xmath40-dependent functional @xmath41 with respect to the sources leads to the effective average action then @xcite : @xmath42 \equiv \gamma_k[g_{\mu\nu } , \bar{g}_{\mu\nu } , a^a_\mu , \bar{a}^a_\mu , \xi^\mu , \bar{\xi}_\mu , \upsilon^a , \bar{\upsilon}^a]\end{aligned}\ ] ] its scale dependence is governed by the functional renormalization group equation ( frge ) @xmath43\end{aligned}\ ] ] where @xmath44 denotes the hessian of @xmath45 with respect to all fluctuation and ghost expectation values , @xmath46 is some suitably chosen generalized laplacian , and @xmath47 is the `` rg time '' . since we deal with two gauge invariances , a remark concerning the notation is in order : we write @xmath48 and @xmath49 for the covariant derivatives that are constructed by means of @xmath50 and @xmath51 , respectively ; the covariant derivative containing both of these connections is denoted by @xmath52 . by adding a bar , we denote their analogues evaluated on the background configurations . since we have to fix two gauge invariances by two gauge conditions , @xmath53 and @xmath54 for the diffeomorphisms and the @xmath10 transformations , respectively , the associated faddeev - popov operator will in general consist of four components . ( here we have already assumed that the diffeomorphism and the @xmath10 gauge condition only involve the metric and the gauge field fluctuation separately . ) the corresponding classical ghost action will then be of the form @xmath55 = - \int\:{\rm d}^d x\:\sqrt{\bar{g}}\:\big(\kappa^{-1}\:\bar{{\cal c}}_\mu \bar{g}^{\mu\nu}\frac{\partial { \rm f}_\nu}{\partial h_{\rho\sigma}}\delta_{\rm d}^{\rm g}({\cal c } ) h_{\rho\sigma}+ \\ & + \kappa^{-1}\:\bar{{\cal c}}_\mu \bar{g}^{\mu\nu}\frac{\partial { \rm f}_\nu}{\partial h_{\rho\sigma}}\delta_{\rm ym}^{\rm g}(\sigma ) h_{\rho\sigma } + \hat{g}\ : \bar{\sigma}^a \frac{\partial { \rm g}^a}{\partial a^b_\mu}\delta_{\rm d}^{\rm g}({\cal c } ) a^b_\mu + \hat g \ : \bar{\sigma}^a \frac{\partial { \rm g}^a}{\partial a^b_\mu}\delta_{\rm ym}^{\rm g}(\sigma ) a^b_\mu \big ) \end{aligned } \label{geistwirk - allg}\ ] ] since we are going to neglect renormalization effects in the ghost sector , the evolution equation for @xmath45 will contain only the classical ghost action but with the quantum ghost fields replaced by their vacuum expectation values , and the full classical fields @xmath56 and @xmath51 identified with their background configurations @xmath16 and @xmath17 , respectively ; stated differently , in this class of approximations the fluctuations @xmath57 and @xmath58 can be set to zero in @xmath28 even before @xmath59 is computed . looking at the @xmath60-@xmath61 part of the resulting ghost action we encounter a serious problem : employing @xmath62 as gauge condition , it is given by @xmath63\right)_{\bar{\upsilon}\xi } = - \int\:{\rm d}^d x\:\sqrt{\bar{g}}\:\big[\bar{\upsilon}^a \bar{{\cal d}}^{\nu\:ab } \big(\xi^\rho \partial_\rho \bar{a}^b_\nu + ( \partial_\nu \xi^\rho ) \bar{a}^b_\rho\big)\big]\ ] ] here the covariant background derivative @xmath64 acts on the ordinary lie derivative of an @xmath10 background connection ; therefore , this part of the ghost action is _ not _ @xmath21-invariant . in order to resolve this issue , it is useful to look at it from a more abstract point of view . to begin with , we consider an arbitrary functional @xmath65 $ ] of the dynamical fields at hand . at this stage the splitting into fluctuations and background configurations has not yet been performed . an infinitesimal gauge transformation of @xmath66 , considered a scalar functional of its arguments , consists of a diffeomorphism along @xmath6 and an @xmath10 transformation with parameters @xmath7 . it can be implemented as @xmath67 \\ & = f[\gamma , { \cal a } , { \cal c } , \bar{{\cal c } } , \sigma , \bar{\sigma } ] - { \cal w}_{\rm d}(v)f[\gamma , { \cal a } , { \cal c } , \bar{{\cal c } } , \sigma , \bar{\sigma } ] - { \cal w}_{\rm ym}(\lambda)f[\gamma , { \cal a } , { \cal c } , \bar{{\cal c } } , \sigma , \bar{\sigma}]\\ & \hspace{0.5 cm } + { \cal o } ( v^2 , v \lambda , \lambda^2 ) \end{aligned}\ ] ] with the corresponding ward operators generating diffeomorphisms , @xmath68 and yang - mills gauge transformations : @xmath69 in these integrals the measure factor @xmath70 cancels against a similar one which would render the functional derivatives tensorial . we define the set of all invariant functionals by @xmath71 computing the algebra of these operators leads to @xmath72 & = & { \cal w}_{\rm d}([v_1 , v_2])\\ \label{ward - l - l}[{\cal w}_{\rm ym}(\lambda_1 ) , { \cal w}_{\rm ym}(\lambda_2 ) ] & = & { \cal w}_{\textrm{ym}}(f \lambda_1\lambda_2)\\ \label{ward - d - l}[{\cal w}_{\rm d}(v ) , { \cal w}_{\rm ym}(\lambda ) ] & = & { \cal w}_{\rm ym}({\cal l}_v \lambda)\end{aligned}\ ] ] where @xmath73 $ ] denotes the lie bracket and @xmath74 . this algebra implies that the total group of gauge transformations , @xmath75 , has the structure of a semi - direct product of the spacetime diffeomorphisms @xmath76 and the local yang - mills transformations @xmath77 , with the latter forming the invariant subalgebra : @xmath78 . whereas the first two relations ( [ ward - d - d ] ) , ( [ ward - l - l ] ) represent the well - known composition laws of diffeomorphisms and yang - mills gauge transformations , the third relation ( [ ward - d - l ] ) lies at the heart of our problem : diffeomorphisms and local gauge transformations do not commute . instead , they close on the lie derivative of the gauge parameter . in particular , this implies that diffeomorphisms do not map @xmath10 tensors onto @xmath10 tensors . what is called for is an @xmath10 covariantization of the ordinary lie derivative . this is tantamount to a different parametrization of @xmath75 that makes the mixed commutator vanish . this can be achieved by defining _ new diffeomorphisms _ which include a @xmath77 transformation with parameter @xmath79 @xcite : @xmath80 loosely speaking , this amounts to shifting a certain @xmath81-dependent part of @xmath77 into the diffeomorphism sector . the notion of an `` invariant functional '' remains unchanged : @xmath82 the algebra relations receive extra contributions now since the ward operators act on the field dependent parameters of the transformations as well . this leads to an algebra of the desired form ( with @xmath83 ) : @xmath84 & = & \widetilde{{\cal w}_{\rm d}}([v_1 , v_2 ] ) - { \cal w}_{\rm ym}(v_1 v_2\cdot f)\\ \label{ward - l - l - new}[{\cal w}_{\rm ym}(\lambda_1 ) , { \cal w}_{\rm ym}(\lambda_2 ) ] & = & { \cal w}_{\rm ym}(f \lambda_1\lambda_2)\\ \label{ward - d - l - new}[\widetilde{{\cal w}_{\rm d}}(v ) , { \cal w}_{\rm ym}(\lambda ) ] & = & 0\end{aligned}\ ] ] if we now split the dynamical fields into fluctuations and background configurations , we have to decide whether the field dependent transformation parameter in ( [ new - diff ] ) should contain the full or the background gauge field only . since , as already mentioned , the metric and gauge field fluctuations do not enter the final form of @xmath85 anyhow , we may safely opt for the latter already at this point : @xmath86 according to the usual distinction between true and background gauge transformations , we now have to consider two classes of ward operators as well . in the background case , the algebraic relations simply carry over , so we have for the ward operators generating background gauge transformations : @xmath87 & = & \widetilde{\widetilde{{\cal w}^{\rm b}_{\rm d}}}([v_1 , v_2 ] ) - { \cal w}^{\rm b}_{\rm ym}(v_1 v_2\cdot \bar{f})\\ \label{b - ward - l - l}[{\cal w}^{\rm b}_{\rm ym}(\lambda_1 ) , { \cal w}^{\rm b}_{\rm ym}(\lambda_2 ) ] & = & { \cal w}^{\rm b}_{\rm ym}(f \lambda_1\lambda_2)\\ \label{b - ward - d - l}[\widetilde{\widetilde{{\cal w}^{\rm b}_{\rm d}}}(v ) , { \cal w}^{\rm b}_{\rm ym}(\lambda ) ] & = & 0\end{aligned}\ ] ] as for the true gauge transformations , the merely _ background _ field dependent parameter @xmath88 of the compensating @xmath77 transformation is not subject to true gauge transformations . the algebra that generates these transformations can be easily computed therefore by taking advantage of the linearity of commutators . we obtain @xmath89 & = & \widetilde{\widetilde{{\cal w}^{\rm g}_{\rm d}}}([v_1 , v_2 ] ) + { \cal w}^{\rm g}_{\rm ym}(v_1 v_2\cdot \bar{f})\\ \label{g - ward - l - l}[{\cal w}^{\rm g}_{\rm ym}(\lambda_1 ) , { \cal w}^{\rm g}_{\rm ym}(\lambda_2 ) ] & = & { \cal w}^{\rm g}_{\rm ym}(f \lambda_1\lambda_2)\\ \label{g - ward - d - l}[\widetilde{\widetilde{{\cal w}^{\rm g}_{\rm d}}}(v ) , { \cal w}^{\rm g}_{\rm ym}(\lambda ) ] & = & { \cal w}^{\rm g}_{\rm ym}(v\cdot\bar{\nabla } \lambda)\end{aligned}\ ] ] by writing @xmath90 we distinguish these ward operators from the modified one that was defined with respect to the undecomposed fields ; in addition , their notation shall remind us of the fact that the gauge field entered their definition only via its background component @xmath91 . thus the actual theory space on which we can define an rg flow consists of the functionals @xmath92\equiv f[g,\bar g , a,\bar a,\xi,\bar \xi,\upsilon,\bar \upsilon]$ ] in @xmath93 finally , we return to our starting point and compute the @xmath60-@xmath61 part of the ghost action , now with the original true diffeomorphism @xmath94 replaced by its modified counterpart , @xmath95 : @xmath96\right)_{\bar{\upsilon}\xi } & = - \int\:{\rm d}^d x\:\sqrt{\bar{g}}\:\big[\bar{\upsilon}^a \bar{{\cal d}}^{\nu\:ab } \big(\widetilde{\widetilde{\delta_{\rm d}^{\rm g}}}(\xi)a^b_\nu\big)\big]|_{a=0}\\ & = - \int\:{\rm d}^d x\:\sqrt{\bar{g}}\:\big[\bar{\upsilon}^a \bar{{\cal d}}^{\nu\:ab } \big(\xi^\rho \partial_\rho \bar{a}^b_\nu + ( \partial_\nu \xi^\rho ) \bar{a}^b_\rho\\ & \hspace{3 cm } -\big(\partial_\nu(\bar{a}^b_\rho \xi^\rho ) + f^{bcd}\bar{a}^b\bar{a}^d_\rho \xi^\rho\big)\big)\big]\\ & = - \int\:{\rm d}^d x\:\sqrt{\bar{g}}\:\big [ - \bar{\upsilon}^a \bar{{\cal d}}^{\nu\:ab } \bar{f}^b_{\nu\rho } \xi^\rho\big ] \end{aligned}\label{ups - xi - geist - neu}\ ] ] this action is obviously invariant under @xmath77 and background diffeomorphisms @xmath97 for background tensorial ghost expectation values . in this section we explicitly evaluate the frge on a truncated theory space which is general enough to allow for an approximate determination of the beta function for the scale dependent yang - mills coupling @xmath98 . our truncation is given by the following ansatz : @xmath99=\gamma^{\textrm{eh}}_k[g]+\gamma^{\textrm{ym}}_k[g , a]+\gamma^{\textrm{gf}}_k[g-\bar g , a-\bar a;\bar g,\bar a]\\ + s_{\textrm{gh}}[g-\bar g , a-\bar a,\xi,\bar \xi,\upsilon,\bar \upsilon;\bar g,\bar a]\label{ansatz } \end{split}\ ] ] here @xmath100=2 \kappa^2 z_n(k)\int { \textrm{d}^d x\:}\sqrt{g}\,\left(-r(g)+2\bar \lambda(k)\right)\ ] ] is a @xmath40-dependent form of the einstein - hilbert action . the corresponding dimensionful running parameters are the cosmological constant @xmath101 and newton s constant @xmath102 where @xmath103 is a fixed reference value . furthermore , @xmath104=\frac{z_f(k)}{4\:\hat{g}^2_{\textrm{ym } } } \int { \textrm{d}^d x\:}\sqrt{g}\ , g^{\mu\rho}g^{\nu\sigma } f^a_{\mu\nu}f^a_{\rho\sigma}\ ] ] is the standard second - order yang - mills action , with a @xmath40-dependent prefactor @xmath105 though . hence the ( dimensionful , except in d=4 ) running gauge coupling is @xmath106 with some constant @xmath107 . finally , @xmath108=\int { \textrm{d}^d x\:}\sqrt{\bar g}\ , \left ( \frac{z_n(k)}{2 \alpha_{\textrm{d } } } \bar g^{\mu\nu } \textrm{f}_\mu\textrm{f}_\nu + \frac{z_f(k)}{2\alpha_{\textrm{ym } } } \textrm{g}^a\textrm{g}^a\right)\ ] ] implements the gauge fixing conditions for the diffeomorphisms , @xmath109 , and the @xmath10 gauge transformations , @xmath110 . as in @xcite and @xcite we factored out the wave function renormalizations @xmath111 and @xmath112 from the gauge fixing parameters @xmath113 and @xmath114 , respectively . in principle the latter are still @xmath40-dependent but we shall neglect their running here . in fact , later on we set @xmath115 . our choice for the gauge conditions complies with the requirements discussed in the previous section : @xmath116 the resulting ghost action reads , with @xmath117 and @xmath118 still , @xmath119=\\ -\int { \textrm{d}^d x\:}\sqrt{\bar{g}}\,\left(\sqrt{2 } \bar \xi_\mu\left(\bar g^{\mu\rho}\bar g^{\sigma\lambda}\bar d_\lambda\left(g_{\rho\nu}d_\sigma+g_{\sigma\nu}d_\rho\right)-\bar g^{\rho\sigma}\bar g^{\mu\lambda}\bar d_\lambda g_{\sigma\nu}d_\rho\right)\xi^\nu \right . \\ + \bar\upsilon^a \bar g^{\mu\nu } \bar{\mathcal d}_\mu ( \bar f^a_{\rho\nu } \xi^{\rho } + \xi^\rho \partial_\rho a^a_\nu + ( \partial_\nu \xi^\rho)a^a_\rho + f^{abc}\bar{a}^b_\rho \xi^\rho a^c_\nu ) + \bar\upsilon^a\left(\bar g^{\mu\rho}\delta^{ab}\bar{\mathcal d}_\mu \nabla_\rho\right)\upsilon^b\big)\end{gathered}\ ] ] it can be checked that @xmath120 of eq . ( [ sghost ] ) is invariant under background gauge transformations : @xmath121 . while this is true even for non - vanishing fluctuations @xmath122 and @xmath11 , in the present calculation we shall need @xmath28 only for @xmath123 . at this point a remark concerning the expected reliability of this truncation ansatz might be in order . as for its gravitational part , all generalizations of the einstein - hilbert truncation explored during the past decade did not change the qualitative picture it gives rise to , at least close to the non - gaussian fixed point . in the yang - mills sector we retained only the first monomial of a systematic derivative expansion . from the analysis in @xcite without gravity we know that this truncation is not only sufficient to reproduce one - loop perturbation theory exactly , but even approximates the two - loop result for the beta - function with a small error of a few percent . therefore we may expect that this truncation , too , is perfectly sufficient as long as @xmath40 is sufficiently large ( well above dynamically generated `` confinement '' scales , say ) . when we insert the truncation ansatz ( [ ansatz ] ) into the exact frge ( [ frge ] ) the supertrace decomposes into a `` bosonic '' and a ghost contribution : @xmath124-\textrm{tr}\left[\frac{\partial_t \mathcal{r}^{\textrm{gh}}_k\big(\mathcal{z}^{-1}_{\textrm{gh } } s^{(2)}_{\textrm{gh}}\big)}{s^{(2)}_{\textrm{gh}}+\mathcal{r}^{\textrm{gh}}_k\big(\mathcal{z}^{-1}_{\textrm{gh}}s^{(2)}_{\textrm{gh}}\big)}\right]\label{erge}\end{aligned}\ ] ] here @xmath125 is the bosonic part of the action and @xmath126 is its hessian . the coarse graining operators in ( [ erge ] ) have the structure @xmath127 where @xmath128 is a `` shape function '' continuously interpolating between @xmath129 and @xmath130 . the constants @xmath131 and @xmath132 are matrices in field space . they will be adjusted in such a way that if in @xmath133 a certain mode has the inverse propagator @xmath134 it becomes @xmath135 when we add @xmath136 to @xmath133 . as we shall see , this requirement is met if @xmath131 and @xmath132 have the following block structure in @xmath137-space : @xmath138^{\mu \nu}_{\ \ \rho\sigma } & = \frac{z_n(k ) \kappa^2}{2}(\delta^\mu_\rho \delta^\nu_\sigma+\delta^\nu_\rho \delta^\mu_\sigma-\bar g^{\mu\nu } \bar g_{\rho\sigma } ) \\ \big[\left(\mathcal z^{\textrm{gh}}_k\right)_{\bar \xi \xi}\big]^\mu_{\ \nu}&=\sqrt{2}\ ; \delta^\mu_\nu \end{aligned } \ \begin{aligned } \big[\left(\mathcal{z}_k\right)_{\bar a\bar a}\big]^{a \mu b}_{\ \ \ \nu } & = \frac{z_f(k)}{\hat{g}^2}\ ; \delta^{ab } \delta^\mu_{\nu}\\ \big[\left(\mathcal z^{\textrm{gh}}_k\right)_{\bar \upsilon \upsilon}\big]^{ab}&=\delta^{ab } \end{aligned } \label{zfactors}\ ] ] note that @xmath132 is actually @xmath40-independent . in setting up eq . ( [ erge ] ) we followed ref . @xcite and opted for the complete hessian operator @xmath133 to play the rle of @xmath46 . more precisely , we set @xmath139 and @xmath140 in the @xmath141- and the ghost - sectors , respectively . the multiplication by the inverse @xmath142 matrices brings @xmath46 closer to an ordinary ( covariant ) laplacian ; symbolically , if @xmath143 , we employ @xmath144 rather than @xmath145 . the most complicated ingredient needed in order to evaluate the traces in the frge is the hessian of the bosonic action @xmath126 , i.e. the matrix of its second functional derivatives with respect to the dynamical fields @xmath146 , or equivalently @xmath147 , at fixed backgrounds @xmath148 . this hessian is most transparently displayed by means of the associated quadratic form @xmath149 which appears in the expansion @xmath150= \breve{\gamma}_k[\bar g , \bar a,\bar g,\bar a]+o(\bar h,\bar a)+\gamma^{\textrm{quad}}_k[\bar h , \bar a;\bar g,\bar a]+{\cal o}(\{\bar h,\bar a\}^3)\ ] ] explicitly , @xmath149 is the sum of the following terms which reflect the block structure of @xmath126 in @xmath141-space : @xmath151 the above quadratic functionals contain the kernels @xmath152 using these formulae it can be checked that the @xmath142-factors ( [ zfactors ] ) are correctly chosen . since we are not going to extract any `` extra '' background field dependence @xcite we may set @xmath153 and @xmath154 after having found the hessian . the truncation contains three running couplings , @xmath155 , @xmath156 and @xmath157 . their beta functions can be found from the frge ( [ erge ] ) by `` projecting out '' the corresponding invariants in the derivative expansion of the traces and equating them to the corresponding field monomials on the lhs of the flow equation . the resulting system of differential equations becomes autonomous if we employ the dimensionless counterparts of @xmath158 , @xmath159 and @xmath160 respectively : @xmath161 in terms of these variables the three coupled rg equations have the structure @xmath162 here we introduced the anomalous dimensions related to the yang - mills and the gravitational field , respectively : @xmath163 in this paper we are only interested in the gravitationally corrected yang - mills beta function @xmath164 . therefore it is sufficient to extract the @xmath165-term from the derivative expansion of the traces . for identifying this monomial and reading off its prefactor we may insert any metric . we shall employ the most convenient choice , @xmath166 . furthermore , we set @xmath115 from now on . the remaining calculation is in principle straightforward , but rather lengthy . one has to expand the traces up to terms with two fields @xmath167 and two derivatives acting on them . because of the built - in background gauge invariance those terms should combine to @xmath168 . as a check we verified that this indeed happens . let us now discuss the result . here we specialize for @xmath169 spacetime dimensions ; for general @xmath170 the reader is referred to the appendix . we present three different formulae for @xmath171 ; they differ with respect to the degree of `` rg improvement '' they take into account . to start with , we `` switch off '' all rg improvements . this means that we discard all terms in @xmath172 on the rhs of the flow equation where @xmath173 hits either a @xmath131-factor or the @xmath133 in the argument of @xmath136 . in this way the evaluation of the frge amounts to a one - loop calculation , with a non - standard regulator though . we find @xmath174 so that @xmath175 here @xmath176 is one of the integrals which were encountered in the pure gravity calculation @xcite already : @xmath177^p}\qquad & ( n&>0)\\ \phi^p_0(w)&=(1+w)^{-p } & ( n&=0 ) \end{aligned}\ ] ] the second contribution on the rhs of ( [ 4d1loop ] ) is the familiar `` asymptotic freedom '' term due to the self - interaction of the gauge bosons , while the first one , due to the virtual gravitons , is new . several comments are in order here . * ( 1 ) * the gravitational correction is manifestly cutoff scheme dependent , i.e. it depends , via @xmath176 , on the shape function @xmath178 . however , for any admissable choice of @xmath178 the constant @xmath176 is positive . as a result , the gravity term has a qualitatively similar impact on @xmath179 as the gauge boson loops , namely to drive @xmath179 smaller at larger @xmath40 . it tends to speed up the approach of asymptotic freedom . for the exponential cutoff @xmath180 , for instance , one finds @xmath181 , while the `` optimized '' one @xcite , @xmath182 , yields @xmath183 . * ( 2 ) * the gravitational correction , in perturbative language , originates from a quadratic divergence or , in frge language , a quadratic running with @xmath40 . for this reason its scheme dependence is by no means surprising or alarming rather , it is the usual situation which is always encountered when the effective average action is applied to matter theories with a quadratic running of parameters , masses , say . however , one should note that the couplings in @xmath45 as such are not observable or `` physical '' quantities . the latter must be @xmath178-independent . this independence comes about by a compensation of the scheme dependence among different running couplings . ( in truncations this compensation might not be perfect . ) in general there will be compensations between effective propagators and vertices , for instance . analogous remarks apply to the gauge fixing dependence . * ( 3 ) * the beta function for @xmath184 depends on all three couplings , @xmath184 , @xmath185 and @xmath186 . in the approximation of ( [ 4d1loop ] ) it happens to be independent of @xmath186 , but it does depend on @xmath187 , the dimensionless newton constant . hence the differential equation for @xmath188 can not be solved in isolation . in principle the full system ( [ sde ] ) should be considered , and this would include the backreaction of the matter fields on the running of the gravitational parameters @xmath185 and @xmath186 . we shall not study this backreaction in the present paper . instead , let us assume that the complete rg trajectory @xmath189 admits a classical regime @xcite in which newton s constant does not run appreciably so that we may approximate @xmath190 this approximation , implicitly , has been made in all perturbative studies @xcite . with ( [ gassumption ] ) , for an abelian field @xmath191 , say , @xmath192 incidentally this beta function has the same general structure as the result by robinson and wilczek @xcite ; it is proportional to @xmath193 and depends explicitly on the energy scale @xmath40 . its @xmath194-dependence indicates that the underlying quantum effect is related to a quadratic divergence . ( [ abelianresult ] ) is easily solved : @xmath195 . here @xmath196 and @xmath197 is the ( ordinary , constant ) planck mass . to first order in the @xmath198-expansion we get @xmath199\label{kmexpansion}\ ] ] we note that to leading order newton s constant itself @xcite has an analogous scale dependence , including the sign of the correction : @xmath200 $ ] . * ( 4 ) * in order to illustrate how the above result fits into the asymptotic safety picture of quantum einstein gravity @xcite we consider a free maxwell field again . it is known that , in the einstein - hilbert truncation , the rg flow of the average action possesses a non - gaussian fixed point for the two gravitational couplings , @xmath201 , both in pure gravity @xcite and in presence of a free maxwell field @xcite . at this fixed point the _ dimensionless _ newton constant equals a positive constant , @xmath202 , while the dimensionful one runs to zero quadratically : @xmath203 for @xmath204 . in this regime , @xmath205 the solution to this equation reads @xmath206 at the fixed point the gauge coupling approaches zero according to a power law with a critical exponent @xmath207 , a positive number of order unity . from existing calculations since the determination of @xmath208 in the einstein - maxwell system in ref . @xcite employs a cutoff different from the present one . ] thus the total system has a non - trivial fixed point of the form @xmath209 . obviously the approach of @xmath210 is much faster than without gravity where @xmath211 note that @xmath188 is a _ relevant _ parameter , it grows when @xmath40 is lowered , hence it contributes one unit to the dimensionality of the fixed point s uv critical manifold . next we present the results for @xmath171 with the rg improvements included . in a first step we retain only the terms which arise when @xmath173 hits the @xmath131-factors in @xmath136 . those terms are proportional to @xmath171 and @xmath212 , respectively . as now @xmath171 appears also on the rhs of the rg equation we obtain an implicit equation for it . its solution reads @xmath213 in this approximation @xmath171 depends not only on newton s but also on the cosmological constant . ( [ 4d1stimproved ] ) resums terms of arbitrary order both in @xmath188 and @xmath185 ; it generalizes a known result @xcite for pure yang - mills theory . if we go on and include also the terms coming from the scale derivative of @xmath133 in the argument of @xmath214 we are led to @xmath215 the eqs . ( [ 4d1stimproved ] ) and ( [ 4d2ndimproved ] ) contain the following integrals involving @xmath178 : @xmath216^p } \qquad & ( n&>0)\\ \tilde{\phi}^p_0(w)&=(1+w)^{-p } & ( n&=0)\\ \tilde{\tilde{\phi}}^p_n(w)&=\frac{1}{\gamma(n)}\int_0^\infty{\textrm{d } z\:}z^{n-1}\frac{r^{(0)}{}'(z)}{[z+r^{(0)}(z)+w]^p}\qquad & ( n&>0)\\ \tilde{\tilde{\phi}}^p_0(w)&=\frac{r^{(0)}{}'(0)}{(1+w)^p}\qquad & ( n&=0 ) \end{aligned}\ ] ] the threshold functions @xmath217 appeared already in the einstein - hilbert truncation of pure gravity @xcite . in eq . ( [ 4d2ndimproved ] ) we abbreviated @xmath218 and @xmath219 , respectively . what is new in ( [ 4d2ndimproved ] ) is the occurrence of the complete beta function of the cosmological constant , @xmath220 . it originates from the differentiation of the @xmath160-term contained in @xmath133 . in the recent literature on the gravitational corrections to the yang - mills beta function @xcite there has been a certain amount of confusion as some of the computations do get a non - zero result while others do nt . however , we believe that different calculations have no reason to yield the same result unless they agree on virtually all details of the regularization and renormalization procedure . the quantum effects of interest are related to quadratic divergences ( or a @xmath194-running ) , and so we should not expect the same high degree of universality as in the case of the familiar gauge boson contribution which is related to a logarithmic divergence . in the present paper we computed the beta function for @xmath179 , defined as a coefficient in the derivative expansion of the effective average action . this approach has two features which are essential here : first , it retains all quadratic divergences ( as opposed to dimensional regularization , say ) , and second , by the background field technique , the regularization ( the cutoff @xmath136 ) preserves gauge invariance . is not gauge fixing independent , though , i.e. we do not use the vilkovisky - dewitt method . ] in this setting , we do get a non - zero gravitational correction . this correction is scheme and gauge fixing dependent but , as we explained , this is by no means unexpected but rather the usual situation . when observable quantities are computed from @xmath45 the scheme and gauge fixing dependences will cancel among the different running couplings involved . among the perturbative calculations only the one by tang and wu @xcite is directly comparable to ours . they employ a regulator which retains quadratic divergences and treats them in a gauge invariant manner . it is gratifying to see that they , too , get a non - zero gravitational correction which has the same structure as ours when we omit the rg improvements . as a first application of our results we mention that , by a standard argument , knowledge about the @xmath40-dependence of wave function normalization constants such as @xmath105 can be used in order to deduce information about the related fully dressed propagator implied by @xmath221 . in the case at hand the running inverse propagator of the gauge field , on a flat background , has the form @xmath222 . at high momenta , if there is no other relevant physical cutoff scale but the momentum itself , the dressed propagator @xmath223 obtains by setting @xmath224 , whence @xmath225 . for the example of eq . ( [ kmexpansion ] ) , for instance , this leads us to expect that the photon propagator gets modified by a @xmath226-term when @xmath227 approaches the planck scale : @xmath228 likewise the fixed point running of ( [ critexp ] ) implies the following behavior for @xmath229 : @xmath230 as @xmath207 is positive the gauge field propagator falls off faster than @xmath231 , thanks to the quantum gravity corrections . in fact , the same argument when applied to the graviton propagator leads to a @xmath232-behavior for @xmath229 @xcite . the asymptotic propagator ( [ propagator ] ) suggests that the quantum gravity corrections improve the finiteness properties of the matter field theory , and this precisely fits into the picture of asymptotic safety . it is also interesting to note that ( [ propagator ] ) leads to a modified static electromagnetic potential @xmath233 of a classical point charge . the 3-dimensional fourier transform of ( [ propagator ] ) yields the potential @xmath234 which , if @xmath207 is large enough , could even be regular at @xmath235 . this makes it obvious that the gravity induced running of the gauge coupling is closely related to the old problem of divergent self energies . in this appendix we display the equation for @xmath171 in @xmath170 spacetime dimensions . with all rg improvements included , it assumes the form : @xmath236 & \qquad+c ( 2\lambda)^{-1}\left(\chi_{\frac{d}{2},0}(-2\lambda)-\chi_{\frac{d}{2},0}(0)\right)\\[-0.1 cm ] & \qquad+d\left\{d^{-d/2}\chi_{\frac{d}{2}+1,2}(-2\lambda d)-\chi_{\frac{d}{2}+1,2}(0)-\frac{d}{2}\left(d^{-d/2}\chi_{\frac{d}{2},1}(-2\lambda d)-\chi_{\frac{d}{2},1}(0)\right)\right\}\bigg]\\[-0.1 cm ] & \quad-{g_{\textrm{ym}}}^2n\left(\frac{26-d}{3}\right)\chi_{\frac{d}{2}-2,0}(0)\\[-0.1 cm ] & \quad + 16 \pi \eta_n g \bigg[a \tilde{\chi}_{\frac{d}{2}-1,0}(-2\lambda)+b d^{-d/2 } \tilde{\chi}_{\frac{d}{2}-1,0}(-2 \lambda d)\\[-0.1 cm ] & \qquad + c \left((2\lambda)^{-1}\tilde{\chi}_{\frac{d}{2},0}(-2\lambda)-(2\lambda)^{-2}\left(\tilde{\chi}_{\frac{d}{2}+1,0}(-2 \lambda)-\tilde{\chi}_{\frac{d}{2}+1,0}(0)\right)\right)\\[-0.1 cm ] & \qquad -d\left\{\frac{d^{-d/2}}{d-1}\tilde{\chi}_{\frac{d}{2}+1,2}(-2\lambda d)+\frac{d^{-d/2 + 1}}{2}\tilde{\chi}_{\frac{d}{2},1}(-2\lambda d)-\frac{1}{d-1}\tilde{\chi}_{\frac{d}{2}+1,2}(0)\right\}\bigg]\\[-0.1 cm ] & \quad + 16 \pi \eta_f g \left[c \left(-(2\lambda)^{-1}\tilde{\chi}_{\frac{d}{2},0}(0)+(2\lambda)^{-2}\left(\tilde{\chi}_{\frac{d}{2}+1,0}(-2 \lambda)-\tilde{\chi}_{\frac{d}{2}+1,0}(0)\right)\right)\right.\\[-0.1 cm ] & \qquad \left .- d\left\{-\frac{d^{-d/2 + 1}}{d-1}\tilde{\chi}_{\frac{d}{2}+1,2}(-2\lambda d)-\frac{d}{2}\tilde{\chi}_{\frac{d}{2},1}(0)+\frac{d}{d-1}\tilde{\chi}_{\frac{d}{2}+1,2}(0)\right\}\right]\\[-0.1 cm ] & \quad+\eta_f{g_{\textrm{ym}}}^2n\left(\frac{24-d}{6}\right)\tilde{\chi}_{\frac{d}{2}-2,0}(0)\\[-0.1 cm ] & \quad-16\pi ( \eta_f-\eta_n ) g \bigg[a \tilde{\tilde{\chi}}_{\frac{d}{2},0}(-2\lambda)+b d^{-d/2 } \tilde{\tilde{\chi}}_{\frac{d}{2},0}(-2 \lambda d)\\[-0.1 cm ] & \qquad+c ( 2\lambda)^{-1}\left(\tilde{\tilde{\chi}}_{\frac{d}{2}+1,0}(-2\lambda)-\tilde{\tilde{\chi}}_{\frac{d}{2}+1,0}(0)\right)\\[-0.1 cm ] & \qquad+d\frac{(d-2)}{2}\left\{\tilde{\tilde{\chi}}_{\frac{d}{2}+1,1}(0)-\tilde{\tilde{\chi}}_{\frac{d}{2}+1,1}(-2\lambda d)d^{-d/2}\right\}\bigg]\\[-0.1 cm ] & \quad+32\pi g ( 2\lambda+\beta_\lambda)\bigg[a \tilde{\tilde{\chi}}_{\frac{d}{2}-1,0}(-2\lambda)+b d^{-d/2 + 1 } \tilde{\tilde{\chi}}_{\frac{d}{2}-1,0}(-2 \lambda d)\\[-0.1 cm ] & \qquad + c \left((2\lambda)^{-1}\tilde{\tilde{\chi}}_{\frac{d}{2},0}(-2\lambda)-(2\lambda)^{-2}\left(\tilde{\tilde{\chi}}_{\frac{d}{2}+1,0}(-2\lambda)-\tilde{\tilde{\chi}}_{\frac{d}{2}+1,0}(0)\right)\right)\\[-0.1 cm ] & \qquad -2d d\left\{\frac{d^{-d/2}}{d-1 } \tilde{\tilde{\chi}}_{\frac{d}{2}+1,2}(-2\lambda d)-\frac{1}{d-1 } \tilde{\tilde{\chi}}_{\frac{d}{2}+1,2}(0)+\frac{d}{2 } d^{-d/2 } \tilde{\tilde{\chi}}_{\frac{d}{2},1}(-2\lambda d)\right\}\bigg ] \end{split}\label{result}\ ] ] eq . ( [ result ] ) contains the following functions of @xmath170 : @xmath237 c&=\frac{4(3d^2 - 10d+4)}{d(d-2 ) } \end{aligned } \qquad \begin{aligned } b&=\frac{(d-4)(d-6)}{2(d-2)}\\[0.5 cm ] d&=\frac{4(d-4)^2}{d(d-2)^2(d-1 ) } \end{aligned}\ ] ] furthermore we introduced the following new threshold functions : @xmath238 when used in eq . ( [ result ] ) with one argument only , the second argument of the @xmath239 functions , @xmath81 , is set to @xmath240 implicitly , for clarity purposes . for @xmath241 and @xmath242 the @xmath239 integrals are independent of @xmath81 and can be expanded in @xmath243 into a _ finite _ sum of @xmath244 integrals , @xmath245 and analogously for @xmath246 and @xmath247 . for @xmath169 only @xmath239 integrals that are expandable in this way occur . for @xmath248 a non - integrable singularity might occur in the integration interval ( depending on the signs of @xmath243 and @xmath81 ) . in this case the definition of the @xmath239 integrals is meant merely as an abbreviation , in the following sense . the integrands of the @xmath239 integrals embraced by curly brackets in ( [ result ] ) should be summed first and only then be integrated over the intersection of their integration intervals . this removes the apparent singularity and we are left with a sum of finite integrals . ( [ result ] ) can be reduced to the corresponding 1-loop result by omitting from its rhs all terms containing @xmath171 , @xmath212 , or any @xmath247 integral . the latter arise from the differentiation of @xmath126 in the arguments of @xmath136 . s. p. robinson and f. wilczek , phys . 96 ( 2006 ) 231601 . s. p. robinson , ph.d . thesis , mit , ( 2005 ) . a. r. pietrykowski , phys . lett . 98 ( 2007 ) 061801 . d. j. toms , phys . d 76 ( 2007 ) 045015 . d. ebert , j. plefka and a. rodigast , phys . b 660 ( 2008 ) 579 . y. tang and y .- l . d. j. toms , phys . lett . 101 ( 2008 ) 131301 . s. weinberg in _ general relativity , an einstein centenary survey _ , + s. w. hawking and w. israel ( eds . ) , cambridge university press ( 1979 ) . s. weinberg , . s. weinberg , . m. reuter , phys . d 57 ( 1998 ) 971 and . d. dou and r. percacci , class . . grav . 15 ( 1998 ) 3449 . o. lauscher and m. reuter , phys . d 65 ( 2002 ) 025013 and m. reuter and f. saueressig , phys . d 65 ( 2002 ) 065016 and o. lauscher and m. reuter , phys . d 66 ( 2002 ) 025026 and o. lauscher and m. reuter , class . grav . 19 ( 2002 ) 483 and o. lauscher and m. reuter , int . j. mod . phys . a 17 ( 2002 ) 993 and w. souma , prog . phys . 102 ( 1999 ) 181 . m. reuter and f. saueressig , phys . d 66 ( 2002 ) 125001 and fortschr . 52 ( 2004 ) 650 and a. bonanno and m. reuter , jhep 02 ( 2005 ) 035 and for reviews see : m. reuter and f. saueressig , ; + o. lauscher and m. reuter in _ quantum gravity _ , b. fauser , + j. tolksdorf and e. zeidler ( eds . ) , birkhuser , basel ( 2007 ) and ; + o. lauscher and m. reuter in _ approaches to fundamental physics _ , + i .- o . stamatescu and e. seiler ( eds . ) , springer , berlin ( 2007 ) . r. percacci and d. perini , phys . d 67 ( 2003 ) 081503 ; + phys . d 68 ( 2003 ) 044018 ; class . quant . grav . 21 ( 2004 ) 5035 . a. codello and r. percacci , phys . 97 ( 2006 ) 221301 ; + a. codello , r. percacci and c. rahmede , int . j. mod . phys . a 23 ( 2008 ) ; + preprint o. zanusso , l. zambelli , g. p. vacca and r. percacci , . d. litim , phys . lett . 92 ( 2004 ) 201301 ; aip conf . 841 ( 2006 ) 322 ; + p. fischer and d. litim , phys . b 638 ( 2006 ) 497 ; + aip conf . 861 ( 2006 ) 336 . p. machado and f. saueressig , phys . d 77 ( 2008 ) 124045 . d. benedetti , p. machado and f. saueressig , mod . lett . a 24 ( 2009 ) 2233 ; + nucl . b 824 ( 2010 ) 168 . a. eichhorn , h. gies and m. scherer , . o. lauscher and m. reuter , jhep 10 ( 2005 ) 050 and m. reuter and j .- m . schwindt , jhep 01 ( 2006 ) 070 and m. reuter and j .- m . schwindt , jhep 01 ( 2007 ) 049 and m. reuter and h. weyer , phys . d 79 ( 2009 ) 105005 and + ; gen . 41 ( 2009 ) 983 and + . m. reuter and h. weyer , phys . rev . d 80 ( 2009 ) 025001 and + . p. f. machado and r. percacci , phys . d 80 ( 2009 ) 024020 . e. manrique and m. reuter , phys . d 79 ( 2009 ) 025008 and + . e. manrique and m. reuter , . e. manrique and m. reuter , . daum and m. reuter , advanced science letters 2 ( 2009 ) 255 and preprint m. niedermaier , phys . lett . 103 ( 2009 ) 101303 . forgcs and m. niedermaier , + m. niedermaier , jhep 12 ( 2002 ) 066 ; nucl . b 673 ( 2003 ) 131 ; + class . grav . 24 ( 2007 ) r171 . for detailed reviews of asymptotic safety in gravity see : + m. niedermaier and m. reuter , living reviews in relativity 9 ( 2006 ) 5 ; + r. percacci , c. wetterich , phys . lett . b 301 ( 1993 ) 90 . m. reuter , c. wetterich , nucl . b 417 ( 1994 ) 181 . m. reuter and c. wetterich , nucl . b 427 ( 1994 ) 291 , + nucl . b 391 ( 1993 ) 147 , nucl . b 408 ( 1993 ) 91 ; + m. reuter , phys . d 53 ( 1996 ) 4430 , mod.phys . a 12 ( 1997 ) 2777 . m. reuter , c. wetterich , phys . d 56 ( 1997 ) 7893 and . h. gies , phys . d 68 ( 2003 ) 085015 . j. berges , n. tetradis and c. wetterich , phys . rep . 363 ( 2002 ) 223 ; + c. wetterich , int . j. mod . phys . a 16 ( 2001 ) 1951 . for reviews of the effective average action in yang mills theory see : + m. reuter , j. pawlowski , ann . phys . 322 ( 2007 ) 2831 ; + h. gies , d. litim , phys . b 486 ( 2000 ) 92 ; phys . d 64 ( 2001 ) 105007 ; + int . j. mod . phys . a 16 ( 2001 ) 2081 . a. bonanno and m. reuter , phys . d 62 ( 2000 ) 043008 and phys . d 73 ( 2006 ) 083005 and phys . d 60 ( 1999 ) 084011 and m. reuter and e. tuiran , in _ proceedings of the eleventh marcel grossmann meeting _ , h. kleinert , r. jantzen , r. ruffini ( eds . ) , world scientific , singapore ( 2007 ) and . a. bonanno and m. reuter , phys . d 65 ( 2002 ) 043508 and ; + m. reuter and f. saueressig , jcap 09 ( 2005 ) 012 and a. bonanno and m. reuter , phys . b 527 ( 2002 ) 9 and + int . j. mod . d 13 ( 2004 ) 107 and + e. bentivegna , a. bonanno and m. reuter , + jcap 01 ( 2004 ) 001 and a. bonanno and m. reuter , jcap 08 ( 2007 ) 024 and ; + j. phys . ser . 140 ( 2008 ) 012008 and . a. bonanno , g. esposito and c. rubano , gen . grav . 35 ( 2003 ) 1899 ; + class . quant . grav . 21 ( 2004 ) 5005 ; + a. bonanno , g. esposito , c. rubano and p. scudellaro , + class . 23 ( 2006 ) 3103 and 24 ( 2007 ) 1443 . m. reuter and h. weyer , phys . d 69 ( 2004 ) 104022 and m. reuter and h. weyer , phys . d 70 ( 2004 ) 124028 and m. reuter and h. weyer , jcap 12 ( 2004 ) 001 and f. girelli , s. liberati , r. percacci and c. rahmede , + class . grav . 24 ( 2007 ) 3995 . d. litim and t. plehn , phys . lett . 100 ( 2008 ) 131301 . j. moffat , jcap 05 ( 2005 ) 2003 ; + j. r. brownstein and j. moffat , astrophys . j. 636 ( 2006 ) 721 ; + mon . not . 367 ( 2006 ) 527 . l. f. abbott , nucl . b 185 ( 1981 ) 189 ; acta phys . b 13 ( 1982 ) 33 ; + b. s. dewitt , phys.rev.162 ( 1967 ) 1195 ; + m. t. grisaru , p.van nieuwenhuizen and c.c . wu , phys . d 12 ( 1975 ) 3203 ; + d. m. capper , j.j . dulwich and m. ramon medrano , nucl . b 254 ( 1985 ) 737 ; + s. l. adler , rev . phys . 54 ( 1982 ) 729 . r. jackiw , phys . lett . 41 ( 1978 ) 1635 . d. litim and j. pawlowski , phys . d 66 ( 2002 ) 025030 . c. a. reuschle , diploma thesis , freiburg , ( 2009 ) .

this paper presents the study of the interstellar medium toward cma , a disk sight - line known for its low neutral gas density . this study uses high and medium resolution hst - ghrs spectra including lines from the following species : hi , di , ni , oi , sii , siii , siii , siiii , siiv , alii , aliii , feii , mgi , mgii , mnii , cii and civ . + the line of sight to cma ( 153 pc ) is dominated by two ionized regions with a velocity difference of 10 kms . the ionized regions account for most of the total hydrogen column density , around 210 , and the neutral gas represents only 10% of the total gas . the two ionized clouds display characteristics of the warm diffuse gas detected in the disk and the halo . their gas - phase abundances indicate that their depletion is low , especially for the more ionized of the two clouds . + special models of photoionization by the two euv - excess stars cma and cma would be needed for a detailed discussion of the ionizing mechanisms of the clouds ; their ionization ratios are nevertheless roughly compatible with collisional ionization at temperatures around 20000 k , substantially higher than the kinetic temperatures derived from the line widths . their characteristics suggest that the clouds may be in the process of cooling down and recombining after having been shocked and ionized by some violent events , possibly related to the local bubble formation .

astro-ph9806178

observations of the interstellar medium toward nearby stars show that the sun is located in an interstellar void , often called the local bubble , filled with a hot ( @xmath5 k ) and low density ( 5 @xmath6 @xmath7 ) gas , responsible for part of the soft x - ray backround ( e.g. cox & reynolds 1987 and references therein ) . the bubble is irregularly shaped ( frish & york 1983 ) , extending to at least 50 pc in most directions . _ copernicus _ observations ( gry et al . 1985 ) have shown an extension of the bubble as far as the star @xmath0 cma ( @xmath8 , @xmath9 ) 153 pc from the sun ( hipparcos catalog ) . they report a neutral hydrogen column density from 1.0 to 2.2 @xmath10 @xmath3 , and a strongly ionized sight - line . welsh ( 1991 ) and welsh et al . ( 1994 ) , have observed the nai line toward many neighbouring stars and confirmed the existence of this low density tunnel , 300 pc long and 50 pc wide , in the direction of @xmath0 cma . more recently , gry et al . ( 1995 ) analysed the uv absorption lines of the interstellar medium in the direction of @xmath4 cma ( @xmath11 , @xmath12 ) , 132 pc from the sun . they have shown that the absorption lines are principally due to local matter : the local interstellar cloud ( lic ) in which the sun is embedded ( see lallement & bertin 1992 ) as well as a cloud already detected toward the very nearby ( 2.7 pc ) star @xmath13 cma ( lallement et al . 1994 ) , and they conclude that beyond three parsecs , the mean gas density is less than 4.5@xmath14 @xmath7 . this makes this line of sight the most devoid region in the nearby interstellar medium . they also detected high ionization species like civ , attributed to the thermal conductive interface with the hot gas , thereby observed for the first time around two nearby diffuse clouds . observations of @xmath4 and @xmath0 cma have been performed in the extreme uv with the _ euve _ spacecraft . due to their location in the tunnel and their euv excess , @xmath4 cma and @xmath0 cma are respectively the first and the second brightest euv sources in the sky , and the principal photoionization sources in the solar neighbourhood : vallerga & welsh ( 1995 ) have found a local interstellar hydrogen photoionization rate for @xmath4 cma seven times greater than previous estimates calculated for all nearby stars ( bruhweiler & cheng 1988 ) . modelling the euv stellar spectra , cassinelli et al . ( 1995 ) and cassinelli et al . ( 1996 ) found interstellar neutral hydrogen column densities of 1.0 @xmath10 @xmath3 and [email protected] 10@xmath16 @xmath3 respectively for @xmath4 cma and @xmath0 cma , confirming the gry et al . ( 1985 ) result for @xmath0 cma . we present a study of the line of sight toward the b1 ii - iii star @xmath0 cma , performed with ultraviolet spectra of the star . our purpose is to give a description of the structure and physical conditions of the ionized gas detected with copernicus ( gry et al . 1985 ) , and thereby discuss the nature of diffuse clouds embedded in the local bubble . section 2 presents the data and describes their processing . in sect . 3 a description of the line of sight toward @xmath0 cma is given based on high and medium resolution data , and column densities and @xmath17-values are derived for each studied element and for each velocity component wherever possible . section 4 deals with the physical conditions in the clouds . the ionization of the clouds is treated in sect . 5 and the gas phase abundances in sect . 6 . in sect . 7 , the nature of the gas is discussed in comparison with other observations . clcc range ( ) & element & wavelength ( ) & @xmath18-value + + 1805.0 1814.6 & siii & 1808.013 & @xmath19 + 1841.6 1875.8 & aliii & 1854.716 & @xmath20 + & aliii & 1862.789 & @xmath21 + 2338.5 2350.7 & feii & 2344.214 & @xmath22 + 2371.8 2383.5 & feii & 2374.461 & @xmath23 + & feii & 2382.765 & @xmath24 + 2573.9 2586.9 & mnii & 2576.877 & @xmath25 + 2792.1 2807.1 & mgii & 2796.352 & @xmath26 + & mgii & 2803.531 & @xmath27 + 2846.7 2860.7 & mgi & 2852.964 & @xmath28 + + 1185.7 1221.9 & siii & 1190.208 & @xmath29 + & siii & 1190.416 & @xmath30 + & siii & 1193.290 & @xmath31 + & ni & 1199.550 & @xmath32 + & ni & 1200.223 & @xmath33 + & ni & 1200.710 & @xmath34 + & siiii & 1206.500 & @xmath35 + & di & 1215.339 & @xmath36 + & hi & 1215.670 & @xmath36 + 1229.8 1265.9 & sii & 1250.584 & @xmath37 + & sii & 1253.811 & @xmath38 + & sii & 1259.519 & @xmath39 + & siii & 1260.422 & @xmath40 + 1300.1 1336.0 & oi & 1302.168 & @xmath41 + & siii & 1304.370 & @xmath42 + & cii & 1334.532 & @xmath32 + 1390.4 1426.2 & siiv & 1393.755 & @xmath43 + & siiv & 1402.770 & @xmath44 + 1524.1 1559.3 & siii & 1526.707 & @xmath45 + & civ & 1548.195 & @xmath46 + & civ & 1550.770 & @xmath47 + 1644.4 1679.3 & alii & 1670.787 & @xmath28 + the observations described here have been obtained in december 1992 with the goddard high resolution spectrometer ( ghrs ) on board the hubble space telescope ( hst ) by vidal - madjar and collaborators . all lines above 1800 have been observed with ech - b , and have a resolution of 85000 giving a velocity resolution of about 3.5 kms@xmath1 . because of the failure of ech - a at that time , the short wavelength observations were made with the g160 m , which gives a resolution of 20000 and a velocity resolution of 15 kms@xmath1 . table 1 lists the observed spectral ranges and the spectral lines analysed . all data were taken with the 0@xmath4825 small science aperture ( ssa ) , the procedure fp - split = 4 and a substepping of 2 samples per diode width ( for details of the instrumentation see duncan 1992 ) . for data processing , standard stsdas procedures of the iraf software were used . wavelengths were assigned from the standard calibration tables . an error of @xmath151 resolution element on the wavelength assignment is expected because of magnetic drift . we made use of an absorption line profile fitting software developed by vidal - madjar et al . ( 1977 ) also described in ferlet et al . ( 1980a , b ) , and we checked our technique by comparing the results obtained with similar software developed by welty ( see welty et al . stellar lines were fitted with a low - order polynomial on both sides of each line and each interstellar absorption component was represented by the convolution of a theoretical voigt profile with the instrumental profile . an iterative procedure which minimizes the quadratic differences between the profile and the data points allows to determine the column density of the absorbing elements _ n _ ( @xmath3 ) , the radial velocity of the cloud ( kms@xmath1 ) and the line @xmath17-value ( kms@xmath1 ) of each interstellar absorption component . the instrumental profile is taken from duncan ( 1992 ) and the atomic parameters from morton ( 1991 ) except for the siii1808 line oscillator strength which comes from bergeson & lawler ( 1993 ) . the high resolution ( @xmath493.5 kms@xmath1 ) of the ech - b data allows us to determine the velocity structure of the sight - line . two main components are clearly visible in all studied lines ( fig . two smaller features are detected only in the four strongest lines feii2344 , feii2382 , mgii2796 and mgii2803 ( fig . these components are labelled a , b , c and d in order of increasing velocity . in the analysis , the velocity shifts between components are constrained to be the same for all lines but the absolute velocities are allowed to vary slightly from one line to another to account for the uncertainty mentioned in sect . the derived velocities all lie within a range of @xmath50 kms@xmath1 . the average of these velocities is therefore adopted , giving @xmath51 , 5.5 , 14.5 and 25 kms@xmath1 for a , b , c and d respectively , and the spectra are shifted accordingly . the above velocities thus define the heliocentric velocities of the components with an absolute precision of 3 kms@xmath1 . the sun is known to be embedded in the local interstellar cloud which is moving relative to it at a velocity of 26 kms@xmath1 towards the direction @xmath52 , @xmath53 ( lallement & bertin 1992 ) . the projection of its velocity vector on the @xmath0 cma direction gives 20.3 kms@xmath1 . this falls between component c and component d : the absorption line due to the lic is hidden and its study is therefore impossible at the resolution of the present data . it is nevertheless taken into account in the fit with the column density and @xmath17-value found by gry et al . ( 1995 ) toward @xmath4 cma . another small component is present in the lines of sight of both @xmath13 cma ( blue component , lallement et al . 1994 ) and @xmath4 cma ( component 2 , gry et al . 1995 ) . because of the vicinity of the @xmath13 cma and @xmath0 cma sight - lines , this cloud is likely to be present in the line of sight to @xmath0 cma and it is also taken into account in the fit with the @xmath4 cma column densities and @xmath17-values . the component column densities and @xmath17-values are determined for all observed elements with the absorption profile fitting program mentioned in sect . all the lines of an ion are fitted simultaneously . the results are shown in table 2 . the error bars given take into account the uncertainties in the stellar continuum modeling , the velocity positions of the components and the parameters of the hidden components ( lic and @xmath4 cma component 2 ) as well as the statistical error of the data . ccccccccc cloud & & & & + vel . & & & & + & _ n _ ( @xmath3 ) & _ b _ ( kms@xmath1 ) & _ n _ ( @xmath3 ) & _ b _ ( kms@xmath1 ) & _ n _ ( @xmath3 ) & _ b _ ( kms@xmath1 ) & _ n _ ( @xmath3 ) & _ b _ ( kms@xmath1 ) + feii & [email protected] 10@xmath54 & [email protected] & [email protected] 10@xmath54 & [email protected] & [email protected] 10@xmath55 & [email protected] & [email protected] 10@xmath55 & [email protected] + mnii & - & - & - & - & [email protected] 10@xmath54 & `` & [email protected] 10@xmath54 & '' + mgii & [email protected] 10@xmath54 & [email protected] & [email protected] 10@xmath56 & [email protected] & [email protected] 10@xmath55 & [email protected] & [email protected] 10@xmath55 & [email protected] + mgi & @xmath573 10@xmath58 & - & @xmath571 10@xmath59 & - & [email protected] 10@xmath59 & `` & [email protected] 10@xmath54 & '' + siii & - & - & - & - & [email protected] 10@xmath55 & [email protected] & [email protected] 10@xmath55 & [email protected] + aliii & - & - & - & - & @xmath60.3 10@xmath54 & `` & @xmath61.6 10@xmath54 & '' + siiii & - & - & - & - & @xmath62 & `` & @xmath63 & '' + sii & - & - & - & - & [email protected] 10@xmath64 & `` & [email protected] 10@xmath64 & '' + ni & - & - & - & - & [email protected] 10@xmath55 & [email protected] & [email protected] 10@xmath55 & [email protected] + the three lines of feii have different oscillator strengths and two of them are not saturated , thus they allow a reliable determination of the column density and @xmath17-value of each component . the interstellar features are superimposed on weak stellar lines which can be easily modeled by polynomials . the mnii2577 line is very weak , but the c and d components are clearly present . it lies in an area of the spectrum free of any stellar lines and the continuum is therefore also very easy to determine . because of the similarity of iron and manganese masses and with the assumption that feii and mnii come from regions of the same temperature and turbulent velocity , we assume _ b_(feii)=_b_(mnii ) for c and d and fit these four lines simultaneously . the mgii lines are strongly saturated . it is thus in principle difficult to derive column densities and @xmath17-values from the line profile fitting . however , with the assumption that _ b_(mgii)=_b_(mgi ) , which is true if mgi and mgii arise from the same regions , the two mgii lines can be fitted simultaneously with the unsaturated mgi line . since the structure of the sight - line is known from the analysis of the feii lines , the positions of all components in the mgii profiles can be determined precisely from the position of component a which is clearly visible in the spectra . mgi and mgii column densities and @xmath17-values could thus be derived for all components from the simultaneous fit of all the lines of the two species . the siii1808 line is weak but the c and d components are clearly identified . it is located in a portion of the spectrum with many narrow stellar lines . it is thus difficult to determine accurately the stellar continuum . but the silicon @xmath17-value must be between that of iron and that of magnesium , i.e. _ b_(siii)=3.5 4.0 kms@xmath1 for cloud c and _ b_(siii)=4.2 4.8 kms@xmath1 for cloud d. with these constraints on the @xmath17-values only the deepest stellar profiles are permitted by the fit and the column density uncertainties are thereby considerably reduced . in both aliii lines , which appear in the same spectrum at 1854.7 and 1862.8 , two small features separated by @xmath65 kms@xmath1 are detected at the bottom of the strong stellar lines . since the atomic mass of aluminium is close to that of silicon , we adopt the siii @xmath17-values to fit the aliii lines . many important species only show absorption lines below 1800 . this is the case for neutrals like ni , oi and hi as well as medium or highly ionized species like sii , cii , siiii , siiv and civ . unfortunately , the ech - a was not available at the time of the observations and the data were taken using the g160 m grating with a lower resolution than the echelle gratings . with a resolution element of about 15 kms@xmath1 , the different components seen in the ech - b spectra are not resolved . an additional difficulty comes from the low rotational velocity of @xmath0 cma ( @xmath66 kms@xmath1 ) which implies relatively narrow stellar lines , sometimes not easily differentiated from the interstellar contribution . nevertheless , for most species the total sight - line column density has been derived by fitting the interstellar absorption with one component and for ni , sii and siiii , a multi - components fit was possible . figure 2 shows some of the interstellar lines observed with g160 m together with the best fits , and table 3 summarizes the derived column densities . ll ni & 0.9 1.1 10@xmath64 + oi & @xmath67 + di & @xmath68 + alii & 5.0 8.0 10@xmath56 + sii & 3.1 3.7 10@xmath64 + cii & @xmath694 10@xmath64 + siiii & 1.5 10 10@xmath64 + siiv@xmath70 & @xmath71 ( @xmath72 ) + siii@xmath73 & 0.9 2.1 10@xmath55 + + + as mentioned in sect . 2 , the g160 m spectra hold an uncertainty in the wavelength position of about one resolution element i.e. @xmath74 kms@xmath1 . however the velocity calibration has been improved with the help of the siii lines , since siii lines are present in one ech - b spectrum ( giving a precision of 3 kms@xmath1 ) and in four of the six g160 m spectra ( see table 1 ) . the convolution of the theoretical profile obtained for the high resolution siii1808 line with the g160 m line spread function gives a blended feature centered at a velocity of 20.5 kms@xmath1 . this velocity is therefore adopted for the center of the interstellar siii absorption lines in the four g160 m spectra , yielding a wavelength calibration of these spectra with the accuracy of the ech - b spectra . the velocity shifts thereby applied to the four g160 m spectra rank from 2.5 to 12.5 kms@xmath1 , which are within the nominal precision of the data . figure 3 shows the central velocity of the absorption ( from the one - component fit of the line ) versus the ionization potential , for all species for which a good calibration has been obtained via the above process . it shows that , as the ionization potential increases from di to siiii , the central velocity of the absorption profile increases from a velocity intermediate to components c and d up to the velocity of component d , with a small departure for oi and cii which are saturated . this implies that component d contributes more than component c to the ionized species absorption and component c contributes more than component d to the neutrals absorption . in the following section we give some details on the analysis of each species special care has been put in trying to extract information on the distribution of matter between the two main components , although it is not always feasible at the g160 m resolution . three ni lines are available in our data at 1199.5 , 1200.2 and 1200.7 . the fact that the three lines have different oscillator strengths , with at least one of them unsaturated , permits an accurate one - component fit , giving the following total line - of - sight ni column density and @xmath17-value : _ n_(ni)@xmath75 @xmath3 , _ b_(ni)@xmath76 kms@xmath1 . but a multi - component fit is also performed , taking into account four components : components c and d , which contribute both to the absorption feature according to sect . 3.2.2 , as well as the lic and component 2 , the column densities of which have been derived in the line of sight of @xmath4 cmawith new ech - a data ( gry & dupin , in prep . ) and represent 25 % of the ni absorption in the @xmath0 cma sight - line . the three ni lines are fitted simultaneously with four components and the following constraints : the velocity of each component is fixed at the known position , the @xmath17-values and column densities for the lic and component 2 are those found for @xmath4 cma and the @xmath17-values for components c and d are between that of mg and that given by the maximum temperature allowed for the clouds ( see sect . we derive _ n_(ni)@xmath77 @xmath3 , _ b_(ni)@xmath78 kms@xmath1 and _ n_(ni)@xmath79 @xmath3 , _ b_(ni)@xmath80 kms@xmath1 for component c and component d , respectively . as the only available oi line ( 1302 ) is saturated , we could not find a unique solution with a multi - component model . therefore we simply derive a column density lower limit for the total sight - line using a one - component model , and taking the largest @xmath17-value that gives a reasonable fit . we find _ n_(oi)@xmath67 @xmath3 for @xmath81 kms@xmath1 . the saturated lyman @xmath13 hi interstellar feature has large wings and it is difficult to separate it from the stellar line . furthermore too many components have to be taken into account to allow reliable results from profile fitting . but the lyman @xmath13 di line contribution can be separated from that of hi and is only slightly saturated . we derive _ n_(di)@xmath68 @xmath3 , @xmath82 kms@xmath1 . five lines of siii at 1190 , 1193 , 1260 , 1304 and 1526 are available but the line at 1190.4 is blended with the siii line at 1190.2 and can not be used . the two - component model derived from the analysis of the ech - b siii1808 line has been applied to the four available lines and gives a good fit for all of them . the g160 m data include three sii lines at 1250 , 1253 and 1259 . the three lines have first been fitted simultaneously with one component only : this gives _ n_(sii)@xmath83 @xmath3 for the total sight - line . as these transitions are weak and the lines are not saturated , most of the absorption must be due to the largest components . the contribution of the lic and component 2 can be estimated from their sii column densities in the @xmath4 cma sight - line : gry & dupin ( 1997 ) have found an upper limit of @xmath84 @xmath3 for the lic , and component 2 appears to be even weaker . these two clouds should therefore not account for more than 3% of the sulfur in the @xmath0 cma sight - line . the three sii lines are thus expected to be strongly dominated by the two largest components c and d. a two - component fit is thus performed , with the siii @xmath17-values , and gives _ n_(sii)@xmath85 @xmath3 for cloud c and _ n_(sii)@xmath86 @xmath3 for cloud d. it is seen that the sum of the two cloud contributions is equal to the one - component fit result , but with a larger uncertainty . the cii1334 line is strong and saturated . each component of the sight - line gives a significant contribution to the interstellar absorption feature and , as for oi , we only derive a column density lower limit : n(cii)@xmath87 @xmath3 with @xmath88 kms@xmath1 . the interstellar cii*1335 line is located at the end of the spectrum and it was not possible to use it . the alii line at 1670 is located in a spectral region which does not include siii lines and thus does not have a precise velocity calibration . it is therefore fitted with a one - component model which gives _ n_(alii)@xmath89 @xmath3 for _ b_@xmath90 kms@xmath1 . the large uncertainty in the column density is due to the fact that we have only one alii line and that the stellar continuum is not well defined at these wavelengths . the siii line at 1190.2 is unfortunately severely blended with the siii line at 1190.4 . thus , no satisfactory stellar continuum could be drawn and this line was not used . the siiii1206 line is saturated but the stellar line is well defined and can be modelled by a parabola . the fit confirms that most of the siiii is in cloud d and with the assumption that _ b_(siiii)=_b_(siii ) , gives _ @xmath92 @xmath3 for cloud d and an upper limit for cloud c of _ n_(siiii)@xmath93 @xmath3 . the large uncertainty on the d column density is due to the line saturation . a spectral region containing the siiv lines at 1393 and 1402 was observed , but no interstellar absorption feature was detected at these wavelengths . only 3@xmath94 upper limits are derived for the column density . if it is assumed that siiv comes from the clouds themselves i.e. _ b_(siiv)=_b_(siii)@xmath95 kms@xmath1 , the strongest line implies _ n_(siiv)@xmath96 @xmath3 . if siivis produced by a high temperature interface between the clouds and the hot surrounding medium , a 3 @xmath94 upper limit gives _ n_(siiv)@xmath97 @xmath3 for @xmath98 k. fig . 4 shows the @xmath0 cma spectrum between 1547 and 1552 containing the civ 1548,1550 doublet . when the stellar lines are represented by fourth order polynomials , an absorption feature appears at the bottom of both stellar lines . although the interstellar nature of these features is not beyond dispute they nevertheless occur in both lines at the same velocity which also happens to be the lic velocity ( @xmath99 kms@xmath1 ) . in their study of the interstellar gas toward @xmath4 cma , gry et al . ( 1995 ) , reported a civ absorption at the lic velocity , with a width corresponding to thermal broadening at @xmath100 k. they suggest that the absorption could be due to the interface between the lic and the hot surrounding medium . in fig . 4 , the civ absorption lines in @xmath0 cma are compared to the model found for the lic absorption in the case of @xmath4 cma ( _ n_(civ)@xmath101 @xmath3 , @xmath102 k , @xmath103 kms@xmath1 ) . the civ data are consistent with the expected lic absorption as determined toward @xmath4 cma . @xmath0 cma has been observed with copernicus ( gry et al . 1985 ) at shorter wavelengths and with a slightly higher resolution than that of g160 m , but not sufficient to resolve the different components . the copernicus oi and siii column densities are compatible with the g160 m values . but the ni , di , sii and siiii column densities derived from the copernicus observations are systematically lower than those we derive from our data . we have therefore reviewed the copernicus background estimates used by gry et al . and reached the conclusion that the contamination due to diffuse light in the copernicus instrument may have been incorrectly estimated . this is clear in the case of the siiii line at 1206.5 which is undoubtedly saturated in our data ( the background level is given by the bottom of the saturated lyman @xmath13 line present in the same spectrum ) while it does not appear saturated in the copernicus spectrum . according to rogerson et al . ( 1973 ) , the background of the copernicus data is roughly 25% of the continuum . if this background contribution has been underestimated , the copernicus column density of unsaturated lines may have been underestimated by 25% . if such a correcting factor is allowed , copernicus and g160 m values are in agreement for ni and sii . since the siiii1206 line is saturated , the column density estimate is very sensitive to any background misplacement and the value derived from copernicus had been more strongly underestimated . thanks to the mass difference between mg and fe , the thermal and turbulent contributions to the @xmath17-values can in principle be disentangled , providing a range of temperatures and turbulent velocities for each cloud . the @xmath17-values are listed in table 2 . for component b , the @xmath17-values found for mgii and feiiare so different that the ranges do not allow a measure of the turbulence . conversely , for components a , c and d , the ranges for mgii and feiioverlap and as the atomic mass of these elements imposes _ b_(mgii)@xmath104_b_(feii ) , this only provides upper limits for the temperatures : @xmath105 k for cloud a , @xmath106 k for cloud c and @xmath107 k for cloud d. the derived ranges for the turbulent velocities are the following : 1.7 kms@xmath1 @xmath108 kms@xmath1 for cloud a , 3.2 kms@xmath1 @xmath109 kms@xmath1 for cloud c and 4.0 kms@xmath1 @xmath110 kms@xmath1 for cloud d. cloud c and cloud d are both detected in mgi and mgii . if ionization equilibrium between mgi and mgii following e.g. frisch et al . ( 1990 ) and lallement et al . ( 1994 ) is assumed , the electron density in the clouds can be derived : @xmath111@xmath112_n_(mgii)/_n_(mgi)@xmath113 ; where the mgii radiative recombination rate ( @xmath114 ) is taken from aldrovandi & pquignot ( 1973 ) , the dielectronic recombination rate ( @xmath115 ) from nussbaumer & storey ( 1986 ) , the mgi ionization rate by charge exchange with protons ( @xmath116 ) from allan et al . ( 1988 ) , and the mgi radiative ionization rate ( @xmath117 ) from phillips et al . _ c _ is the mgi collisional ionization rate but it is negligible at the temperatures involved . for a warm cloud typical temperature of @xmath118 k , the range of _ n_(mgii)/_n_(mgi ) values derived from the observations implies @xmath111=0.18 0.54 @xmath7 for cloud c and @xmath111=0.38 1.82 @xmath7 for cloud d. lllllll & & + ratios & c & d & @xmath119=15000 k & @xmath119=20000 k & @xmath119=25000 k & @xmath119=30000 k + hi / hii & @xmath120 & @xmath121 & 1.79 & 7.2 @xmath122 & 1.2 @xmath122 & 3.4 @xmath6 + siii / siiii & @xmath123 & 0.07 0.52 & 47 & 1.1 & 0.17 & 8.5 @xmath122 + siiii / siiv & - & @xmath124 & - & 3.8 @xmath125 & 648 & 60 + siii / siiv & @xmath126 & @xmath127 & - & 4.1 @xmath125 & 134 & 5.46 + sii / siii & - & 8.6 33 & 5300 & 30.4 & 3.1 & 1.0 + feii / feiii & @xmath128 & @xmath129 & 19.6 & 0.52 & 0.13 & 7 @xmath122 + mgii / mgiii & @xmath130 & @xmath130 & 7.7 & 0.24 & 3.3 @xmath122 & 9.3 @xmath6 + ni / nii & @xmath131 & @xmath132 & 18.8 & 0.18 & 2.9 @xmath122 & 1.2 @xmath122 + alii / aliii & @xmath133 & @xmath134 & 2.1 @xmath135 & 62.4 & 8.8 & 2.38 + sulfur is a good element to use to derive the total hydrogen abundance because it is only slightly ( harris & mas hesse 1986 ) or not ( sofia et al . 1994 ) depleted in the interstellar gas . we thus apply the solar abundance given by anders & grevesse ( 1989 ) , @xmath136=-4.73 $ ] , to our determined sulfur column densities to derive the total hydrogen column density in the main components c and d. the siii column density is taken from gry et al . ( 1985 ) , _ n_(siii)=0.9 2.1 10@xmath55 @xmath3 , after correction for the likely background misplacement as explained in sect . 3.3 . most of the siii is assumed to be in component d , as is siiii ( see sect . 3.2.5 ) , thus for cloud d , the total column density is derived from the sum of sii and siii column densities : @xmath137 @xmath138 @xmath3 . for cloud c , the siicolumn yields @xmath139 @xmath140 @xmath3 . for the total sight - line , the sum of the total sight - line sii column density and the above siii column density gives @xmath141 @xmath142 @xmath3 . cassinelli et al . ( 1996 ) have derived _ n_(hei)@xmath143 @xmath3 from the hei504 line . with a cosmic abundance of 0.1 for he this implies @xmath144 @xmath3 which is in agreement with our value . the neutral hydrogen column density toward @xmath0 cma has been estimated by gry et al . ( 1985 ) by fitting the lyman-@xmath0 profile observed with copernicus . they found _ n_(hi)@xmath145 @xmath3 . recently , cassinelli et al . ( 1996 ) , using euve observations , have modelled the stellar spectrum and found an interstellar neutral hydrogen absorption of _ n_(hi)@xmath146 @xmath3 , which is in good agreement with the previous estimate . if we adopt the wider range for _ n_(hi ) , the neutral fraction is _ n_(hi)/_n_(h@xmath147)=0.05 0.13 for the total line of sight . this implies that about 90% of the matter of the sight - line is ionized . the share of neutral hydrogen between the main components is difficult to assess . as seen in table 2 , the way ni is distributed over the main components is very different from the way mgii is distributed , although the ionization potentials of the two elements are close ( 14.5 and 15.0 ev ) : the column density ratio between c and d is close to 2 for ni and close to 1/2 for mgii . this suggests that the share of neutral hydrogen between c and d can not be derived from that of neutral nitrogen . nevertheless , since component d is in general the more ionized , and since the ni ionization potential is slightly larger than that of hi , it is likely that the fraction of hi contained in d does not exceed the fraction of ni , which is 20% to 30% of the total neutral nitrogen in the line of sight ( sect . it is thus likely that no more than 30% of the total neutral gas lies in cloud d : _ n_(hi)@xmath148 @xmath3 . this gives _ n_(hi)/_n_(h@xmath147)@xmath149% for d. as for component c , what we can safely say is that it does contain a good fraction of the neutral gas since the di and oi features are centered at a velocity close to that of component c , and that it contains no more hi than that present in the total line of sight . this implies that cloud c is partially ionized with _ n_(hi)/_n_(h@xmath147)@xmath150% . table 4 summarizes all the information available on ionization ratios in the clouds or in the line of sight as a whole . the case of hydrogen has been discussed above . for sulfur , since all siii is assumed to be in cloud d , _ n_(sii)/_n_(siii ) is derived for cloud d only . for iron and magnesium , upper limits for their total gas phase abundances _ n_(x)@xmath151 were derived by adopting the minimum depletion values found in the interstellar medium as given by fitzpatrick ( 1996 , 1997 ) . then upper limits on their second ion column density were derived by _ n_(xiii ) = _ n_(x)@xmath151 - _ n_(xii ) . this yields lower limits for the ratios _ n_(xii)/_n_(xiii ) . for nitrogen , we proceed similarly but with the first and second ion column densities . for aluminium , upper limits on the _ n_(alii)/_n_(aliii ) ratio are derived taking the total sight - line alii column density as an upper limit on the individual cloud alii column densities . for silicon , the ionization ratios are derived directly from the data , taking the same upper limit for the siiv column densities of the individual clouds as that derived for the total siiv column density . in principle these ionization ratios allow a discussion of the ionization mechanisms involved in the gas . the most straightforward mechanism is photoionization by the star @xmath0 cma as well as its neighbour @xmath4 cma which is , as already mentioned , the strongest ionizing source in the solar neighbourhood ( vallerga & welsh 1995 ) . but the comparison of our ionization ratios with models and other observations found in the litterature is not really conclusive . in general , hi regions are dominated by singly ionized species like siii , sii and alii while photoionized hii regions show strong absorptions in siiv , siii and aliii ( e.g. fitzpatrick & spitzer 1994 , spitzer & fitzpatrick 1995 , jenkins & wallerstein 1996 , savage & sembach 1996a , b ) . the @xmath0 cma line of sight main components are thus atypical because , while hydrogen is substantially ionized , sii , alii and siii ( for cloud c ) or siiii ( for cloud d ) are still dominant and siiv is too weak to be detected . cowie et al . ( 1981 ) , in a study of the origin and the distribution of civ and siiv with 46 o - b stars observed with iue , found that almost all stars are associated with civ and siiv column densities of @xmath152 @xmath3 . they conclude that most of this absorption arises in photoionized hii regions . they also present a simple photoionization model predicting the hii regions civ and siiv column densities . for a 25000 k effective temperature star and an electron density of 1 @xmath7 it gives _ n_(civ)@xmath153 @xmath3 and _ n_(siiv)@xmath154 @xmath3 the latter being a factor of 10 higher than the @xmath0 cma upper limit . the model gives values compatible with the @xmath0 cma data for electron densities of 0.01 @xmath7 or less , much smaller than the range of values we derive ( see sect . york ( 1983 ) found evidence for hii regions in the line of sight to @xmath155 sco ( b2 iv ) , for which he compared the ionization ratios with those given by photoionization models with radiation fields of stars of different effective temperatures . for the effective temperature of @xmath0 cma(@xmath156 k , cassinelli et al . 1996 ) , he quotes the results from thuan ( 1975 ) for an electron density of 0.1 @xmath7 : sii / siii@xmath157 and siii / siiii@xmath158 . in component c , siii is clearly less ionized than in the model . for component d sii / siii is compatible with the quoted model value and siii is somewhat more ionized , which could be explained in view of the euv excess of @xmath0 cma and @xmath4 cma . on the other hand , spitzer & fitzpatrick ( 1995 ) have identified a photoionized hii region toward hd149881 , a halo star of similar spectral type ( b0.5iii ) , where sii / siii is less than 1 , i.e. similar to that of typical hii regions around hotter stars ( e.g. the three hii regions detected in the line of sight of the o9i+wc8 star @xmath159 velorum by fitzpatrick & spitzer 1994 ) and significantly lower than the ratio we measure for component d. clearly the comparison of all these observations shows that a more detailed ionization model by the stars is required before any conclusion can be drawn . however , detailed photoionizing models by these two sources which take into account their euv excess evidenced by euve ( cassinelli et al . 1995 , cassinelli et al . 1996 ) is beyond the scope of this paper . lcccccccc & [ c / h ] & [ n / h ] & [ o / h ] & [ mg / h ] & [ si / h ] & [ al / h ] & [ mn / h ] & [ fe / h ] + cloud c & - & @xmath160 & - & @xmath161 & @xmath162 & @xmath163 & @xmath164 & @xmath165 + cloud d & - & @xmath166 & - & @xmath161 & @xmath167 & @xmath168 & @xmath169 & @xmath170 + total & @xmath171 & @xmath172 & @xmath173 & @xmath174 & @xmath175 & @xmath176 & @xmath177 & @xmath178 + our ionization ratios can also be compared with those given by collisional ionization . the comparison in table 4 with the ratios given for various equilibrium temperatures in the model of sutherland & dopita ( 1993 ) shows that for component d there is an agreement for all ratios for temperatures around @xmath179 k. for component c , although the temperature ranges derived from siii / siiii and alii / aliii do not overlap , the ratios point to temperatures ranging from about @xmath180 k to about @xmath181 k. for both components , these temperatures are significantly higher than the kinetic temperatures of the clouds derived from profile fitting ( sect . . thus if the clouds are collisionally ionized , they can not be in equilibrium , which indicates that they are still in a cooling and recombining phase . trapero et al . ( 1996 ) in their study of halo high velocity clouds ( hvcs ) with the ghrs also proposed that the ionization of several hvcs in the line of sight to 23ori and @xmath182cma is due to collisional ionization . as in the case of the clouds in the @xmath0 cma sight - line they are not in ionization equilibrium and are overionized with respect to their kinetic temperatures : the ionization balance corresponds to a collisional ionization equilibrium near @xmath183 k while the temperature derived from the mgii lines is @xmath184 k. trapero et al ( 1996 ) suggest that the hvcs consist of already cooling warm ionized gas , originally probably collisionally ionized and heated by a shock . our results tend to support this interpretation for components c and d , in view of the depletion results which are discussed in sects . 6 and 7 . the gas phase abundance of an element x relative to h is defined by @xmath185 = \log\left[\frac{n({\rm x})}{n({\rm h})}\right ] - \log\left[\frac{n({\rm x})}{n({\rm h})}\right]_{\rm cosmic}\ ] ] where @xmath186_{\rm cosmic}$ ] is the solar abundance as given by anders & grevesse ( 1989 ) . all available [ x / h ] derived from our study are listed in table 5 . when possible their values for components c and d are listed separately . all relevant ionization stages are considered and when applicable ( si and al ) we use the sum of the various stage column densities . _ n_(oi ) is compared to _ n_(hi ) , as is _ n_(ni ) for consistency with other studies although in the case of ni , as mentioned in sect . 5.1 , some ionization effects might be involved . all other element column densities are compared to total hydrogen column density . as siiii is the dominant stage in component d , it is probable that feiii , mgiii and mniii are also significant . this is why in table 5 the gas phase abundance of their singly ionized form is listed as a lower limit only . it appears that in cloud d silicon is not at all , or only slightly , depleted . its gas phase abundance is at least 0.44 dex higher than the value given in the table 5 of jenkins & wallerstein ( 1996 ) for the depleted low velocity gas in the disk . in their fig . 6 , savage & sembach ( 1996a ) have plotted the mean gas phase abundances of several elements for different ism components found in the disk and in the halo . the value we find for silicon is the one they quote for the warm diffuse gas in the halo . in this component , most of the silicon is in the double ionized form ( siiii / siii@xmath187 14 ) . the ionization potentials of feii , mnii and mgii being lower than that of siii , it is expected that in cloud d , feiii , mniii and mgiii are at least two times more abundant than feii , mnii and mgii . the gas phase abundances we derive for iron , manganese and magnesium are thus , as for silicon , the same as those quoted for the halo gas in the fig . 6 of savage & sembach ( 1996a ) . in cloud c , the silicon and the iron gas phase abundances are at least 0.12 and 0.13 dex higher respectively than the values of the depleted low velocity gas in the disk ( jenkins & wallerstein 1996 ) . the silicon gas phase abundance corresponds to that of the warm gas in the disk in fig . 6 of savage & sembach ( 1996a ) and is within the error bars of that of the halo gas . in cloud c , silicon is mostly in its singly ionized form . we thus assume the same is true for iron , manganese and magnesium , for which as silicon , the gas phase abundances thus also correspond to those of the warm gas in the disk and are within the error bars of those of the halo gas . for the other elements , gas phase abundances were derived for the total sight - line only . the abundance of nitrogen , [ n / h]@xmath188 , is in good agreement with previous values derived from copernicus observations toward many stars : @xmath189 ( ferlet , 1981 ) , and @xmath190 ( york et al . , 1983 ) , as well as with the recent determination by meyer et al . ( 1998a ) with the ghrs at high resolution : @xmath191 . for oxygen , our lower limit [ o / h]@xmath173 is compatible with copernicus results of [ o / h]@xmath192 ( york et al . 1983 , keenan et al . 1985 ) and very close to the meyer et al . ( 1998b ) accurate determination of @xmath193 , which is only @xmath194 higher . if meyer s value is applicable to the @xmath0 cma sight - line this implies that its oi column density should not exceed @xmath195 @xmath3 . for carbon , the lower limit derived from the strongly saturated ciiline does not give any interesting constraint . the aluminium abundance in the sight - line of @xmath0 cma is known with a rather good precision . aluminium is found to be significantly depleted , but less than toward @xmath13 vir ( york & kinahan 1979 ) or toward @xmath155 sco ( york 1983 ) . jenkins & wallerstein ( 1996 ) show in their table 5 the aluminium abundance in the gas phase for the depleted low velocity gas in the disk for @xmath111=3 @xmath7 and @xmath111=0.1 @xmath7 . these values are respectively @xmath196 dex and @xmath197 dex lower than that derived from our data . conversely the @xmath0 cma value agrees with those found toward halo stars like hd 18100 ( savage & sembach , 1996b ) or hd 72089 , hd 22586 , hd 49798 and hd 120086 ( jenkins & wallerstein , 1996 ) . therefore it appears that both clouds have gas phase abundances corresponding to warm gas . component c abundances are compatible with those of warm clouds in the disk , and with those of halo clouds if the amounts of feiii , mgiii and mniii are significant . component d shows a depletion pattern similar to that of the halo clouds . table 6 summarizes the characteristics of components c and d derived from the observations . lll cloud & c & d + @xmath198 ( kms@xmath1 ) & 14.5 & 25 + @xmath199 ( kms@xmath1 ) & @xmath200 & 5.5 + ( k ) & @xmath201 & @xmath202 + @xmath203 ( kms@xmath1 ) & 3.2 4.0 & 4.0 4.8 + @xmath204 ) ( @xmath3 ) & 3.7 8.0 10@xmath16 & 1.0 1.6 10@xmath2 + ( hi)/_n_(h@xmath147 ) & @xmath205 & @xmath206 + @xmath111 ( @xmath7 ) @xmath70 & 0.18 0.54 & 0.38 1.82 + + @xmath70 from mgi / mgii , assuming ionization equilibrium and _ t_=7000 k warm diffuse ionized gas is a major component of the interstellar medium in our galaxy . it is detected mostly through pulsar dispersion measures and h@xmath13 emission , especially at high galactic latitudes . it is warm ( @xmath207 k ) and like our components it presents line ratios which differ from that of classical hii regions ( e.g. reynolds 1991 ) . the warm ionized gas is often seen to be associated with neutral gas as in the line of sight of hd93521 where spitzer & fitzpatrick ( 1993 ) observed clouds where neutral and ionized gas seem to be mixed in partially ionized clouds , as in component c. reynolds et al . ( 1995 ) have compared h@xmath13 and 21 cm emission in a @xmath208 region of the sky away from the galactic plane and found that 30% of the h@xmath13 emitting gas is spatially and kinematically associated with 21 cm emitting gas but show that in this case h@xmath209 and h@xmath210 gas occupy close but separate regions . here again the temperature of these `` h@xmath13 emitting hi clouds '' is about 8000 k , with electron densities around 0.2 @xmath7 and hydrogen ionization ratios ( _ n_(hi)/_n_(h@xmath147 ) ) ranging from 0.5 to 0.9 . all these characteristics are comparable to those derived here , especially for component c. the ionization mechanism of the diffuse ionized gas is still unclear ( e.g. reynolds et al . many authors ( domgrgen & mathis 1994 , miller & cox 1993 , reynolds & ogden 1979 , mathis 1986 ) invoke photoionization by o stars but often note that the fact that the uv photons can reach the diffuse gas at high latitude many parsecs away without being absorbed remains a mistery . ogden & reynolds ( 1985 ) use a model of a shock propagating through the ambient medium to explain the ionization of the long filament they have observed . spitzer & fitzpatrick ( 1993 ) mention penetrating radiation ( energetic particles , x - rays ) or in situ sources ( dark matter decay , sciama 1990 ) to explain the ionization of the h@xmath210 and h@xmath209 mixed clouds . whether components c and d observed in the line of sight of @xmath0 cma are of comparable nature as the widespread warm ionized gas remains an open question . our components can also be compared to other interstellar constituents by their elemental abundances . component d shows abundances similar to that of the high velocity halo clouds and to that of supernova remnants observed toward hd 72089 ( jenkins & wallerstein 1996 ) which are known to have been shocked . as noted by savage & sembach ( 1996a ) , shocks are the most important process in the destruction of dust grains , and the low depletion in the halo clouds should result from the processes that inject the clouds into the halo . we suggest that cloud d could have been subjected to shocks in the past . this suggestion is supported by the ionization ratios which are compatible with collisional ionization and recombination after cooling ( as in trapero et al . 1996 ) . for cloud c the situation is somewhat less clear , as it is slightly more depleted than cloud d and less ionized . however cloud c has abundances similar to that of the warm disk and halo gas . it could also have been shocked , as it has ionization ratios close to those given by collisional ionization at a temperature lower than that derived for cloud d. the characteristics of cloud c and d are also comparable to those of other local interstellar clouds . the electron density of the lic derived by gry et al . ( 1995 ) from the mgii / mgi ratio is @xmath211 @xmath7 and the central value corresponding to the prefered temperature of @xmath118 k implies a hydrogen ionization fraction close to 50% . wood & linsky ( 1997 ) have derived a similar range for the electron density of the lic in the line of sight of capella , @xmath212 @xmath7 , from the excited state cii column density , confirming that hydrogen is substantially ionized in the lic . a study of elemental abundances in the lic has been performed through the analysis of new ghrs ech - a data for @xmath4 cma by gry & dupin ( 1997 ) and suggests that the local cloud shows very little depletion , implying that most of the matter has been returned to the gas phase by some dust sputtering process . another small component in the line of sight of @xmath4 cma has been analysed with the same data and its preliminary study has been presented by dupin & gry ( 1997 ) . as in the case of our component d in the line of sight of @xmath0 cma , this component is warm ( @xmath213 k ) , strongly ionized ( more than 95% ) , and only slightly depleted . it also has ionization ratios compatible with those given by collisional equilibrium at @xmath214 k. these results show that this overall region of the local interstellar medium is substantially ionized and undepleted . the characteristics of the components and their similarities with diffuse high velocity halo clouds lead us to suggest that the canis majoris tunnel was affected by violent events in the past , perhaps related to the local bubble formation . we have studied the nature of the ionized gas that dominates the @xmath0 cma sight - line . there appears to be two distinct components ( named c and d ) with ionization fractions greater than 40% and 90% respectively . constraints have been derived for the depletion which are similar to those for the warm diffuse gas in both the disk and halo for component c , and to those for the high velocity halo clouds for component d. silicon seems to be undepleted in cloud d and only slightly depleted in cloud c , which is also the case for the other elements studied . the ionization ratios observed are roughly compatible with those given by theoretical calculations of collisional ionization equilibrium at temperatures close to 20000 k. as the kinetic temperatures of the clouds are actually a factor of two lower , we conclude that if the clouds are collisionally ionized , they are out of equilibrium . nevertheless photoionization by @xmath0 cma and @xmath4 cma has not been ruled out and the data we present should be analysed with more complete photoionization models taking into account the strong euv excess of these stars . the processes which have led to this physical state could be the same as those which have given rise to the high velocity clouds detected in the halo . new obervations of the interstellar medium in the sight - line to @xmath4 cma lead to similar conclusions for the lic and another component detected toward this star . a shock which travelled in the past through these clouds could have ionized them and disrupted the dust grains giving rise to the ionization and low depletion values found ; the clouds would now be recombining after cooling . thus , the main components detected towards @xmath0 cma could result from the blast - wave that created the so - called local bubble . we thank alfred vidal - madjar for making the data available to us and dan welty for providing his profile fitting software . we are grateful to alan harris and to our referee k. de boer for suggestions which led to substantial improvements on the manuscript . + allan r.j . , clegg r.e.s . , dickinson a.s . , flower d.r . , 1988 , mnras , 235 , 1245 anders e. , grevesse n. , 1989 , geochim . acta 53 , 197 aldrovandi s.m.v . , pquignot d. , 1973 , a&a 25 , 137 bergeson s.d . , lawler j.e . , 1993 , apj 414 , l137 bruhweiler f.c . , cheng k.p . , 1988 , apj 355 , 188 cassinelli j.p . , cohen d.h . , mac farlane et al . , 1995 , apj 438 , 932 cassinelli j.p . , cohen d.h . , mac farlane et al . , 1996 , apj 460 , 949 cowie l.l . , taylor w. , york d.g . , 1981 , apj 248 , 528 domgrgen h. , mathis j.s . , 1994 , apj 428 , 647 cox d.p . , reynolds r.j . , 1987 , ara&a 25 , 203 duncan d. , 1992 , goddard high resolution spectrograph instrument handbook , version 3.0 , space telescope science institute dupin o. , gry c. , 1997 , in iau coll . 166 , the local bubble and beyond , eds breitschwerdt d. & freyberg m. ( springer , heidelberg ) , lect . notes in phys . 506 ferlet r. , laurent c. , vidal - madjar a. , york d.g . , 1980a apj , 235 , 478 ferlet r. , vidal - madjar a.,laurent c. , york d.g . , 1980b apj , 242 , 576 ferlet r. , 1981 , a&a 98 , l1 fitzpatrick e.l . , 1996 , apj 473 , l55 fitzpatrick e.l . , 1997 , apj 482 , l199 fitzpatrick e.l . , spitzer l. , 1994 , apj 427 , 232 frisch p.c . , york d.g . , 1983 , apj 271 , l59 frisch p.c . , welty d.g . , york d.g . , fowler j.r . , 1990 , apj 357 , 514 gry c. , york d.g . , vidal - madjar a. , 1985 , apj 296 , 593 gry c. , lemonon l. , vidal - madjar a. , lemoine m. , ferlet r. , 1995 , a&a 302 , 497 gry c. , dupin o. , 1997 , in iau coll . 166 , the local bubble and beyond , eds breitschwerdt d. & freyberg m. ( springer , heidelberg ) , lect . notes in phys . 506 harris a.w . , mas hesse j.m . , 1986 , apj 308 , 240 jenkins e.b . , wallerstein g. , 1996 , apj 462 , 758 keenan f.p . , hibbert a. , dufton p.l . , 1985 , a&a 147 , 89 lallement r. , bertin p. , 1992 , a&a 266 , 479 lallement r. , bertin p. , ferlet r. , vidal - madjar a. , bertaux j.l . , 1994 , a&a 286 , 898 mathis j.s . , 1986 , apj 301 , 423 meyer d.m . , cardelli j.a . , sofia u.j . , 1998a , apj ( in press ) meyer d.m . , jura m. , cardelli j.a . , 1998b , apj ( in press ) miller w.w . , cox d.p . , 1993 , apj 417 , 579 morton d.c . , 1991 , apjs 77 , 119 nussbaumer h. , storey p.j . , 1986 , a&as 64 , 545 ogden p.m. , reynolds r.j . , 1985 , apj 290 , 238 phillips a.p . , gondhalekar p.m. , blades j.c . , 1981 , mnras 195 , 485 reynolds r.j . , ogden p.m. , 1979 , apj 229 , 942 reynolds r.j . , 1991 , in iau symp . 144 , the interstellar disk - halo connection in galaxies , ed bloemen h. , ( kluwer , dortretch ) , 67 reynolds r.j . , tufte s.l . , kung d.t . , mccullough p.r . , heiles c. , 1995 , apj 448 , 715 rogerson j.b . , spitzer l. , drake j.f . , et al . , 1973 , apj 181 , l97 savage b.d . , sembach k.r . , 1996a , ara&a 34 , 279 savage b.d . , sembach k.r . , 1996b , apj 470 , 893 sciama d.w . , 1990 , apj 364 , 549 sofia u.j . , cardelli j.a . , savage b.d . , 1994 , apj 430 , 650 spitzer l. , fitzpatrick e.l . , 1993 , apj 409 , 299 spitzer l. , fitzpatrick e.l . , 1995 , apj 445 , 196 sutherland r.s . , dopita m.a . , 1993 , apjs 88 , 253 thuan t.x . , 1975 , apj 198 , 307 trapero j. , welty d.e . , hobbs l.m . , et al . , 1996 , apj 468 , 290 vallerga j.v . , welsh b.y . , 1995 , apj 444 , 702 vidal - madjar a. , laurent c. , bonnet r.m . , york d.g . , 1977 , apj 211 , 91 welsh b.y . , 1991 , apj 373 , 556 welsh b.y . , craig n. , vedder p.w . , vallerga j.v . , 1994 , apj 437 , 638 welty d.e . , hobbs l.m . , york d.g . , 1991 , apjs 75 , 425 wood b.e . , linsky j.l . , 1997 , apj 474 , l39 york d.g . , 1983 , apj 264 , 172 york d.g . , kinahan b.f . , 1979 , apj 228 , 127 york d.g . , spitzer l. , bohlin r.c . , hill j. , jenkins e.b . , savage b.d . , snow t.p , 1983 , apj 266 , l55

we present the results of an ongoing weak lensing survey conducted with the subaru telescope whose initial goal is to locate and study the distribution of shear - selected structures or _ halos_. using a suprime - cam imaging survey spanning 21.82 deg , we present a catalog of 100 candidate halos located from lensing convergence maps . our sample is reliably drawn from that subset of our survey area , ( totaling 16.72 deg ) uncontaminated by bright stars and edge effects and limited at a convergence signal to noise ratio of 3.69 . to validate the sample detailed spectroscopic measures have been made for 26 candidates using the subaru multi - object spectrograph , focas . all are confirmed as clusters of galaxies but two arise as the superposition of multiple clusters viewed along the line of sight . including data available in the literature and an ongoing keck spectroscopic campaign , a total of 41 halos now have reliable redshifts . for one of our survey fields , the xmm lss field , we compare our lensing - selected halo catalog with its x - ray equivalent . of 15 halos detected in the xmm - lss field , 10 match with published x - ray selected clusters and a further 2 are newly - detected and spectroscopically confirmed in this work . although three halos have not yet been confirmed , the high success rate within the xmm - lss field ( 12/15 ) confirms that weak lensing provides a reliable method for constructing cluster catalogs , irrespective of the nature of the constituent galaxies or the intracluster medium .

0707.2249

clusters of galaxies represent the most massive bound systems in the cosmos . although they result from non - linear structure evolution , the departure from linear growth is modest compared to that for less massive objects . as a result , simple analytic models can provide an accurate indication of their expected number density at various redshifts . this is the primary reason why cluster of galaxies are considered to be valuable cosmological probes . cosmological attention has focused on the redshift - dependent number of clusters , @xmath1 , whose mass exceeds a certain threshold . this is one of the most straightforward observables , and is a function of the cluster mass function , @xmath2 , and the evolution of comoving volume @xmath3 . the mass function is obtained from the growth rate of density fluctuations , @xmath4 , numerically @xcite under the assumption of a particular theory of structure formation , e.g. the currently - popular cold dark matter ( cdm ) model . since both @xmath4 and @xmath3 are dependent on the cosmological model , useful constraints could be estimated by comparing @xmath1 with various model predictions . to make progress , e.g. on the dark energy equation of state parameter @xmath5 , data on several thousand clusters to @xmath61 is thought to be required @xcite , and maintaining an accurate and uniform mass threshold is critical . most early work focused on selecting clusters optically , with detection techniques that have improved over the decades : e.g. matched - filter @xcite , red - sequence @xcite , cut - and - enhance @xcite . optically selected samples have traditionally suffered from uncertainties in the optical richness - mass relation , although there has been recent progress in calibrating the closely related richness - velocity dispersion relation using the large maxbcg sample of clusters identified in the sloan digital sky survey @xcite . so far , x - ray samples have been the most popular cosmological probes e.g. @xcite , @xcite , @xcite . luminosity ( @xmath7 ) or temperature ( @xmath8)-limited samples offer simpler selection functions because the observables , @xmath7 , @xmath8 , are a fair estimate of the cluster mass , calibrated through empirical scaling relations . the derived mass does depend , however , on the assumed dynamical state of the system . unrelaxed clusters , arising for example from recent mergers , will introduce scatter in the scaling relation . a recent study by @xcite points out that at least half of 10 z @xmath9 0.2 cluster cores show unrelaxed features and a scatter of @xmath10 around the mean scaling relation . weak gravitational lensing , which analyses the coherent shear pattern of background galaxies , can potentially provide estimates of the cluster mass _ regardless of its dynamical state or the properties of the constituent galaxies_. for some years , the method has been used to calibrate mass obtained from x - ray data @xcite . @xcite concluded that x - ray mass measurements are consistent with those from weak lensing , particularly for relaxed systems which thus offer a useful cosmological probe@xcite . a natural extension of this progress is thus to consider selecting clusters directly from weak lensing signals . the development of panoramic imaging surveys has now made this a practical proposition . @xcite reported the first discovery of a cluster located from a weak lensing convergence map , during the course of conducting the deep lens survey @xcite . @xcite later undertook a systematic search of mass concentrations on a 2 deg@xmath0 field using the suprime - cam imager on subaru . they detected several significant ( s / n@xmath11 ) candidates , one of which was later spectroscopically identified as a cluster at @xmath12 = 0.41 . @xcite investigated 50 randomly - selected vlt fors1 fields , spanning 0.64 deg@xmath0 in total , and found 5 shear - selected candidates , each associated with an overdensity in luminosity . the first results from the deep lens survey , based on an area of 8.6 deg@xmath0 have also recently emerged @xcite . the above pioneering studies have demonstrated that clusters can be located directly via weak lensing . however , key issues , including the optimum selection threshold , the rate of spurious detection and the degree of mass completeness at a given redshift , crucial for any eventual cosmological application , remain unresolved . at the present time , theoretical studies offer the only insight into these issues . projection is one of the most troublesome aspects of a weak lensing survey , because of the relatively broad window function . unrelated structures contributing to the signal would lead to an overestimate of the cluster mass . moreover , as the noise in the convergence map arises largely from shot noise in the ellipticity distribution of background galaxies , some fraction of genuine clusters might be missed in a shear selected catalog . n - body and ray - tracing simulations @xcite have concluded that , for systems whose convergence signal lie above a 4 standard deviation ( @xmath13 ) threshold , 60 - 75% of clusters can reliably recovered ( completeness ) . likewise , for peaks detected in the simulated data using typically - used algorithms , 60 - 75 % represent genuine clusters ( efficiency ) . the difference in these figures between the various studies is largely due to differences in the lower mass limit adopted in the studies . this series of papers is motivated by the need to address these key issues observationally . the survey we describe is a natural and ongoing extension of the 2 deg@xmath0 survey of @xcite ; to date a total field of 21.82 deg@xmath0 has been imaged . this first paper describes the imaging survey and discusses the validation of the candidates found , both via spectroscopic verification and comparison with clusters located via x - ray techniques . later papers in the series will extend the spectroscopic follow - up to the full sample and will consider the feasibility of deriving cosmological constraints from both this survey and future enhanced versions . we note that a similary motivated program has been initiated by @xcite , @xcite and @xcite , all of which made used of the imaging data taken by 2.5 m vlt survey telescope . we compare the their conclusions with our own in this paper . we note that a similarly motivated program has been initiated by @xcite and @xcite whose conclusions we compare with our own in this paper . a plan of the paper follows . we discuss the imaging strategy and data analysis in 2 , including construction of the cluster catalog and its reliability . in 3 , we describe our initial spectroscopic follow - up with subaru and address the completeness by comparing x - ray selected clusters on one of our survey area where relevant x - ray data is available . we summarize our conclusions in 4 . [ cols="<,<,<,^,^,^,^,^,^,^ " , ] in order to evaluate the efficiency of our weak lensing survey for locating cluster halos , we considered that a comparison with a sample of x - ray selected clusters would be highly advantageous @xcite . therefore , our survey fields were primarily selected to contain x - ray data as shown in table [ tab : survey_field ] . we set a minimum rosat exposure time , @xmath14 , of 25 ksec ensuring a detection limit of @xmath15erg / s at @xmath60.5 ( for @xmath16,@xmath17 , @xmath18 ) . this corresponds to @xmath19 , which is well matched to the likely mass detection limit of our weak lensing survey @xcite . more sensitive x - ray missions , xmm - newton and chandra , have been surveying rosat fields , in part , to deeper limits . these include the `` lynx '' , `` lockman hole '' and `` uk 13 hr deep field '' in our target list . xmm is also actively involved in international campaigns of multi - wavelength wide field ( @xmath20 ) observations such as the `` cosmos '' , `` xmm - lss '' and `` sxds '' fields which we also included . among these , the xmm - lss field @xcite is of particular interest given its panoramic area and published cluster catalogs @xcite . finally , we included the deimos deep2 survey fields @xcite where spectra of @xmath9 50,000 faint galaxies will eventually become available . this will enable close correlations of lensing mass and various measures of the luminosity density as was recently pioneered for the cosmos field @xcite . although x - ray data is not available for these deep fields at the current time , such observations are planned and will likely become available soon . the imaging observations were largely carried out on may 1 - 4 and september 27 , 2003 as part of an `` intensive program '' of subaru telescope . the lynx field was observed on january 29 - 30 , 2003 , sxds data was obtained from the public archive and the cosmos field observed on february 18 and 21 , 2004 . the suprime - cam field size is 0.25 deg@xmath0 and all observations were conducted with the @xmath21 filter ( except the cosmos field which was observed in @xmath22-band to enable direct comparison with acs@xmath23hst s f814w images ) . the total exposure time was 30 minutes for each pointing , taken via four 7.5 minute exposures in a dither pattern of spacing @xmath9 1 arcmin . [ fig : fwhmee ] ( top ) shows the seeing ( fwhm ) of each 7.5 minute exposure over a typical five night observing sequence . the bottom panel shows ellipticities of moderately bright unsaturated stars which provide a measure of the psf anisotropy ; the raw ellipticity is mostly 2 - 4 % ( and occasionally @xmath24 5 % ) . we discuss the question of the psf anisotropy in appendix [ sec : imagequality ] . thus far , we have surveyed over 21.82 deg@xmath0 as summarized in table [ tab : survey_field ] . the useful field area excludes the surroundings of bright stars and galaxies and field boundaries . individual pointings whose seeing was worse than 0.9 arcsec were also excluded ; this occurred for only 5 % of the clear time ( see section [ sec : wlana ] ) . the data reduction procedures for the present survey closely followed those described in section 3.1 of @xcite , enhanced as discussed below . normally , with suprime - cam images , each ccd exposure is `` mosaic - stacked '' to yield a single image of a particular pointing . a simple geometrical model is used for the focal plane astrometry whose parameters include the effects of camera distortion , the displacement and rotation of each detector from a defined fiducial location and the offset and the rotation of the dithered exposures . the best fit parameters are obtained by minimizing the positional difference of control stars ( 70@xmath9100 stars per ccd ) held common for each exposure . the residual alignment error in this procedure is @xmath9 0.5 pixel rms ( 0.1 arcsec rms ) . such a residual is sufficiently small for most imaging applications . however , in seeing better than 0.6 arcsec ( fwhm ) , a misalignment of 0.1 arcsec between two images introduces a @xmath252 % ellipticity on the stacked image which is a serious issue for weak lensing studies . a further improvement is thus necessary . the residual @xmath26 from a reference frame is parameterized as a polynomial function of field position @xmath27 as : @xmath28 pixel values are estimated via linear interpolation of neighboring four pixels . the coefficients @xmath29 and @xmath30 are then obtained by minimizing the variance of the residuals . a sixth order polynomial is usually sufficient for this purpose . each individual ccd image is ` warped ' using this polynomial correction prior to stacking . this process reduces the alignment error to @xmath9 0.07 pixel ( 0.014 arcsec ) and is similar to the `` jelly ccd '' model described in @xcite . although an external stellar catalog would ideally be used for accurate astrometry , no such data is available at the relevant faint limits ( r @xmath24 22 ) . we therefore employ the first exposure , corrected by the simple geometrical model discussed above , as the basic reference frame . we noticed that this refined procedure still introduces some artificial deformation . [ fig : warp](a ) represents the raw image of one ccd whereas fig . [ fig : warp](b ) is that slightly rotated by @xmath31 rad , a typical value , using the mapping described above . clearly some degree of artificial deformation is introduced . after some experimentation , we found this deformation arises from the undersampled nature of the suprime - cam images . by adopting 2@xmath322 oversampling prior to rotation , the resulting ellipticity field shows no sign of image deformation . however , this is a computationally a very time - consuming solution . accordingly , in our final analysis , instead of using oversampled images , we modified the mapping procedure itself . we estimate the pixel values of the target images from 3rd order bi - linear polynomial interpolation of 4@xmath324 source pixels rather than the linear interpolation four neighboring pixels in the previous procedure . this mapping process avoids introducing image deformation and is significantly quicker computationally . object finding and shape measurement was executed on the mosaic - stacked images using the _ imcat _ software suite developed by nick kaiser . a threshold _ nu_=10 was adopted . photometric calibration used landolt standard stars @xcite and the faint standards of @xcite . we adopt the vega magnitude system in the following . galaxies are distinguished from stars via their half light radius , @xmath33 , viz : @xmath34 where @xmath35 and @xmath36 are the half light radius of a stellar image and its rms respectively . the galaxy size distribution is shown in fig . [ fig : size_gd140 ] . [ fig : ncnt_gd140 ] shows the cumulative number density of galaxies as a function of @xmath21-band magnitude . the surface density exceeds 50 arcmin@xmath37 when the seeing is superb ( 0.47 arcsec ) and is @xmath38 arcmin@xmath37 in those poor seeing images ( @xmath240.9 arcsec ) discarded from our analysis ( fig . [ fig : seeing_gdens_pgamma](a ) ) . table [ tab : survey_field ] lists the seeing and the galaxy density for each field . finally , we masked all objects close to bright stars ( within 18 arcsec for @xmath39 , 90 arcsec for @xmath40 ) . light halos around bright stars can introduce spurious galaxies . object shapes are represented by the ellipticities , @xmath41 where @xmath42 are gaussian - weighted quadrupole moments of the surface brightness distribution . the point spread function ( psf ) of the images is usually smeared by various instrumental effects such as optical aberrations and the tracking error of the telescope . the psf anisotropy is estimated based on images of stars , and the galaxy images are corrected so that images of neighboring stars are re - circularized . galaxy ellipticities are then corrected as : @xmath43 where the asterisk designates a stellar value , @xmath44 is the smear polarisability tensor and is mostly diagonal @xcite . @xmath45 is evaluated using stars in the field of view and modeled as 5th order bi - polynomial function of position . eqn.[psmcorrection ] then applies this for the galaxy images . this correction is carried out independently on each pointing . we further justify the correction procedure in appendix [ sec : imagequality ] . the shear induced by gravitational lensing , @xmath46 , is diluted by atmospheric seeing . @xcite developed a prescription to convert the observed ellipticities to a pre - seeing shear as @xmath47 where @xmath48 is the pre - seeing shear polarisability tensor defined as @xmath49 @xmath50 is the shear polarisability tensor defined in @xcite , and @xmath51 is the stellar shear polarisability tensor . note that ( the inverse of ) @xmath48 represents the degree of dilution . since the @xmath50 and @xmath44 are mostly diagonal , we replace the tensors in eqn . [ pgamma ] with their trace and evaluate @xmath48 as a scalar . the average value , @xmath52 , over all galaxies ( @xmath53 ) of each pointing is shown in fig [ fig : seeing_gdens_pgamma](b ) . @xmath52 decreases slightly as the seeing worsens but the change is not very large ( 0.4 to 0.3 ) . this is because we only select larger galaxies compared with the seeing size . thus , the dilution factor is 30@xmath940 % regardless of the seeing . in the mean time , we compare the first component of galaxy ellipticities , @xmath54 , of suprime - cam and acs / hst images taken in the cosmos field , and the result is shown in fig [ fig : suprime_acs ] . the ellipticities observed by suprime - cam are in fact diluted by 36 % compared with those of acs , which is consistent with fig [ fig : seeing_gdens_pgamma](b ) . to calculate @xmath48 as a function of position we employed a `` smoothing '' scheme @xcite . we took the median @xmath48 for 20 neighboring galaxies on the @xmath55-magnitude plane ( where @xmath55 is a measure of object size adopted in the _ imcat _ suite ) . in deriving the mean , the weight @xmath5 on an individual measure is taken to be : @xmath56 where @xmath57 is the variance of the raw @xmath58 of those 20 neighbors , obtained using the raw @xmath48 . @xmath59 is the variance of all of the galaxies in the catalog ( @xmath9 0.4 ) . in general , the weighted value of a quantity @xmath60 is calculated as @xmath61 . the method we adopt is based on that adopted by @xcite . more sophisticated methods have since been developed and the variants are summarized by @xcite . in their notation , our method is very similar to the procedure termed `` lv '' . based on the results of the step simulation study , @xcite concluded that both the `` ksb+ '' method , modified by @xcite and implementations of `` bj02 '' method @xcite are able to reconstruct input shears to a few percent level . to calibrate our method , we analyzed the simulated data provided by @xcite . for the model designated `` psf3 '' , we underestimate the input shear by 5% for @xmath62 = 0.1 and 0.05 , whereas for @xmath63 , the difference , @xmath64 is insignificant . an error of 5 % in the recovered shear is competitive with most of the methods discussed by @xcite ( @xmath9 7 % is a typical error ) . a 5% shear error would induce a similar uncertainty in the mass estimate of a typical halo . such an error is considered adequate for the applications envisaged . the dimensionless surface mass density , @xmath65 , is estimated from the shear field @xmath66 by the @xcite inversion algorithm as : @xmath67,\ ] ] where @xmath68 is defined as @xmath69 which is a complex convolution kernel for @xmath70 to obtain the shear @xmath58 . in the actual implementation , we smoothed to avoid the effect of noise . a smoothing scale of @xmath71 arcmin was chosen following the discussion given by @xcite . we adopt 15@xmath3215 arcsec@xmath0 square grids and then calculate @xmath65 on each grid using eqn.[eqn : ks93 ] to obtain the @xmath70 map . in order to estimate the noise of the @xmath70 field , we randomized the orientations of the galaxies in the catalog and created a @xmath72 map . we repeated this randomization 100 times and computed the rms value at each grid point where @xmath70 is computed . assuming the @xmath70 error distribution is gaussian , this rms represents the 1-@xmath13 noise level , and thus the measured signal divided by the rms gives the signal / noise ratio , @xmath73 , of the @xmath70 map at that point . the gaussian - smoothed signal / noise map is then searched for mass concentrations . @xcite concluded that a truncated nfw filter applied to the aperture mass map , @xmath74 , is the most efficient detection technique . such optimization may be necessary to improve the efficiency of future very wide field surveys where thousands of clusters are sought . here we adopt a simpler technique in order to evaluate its effectiveness in a direct comparison of lensing and x - ray techniques . figure [ fig : saclay_halomap ] shows the results of a halo search in one of our fields : the xmm - wide field . the red contour shows the @xmath70 s / n map where the threshold and increment are set at 2 and 0.5 , respectively . the blue contour shows the surface number density of moderately bright galaxies ( @xmath75 ) . local peaks are searched on the @xmath70 s / n map and their positions are marked as open and filled circles . figure [ fig : deep02_halomap ] to [ fig : deep23_halomap ] show the @xmath70 s / n maps for the remainder of our survey fields . while visually inspecting galaxy concentrations around the detected halos , we noticed that less concentrated halos tend to occur preferentially near bright stars and field boundaries . since regions near bright stars are masked ( section [ sec : galaxycatalogs ] ) conceivably the discontinuity in the faint background galaxy distribution could cause spurious peaks . to avoid this , we reject halos occurring within a 4 arcmin radius of bright stars ( @xmath76 @xmath77 11 ) and within 2.3 arcmin of the field boundary . these restrictions reduce the survey area by 23 % to what we will refer to as the _ secure survey area _ ( 16.72 deg@xmath0 ) . halos found within the secure area are termed the _ secure sample_. it is certainly possible that a significant fraction of halos lying in the non - secure area are genuine clusters . we will discuss this further in a later paper concerned with their spectroscopic follow - up ( green et al , in preparation ) . detailed inspection of the halo candidates and the spectroscopic follow - up discussed below revealed our completeness is high to a limiting signal to noise in the convergence map of 3.69 . table [ tab : halocat ] , [ tab : halocat2 ] lists the secure sample with s / n @xmath24 3.69 . in this table , n@xmath78 represents the number of moderately bright ( @xmath79 ) galaxies within 2 arcmin , indicative of the galaxy concentration . concerning the optimum threshold for the significance , decreasing it will increase the halo sample but likely introduce more false detections . the optimum value should be set based on the spectroscopically - observed true / false rate . investigating the rate is a major goal of our study . in this work , the least significant spectroscopically - identified halos have s / n = 3.69 ( xmm - wide n=23 ) . we adopted the threshold of 3.69 for this work so that all the spectroscopically followed - up samples are included in the table . we adopted significance maps for the selection of candidates rather than kappa maps . this is because we would like to minimize contamination by the false peaks . however , the effective kappa threshold varies over the field , so we may encounter a `` completeness '' problem ; i.e. halos that have high kappa value are lost from the list . we investigated such omissions in the xmm - wide field ( figure [ fig : saclay_halomap ] ) , and only one halo is found in this category , with @xmath80 & @xmath81 where the @xmath82 is calculated as @xmath83 . the `` @xmath84 '' is estimated globally over the entire xmm - wide field kappa map , and is 0.018 here . this halo is lost because the local noise is as high as 0.022 . in practice , it will be very hard to generate completely uniform data sets over the entire field of a survey because weather and seeing conditions will vary . we will have to optimize the @xmath82 on a field by field basis based on the data quality of each field . therefore , our strategy is the following : at first , we collect reliable cluster samples based on the significance , and then users of the catalog can set their own kappa threshold or mass threshold to carry out their studies . this work represents the results of the first step above . we list the kappa values in table [ tab : halocat ] , [ tab : halocat2 ] for reference . [ tab : halocat ] llllllllllll field & n & i d & ra & dec & @xmath70s / n & @xmath70 & n@xmath78 & focas & known & nedg & note + deep02 & 00 & - & 37.32 & 0.63 & 4.44 & 0.099 & 16 & - & - & 1.35 & + & 01 & - & 37.87 & 0.51 & 4.39 & 0.083 & 24 & - & - & - & + & 02 & - & 37.38 & 0.41 & 4.25 & 0.120 & 19 & - & - & 0.73 & + & 04 & - & 37.15 & 0.73 & 4.11 & 0.086 & 25 & - & - & 0.10 & + & 05 & - & 37.31 & 0.44 & 4.07 & 0.131 & 20 & - & - & 1.03 & + & 06 & - & 37.72 & 0.69 & 3.97 & 0.081 & 18 & - & - & 0.86 & + & 07 & - & 37.86 & 0.57 & 3.95 & 0.074 & 18 & - & - & 0.92 & + & 08 & sl j0228.4 + 0030 & 37.12 & 0.51 & 3.93 & 0.085 & 49 & - & 0.46(p ) & - & vgcf 46 + & 09 & sl j0228.2 + 0033 & 37.07 & 0.55 & 3.84 & 0.102 & 33 & - & 0.50(p ) & - & sdss ce j037.099808 + 00.540769 + sxds & 00 & - & 34.29 & -5.59 & 5.33 & 0.057 & 26 & - & - & - & + & 01 & - & 34.38 & -4.86 & 4.14 & 0.084 & 24 & - & - & - & + & 02 & - & 34.61 & -4.41 & 3.96 & 0.059 & 16 & - & - & - & + & 03 & - & 34.74 & -4.70 & 3.90 & 0.044 & 19 & - & - & - & + & 04 & - & 34.96 & -5.12 & 3.86 & 0.069 & 21 & - & - & - & + & 05 & - & 34.41 & -4.50 & 3.77 & 0.056 & 26 & - & - & - & + xmm - wide & 00 & sl j0221.7 - 0345 & 35.44 & -3.77 & 8.15 & 0.156 & 72 & - & 0.43 & - & xlssc 006 + & 01 & sl j0225.7 - 0312 & 36.43 & -3.21 & 5.72 & 0.108 & 41 & 0.14 & - & - & lris z = 0.14 + & 02 & sl j0224.4 - 0449 & 36.10 & -4.82 & 5.06 & 0.074 & 40 & 0.49 & - & - & + & 04 & - & 35.34 & -3.50 & 4.91 & 0.082 & 21 & - & - & - & + & 08 & sl j0222.3 - 0446 & 35.48 & -3.80 & 4.33 & 0.081 & 29 & - & - & - & lris z = 0.41 + & 10 & - & 36.25 & -4.25 & 4.20 & 0.062 & 23 & - & - & - & + & 12 & sl j0224.5 - 0414 & 36.13 & -4.24 & 4.06 & 0.057 & 70 & 0.26 & - & - & lris z = 0.26 + & 15 & sl j0225.3 - 0441 & 36.34 & -4.70 & 3.94 & 0.091 & 34 & 0.26 & - & - & + & 16 & sl j0228.1 - 0450 & 37.03 & -4.84 & 3.94 & 0.072 & 31 & 0.29 & - & - & + & 17 & sl j0226.5 - 0401 & 36.63 & -4.02 & 3.90 & 0.079 & 37 & - & 0.34 & - & xlssc 014 + & 19 & sl j0227.7 - 0450 & 36.94 & -4.85 & 3.81 & 0.064 & 43 & - & 0.29 & - & pierre et al . ( 2006 ) + & 20 & - & 35.98 & -3.77 & 3.81 & 0.048 & 20 & - & - & - & + & 21 & sl j0228.4 - 0425 & 37.12 & -4.43 & 3.80 & 0.055 & 49 & - & 0.43 & - & xlssc 012 + & 22 & sl j0225.4 - 0414 & 36.36 & -4.25 & 3.72 & 0.073 & 43 & 0.14 & - & - & + & 23 & sl j0222.8 - 0416 & 35.71 & -4.27 & 3.69 & 0.049 & 52 & 0.43,0.19,0.23 & - & - & + lynx & 00 & - & 131.91 & 44.80 & 5.84 & 0.121 & 20 & - & - & - & + & 01 & - & 132.59 & 44.07 & 5.01 & 0.083 & 43 & - & - & - & + & 03 & - & 131.83 & 44.86 & 4.57 & 0.139 & 23 & - & - & - & + & 05 & - & 131.77 & 44.85 & 4.37 & 0.110 & 13 & - & - & - & + & 07 & - & 132.69 & 44.95 & 4.15 & 0.105 & 31 & - & - & - & + & 08 & sl j0850.5 + 4512 & 132.64 & 45.20 & 4.02 & 0.085 & 53 & 0.19 & 0.24(p ) & - & nsc j085029 + 451141,lris@xmath85 + & 09 & - & 131.47 & 44.96 & 4.02 & 0.076 & 26 & - & - & - & + & 10 & - & 133.02 & 44.14 & 4.00 & 0.108 & 23 & - & - & - & + & 12 & - & 132.37 & 44.38 & 3.90 & 0.081 & 36 & - & - & - & + & 13 & - & 132.41 & 44.37 & 3.90 & 0.072 & 31 & - & - & - & part of n=12 + & 14 & - & 132.54 & 44.07 & 3.86 & 0.066 & 48 & - & - & - & + & 15 & - & 132.81 & 44.35 & 3.77 & 0.077 & 39 & - & - & - & + & 16 & - & 132.31 & 44.30 & 3.75 & 0.072 & 44 & - & - & - & + & 17 & - & 131.40 & 44.94 & 3.74 & 0.084 & 25 & - & - & 0.15 & + cosmos & 00 & sl j1000.7 + 0137 & 150.19 & 1.63 & 6.11 & 0.113 & 64 & 0.22 & 0.20(p ) & - & nsc j100047 + 013912 + & 01 & sl j1001.4 + 0159 & 150.35 & 1.99 & 5.64 & 0.098 & 32 & - & 0.85(p ) & - & @xcite + & 02 & sj j0959.6 + 0231 & 149.92 & 2.52 & 4.74 & 0.067 & 83 & - & 0.73(p ) & - & @xcite + & 05 & - & 149.65 & 1.55 & 3.92 & 0.078 & 47 & - & - & - & + & 07 & - & 150.19 & 2.01 & 3.88 & 0.070 & 36 & - & - & - & + [ tab : halocat2 ] llllllllllll field & n & i d & ra & dec & @xmath70s / n & @xmath70 & n@xmath78 & focas & known & nedg & note + lockman & 00 & sl j1057.5 + 5759 & 164.39 & 58.00 & 6.28 & 0.109 & 68 & 0.60 & - & - & + & 03 & sl j1051.5 + 5646 & 162.88 & 56.77 & 4.97 & 0.082 & 31 & 0.33 , 0.35 & - & - & + & 05 & sl j1047.3 + 5700 & 161.84 & 57.01 & 4.56 & 0.103 & 56 & 0.30 , 0.24 & - & - & + & 06 & sl j1049.4 + 5655 & 162.35 & 56.93 & 4.51 & 0.095 & 47 & 0.42 & - & - & lris z = 0.31 + & 09 & sl j1055.4 + 5723 & 163.86 & 57.38 & 4.22 & 0.086 & 20 & - & - & - & lris z = 0.38 + & 10 & sl j1051.6 + 5647 & 162.92 & 56.78 & 4.20 & 0.068 & 49 & 0.33 , 0.25 & - & 0.05 & part of sl j1051.5 + 5646 + & 11 & sl j1053.4 + 5720 & 163.35 & 57.34 & 4.08 & 0.064 & 50 & - & 0.34 & - & rx j1053.3 + 5719 + & 12 & - & 163.69 & 57.55 & 4.07 & 0.068 & 26 & - & - & - & + & 13 & - & 162.91 & 58.02 & 4.04 & 0.089 & 21 & - & - & 0.08 & + & 14 & - & 162.54 & 57.28 & 3.93 & 0.070 & 38 & - & - & - & + & 15 & sl j1048.1 + 5730 & 162.04 & 57.51 & 3.89 & 0.071 & 35 & 0.32 & - & - & + & 16 & - & 163.85 & 57.95 & 3.83 & 0.072 & 24 & - & - & 0.02 & + & 18 & - & 163.16 & 57.88 & 3.77 & 0.069 & 26 & - & - & - & + & 19 & - & 164.21 & 57.70 & 3.77 & 0.070 & 22 & - & - & - & + & 20 & - & 163.23 & 57.84 & 3.73 & 0.102 & 29 & - & - & - & + & 21 & - & 163.14 & 57.82 & 3.72 & 0.111 & 19 & - & - & - & + gd140 & 00 & sl j1135.6 + 3009 & 173.91 & 30.16 & 4.98 & 0.126 & 35 & 0.21 & - & - & + & 01 & - & 173.89 & 30.21 & 4.19 & 0.086 & 23 & - & - & - & + & 02 & - & 173.96 & 29.81 & 4.09 & 0.069 & 17 & - & - & - & + & 03 & sl j1136.3 + 2915 & 174.09 & 29.26 & 4.03 & 0.100 & 24 & - & - & - & lris z = 0.20 + & 05 & - & 174.77 & 29.89 & 3.86 & 0.111 & 26 & - & - & - & + & 06 & - & 174.86 & 30.33 & 3.83 & 0.080 & 21 & - & - & - & + pg1159 - 035 & 05 & sl j1201.7 - 0331 & 180.44 & -3.53 & 4.42 & 0.077 & 49 & 0.52 & - & - & + & 06 & - & 180.99 & -3.09 & 3.90 & 0.119 & 21 & - & - & 0.09 & + & 08 & - & 181.76 & -3.27 & 3.71 & 0.102 & 37 & - & - & - & + 13 hr field & 00 & sl j1334.3 + 3728 & 203.60 & 37.47 & 4.33 & 0.128 & 74 & 0.30 & 0.48(p ) & - & nscs j133424 + 372822 + & 01 & sl j1335.7 + 3731 & 203.94 & 37.53 & 4.10 & 0.091 & 65 & 0.41 & - & - & + & 04 & sl j1337.7 + 3800 & 204.43 & 38.01 & 3.85 & 0.080 & 34 & 0.18 & - & - & + & 06 & - & 203.85 & 37.90 & 3.78 & 0.068 & 27 & - & - & - & + & 07 & - & 204.22 & 37.54 & 3.77 & 0.078 & 30 & - & - & - & + gto 2deg@xmath0 & 00 & sl j1602.8 + 4335 & 240.72 & 43.59 & 6.65 & 0.110 & 56 & 0.42 & - & - & + & 01 & sl j1603.1 + 4245 & 240.78 & 42.76 & 5.47 & 0.106 & 57 & - & - & - & lris z = 0.18 + & 02 & - & 241.82 & 43.19 & 5.17 & 0.133 & 21 & - & - & - & + & 04 & - & 241.95 & 43.60 & 4.49 & 0.093 & 17 & - & - & - & + & 07 & sl j1604.1 + 4239 & 241.04 & 42.65 & 4.19 & 0.083 & 39 & - & - & - & lris z = 0.30 + & 08 & - & 241.63 & 43.61 & 4.16 & 0.093 & 25 & - & - & - & + & 09 & sl j1605.4 + 4244 & 241.36 & 42.74 & 4.09 & 0.064 & 36 & 0.22 & - & - & + & 10 & - & 241.18 & 43.46 & 3.85 & 0.092 & 29 & - & - & - & + & 11 & - & 241.70 & 43.64 & 3.84 & 0.077 & 33 & - & - & - & + & 12 & sl j1603.1 + 4243 & 240.78 & 42.72 & 3.82 & 0.091 & 37 & - & - & - & part of sl j1603.1 + 4245 + cm dra & 04 & sl j1639.9 + 5708 & 249.98 & 57.15 & 4.15 & 0.073 & 32 & 0.20 & - & - & + & 06 & sl j1634.1 + 5639 & 248.55 & 56.66 & 3.97 & 0.104 & 31 & 0.24 & - & - & + deep16 & 00 & sl j1647.7 + 3455 & 251.94 & 34.93 & 4.30 & 0.084 & 42 & 0.26 & - & - & + & 01 & - & 253.54 & 34.98 & 4.24 & 0.078 & 24 & - & - & - & + & 02 & - & 252.36 & 35.02 & 3.75 & 0.123 & 25 & - & - & - & + & 03 & - & 251.79 & 35.04 & 3.72 & 0.087 & 21 & - & - & - & + deep23 & 00 & sl j2326.4 + 0012 & 351.61 & 0.20 & 4.41 & 0.076 & 19 & 0.28 & - & - & + & 01 & - & 351.75 & 0.00 & 4.37 & 0.095 & 20 & - & - & - & + & 02 & - & 352.33 & 0.15 & 4.27 & 0.107 & 33 & - & - & 1.38 & + & 03 & - & 352.47 & -0.06 & 4.11 & 0.079 & 34 & - & - & 0.07 & + & 04 & - & 352.22 & 0.09 & 3.87 & 0.084 & 21 & - & - & 1.37 & + & 05 & - & 353.09 & 0.04 & 3.70 & 0.066 & 22 & - & - & - & + here we compare the detected halo number density [ deg@xmath37 ] with the prediction of numerical simulations done by @xcite , who have attempted to reproduce a survey such as ours . figure [ fig : seeingdens ] shows the halo density of our thirteen survey fields . the horizontal axis shows the density of faint galaxies ( @xmath87 ) used for weak lensing analysis . the error bars are based only on @xmath88 estimates and do not include the effects of cosmic variance . solid lines in figure [ fig : seeingdens ] show the prediction of the simulation where three different mean redshifts for the background galaxies are assumed ( from bottom to top : 0.8 , 1.0 and 1.2 ) . although the scatter is large , there is reasonable agreement between the observations and the prediction . we also see a gradual increase of the observed number density over the @xmath87 range sampled as expected . these comparisons validate our observational procedures . we have , however , two outliers in figure [ fig : seeingdens ] . these could be caused by cosmic variance or may arise from some other reasons . further follow - up studies will be important to clarify the issue . armed with the shear - selected halo catalog , we now discuss the tests we have made to verify its reliability , using both spectroscopic observations and comparisons with x - ray data in the fields where the overlap of targets can be studied . because the lensing kernel ( or window function ) has a relatively broad redshift range , the superposition of multiple low mass halos with different redshifts could yield highly significant weak lensing signals . such ` superposition halos ' are unwelcome in a mass - selected cluster catalog . in order to identify such superposition halos we undertook a more comprehensive spectroscopic survey using a multi - object spectrograph for selected candidates . this also gives us the velocity dispersion of member galaxies , which is an estimate of dynamical mass of clusters . by comparing dynamical mass with weak lensing mass we will be able to discuss the dynamical state of the clusters ( hamana et al . in preparation ) . we used the focas spectrograph on subaru whose multi - object mode permits the simultaneous observation of 25 - 30 galaxies in the field of a particular halo . we used the 150/mm grating and a sy47 order sorting filter . this configuration spans the wavelength range @xmath89 @xcite . target selection was based primarily on apparent magnitude and the color information was used when available . the exposure time was 45 @xmath9 70 minutes depending on the magnitude of the selected galaxies and the observing conditions . the spectroscopic data was reduced using standard iraf procedures ( hamana et al . in preparation ) . figure [ fig : focasobs ] shows a typical focas observation where the target is identified as a @xmath12 = 0.6 cluster . since may 2004 we have observed 26 halos from the secure sample with focas . higher priority was given to more significant halos , except in the early stages of program ( for example in the lynx and pg1159 - 035 fields ) when the follow - up strategy was still being evaluated . each of the 26 halos has been reliably identified with a cluster of galaxies . the redshift so determined is shown in the column labeled `` focas '' of table [ tab : halocat],[tab : halocat2 ] . in parallel with the focas follow - up , long slit observations with keck / lris have been carried out to enlarge the sample of redshifts as quickly as possible . we have to keep in mind that this method can not discriminate the projected samples , and some statistical consideration is necessary in dealing with the data for further studies . a complete discussion of that aspect of our program is discussed in a separate paper ( green et al , in prep . ) . in the `` note '' column of table [ tab : halocat],[tab : halocat2 ] , we list the preliminary keck / lris results as `` lris z = zvalue '' whose identification is already robust ( e.g. at least two galaxy redshifts agree ) including one of the halos in the xmm - wide field ( sl j0222.3 - 0446 , @xmath12=0.41 ) where we discuss the reliability of our catalog ( see section [ sec : reliability ] ) . in addition to undertaking our own spectroscopic observations , we also searched the nasa extragalactic database ( ned ) for clusters and groups with published redshifts close to the position of our detected halos . we included recently - published cluster catalogs in the xmm - wide survey by @xcite , @xcite and @xcite ( * vwp * data hereafter ) . in making assignments , we considered a cluster to be associated with a halo if the angular separation was less than 2 arcmin and the 3 dimensional distance less than 1 @xmath90 mpc . 13 out of 100 halos can be identified in this way with rich clusters in the literature and the published redshift is shown in the column labeled `` known '' of table [ tab : halocat],[tab : halocat2 ] . three of these thirteen clusters were also observed by focas in multi - object mode and we note that the mean focas redshift is different from the published value for these systems ( @xmath91 ) . these discrepancies can be explained by the fact that ned redshift of these clusters are all photometric , which have larger uncertainties . therefore , we adopted the focas redshifts for these three cases . we searched the ned galaxy group catalog in the same manner , but no group matched any of our halos . next , we searched for individual galaxies with published redshifts within 3 arcmin of our halos . if suitable data was found , we grouped them together using a `` friend of friend '' algorithm with a linking length of 1 @xmath90 mpc . if we found a clustering of more than two members , we calculated the average redshift and the mean astrometric position . using the same proximity criterion above , we assigned a redshift to 14 further halos . these are shown in the column labeled `` nedg '' . in order to estimate the probability of a chance coincidence of the matching procedures , we first randomized the position of the detected halos in the field , and applied the same matching procedure described above to the randomized halo catalog . this process suggests that the probability of a chance coincidence is roughly 10 % and 25 % for the cluster and galaxy clustering searches , respectively . clearly care should be taken in analyzing such data . the total number of reliably identified clusters are 41 ( out of 100 in table [ tab : halocat],[tab : halocat2 ] ) where we do not include those identified by `` nedg '' galaxies because the chance coincidence is not negligible . we assign i d labels only for these reliable halos in the tables . we now turn to the important question of the evaluating the reliability of our shear - selected catalog . we will do this by examining both the success rate of our identifications and by comparing with x - ray samples obtained via the xmm - wide field survey . vwp have confirmed spectroscopic redshifts for 28 x - ray selected clusters in the 3.5 deg@xmath0 survey field . their locations are plotted as squares on figure [ fig : saclay_halomap ] . our secure sample of 15 shear - selected halos with s / n @xmath24 3.69 is plotted as solid circles . those we have spectroscopically confirmed using subaru s focas and/or keck lris are marked by large open circles . we find that three halos ( 1 , 16 and 23 ) are confirmed as clusters by focas and are not reported in vwp . among our 15 shear - selected halo samples , only three ( 4 , 10 and 20 ) have yet to be identified . this identification success rate ( 80 % ) can be regarded as a lower limit given future observations may yet locate an associated cluster . although affected by small number statistics , this minimum efficiency is already higher than expected by simulation studies@xcite . it is interesting to compare our identification success rate with those found in the _ gabods _ survey . @xcite found 14 significant halos in their 18 deg@xmath0 area . among them , 5 halos turn out to be _ known _ clusters of galaxies , 2 seem to have associated _ light _ concentrations but no spectroscopic confirmation , and the remaining _ uncertain _ 7 ( 50 % ) had no apparent counterpart in either optical nor x - ray data . @xcite undertook a search on the same data adopting a different peak selection algorithm and a lower detection threshold . they found 158 `` possible mass concentration '' on the 18 deg@xmath0 field . if those halos are divided into the same classes as above ( _ known , light , uncertain _ ) , the ratio is almost the same as @xcite . regardless , almost half of the _ gabods _ halos could be considered uncertain at this point . meanwhile , we find 15 halos in a 2.24 deg@xmath0 field ( xmm - wide ) and show that 80 % of have been already identified as clusters . it is too early to make any definite conclusion because their spectroscopic is underway . however , our ( tentatively ) higher success rate could be explained at least in part by the larger number density of faint galaxies , @xmath87 , usable for the weak lensing analysis owing to larger aperture and better average seeing . in the case of _ gabods _ , @xmath87 spans from 6 to 28 arcmin@xmath37 depending on the field ; the average value is 11 arcmin@xmath37 . this is generally smaller than our @xmath87 shown in table [ tab : survey_field ] . as a result one can expect a lower angular resolution of the @xmath70 map , reducing the s / n ratio for a fixed smoothing scale and possibly an increased degree of contamination in the resulting halo catalog . another possibility to explain the increased contamination is that their sample consists of a combination of different sets of catalogs , each of which is selected by different methods . this could decrease the significance threshold effectively , and could introduce more false peaks . @xcite estimated , based on their simulation , that the halo superposition rate in survey such as ours should be roughly 3 % - a small but not negligible effect . a longslit spectroscopic survey , such as that undertaken with lris ( green et al , in prep ) might be poorly - suited for locating such cases . however , the focas multi - object survey reported here should reliably find them . in fact , we have found three apparent superposition halos ( sl j0222.8 - 0416 , sl j1051.5 + 5456 and sl j1047.3 + 5700 ) out of 26 examined . the overlap rate is broadly consistent with expectation considering the small number so far sampled . we have introduced a new subaru imaging survey and described techniques for locating and verifying shear - selected halos . across a search area of 16.72 deg@xmath0 we have found 100 halos whose @xmath70 s / n exceeds 3.69 . we have described the first phase of a detailed follow - up campaign based on multi - object spectroscopy of 26 halos using focas on the subaru telescope . a later paper in this series ( green et al , in prep ) will extend the spectroscopic survey to the full sample using a longslit approach . detailed studies on one of our fields , the xmm - wide field , show that 80% of the shear selected 15 halos in the 2.2 deg@xmath0 area can be confirmed as genuine clusters of galaxies . 10 overlap with x - ray detections and two are new systems confirmed spectroscopically . the overall success rate and reliability of our sample provides convincing proof that , with care , a weak lensing survey can provide a large sample of mass - selected halos . we compare our success rate and the reliability of our catalog with that of the _ gabods _ survey and conclude a major advantage of our approach is the superior imaging depth which leads to a high surface density of usable galaxies . this suggests future , more ambitious , surveys for shear - selected halos will be more effective if undertaken with large aperture telescopes . it is interesting to use our results to estimate the requirements for a future survey motivated by the need to constrain dark energy . @xcite discuss hypothetical missions which would aim to analyze the redshift distribution @xmath1 of 10,000 clusters . figure [ fig : seeingdens ] shows that our survey technique typically finds 5 halos deg@xmath37 . because the field size of suprime - cam is @xmath9 0.25 deg@xmath0 , it takes two hours to cover 1 deg@xmath0 assuming the exposure time of 30 minutes of each field as adopted in this study . a survey of 10,000 clusters would be prohibitive even in terms of imaging alone ( @xmath9 500 clear nights ) even before contemplating the follow - up spectroscopy . the hyper suprime - cam ( hsc ) project @xcite has been proposed to remedy this shortcoming . hsc is expected to have a field of view ten times larger than suprime - cam while maintaining the same image quality . this new facility will makes the 2000 deg@xmath0 scale imaging survey within a reasonable number of clear nights . spectroscopic follow - up might be enabled by the proposed prime focus multi object optical spectrograph wfmos ( wide field fiber multi object spectrograph ) whose field is 1.5 deg@xmath0 in diameter . typically we can expect @xmath2510 shear - selected clusters in each spectroscopic field . with only 20 targets per halo , only a small fraction of the several thousand fibers envisaged for wfmos need be allocated to the halo verification program . we are very grateful to subaru astronomers : y. oyama , k. aoki and t. hattori for their dedicated supports of the focas observing . numerical computations presented in this paper were carried out at the astronomical data center ( adc ) and at the center for computational astrophysics ( cfca ) of the national astronomical observatory of japan . this work is supported in part by grant - in - aid for scientific research ( kaken - hi ) of japan society for the promotion of science ( jsps ) : project number 15340065 ( sm&th ) and 17740116 ( th ) . 999 allen , s.w . , schmidt , r.w . , fabian , a.c . , ebeling , h. 2003 342 , 287 allen , s.w . , schmidt , r.w . , ebeling , h. , fabian , a.c . , van speybroeck , l. 2004 , 353 , 457 becker , m.r . 2007 submitted to apj ( astro - ph/0704.3614 ) bernstein , g. & jarvis , m. 2002 , , 123 , 583 bhringer , h. et al . 2001 a&a , 369 , 826 davis , m. , faber , s.m . , newman , j. et al . 2003 , spie , 4834 , 161 dietrich , j.p . , erben , t. , lamer , g. , schneider , p. , schwope , a. , hartlap , j. , maturi , m ( 2007 ) a&a in press ( astro - ph/0705.3455 ) erben , t. , van waerbeke , l. , bertin , e. , mellier , & y. & schneider , p. 2001 , a&a , 366 , 717 finoguenov , a. , guzzo , l. , hasinger , g. , scoville , n.z . et al . ( 2006 ) apjs in press ( astro - ph/0612360 ) gladders , m.d . & yee , h.k.c . 2000 , , 120 , 2148 goto , t. et al . 2002 , , 123 , 1807 hamana , t. , miyazaki , s. et al . 2003 , , 597 , 98 hamana , t. , takada , m. , yoshida , n. 2004 , , 350 , 893 hennawi , j.f . & spergel , d.n . 2005 , 624 , 59 henry , j.p . 2000 , , 534 , 565 hetterscheidt , m. , erben , t. , schneider , p. , maoli , r. , van waerbeke , l. , mellier , y. 2005 , a&a , 442 , 43 . heymans , c. et al . 2006 , , 368 , 1323 hokekstra , h. , franx , m. , kuijken , k. , squires , g. 1998 , 504 . 636 hoekstra , h. 2004 , , 347 , 1337 ikebe , y. , reiprich , t.h . , bhringer , h. , tanaka , y. , kitayama , t. 2002 a&a , 383 , 773 jenkins , a. , frenk , c. s. , white , s. d. m. , colberg , j. m. , cole , s. , evrard , a. e. , couchman , h. m. p. & yoshida , n. 2001 , , 324 , 450 kaiser , n. & squires , g. 1993 , , 404 , 441 kaiser , n. , squires , g. & broadhurst , t. 1995 , , 449 , 460 kaiser , n. , wilson , g. , luppino , g. , dahle , h. 1999 submitted to pasp ( astro - ph/9907229 ) kashikawa , n. et al . 2002 , , 54 , 819 kolb et al . 2006 , us dark energy task force report landolt , a.u . 1992 , , 104 , 340 levine , e.s . , schulz , a.e . , white , m. 2002 , 577 , 569 luppino , g.a . & kaiser , n. 1997 , , 475 , 20l majewski , s.r . , kron , r.g . , koo , d.c . , bershady , m.a . 1994 , , 106 , 1258 massey , r.j , rhodes , j. , ellis , r.s . et al 2007 , nature , 445 , 286 maturi , m. , schirmer , m. , meneghetti , m. , bartelmann , m. , moscardini , l. 2006 , a&a in press ( astro - ph/0607254 ) miyazaki , s. , hamana , t. , shimasaku , furusawa , h. , doi , m. , hamabe , m. , imi , k. , kimura , m. , komiyama , y. , nakata , f. , okada , n. , okamura , s. , ouchi , m. , sekiguchi , m. , yagi , m. , yasuda , n. 2002a , 580 , l97 miyazaki , s. , komiyama , y. , okada , n. , imi , k. , yagi , m. , yasuda , n. , sekiguchi , m. , kimura , m , doi , m. , hamabe , m. , nakata , f. , shimasaku , k. , furusawa , h. , ouchi , m. & okamura , s. 2002b , , 54 , 833 miyazaki , s. , komiyama , y. , nakaya , h. , doi , y. , furusawa , h. , gillingham , p. , kamata y. , takeshi , k. , nariai , k. 2006 , spie , 6269 , 9 pierre , m. et al . 2004 j. cosmol . phys , 09 , 011 pierre , m. , pacaud , f. et al . 2006 , , 372 , 591 postman , m. , lubin , l.m . , gunn , j.e . , oke , j.b . , hoessel , j.g . , schneider , d.p . , christensen , j.a . 1996 , , 111 , 615 press , w.h . , flannery , b.p . , teukolsky , s.a . , vetterling , w.t . 1993 , `` numerical recipes in c '' , isbn-13 : 9780521431088 reprich , t.h . & bhringer , h. 2002 , 567 , 716 schirmer , m. , erben , t. , hetterscheidt , m. , schneider , p. 2006 , submitted to a&a ( astro - ph/0607022 ) smail , i. , ellis , r.s . , dressler , a. et al . 1997 , 479 , 70 smith , g.p . , kneib , j. , smail , i. , mazzotta , p. , ebeling , h. , czoske , o. 2005 , , 359 , 417 valtchanov , i. et al . 2004 , a&a , 423 , 75 wang , s. , khoury , j. , haiman , z. may , m. 2004 phrvd , 70 , 123008 van waerbeke , l. , mellier , y. , erben , t. et al . 2000 , a&a , 358 , 30 white , m. , van waerbeke , l , mackey , j. 2002 , , 575 , 640 willis , j.p . , pacaud , f. , valtchanov , i. et al 2005 , , 363 , 675 wittman , d. , tyson , j.a . , margoniner , v.e . , cohen , j.g . , dellantonio , i.p . 2001 , 557 , l89 . wittman , d , tyson , j.a . , dellantonio , i.p . 2002 , proc , spie , 4836 , 73 wittman , d , dellantonio , i.p , hughes , j.p . et al . 2006 , , 643 , 128 the shear induced by massive cluster of galaxies is expected to be 7@xmath9 10 % . this is actually larger than raw ellipticities due to psf anisotropy ( 2 @xmath9 4 % refer to fig . [ fig : fwhmee ] ) , but the difference is not so significant . therefore , the correction of the anisotropy is very important to make a precise kappa map . here , we examine the psf anisotropy suprime - cam , and evaluate the effect of imperfect correction in this work . in order to characterize the anisotropy of the suprime - cam images , we obtained sequences of i-band short exposures ( 1 min ) of dense stellar fields with various telescope pointings over one night long . the seeing was mostly @xmath9 0.65 arcsec ( fwhm ) . the psf anisotropy is approximated as an elliptical , and the field position dependence of the ellpticities are investigated . two examples of such ellipticity fields are shown in fig . [ fig : starfield_e_examples ] . general tendency that we note is that the fields can be represented as a super - position of ( 1 ) the radial field at four corners ( almost invariant ) and ( 2 ) almost unidirectional field ( variable ) . the variable components is most likely due to the telescope shaking whereas the invariant component can be explained by the optical aberration of the corrector . slight asymmetry of the radial component is visible ( i.e. ellipticities near the lower left corner is larger than other corners ) , and this would be a sign of imperfect optical alignment . we also note that the discontinuity of ellipticities is not visible across the boundary of the ccds , and the ellipticities can be represented as a single continuous function of field position . we adopt 5-th order polynomial function here . in order to simulate the actual science field analysis , we randomly select controls stars and `` galaxy role '' stars from the star catalogs with an appropriate density ; 2 arcmin@xmath92 and 40 arcmin@xmath92 , respectively . using the control stars , we obtain the best fit coefficients of the polynomial , and the anisotropy of `` galaxy role '' stars is corrected using eqn . [ psmcorrection ] . because the intrinsic ellipticities of `` galaxy role '' stars is all zero , the residual ellipticities after the correction is a estimate of incompleteness of the correction . fig [ fig : radreal](a ) shows the azimuthally averaged ellipticities of stars ; before the correction , the typical 3 % raw ellipticity is seen . it might be interesting to note that the ellipticities of 3 % . can be induced by rms pointing error of merely 0.1 arcsec under the seeing of 0.65 arcsec . the ellipticities is reduced down to about 0.75 % after the correction . beyond the field angle of 18 arcmin , the correction does not work fine and we decided to eliminate the field r @xmath24 18 from lensing analysis ; which , however , results in only a few percent loss of fov . we estimated a psf anisotropy caused by the optical aberration using a ray - tracing code , _ zemax_. the calculated psf is convolved with 0.65 arcsec fwhm gaussian , and the shape is evaluated by elliptical . the result is shown in filled square in fig [ fig : radreal](a ) . compared with this ideal case , the observed ellipticities is large even after the correction . this shows a limit of the adoption of single polynomial function as a representatives of the psf anisotropy , where the residual is still locally correlated correlated and not completely random . we now want to evaluate the impact of the incompleteness onto the @xmath70 s / n map . the map is created based by the residual ellipticities after the correction . we simulate the galaxy ellipticities using the following `` conversion '' formula @xcite , @xmath93 which is essentially a sensitivity correction against the psf anisotropy . this is necessary because galaxies are larger and their shape is more insensitive to the deformation compared with stars . we evaluate @xmath94 with @xmath95 where we adopt @xmath96 which is a typical value under the typical 0.7 arcsec seeing . since the @xmath97 here , the conversion factor is roughly @xmath98 . we omit @xmath48 correction because it cancels out in this case when we calculate the s / n . we created twelve such maps from independent exposures , and co - added the peak distribution functions to obtain fig [ fig : radreal](b ) . it is obvious that the incompleteness of the anisotropy correction is quite unlikely to create any significant ( say s / n @xmath24 3 ) fake peaks .

we have tabulated lists of upper red giant , horizontal , and asymptotic giant branch ( rgb , hb , agb ) stars in the globular cluster m5 that are complete to over from the core for the rgb and agb samples , and for the hb sample . the large samples give us the most precise value of to date for a single globular cluster ( ) . this is incompatible with theoretical calculations using the most recent physical inputs . the discrepancy can probably be attributed to the dependence of observed values on horizontal branch morphology . we identify the cluster m55 as being another possible example of this effect . samples of hb and agb stars in populous clusters may provide a means of calibrating the masses of horizontal branch stars in globular clusters . the cumulative luminosity function of the upper red giant branch shows an apparent deficit of observed stars near the tip of the branch . this feature has less than a 2% chance of being due to statistical fluctuations . the slope of the cumulative luminosity function for agb stars is consistent with the theoretically predicted value from stellar models when measurement bias is taken into account . we also introduce a new diagnostic that reflects the fraction of the agb lifetime that a star spends in the agb clump . for m5 , we find , in marginal disagreement with theoretical predictions . finally , we note that the blue half of m5 s instability strip ( where first overtone rr lyraes reside ) is underpopulated , based on the large numbers of fundamental mode rr lyraes and on the nonvariable stars at the blue end of the instability strip . this fact may imply that the evolutionary tracks ( and particularly the colors ) of stars in the instability strip are affected by pulsations .

astro-ph0404447

evolved - star populations in clusters have long provided important tests of stellar evolution theory because they present us with coeval ensembles of stars having nearly identical chemical compositions . because the ratio of the numbers of stars in two different evolved phases relates to the the ratio of the intervals of time that a star spends in the two phases , we can get a glimpse at the stellar evolution clock . however , it is only in the most massive clusters that we can expect to sample , and therefore test model timescales for , stars in the shortest evolutionary phases . even in the most massive clusters , the huge range of stellar densities requires that wide - field data be taken to cover the extended halo of the cluster , while high spatial - resolution ( often space - based ) photometry is gathered for the crowded cores . this study is an analysis of evolved stars in the massive milky way globular cluster m5 . we have compiled large samples of relatively rare populations on the upper red giant branch ( rgb ) and the asymptotic giant branch ( agb ) . our intent is to test whether the observed evolution conforms to our expectations from stellar models with current physics inputs . in 2 , we describe the observational material we used in identifying and the stars , and the procedures used to classify them based on their photometric properties . in 3 , we describe various diagnostics of the bright populations of globular clusters , including the population ratio @xmath4 , the distribution of hb stars , cumulative luminosity functions , and a new population ratio @xmath5 . in order to collect a nearly complete sample of bright evolved stars in m5 , we used data from a variety of sources . the primary source of ground - based photometry was the study of sandquist et al . ( 1996 ; hereafter s96 ) . that dataset was composed of @xmath7 photometry taken at the cerro tololo interamerican observatory ( ctio ) 4 m telescope and @xmath8 images of the cluster core taken using the high resolution camera ( hrcam ) on the 3.5 m canada - france - hawaii telescope ( cfht ) . the cfht data were obtained during very good seeing conditions so that there was little ambiguity as to the population identifications of cluster stars . since the 1996 study , the cfht data have been calibrated against the ctio 4 m data . [ cfhtcomp ] presents a comparison of the photometry for the two datasets . because stars near the core in the ctio sample are quite likely to be affected by contaminating light from other nearby stars , there is a substantial bias toward brighter magnitudes . however , the lower envelope of values in the upper 4 plots indicates that the best measured stars have small residuals and little or no trend with color . we have also used hubble space telescope data from the cluster core reported in @xcite . the f439w and f555w bands were calibrated to @xmath9 and @xmath10 by that group , and we present a comparison with our cfht data in the core in fig . [ hstcomp ] . the @xmath9 magnitudes show evidence of either a linear trend with @xmath9 or a second - order trend with color . in the course of examining the hst dataset , we found systematic differences in the position of the rgb in color - magnitude diagrams ( cmds ) derived from different ccds of the wfpc2 camera . for this reason , we found that it was more reliable to make population identifications ( rgb , hb , or agb ) based on relative position in the cmd for individual chips . although this reduces the number of stars on the rgb in the cmds for each chip and makes it harder to determine where the rgb fiducial line is , the selection of agb stars is generally cleaner . we have also found that the hst tabulation of piotto et al . apparently missed some stars whose positions fell near where different chips overlapped . identification numbers for the stars in the hst dataset that are given in our final tables were composed of the four digit i d number from the piotto listing , with a leading digit identifying the ccd chip used in the measurement ( 1 = pc , 2 - 4 = wf2 - 4 ) . positions in hst frames were derived using the iraf routine metric in the stsdas package , which converts pixel coordinates to sky coordinates , corrects for geometric distortions of individual chips , and puts coordinates from different chips on the same coordinate system . lists from different datasets were matched by position using the program daomaster ( p. b. stetson , private communication ) , which determines six - coefficient coordinate transformations . the positions we tabulate are given in arcseconds from the center of the cluster ( taken to be @xmath11 , @xmath12 , epoch 2000.0 ; harris 1996 ) , with the x position being the @xmath13 offset and the y position being the @xmath14 offset ( positive being higher @xmath13 and @xmath14 ) . proper motion information from rees ( 1993 ) was used to eliminate field stars from the evolved star samples . overall this did not impact our samples greatly because m5 is quite far from the galactic plane ( @xmath15 ) . the biggest impact of the proper motion information was to verify that most stars bluer than the main line of the agb were field stars . because the rees dataset extends well beyond the bounds of the s96 ctio fields , we have used his photographic data in the outer cluster regions to identify additional stars to include in the sample . [ pmcomp ] shows a comparison between rees photometry and our ctio photometry . overall , the photometry agrees well , with the exception of slight trends with color for the brightest giants in @xmath9 and the faintest hb stars in @xmath10 . for the rgb and agb samples , the inclusion of the rees dataset allows us to determine complete samples from the center of the cluster to well over @xmath0 . because the rees data do not reach the faintest hb stars , the hb sample is only complete from the core to approximately @xmath1 from the center . although figs . [ hstcomp ] and [ pmcomp ] make it clear that there are systematic differences in the calibration of the photometry in these different sets , the residuals are less than 0.1 mag in all cases [ with the exception of the tip of the rgb in @xmath9 in the rees ( 1993 ) data ] and therefore this is unlikely to make any substantial changes to our conclusions . @xcite indicate that there are some calibration issues in the s96 dataset ( mean differences of @xmath16 and @xmath17 , and position - dependent residuals ) . however , because the offsets for the brightest stars are still only at the level of a few hundredths of a magnitude , and are considerably less serious for stars belonging to the brightest populations ( see their fig . 6 ) , we believe that this is not a serious consideration for this study . identification of the evolutionary stages of individual stars are made relative to other stars in the same photometric dataset where possible , and there are few cases in which the identification is questionable . the tests of evolutionary theory below are not strongly dependent on the brightness of individual stars because of the samples we have collected are relatively large . the assignment of a given star into the rgb , hb , or agb groups was based on position in the cmd . for different regions of the cluster , we used the dataset and color - magnitude combination that provided the cleanest separation of the evolved sequences . the most challenging tasks were distinguishing between upper rgb and agb stars , and identifying hb stars at all brightness levels with almost no field or cluster star contaminants . in all cases , rgb , hb , and agb samples were first selected from each individual dataset based solely on cmd position in _ that _ dataset . the lists from different datasets were then compared and discrepancies were reconciled , generally in favor of the identification from the highest resolution , least noisy dataset . in each case , we conducted positional searches near the positions of candidates to try and ascertain the reason for the discrepancy ( such as blending and ccd flaws ) . in the core of the cluster the only practical choices were the hst or cfht datasets . the cfht photometry was given higher weight than the hst photometry in regions of overlap because there was smaller amount of scatter in the cfht photometry on the giant branches . high - resolution photometry of the core of the cluster is important for selecting agb stars because in lower - resolution ground - based data , blends of rgb and hb stars or two rgb stars can produce objects that appear to be agb stars . although this possibility exists in the outskirts of a cluster ( meaning that we can not be certain that our samples are completely free of such blends ) , the probability of overlap is much smaller . the stars in the three samples are presented in tables [ agb ] - [ hb ] . the first line for each star contains the new identification number from this study , the offsets from the cluster center in arcsec , and the membership probability @xcite when available . subsequent lines present identifications and photometry from the different catalogs : hst @xcite , cfht ( s96 , calibrated for this study ) , ctio ( s96 , from either the @xmath18 or @xmath8 calibrations ) , and pm @xcite . notes on some stars are provided in the table [ notes ] . the rgb star sample is truncated at the @xmath10 magnitude of the hb ( @xmath19 ) . this falls slightly below the rgb bump . we also must note a practical aspect to the identification of post - hb stars . we identify agb stars as being those stars falling close in color to a track nearly paralleling the rgb . theoretical models predict , however , that stars originating on the bluest portions of the hb never reach the agb @xcite . these stars actually spend longer times ( up to about 100% longer ) in a double fusion shell phase than do agb stars that come from the reddest portions of the hb , but maintain considerably higher surface temperatures . as a result , proper motion information is important for the separation of these `` agb - manqu '' stars from foreground stars that fall in the same portion of the cmd . as can be seen from figs . [ cfht ] and [ hst ] , m5 appears to be largely devoid of agb - manqu stars , which plays a role in its @xmath4 value , as we will see in the next section . when photometry in multiple bandpasses was available , we examined cmds using different combinations of magnitude and color to help verify identifications . often the use of the shortest wavelength filter band on the magnitude axis of the cmd was most helpful since this gives both the hb and agb / rgb a horizontal appearances in the cmd . this can be seen in @xmath20 , @xmath21 cmds of the cluster ( e.g. fig . 2 of markov , spassova , & baev 2001 ) . cmds of the different samples are shown in figs . [ cfht ] - [ pm ] . we put the cmds in order of importance in the selection . the ctio @xmath8 dataset is given precedence over the @xmath18 dataset due to poorer seeing during the @xmath10 photometry . we attempted to identify rr lyrae variable stars from the literature in order to avoid the accidental elimination of these stars from the hb counts . the primary source of information on these stars was the catalog of variable stars in globular clusters ( hereafter , cvsgc ; described in clement et al . we have also gathered variability information on rr lyrae variables for the cluster from storm , carney , & beck ( 1991 ; @xmath18 ; intensity - weighted colors ) , brocato , castellani , & ripepi ( 1996 ; @xmath18 ; magnitude averages ) , reid ( 1996 ; @xmath22 ; magnitude - weighted colors ) , and caputo et al . ( 1999 ; @xmath18 ; corrected colors ) . to insure that the resulting `` static '' colors were on the same scale , we corrected the storm et al . colors using values from bono , caputo , & stellingwerf ( 1995 ) . in the process of compiling this list , we discovered that the variables v162 and v163 @xcite were actually the previously discovered variables v90 and v17 respectively . hb stars represent the sample from which the agb stars are born and we observationally identify the portions of the hb that contribute the most stars to the observed agb by examining the distribution of hb stars . a detailed analysis of this question will require comparison of observations with monte carlo simulations of hb and agb populations using new tabulations of theoretical models with updated compositions and physics . for the time being , we restrict ourselves to characterizing the hb population . numerous previous studies have computed hb star distributions where the independent variable is a cmd length coordinate running from the red end of the hb to the blue end ( e.g. , crocker , rood , & oconnell 1988 ; fusi pecci et al . 1993 ) . computer implementations of this idea have generally used scale factors to set the relative contributions of increments in color and in magnitude . ideally though , we would like to be able to relate the derived distribution function to the distribution of hb star masses in a straightforward manner . although this transformation is complicated by evolution of the stars and uncertainties in chemical composition , these uncertainties mostly affect the average hb mass that is derived , and not the shape of the distribution @xcite . because position on the hb is a nonlinear function of total mass , we decided to attempt to remove this nonlinearity using the procedure of @xcite , projecting the stars onto a theoretical zahb locus for an appropriate chemical composition . this has the advantage of producing a distribution that will have some physical meaning ( something which can not as easily be said about @xmath23 distributions that use fixed color and magnitude scale factors ) . we elected to use our @xmath9 and @xmath24 photometry in computing mass distributions because the @xmath8 datasets cover the vast majority of the stars , have the greatest photometric accuracy , and the @xmath25 color is sensitive to temperature over the full range covered by m5 hb stars . we have incorporated rr lyrae variables using `` static '' colors derived from studies listed in [ obs ] . since these studies did not use @xmath25 colors , we used our ctio datasets to do a linear interpolation across the instability strip to convert from @xmath26 or @xmath27 colors to @xmath25 . for the color distribution shown in fig . [ hbcol ] , we scaled the rrab and rrc portions of the histogram separately ( by 1.771 and 1.38 , respectively ) to account for known variables ( cvsgc ) of the two types that do not have measured static colors . for the mass distributions computed below , we randomly chose a @xmath25 color from within the range covered by variables with the same pulsation mode ( @xmath28 for rrab variables , and @xmath29 for rrc variables ) if the variable did not have a tabulated static color . the pulsation mode was taken from the cvsgc tabulation . because the rr lyrae instability strip covers a relatively small range of mass in the theoretical models ( @xmath30 ) , the exact method used is not critical to the overall shape of the distribution . we plot computed hb mass distributions in fig . [ hbm ] using zahb models from @xcite for enhanced @xmath13-element abundances using two [ fe / h ] values roughly corresponding to tabulated values from @xcite and @xcite and a reddening e@xmath31 ( schlegel , finkbeiner , & davis 1998 ) . we note that with this reddening the blue tail of the hb observations are not a good color match to the models . partly the discrepancy seems to result from calibration errors for the cfht data at the blue end of the hb . this should not be an issue for the mass distributions because the distributions are derived from _ projections _ to the zahb . for [ fe / h ] @xmath32 , we find @xmath33 with @xmath34 , while for [ fe / h ] @xmath35 , we find @xmath36 with @xmath37 . the @xmath38 values are somewhat misleading though because the distributions are asymmetric . the identification of the portion of the hb containing the peak of the mass distribution and the majority of stars is a robust result of this procedure , independent of the exact values of the star masses . the peak of the distribution falls at @xmath39 ( right at the blue edge of the instability strip ) , while the median color value is @xmath40 . we can characterize the width of the hb color distribution in several ways as well . from looking at the hb mass distribution , and identifying by eye where the distribution drops off most quickly , we can crudely bracket the range of colors containing the majority of stars . in this way , we find that the red edge of the mass distribution at @xmath41 ( essentially the red end of the instability strip ) , and the blue edge falls on the blue tail at @xmath42 ( although this is more poorly defined ) . the stars on the red side of the instability strip form a tail in the mass distribution largely because hb color changes much more slowly with mass there . alternately we can identify the middle 68.3% of the hb sample ( corresponding to @xmath43 points of a gaussian distribution ) . in this way we find a red bound at @xmath44 in the blue half of the constant red hb , and a blue bound at @xmath45 on the blue tail . these two methods give similar results . the color distribution emphasizes a peculiar property of m5 : a majority of m5 s hb stars fall on the nonvariable blue hb , but m5 is an oosterhoff group i cluster with the expected excess of rrab type variables ( @xmath46 ; cvsgc ) , which reside on the red side of the instability strip . in fact , m5 has one of the bluest hbs of the oosterhoff group i clusters in the galaxy . recently @xcite presented evidence that the dominant oo i population in m3 is composed of stars in the early stages of post - zahb evolution , and that stars that have begun to evolve redward appear to have an oo ii - like population . jurcsik et al . reference the hysteresis hypothesis of @xcite that proposed that switches between fundamental and overtone pulsation modes occur at different temperatures depending on the direction of the star s evolution in color . this may also be the case for m5 : its oo i variable population derives from the fact that the zahb in the instability strip _ is _ populated as a result of the cluster s large dispersion in @xmath47 . however , this does not explain why the distribution of static colors for the variables is biased toward the red half of the instability strip . m5 is one of the only oo i clusters for which the number of stars per color interval should be rising as you go from the red end of the instability strip to the blue end , but the cluster still has many fewer rrc variables than rrab variables . theoretical evolution tracks do not diverge from this portion of the hb toward the red and blue . this seems to imply that hb stars are being kept out of the range of colors that are normally occupied by rrc variables . in other words , the presence of pulsation seems to affect the evolutionary track of the star . our discussion has focussed on published values of static colors ; however , there is no reason to believe that this affects the conclusion . the larger number of fundamental mode pulsators is undeniable , and the vast majority of those rr lyrae stars having static colors fall in two distinct color ranges sorted by the pulsation mode . the fraction of variables pulsating in the first overtone in other globular clusters also shows very little dependence on hb morphology ( e.g. castellani , caputo , & castellani 2003 ) . in order to facilitate the identification of unusual stars , we plot a cmd for the combination of the cfht and ctio datasets in fig . [ bothbi ] . this is the largest sample with a common photometric calibration ( although a slight mismatch in the colors of the bluest hb stars in the two datasets is evident ) . in examining hb stars , we have found several groups of stars with unusual magnitudes and colors even in the best datasets . there are two separate groupings of stars redder and brighter than the most populated portion of the red hb in m5 . these two groups may correspond to groups in m3 that were given the classification `` er '' by @xcite . the brighter of the two groups ( @xmath48 ) contains 11 stars ( ids 13 , 28 , 30 , 53 , 110 , 118 , 137 , 141 , 168 , 239 , 392 ) and is at the extreme red end of the hb , noticeably disconnected from the rest of the hb . the fainter of the two groups ( @xmath49 ; see fig . [ hbzoom ] ) is less noticeably separated from the red hb and contains 14 stars ( ids 34 , 42 , 62 , 63 , 67 , 76 , 169 , 198 , 360 , 389 , 412 , 425 , 434 , 444 ) . the two groups contain about 4.5% of the total hb sample . @xcite hypothesized that the er hb stars are the progeny of blue straggler stars because blue stragglers are believed to have mass larger than turnoff mass stars , when they evolve to the core he fusion phase , they would tend to have higher masses and redder colors than the majority of hb stars . @xcite provide some evidence that the radial distribution of er stars in m3 parallels that of the stragglers . @xcite find than the cumulative radial distribution of the er hb stars parallels that of the blue stragglers , and both are more centrally concentrated than the combined rgb and hb sample . for m5 , we find that the cumulative radial distribution of the two groups of er hb stars is actually _ less _ concentrated than the hb stars ( fig . [ crds ] ) . from a kolmogorov - smirnov test , there is only a probability of 0.0014 that the two samples come from the same distribution . ( the redder of the two groups of er hb stars is too small to make a statistically significant comparison . ) however , 10 of these stars are concentrated of these stars within about @xmath50 of the edge of the high resolution hst and cfht fields , so that there is a good chance some of these are unresolved blends of an hb and an rgb star . while blends are unlikely to explain other identified er hb stars ( particularly when the ones forming a tight group in the cmd ) , higher resolution photometry of the near - core region is needed . we have also identified 13 stars fainter than the horizontal portion of the hb and redder than the blue tail ( see fig . [ hbzoom ] ) . all of these stars are found within @xmath51 of the core . as with the er stars , several ( ids 81 , 95 , 302 , 317 ) are found just outside the hst and cfht fields , and so may be blends of a blue hb star and a star on the faint half of the rgb . others ( ids 177 , 243 , 316 ) are found near the cluster core and were only measured in the cfht field , so that they may also be blends . however , others ( ids 107 , 201 , 221 , 250 , 268 , 276 ) were measured in both the hst and cfht fields , and deviate in similar ways in both datasets . blends or unresolved binaries involving a blue hb star and a star on the lower rgb are one possibility . in a few of the cases , it is also possible that the stars are extremely bright blue stragglers . a less likely possibility for a subset of these stars ( ids 268 and 317 ) is that there may be unidentified rr lyrae stars caught in a particular part of the pulsation cycle . two stars are found in the rr lyrae instability strip in different datsets . hb star 340 falls within the instability strip in the hst dataset , and within the area of the cmd covered by rr lyrae stars measured in the cfht datset . as a result , we flag it as possible rr lyrae star . star 132 is relatively near the core , but falls within the instability strip in the ctio dataset . it may be a blend of an blue hb star and and giant star . finally we note that there is one star in the hst and cfht datasets ( i d 341 ) several magnitudes below the blue end of the hb that may be an extreme blue hb star . in spite of the fact that it is observed in the core of the cluster , it may still be a field star , or a cluster member in an evolutionary phase following the agb . the star is substantially bluer than the main sequence in both the @xmath52 and @xmath53 colors . we can not distinguish between these possibilities with the datasets discussed here . the population ratio @xmath54 is a diagnostic of the relative durations of the helium fusion phases of stellar evolution ( buzzoni et al . 1983 ) . as discussed by several groups ( e.g. straniero et al . 2003 ; renzini & fusi pecci 1988 ) , this ratio is most affected by the physical processes occurring in the cores of stars during the horizontal branch ( hb ) phase . the largest uncertainties in theoretical predictions for this ratio are the rate for the @xmath55 reaction and details of the mixing processes occurring at the outer boundary of the convective core . even with these uncertainties , observed values of @xmath4 are sufficient to rule out the possibility of the so - called `` breathing pulses '' seen in early models of agb stars . breathing pulses mix fresh helium into the core and hence prolong the hb phase , resulting in low @xmath4 values . as a result of @xmath4 measurements , ad hoc algorithms have been introduced into the models to suppress these pulses ( e.g. cassisi et al . 2001 ) . there are observational influences on the @xmath4 ratio that have been mostly ignored up to this point . typically , a hb star becomes brighter and redder as it evolves toward the agb ( although the star may make smaller excursions blueward and redward in color before leaving the hb entirely ) . theoretical models ( e.g. dorman , rood , & oconnell 1993 ) showed that stars evolving off of the bluest extensions of a hb may turn blueward before they reach the portions of the cmd most heavily populated by agb stars . generally only stars less massive than about @xmath56 will completely avoid the agb , although the exact value depends on chemical composition . slightly redder hb stars reach the upper agb , but turn off toward the white dwarf cooling sequence before reaching the tip of the agb . hb stars that are slightly redder still start the agb phase in the agb clump , which results in a large increase in agb lifetime . clearly the hb morphology of a globular cluster will affect the number of agb stars that are present . an important point of this discussion is that `` second parameter '' effects on the hb morphology can cause differences in the @xmath4 ratio for globular clusters of the same metallicity . as illustrations , we present sample single - star @xmath4 values calculated from the models of dorman , rood , & oconnell ( 1993 ) in fig . [ rvals ] . we should keep in mind that the dorman et al . models have physical inputs that are somewhat out of date : oxygen - enhanced compositions ( rather than @xmath13-element enhanced ) , a somewhat low initial @xmath57 [ compared to catelan et al . ( 1998 ) , for example ] , and potentially an initial envelope helium abundance @xmath58 that is not consistent with other observational constraints @xcite . however , the general features of this figure should not change . this leads to at least two interesting questions . first , is it possible to observe differences between the agb populations of clusters whose hb stars show second parameter effects ? second , is it possible to predict a cluster s @xmath4 value given its metallicity and particular hb morphology , or does the second parameter affect the agb as well ? from our dataset , we find @xmath59 and @xmath60 for the total sample , excluding stars found only in the @xcite dataset . thus , we have @xmath61 , where the error is computed using @xmath62 this error formula was derived under the assumption that stars that have passed the he flash can only have two identifications ( hb or agb ) , and therefore are described by the binomial distribution with the probability of being in the agb phase @xmath63 . the resulting error @xmath64 is smaller than what is derived from poisson errors by a factor of @xmath65 . the large samples of stars have allowed us to substantially reduce the error on the determination . the maximum single - star @xmath4 values from @xcite do agree with the observed value for m5 . however , more recent models by @xcite with updated @xmath55 rate and equation of state produce a value @xmath66 , significantly lower than the @xcite and @xcite models . the newer theoretical results are out of agreement with the @xmath4 value for m5 by more than @xmath67 , although they are in better agreement with values from the majority of clusters with reasonably large measured samples . in the tabulation of @xcite , there were 8 clusters with more than 200 measured hb stars ( see fig . [ robs ] ) . excluding m5 , we compute @xmath68 . for the four clusters with redder hb morphologies than m5 , @xmath69 . so , while the most recent theoretical models are in agreement with values from the globular clusters with the largest tabulated samples , m5 has a value that appears to be out of agreement with other clusters and with the theoretical values . in fact , the @xmath4 value is one of the highest seen for a globular cluster to date . that m5 has a high @xmath4 value has been apparent even in smaller subsamples ( see table 7 of s96 ) . the effect that hb morphology has on numbers of agb stars can be crudely seen using the ratio @xmath70 ( lee , demarque , & zinn 1994 ) , where @xmath9 , @xmath10 , and @xmath71 are numbers of blue , variable , and red hb stars . m5 has @xmath72 ( this paper ) , while the slightly more metal - poor cluster m3 has @xmath73 . a preliminary determination of @xmath4 for m3 gives a value of @xmath74 . overall m3 has a redder hb morphology than the more metal - rich m5 , and has an @xmath4 value that is more consistent with typical theoretical predictions . however , the differences in @xmath4 values are not necessarily as simple as a shift in the peak color of the hb distribution because the color distributions of the two clusters overlap to a large extent . in addition , the competing effects of hb morphology and metallicity mean that the masses of hb stars in these two clusters are likely to be similar . ( higher metallicity makes the zahb mass at a given color smaller , while a bluer hb morphology at a given metallicity generally means a lower mean hb mass . ) catelan ( 2000 ) computed synthetic hbs for m5 , and found that a mean mass @xmath75 and a mass dispersion @xmath76 . in a similar study for m3 , @xcite finds @xmath77 and @xmath78 ( although they state that there appears to be a systematic difference in @xmath79 as a function of cluster radius ) . so although the mean mass and mass dispersion for m3 and m5 appear to be similar , m5 has a bluer hb and more agb stars . theoretical models predict that the bluest hb stars that reach the agb spend a somewhat longer time in the agb phase @xcite than redder hb stars due to smaller envelope mass and weaker fusion shell sources . [ this last point was mentioned briefly by @xcite . ] models indicate that the stars with maximal agb lifetimes originate on the blue tail of the hb , where the hb luminosity has dropped by about a factor of 2 from what it is in the instability strip . as a result , there should also be an enhancement of the numbers of agb stars relative to numbers of stars in other phases of evolution . using the numbers of rgb stars tabulated in @xcite , we examine in fig . [ robs ] the population ratio @xmath82 , where @xmath83 is the number of rgb stars more luminous than the hb . the relative positions of the clusters in a plot of @xmath84 versus @xmath85 do not change m5 still has a higher value than clusters with redder morphologies , which supports the idea that m5 stars spend longer times in the agb phase than those in clusters with redder hb morphologies . this indicates that m5 may be anomalous in some way , which would mean that input physics for the stellar models is _ not _ the reason for the difference . we note that the two clusters with the most extended blue hbs ( ngc 2808 and ngc 6752 ) have @xmath4 @xmath86 , consistent with the idea that a significant fraction of the stars are avoiding the agb phase . the cluster m30 also has a low @xmath4 value , but a much more compact blue hb . m30 s metallicity is low , however , which makes it more likely that some of its bluest hb stars would also miss the agb ( see fig . [ rvals ] ) . the cluster m55 stands out as having high @xmath84 and @xmath4 values that are similar to those found for m5 . overall , m55 has a compact blue hb with the hb distribution peaking near the bright end of the blue tail . it also has a considerably lower metallicity than m5 , which probably means that the change in hb track morphology will occur at higher temperatures as well . this should be checked with a study using synthetic hb and agb populations . unfortunately , there are no sets of hb and agb star models tabulated in the literature for the most recent physical inputs . as a result we are unable to gauge the exact effect of the hb morphology of m5 on the @xmath4 value . if the @xcite models are any guide , then the @xmath4 value would be increased if the majority of the hb stars have masses giving them proportionately long agb phases . the large @xmath4 value found for m5 indicates that only a small fraction of the observed hb stars are likely to avoid having an agb phase , which makes it possible to place a color constraint on where the morphology of the post - hb evolutionary tracks change . stars that do not reach the agb still spend a large amount of time brighter than the hb and bluer than the agb . because we have found _ no _ cluster members significantly to the blue of the heavily - populated agb , the post - hb evolutionary tracks are not likely to change until @xmath87 , at the observed tip of m5 s hb . in addition , because the time spent on the agb changes rapidly with decreasing mass , it might be possible in the future to calibrate hb star masses using this feature . it may also be possible to empirically determine which hb stars have the maximum relative agb lifetime @xmath88 . the examination of large agb star samples gives us a way of testing the morphology of hb tracks to see whether stars from the most populated portions of the hb really do reach the agb and whether they spend the predicted amount of time there ( relative to the time spent on the hb ) . by examining clusters with bluer hb morphologies or lower metallicity , we can attempt to empirically identify the point on the hb at which the stars no longer have an agb phase . although there should be smaller differences in @xmath4 values between clusters with red hbs and those with intermediate morphologies , it may also be possible to identify those stars which have agb phases of maximum duration . the luminosity function ( lf ) of red giant branch stars is sometimes used to test theoretical predictions for evolution timescales of evolved stars . although with agb stars we are dealing with smaller samples than are available for rgb stars , we can extract evolutionary information using the cumulative lf . it is well - known that the logarithm of the cumulative lf for stars on the upper rgb is linear with magnitude @xcite . the structure of agb stars is similar to that of rgb stars with the hydrogen fusion shell providing the majority of the luminosity during most of the evolution . we examined the models of @xcite and found nearly linear relationships between @xmath89 and @xmath90 for agb evolution between the luminosity minimum in the agb clump and the luminosity maximum before the first thermal pulse . ( @xmath91 is the age of the star at the luminosity maximum . ) in addition , the slope of the relation is nearly identical for stars of very different mass , and is also fairly insensitive to metallicity . the situation is somewhat complicated by the agb clump ( see [ sclump ] below ) and by thermal pulses ( involving interactions between the he and h fusion shells ) , but if we restrict ourselves to agb stars brighter than the clump , the logarithm of the cumulative lf should also be linear with magnitude . because thermal pulses last a very small portion of the total agb evolution ( 10% or less ) , they can be safely neglected here . for cool stars , magnitudes derived from redder filter bands are most linearly related to @xmath90 . for our rgb and agb samples we have selected the best @xmath24 photometry for each star . we have chosen from the cfht dataset where available , and from the ctio data in all other cases ( except for stars outside the ctio field for which no @xmath24 photometry was available ) . these two sets have the least scatter in the respective cmds and are also very nearly on the same photometric system . the cumulative lf in @xmath24-band for m5 is presented in fig . there are two features in this figure that we wish to discuss : the behavior of the rgb lf at the bright end , and the slopes of the rgb and agb lfs . in the figure we compare with theoretical lfs for the rgb from kim et al . ( 2002 ; hereafter , y@xmath92 models ) for two different color-@xmath93 transformations . the distance modulus used in these comparisons was @xmath94 ( s96 ) . while the models reproduce the observed magnitude difference between the rgb tip and the rgb bump to within 0.1 mag and the slope of the cumulative lf immediately brighter than the bump ( @xmath95 ) , there appear to be too few stars observed near the red giant tip . systematic errors in the color transformations or underestimation of the neutrino emission rates used in the stellar models ( e.g. haft , raffelt , & weiss 1994 ) might be able to account for at least part of the difference . [ changes to neutrino emission rates have other consequences , however : more than a 50% change to the rates would result in greater than a 0.1 mag increase in the brightness of the rgb tip , which would probably be inconsistent with the observations ( raffelt & weiss 1992 ) . ] it should be kept in mind that identification of rgb and agb stars becomes most uncertain near the tip of the rgb where the two branches are closest together in color . however , the number of `` missing '' rgb stars appears to be about 8 , which is as much as 50% of the predicted sample at @xmath96 . small horizontal shifts in magnitude do not affect this discrepancy noticeably . if we use models with [ fe / h ] corresponding to the @xcite scale , the models which match the magnitude level of the rgb bump also fit most of the upper rgb well , and the slope of the rgb in the cmd is a closer match to the observations . however , the fit is unsatisfactory because the tip of the rgb is too faint by 0.4 mag ( 9 observed stars fall above the theoretical prediction ) , and it requires a distance modulus @xmath97 ( well outside the range determined from subdwarf fitting to the main sequence ) . to gauge the significance of this difference , we conducted monte carlo simulations . in each trial we randomly chose 300 stars ( the number of observed stars with @xmath98 ) from a cumulative probability distribution derived from the theoretical cumulative lf from y@xmath92 models ( with green et al . 1987 color-@xmath93 transformations ) . we then measured the largest difference in @xmath99 between the simulated and theoretical values at the @xmath24 magnitude of each simulated star . the largest deviations in number @xmath100 tend to occur near the faint limit of the sample , but the largest fractional differences ( or differences in @xmath99 ) occur closest to the tip in the simulations . even so , in a run of 100,000 trials , we found 1168 trials with @xmath101 . ( the rms value was 0.199 . ) if we look for trials having @xmath102 at @xmath103 , we found 1849 trials of 100,000 meeting the conditions . as a result , the low number of rgb tip stars observed in m5 is marginally significant , with a probability of less than 2% of being due to statistical fluctuations . in view of the importance to the understanding of the physics of the upper rgb , other well - populated clusters should be examined . the fitted slopes of the agb and rgb cumulative lfs are significantly different : @xmath104 for the rgb above the bump ( @xmath105 ) , and @xmath106 for the agb ( @xmath107 ) . the shorter evolutionary timescale for agb stars is responsible for differences in the total numbers of stars , but the slope relates to the change in the evolutionary timescale as the brightness changes . the rgb slope agrees closely with predictions from stellar models @xmath108 where @xmath109 is the cumulative luminosity function for rgb stars . the comparison with predictions for the agb is somewhat more complicated because the slope has a slight dependence on star mass , with the slope being minimum for the bluest hb stars that have an agb phase . from @xcite models , the predicted slope is still too high : @xmath110 where @xmath111 is the cumulative luminosity functions for agb stars . to test whether the difference between the observed and predicted values for the slope of the agb cumulative lf is significant , we conducted additional monte carlo simulations . we ran 50,000 trials selecting 48 stars ( the number of agb stars in m5 that were fit to determine the slope ) from a cumulative probability distribution having a slope @xmath112 , and determined an observed slope from each simulated sample . the trials ( see fig . [ agba ] ) indicate that there is a significant bias toward measuring a shallower slope ( the average was 0.489 , and the mode was 0.47 ) , and the hwhm of the distribution of slopes was 0.09 . the bias comes from stars at the bright end , which are given somewhat more weight due to the nature of the cumulative lf : the stars are less abundant , and contribute at a place in the lf where the addition of a star makes a large fractional change . based on the simulations , our observed slope is quite consistent with the theoretical prediction , and is in fact almost exactly what is expected when the bias in measurement is factored in . however , the simulations indicate that the cumulative lfs are not likely to be able to give us a strong constraint on agb evolution in a practical sense . [ agba ] shows the distribution of results for monte carlo trials involving 100 stars on the upper part of the agb . the mode of the distribution ( 0.50 ) is closer to the input value ( 0.53 ) , and the hwhm of the distribution is smaller ( 0.06 ) . however , there are only a handful of clusters in the milky way that have large enough populations of bright agb stars . with the large agb sample , we can introduce new diagnostics to test our understanding of the physics affecting the evolution of these stars . the agb phase is characterized by the progression of a helium fusion shell source through a helium profile set up during the hb phase . as discussed by @xcite , this means than the distribution of stars on the agb is a reflection of the helium abundance of the gas being consumed by the shell . the spatial resolution of the helium fusion shell as a probe of the chemical profile is poorer than the resolution of the hydrogen fusion shell during the first ascent of the giant branch because the helium shell is thicker . the red giant branch ( rgb ) bump in globular clusters provides us with a way of measuring the depth to which the convective envelope reaches . a similar feature appears at the base of the agb , called the `` agb clump '' . during hb evolution , the action of semiconvective mixing leaves a region of changing helium abundance that the helium fusion shell processes at the beginning of the agb phase . while this happens and the shell stabilizes with the nearly pure helium mixture outside the mixed region , an agb star remains in the clump . ferraro et al . ( 1999 ) discuss a magnitude difference diagnostic involving the agb clump . @xmath113 is defined in analogy to a more widely used indicator for the rgb bump : @xmath114 . ferraro et al . found good agreement with theoretical models using the franec code @xcite for nine globular clusters , although this required careful simulation of the hb and determination of the level of the zahb . because there are no recent sets of hb / agb models available , we will not discuss this indicator further here . in addition , a earlier comparison of the theoretically predicted colors of agb clump stars with observations indicated that the theoretical predictions were too blue ( see fig . 25 of s96 ) . again though , this kind of comparison is somewhat sensitive to composition and color-@xmath115 transformations , and so we have chosen not to discuss this here . the number of stars found in the clump is related to the richness of the helium in what was the outer convective core . to attempt to test the evolution of the clump stars , we introduce a new diagnostic @xmath116 , where @xmath117 is the number of stars in the agb clump ( between the base of the clump and a point 0.25 @xmath10 magnitudes brighter ) . such a definition avoids the need to attempt to identify the upper boundary of the clump in the agb samples we gather . the 0.25 mag size is approximately what is predicted theoretically for the height of the clump , however . the base of the agb clump can be identified using the cumulative lf presented in [ clf ] . the base of the agb clump seems to fall at @xmath118 in the ctio dataset . there are a small number of stars fainter than this , but which appear to be better classified as agb stars than hb stars . we include these stars in the `` clump '' category , and find @xmath119 , @xmath120 , and @xmath121 . we have used @xmath122 to make a more direct connection with theoretical models , we also create a similar operational definition for @xmath24-band photometry : stars within 0.30 mag of the base of the agb clump . we find that the base of the clump is at @xmath123 in the m5 data , and using this we get @xmath124 , @xmath125 , and @xmath126 since the outermost agb stars do not have @xmath24 photometry . from the models of @xcite shown in fig . [ rvals ] , we see a couple of useful features of this indicator : it is roughly constant with hb mass for @xmath127 , and the values for the two most metal - rich compositions tabulated are quite similar : @xmath128 . from this comparison , we find that the observed value for m5 is a little over @xmath129 too low compared to theory . once again , we must remind the reader that the @xcite models have somewhat dated physical inputs that may result in some systematic errors . until updated models are publicly available , we merely point out areas where it may be possible to learn new details about the stellar interiors . as a final note , the models of @xcite indicate that the inclusion of breathing pulses in theoretical models decreases the agb lifetime by more than 25% , primarily by considerably shortening the time spent in the clump . our results for @xmath4 and @xmath5 add more support for the idea that breathing pulses should somehow be suppressed in models of hb stars . we have compiled a nearly complete list of bright rgb , hb , and agb stars for the globular cluster m5 reaching from the core of the cluster to @xmath130 from the center . we have used these samples to conduct a thorough comparison with theory in order to test stellar interior physics under the conditions prevalent in these bright stars . we have introduced a new diagnostic @xmath5 for evaluating the evolutionary timescale in the early part of the agb phase , finding that there is a marginal disagreement between the observations and the models of @xcite . on the whole we find good agreement between observations and theory for the cumulative lfs of the rgb above the rgb bump and for the agb , indicating that the evolutionary timescales in both phases are predicted accurately . an apparent exception appears near the tip of the rgb , where there appear to be too few giants compared to theoretical predictions . the chance that this is a statistical fluctuation is less than 2% . this may be an indication that the neutrino emission rates in the cores of these bright stars are underestimated . the most significant result is the large value for the population ratio @xmath4 compared to recent theoretical values . the high value is probably the result of the particular hb morphology of the cluster and not continuing uncertainties in physical inputs like the cross section for the @xmath55 reaction or the core mixing algorithm . we encourage new calculations of hb and agb phases using updated physics because agb stars provide a means of testing predictions for the morphology of evolutionary tracks for stars fusing he into c and o and of constraining the @xmath55 reaction rate . in particular , they stand the greatest chance of identifying the color at which hb stars change from going into or not going into a typical agb phase , and the range of colors for which the agb phase has its maximum duration . we also call attention to a peculiarity of the distribution of stars on m5 s hb . we find that the mass distribution for hb stars peaks at a position corresponding to the blue edge of the instability strip . the instability strip is heavily populated due to the large dispersion in hb masses ( @xmath131 ) . however , the distribution of stars within the instability strip is heavily biased toward the red half of the instability strip . this is consistent with m5 being an oosterhoff group i cluster , but it means that the first overtone instability strip is underpopulated compared to the fundamental strip to the red and compared to the nonvariable stars to the blue . because m5 has one of the bluest hb morphologies of the oo i clusters ( including the more metal - poor m3 ) , it is a severe test of potential explanations of the oosterhoff dichotomy . some mechanism for reducing the evolutionary timescale of stars in the first overtone instability strip seems to be needed . the blue half of m5 s instability strip _ should be _ more heavily populated the big question is why it is nt . bakos , g. a. , benko , j. m. , & jurcsik , j. 2000 , aca , 50 , 221 brocato , e. , castellani , v. , & ripepi , v. 1996 , , 111 , 809 buzzoni , a. , fusi pecci , f. , buonanno , r. , & corsi , c. e. 1983 , , 123 , 94 caputo , f. , castellani , v. , marconi , m. , & ripepi , v. 1999 , , 306 , 815 carretta , e. & gratton , r. g. 1997 , , 121 , 95 cassisi , s. , castellani , v. , deglinnocenti , s. , piotto , g. , & salaris , m. 2001 , , 366 , 578 cassisi , s. , salaris , m. , & irwin , a. w. 2003 , , 588 , 862 castellani , m. , caputo , f. , & castellani , v. 2003 , , 410 , 871 catelan , m. , borissova , j. , sweigart , a. v. , & spassova , n. 1998 , , 494 , 265 catelan , m. 2002 , , 531 , 826 clement , c. m. et al . 2001 , , 122 , 2587 , 122 , 3183 crocker , d. a. , rood , r. t. , & oconnell , r. w. 1988 , , 332 , 236 dorman , b. & rood , r. t. 1993 , , 409 , 387 dorman , b. , rood , r. t. , & oconnell , r. 1993 , , 419 , 596 ferraro , f. r. , carretta , e. , corsi , c. e. , fusi pecci , f. , cacciari , c. , buonanno , r. , paltrinieri , b. , & hamilton , d. 1997 , , 320 , 757 ferraro , f. r. et al . 1997 , , 324 , 915 ferraro , f. r. , messineo , m. , fusi pecci , f. , de palo , m. a. , straniero , o. , chieffi , a. , & limongi , m. 1999 , , 118 , 1738 ferraro , f. r. , paltrinieri , b. , rood , r. t. , & dorman , b. 1999 , , 522 , 983 fusi pecci , f. , ferraro , f. r. , bellazzini , m. , djorgovski , s. , piotto , g. , & buonanno , r. 1993 , , 105 , 1145 fusi pecci , f. , ferraro , f. r. , corsi , c. e. , cacciari , c. , & buonanno , r. 1992 , , 104 , 1831 fusi pecci , f. , ferraro , f. r. , crocker , d. a. , rood , r. t. , & buonanno , r. 1990 , , 238 , 95 green , e. m. , demarque , p. , & king c. r. 1987 , the revised yale isochrones and luminosity functions ( new haven : yale univ . haft , m. , raffelt , g. , & weiss , a. 1994 , , 425 , 222 harris , w. e. 1996 , , 112 , 1487 johnson , j. a. & bolte , m. 1998 , , 115 , 693 jurcsik , j. , benk , j. m. , bakos , g. . , szeidl , b. , & szab , r. 2003 , , 597 , l49 kim , y .- c . , demarque , p. , yi , s. k. , & alexander , d. r. 2002 , , 143 , 499 kunz , r. , fey , m. , jaeger , m. , mayer , a. , hammer , j. w. , staudt , g. , harrissopoulos , s. , & paradellis , t. 2002 , , 567 , 643 lee , y .- w . , demarque , p. , & zinn , r. 1994 , , 423 , 248 lejeune , th . , cuisinier , f. , & buser r. 1998 , a&a , 130 , 65 markov , h. s. , spassova , n. m. , & baev , p. v. 2001 , , 326 , 102 olech , a. , wozniak , p. r. , alard , c. , kaluzny , j. , & thompson , i. b. 1999 , , 310 , 759 piotto , g. et al . 2002 , , 391 , 945 raffelt , g. & weiss , a. 1992 , , 264 , 536 rees , r. f. 1993 , , 106 , 1524 reid , n. 1996 , , 278 , 367 renzini , a. & fusi pecci , f. 1988 , , 26 , 199 sandquist , e. l. 2000 , , 313 , 571 sandquist , e. l. , bolte , m. , stetson , p. b. , & hesser , j. e. 1996 , , 470 , 910 ( s96 ) schlegel , d. j. , finkbeiner , d. p. , & davis , m. 1998 , , 500 , 525 storm , j. , carney , b. w. , & beck , j. a. 1991 , , 103 , 1264 straniero , o. , chieffi , a. , & limongi , m .. 1997 , , 490 , 425 straniero , o. , dominguez , i. , imbriani , g. , & piersanti , l. 2003 , , 583 , 878 sweigart , a. v. & catelan , m. 1998 , , 501 , l63 van albada , t. s. & baker , n. 1971 , , 169 , 311 vandenberg , d. a. , swenson , f. j. , rogers , f. j. , iglesias , c. a. , & alexander , d. r. 2000 , , 532 , 430 zinn , r. & west , m. j. 1984 , , 55 , 45 rrrcccl 1 & @xmath132 & @xmath133 & 0.99 & v42 + & 1 & & 12.81 & 10.93 & & pm + 2 & 7.296 & @xmath134 & + & 1 & & @xmath135 & & @xmath136 & ct + 3 & 6.125 & 86.957 & 0.99 + & 2 & & @xmath137 & & @xmath138 & ct + & & & @xmath139 & @xmath140 & & ct + & 2 & & 13.98 & 12.02 & & pm + 4 & 28.847 & @xmath141 & + & 5 & & @xmath142 & & @xmath143 & ct + 5 & 7.478 & @xmath144 & + & 6146 & & @xmath145 & & @xmath146 & cfh + 6 & 35.655 & @xmath147 & + & 6665 & & @xmath148 & & @xmath149 & cfh + 7 & @xmath150 & @xmath151 & 0.99 + & 14 & & @xmath152 & & @xmath153 & ct + & & & @xmath154 & @xmath155 & & ct + & 10 & & 13.87 & 12.37 & & pm + 8 & 30.902 & @xmath156 & + & 15 & & @xmath157 & & @xmath158 & ct + & & & @xmath159 & @xmath160 & & ct + 9 & 53.961 & 69.500 & 0.99 + & 32456 & & @xmath161 & @xmath162 & & hst + & 603 & & @xmath163 & & @xmath164 & cfh + & 22 & & @xmath165 & & @xmath166 & ct + & & & @xmath167 & @xmath168 & & ct + & 15 & & 13.90 & 12.48 & & pm + 10 & 114.220 & 61.938 & 0.99 & i-20 + & 28 & & @xmath169 & & @xmath170 & ct + & & & @xmath171 & @xmath172 & & ct + & 20 & & 14.02 & 12.62 & & pm + [ agb ] rrrcccl 1 & @xmath173 & @xmath174 + & 4873 & & @xmath175 & & @xmath176 & cfh + & 4 & & @xmath177 & & @xmath178 & ct + & & & @xmath179 & @xmath180 & & ct + 2 & 33.925 & 99.017 + & 32361 & & @xmath181 & @xmath182 & & hst + & 6 & & @xmath183 & & @xmath184 & ct + & & & @xmath185 & @xmath186 & & ct + 3 & 15.390 & @xmath187 + & 6680 & & @xmath188 & & @xmath189 & cfh + & 3 & & @xmath190 & & @xmath191 & ct + 4 & @xmath192 & 0.678 + & 210174 & & @xmath193 & @xmath194 & & hst + & 3869 & & @xmath195 & & @xmath196 & cfh + 5 & 442.242 & 184.036 & 0.99 + & 3 & & 13.86 & 12.16 & & pm + 6 & 167.531 & @xmath197 & 0.99 & iv-81 + & 8 & & @xmath198 & & @xmath199 & ct + & & & @xmath200 & @xmath201 & & ct + & 4 & & 13.89 & 12.20 & & pm + 7 & @xmath202 & 45.865 & 0.99 + & 23898 & & @xmath203 & @xmath204 & & hst + & 9 & & @xmath205 & & @xmath206 & ct + & & & @xmath207 & @xmath208 & & ct + & 11 & & 13.93 & 12.38 & & pm + 8 & @xmath209 & @xmath210 & 0.99 & iii-122 + & 10 & & @xmath211 & & @xmath212 & ct + & & & @xmath213 & @xmath214 & & ct + & 5 & & 13.83 & 12.25 & & pm + 9 & @xmath215 & 271.834 & 0.99 + & 6 & & 13.86 & 12.26 & & pm + 10 & 54.805 & 93.936 + & 12 & & @xmath216 & & @xmath217 & ct + & & & @xmath218 & @xmath219 & & ct + [ rgb ] rrrccccccl v1 & 25.226 & 150.907 & 0.99 & & & & 15.107 & 0.493 + v2 & @xmath220 & @xmath221 & 0.99 + v3 & 156.521 & 99.011 & 0.99 & & & & 15.063 & 0.541 + v4 & @xmath222 & 65.207 & & & & & 15.070 & 0.413 & hst 23554 + v5 & @xmath223 & 43.050 & & & & & 15.103 & 0.475 & hst 22389 + v6 & 18.312 & @xmath224 & & & & & 15.011 & 0.515 + v7 & @xmath225 & @xmath226 & 0.99 & 15.035 & 0.344 & s + v8 & 122.626 & @xmath227 & 0.99 & 15.085 & 0.334 & s & 15.072 & 0.514 + v9 & 190.772 & 72.110 & 0.99 + v10 & 111.964 & 369.089 + [ rr ] rrrcccl 1 & @xmath228 & 147.750 & 0.99 & ii469 + & 206 & & 15.32 & 14.58 & & pm + 2 & @xmath229 & 183.246 & 0.99 & ii493 + & 342 & & 15.56 & 15.05 & & pm + 3 & @xmath230 & @xmath231 & 0.99 & ii210 + & 465 & & 15.36 & 15.26 & & pm + 4 & @xmath232 & @xmath233 & 0.99 & ii213 + & 274 & & 15.04 & 14.95 & & pm + 5 & @xmath234 & 132.187 & 0.99 & ii429 + & 263 & & 15.51 & 14.94 & & pm + 6 & @xmath235 & @xmath236 + & 1749 & & @xmath237 & & @xmath238 & ct + & & & @xmath239 & @xmath240 & & ct + 7 & @xmath241 & 128.281 & 0.99 & ii425 + & 759 & & @xmath242 & & @xmath243 & ct + & & & @xmath244 & @xmath245 & & ct + & 399 & & 15.62 & 15.12 & & pm + 8 & @xmath246 & @xmath247 & 0.99 + & 773 & & @xmath248 & & @xmath249 & ct + & & & @xmath250 & @xmath251 & & ct + & 381 & & 15.54 & 15.10 & & pm + 9 & @xmath252 & 295.727 & 0.99 & i43 + & 1075 & & @xmath253 & & @xmath254 & ct + & & & @xmath255 & @xmath256 & & ct + & 431 & & 15.24 & 15.15 & & pm + 10 & @xmath257 & @xmath258 & 0.99 + & 454 & & 15.27 & 15.23 & & pm + [ hb ] rl + 1 & ( v42 ) post - agb star ; pm only + 12 & hst identification ambiguous + 14 & hst identification ambiguous + 22 & hst identification ambiguous , blends with rgb 99 + 23 & hst identification ambiguous + 34 & hst identification ambiguous + 46 & rgb 310 nearby + 60 & rgb star ( hst 10218 ) nearby + [ notes ]

the next generation health study investigates the dating violence of adolescents using a survey questionnaire . each student is asked to affirm or deny multiple instances of violence in his / her dating relationship . there is , however , evidence suggesting that students not in a relationship responded to the survey , resulting in excessive zeros in the responses . this paper proposes likelihood - based and estimating equation approaches to analyze the zero - inflated clustered binary response data . we adopt a mixed model method to account for the cluster effect , and the model parameters are estimated using a maximum - likelihood ( ml ) approach that requires a gaussian hermite quadrature ( ghq ) approximation for implementation . since an incorrect assumption on the random effects distribution may bias the results , we construct generalized estimating equations ( gee ) that do not require the correct specification of within - cluster correlation . in a series of simulation studies , we examine the performance of ml and gee methods in terms of their bias , efficiency and robustness . we illustrate the importance of properly accounting for this zero inflation by reanalyzing the next data where this issue has previously been ignored . , ,

1506.00360

in public health studies , clustered or longitudinal binary responses may be collected on a group of individuals where only a subset of these individuals are susceptible to having a positive response . for example , questionnaires may ask teenagers who are dating to answer a series of questions about dating violence . as in the next generation health study , a larger proportion of all zero responses are observed than would occur by chance ; presumably many individuals who are not dating filled in all zeros on the questionnaire ( also known as `` structural zeros '' ) . while there may be alternative reasons for structural zeros , for example , participants giving socially desirable responses , we believe this accounts for only a small fraction of zero inflation . interest is in making inference about the correlated binary responses for those who are susceptible ( i.e. , inference about dating violence among individuals who were dating ) . there is an extensive literature on zero - inflated poisson and binomial models [ @xcite ; @xcite ] that provide early references , along with more recent work on zero - inflated ordinal data [ @xcite ] and zero - inflated sum score data with randomized responses [ @xcite ] . @xcite reviewed various statistical models incorporating zero inflation in both discrete and continuous outcomes for cross - sectional data . @xcite discussed cross - sectional binary regression with zero inflation , and proved the model identifiability when at least one covariate is continuous . @xcite first considered longitudinal or clustered data with zero - inflated binomial or poisson outcomes . they incorporated a random effect structure to model the within - subject correlation and proposed an em algorithm to estimate the parameters . @xcite extended the work of @xcite by proposing a generalized estimation equation ( gee ) approach to model several zero - inflated distributions in a longitudinal setting . @xcite presented a hurdle model with random effects for repeated measures of zero - inflated count data . there has been no work , however , on zero - inflated clustered binary data . a component of the next generation health study examines the prevalence and correlates of dating violence among 2787 tenth - grade students , following them over seven years . dating violence is common among adolescents , may impact adolescent expectations regarding adult intimate relationships [ @xcite ] , and has been found to be associated with increased risk of depression and engagement in high - risk behaviors [ @xcite and @xcite ] . thus , dating violence among adolescents merits interest from both developmental and public health perspectives [ @xcite ] . investigators involved in the next study are primarily interested in identifying the risk factors associated with dating violence . @xcite found a relationship between high - risk behaviors ( i.e. , depressive symptoms , alcohol use , smoking and drug use ) , gender and the prevalence of dating violence victimization . a total of 10 questions were asked about dating violence . five of the questions were on dating violence victimization : did your partner ( 1 ) insult you in front of others , ( 2 ) swear at you , ( 3 ) threaten you , ( 4 ) push or shove you , or ( 5 ) throw anything that could hurt you ; the other five were similar questions on perpetration : did you ( 1 ) insult your partner in front of others , ( 2 ) swear at your partner , ( 3 ) threaten your partner , ( 4 ) push or shove your partner , or ( 5 ) throw anything that could hurt your partner ? as illustrated in figure [ fig1 ] , the distribution of the number of `` yes '' responses is clumped at zero . when we fit the frequencies with a zero - inflated binomial distribution , the zero - inflation probability is estimated to be about 58% . the binomial distribution yields a poor fit to the frequencies for two reasons . first , the prevalence of `` yes '' responses is unequal across different questions ; second , the responses from the same subject are correlated . but this only serves as an intuitive visualization of zero inflation . one can argue that the clump of zeros might be due to the high correlation of the binary responses within the same subject ; and , therefore , we also fit the generalized linear mixed model ( glmm ) and plot the fitted frequencies in figure [ fig1 ] . glmm attempts to fit the spike at 0 , and hence tends to overestimate the within - subject correlation . in this paper , we hope to explore whether zero inflation exists while allowing for the cluster effects . we propose maximum - likelihood ( ml ) and gee approaches to simultaneously account for the zero inflation and clustering in the multiple binary responses . the major difference between our work and the previous work is that @xcite , @xcite , and @xcite all considered the zero inflation at the observation level , while in our paper the zero inflation is at the subject level ( meaning that with a structural zero , all the binary responses from a subject are zero ) . for our dating violence example , subjects have all zero responses because they are not susceptible to the condition ( e.g. , in a relationship ) . the proposed methods are evaluated and compared in simulation studies . we then reexamine the relationships between high - risk behaviors and dating violence among teenagers using the proposed analysis strategy accounting for zero inflation . in section [ sec2 ] we present both maximum - likelihood and gee approaches for parameter estimation . section [ sec3 ] discusses the identifiability of the proposed model and proposes a likelihood ratio test for zero inflation . simulation study results are presented in section [ sec4 ] . the next dating violence data is analyzed in section [ sec5 ] , and a discussion follows in section [ sec6 ] . let @xmath0 be the multivariate binary outcome for subject @xmath1 ( @xmath2 ) , and @xmath3 be the corresponding matrix of covariates . let @xmath4 be the latent class , so that @xmath5 always takes the value of @xmath6 ( structural zero ) if @xmath7 , and @xmath5 follows a multivariate binary distribution with density @xmath8 if @xmath9 , where @xmath10 is a vector of parameters . we suppress the subscript @xmath1 when there is no confusion . let @xmath11 be the prevalence of the latent class 1 . in our example , @xmath9 indicates that subject @xmath1 is susceptible to the possibility of dating violence ( i.e. , potential of answering the dating violence questions in a positive fashion ) , while @xmath7 indicates that the subject is not susceptible . if both @xmath12 and @xmath13 are observed , the individual contribution to the full data likelihood is @xmath14 the observed likelihood of @xmath12 is then given by @xmath15 here we assume that the zero - inflation probability @xmath16 is the same across all the subjects in the sample . this could easily be extended to allow @xmath16 to depend on covariates , for example , with a logistic regression model . we use a generalized linear mixed effects model ( glmm ) to describe the multivariate distribution , @xmath17 : @xmath18 where @xmath19 , @xmath20 is the vector of random effects following the multivariate normal distribution @xmath21 , @xmath22 is the design matrix of the random effects , and @xmath23 is the known link function . the parameter vector @xmath10 consists of the parameter of interest @xmath24 and the nuisance parameters in the variance component @xmath25 . assume @xmath26 s are mutually independent given @xmath27 and @xmath20 , and let @xmath28 be the probability density function of @xmath20 . then the likelihood for subject @xmath1 becomes @xmath29 the integral with respect to the random effects can be approximated by gaussian hermite quadrature as @xmath30 where @xmath31 is the @xmath32th quadrature grid point and @xmath33 is the associated weight [ @xcite ] . the parameter estimation for @xmath16 and @xmath10 can be found by maximizing the log - likelihood for all @xmath34 subjects , @xmath35 . the variance estimation is calculated from the inverse of the observed information : @xmath36 and can be implemented by the ` optim ` function in ` r ` [ @xcite ] . likelihood - based inference makes full distributional assumptions on @xmath37 . when these assumptions are correct , the estimator gains efficiency ; otherwise , classical inference has poor statistical properties . we explore the estimating equations approach [ @xcite ] that only specifies a structure for the conditional mean @xmath38 . suppose @xmath39 where @xmath23 is the known link function and @xmath40 is the regression coefficients of interest . unconditional on @xmath4 , the `` marginal '' mean of @xmath5 is given by @xmath41 the estimating equations can then be written as @xmath42 where @xmath43 and @xmath44 is the working covariance matrix for @xmath5 [ @xcite ] . we can decompose @xmath44 as @xmath45 with @xmath46 being the diagonal matrix of the variance of @xmath26 [ which is @xmath47 and @xmath48 being the working correlation matrix specified by some nuisance parameter @xmath49 . if the mean model ( [ eq0-mean ] ) is correct , the estimating equations ( [ eq1-gee1 ] ) are always consistent regardless of the working correlation , and choosing an approximately correct working correlation generally leads to improved efficiency . in the context of zero - inflated regression , we propose two ways to specify the working correlation : marginal and conditional specification . the marginal correlation directly makes assumptions on @xmath48 , which is similar to the standard gee : the _ marginal independent correlation _ assumes @xmath50 , the @xmath51-dimensional identity matrix ; the _ marginal exchangeable correlation _ assumes that @xmath52 , where @xmath53 is the @xmath54 square matrix of ones . we refer to these two different approaches as gee - mi and gee - me , respectively . the conditional correlation exploits the zero - inflated structure and utilizes the conditional covariance , @xmath55 , to derive the unconditional covariance @xmath56 . a similar idea was first used by @xcite to derive their gee estimator for observation - level zero inflation . by the law of total covariance , for @xmath57 , @xmath58 where @xmath59 is the @xmath60 element of @xmath61 , and @xmath62 is the @xmath63th element of @xmath64 . the variance of @xmath26 is given by @xmath65 . conditional independence correlation _ assumes that @xmath66 , so the working correlation is @xmath67 with the @xmath60 element as @xmath68 the _ conditional exchangeable correlation _ assumes that @xmath69 that is , a correlation of @xmath70 between any @xmath26 and @xmath71 _ given _ @xmath72 . therefore , the @xmath60 element of the working correlation @xmath73 is @xmath74 we refer to these conditional gee approaches as gee - ci and gee - ce , respectively . similar to the ordinary gee , an unstructured working correlation can be assumed that allows for distinct correlations for each pair of outcomes . with the unstructured gee , the marginal and conditional specification of working correlation are equivalent , that is , @xmath75 we refer to this approach as gee - un . with each of the five forms of working correlation matrices , we could solve ( [ eq1-gee1 ] ) using the newton raphson method to obtain the corresponding parameter estimates @xmath76 . with the exchangeable or unstructured correlation structure , we iteratively update @xmath70 from its moment estimator and @xmath40 from equation ( [ eq1-gee1 ] ) [ @xcite ] . according to the standard theory of gee , the variance of the estimated @xmath76 has the usual sandwich form @xmath77 , where @xmath78 we note that the regression parameters in the glmm and gee are not directly comparable as they have different interpretations . the former is interpreted as the `` subject - specific effect '' conditional on a subject @xmath1 , while the latter is the `` population - averaged effect '' or `` marginal effect '' [ @xcite ] . thus , glmm and gee are not compatible for nonidentity link functions . in other words , if the glmm is true , the marginal expectation by integrating out the random effects @xmath20 may not preserve the linear additive form of the covariates . however , for binary regression with a probit link and random intercept , glmm and gee are compatible . we adopt a probit random effects model for both the simulations and example analysis . let @xmath79 and @xmath80 be the c.d.f . and p.d.f . of the standard normal distribution . consider the generalized linear mixed effects model with a probit link and a random intercept only , @xmath81 by integrating out @xmath82 , the marginal probability of @xmath26 is computed as follows : @xmath83 while glmm estimates @xmath84 , gee estimates @xmath85 . the latter is a probit regression model as well , with the regression coefficients , @xmath86 . this allows us to compare the performance of glmm and gee by comparing the marginal effects of the covariates , which is our interest in the dating violence analysis of the next study . in general , zero - inflated models are mixtures of two parametric parts , a point mass at zero ( equivalently , a binary distribution with @xmath87 ) and a parametric distribution for the nonstructural zero part . typically , zero - inflated models are identified by observing a larger number of zeros than would be consistent with the parametric model . for example , with poisson or binomial outcomes , one can observe excessive proportion of zeros with a histogram . for a single binary outcome , zero inflation can not be distinguished from rare events , unless covariate dependence is introduced . when there is a continuous covariate @xmath88 , zero inflation is identified because of the linear effect of @xmath88 on the binary response through a known link function . @xcite proved a weaker sufficient condition for identifiability when covariates are all categorical : to identify a two component mixture of logistic regressions with a binary response , the covariate vector needs to take at least 7 distinct values . @xcite also used the same argument to prove the identifiability of zero - inflated ordinal regression . single binary outcome can be seen as a special case of our proposed model with @xmath89 and @xmath90 . as more information is available with @xmath91 , our model is also identified under follmann and lambert s condition . @xcite proved the model identifiability for the zero - inflated binary regression with at least one continuous covariate . using a similar technique , we can prove our model identifiability . for gee with a probit link , consider @xmath92 and @xmath93 to be two parameter vectors that yield the same conditional mean @xmath94 , that is , @xmath95 equivalently , @xmath96 . suppose the @xmath97th component of @xmath98 ( i.e. , @xmath99 ) is continuous , then we can take the partial derivative with respect to @xmath99 , which yields @xmath100 taking the partial derivative on both sides of ( [ eq2-iden ] ) with respect to @xmath99 , and with some algebra , it follows that @xmath101 , and hence @xmath102 . from ( [ eq1-iden ] ) , we further get @xmath103 . this proves the identifiability gee - ci and gee - mi . in gee - un , the association parameters are indeed obtained from a moment estimator of the correlation between @xmath26 and @xmath71 . since the mean model is identified , the variance and correlation are also identified . for exchangeable working correlation , the association parameter is the `` average '' correlation between all @xmath26 and @xmath71 pairs with @xmath57 , which is identifiable as well . we now prove the identifiability of the random effects model ( [ eq0-glmm ] ) with a probit link . with normally distributed random effects , the mean of @xmath26 could be marginalized as @xmath104 , where @xmath25 is the variance covariance matrix of random effects @xmath20 . we further assume that a continuous covariate is contained in @xmath98 but not in @xmath105 . then the same argument of ( [ eq2-iden ] ) still applies by denoting @xmath106 as @xmath40 , which proves the identifiability of the regression coefficients up to a scale . now it suffices to prove the identifiability of @xmath25 . denote @xmath107 as the correlation coefficient of @xmath26 and @xmath71 ( @xmath57 ) given @xmath72 . note that from section [ sec2.2 ] , we have @xmath108 since @xmath40 is identifiable , @xmath109 is also identifiable . therefore , if two parameter vectors @xmath110 and @xmath111 lead to the same @xmath112 and @xmath113 , @xmath107 must be the same . furthermore , the regular glmm is identifiable , suggesting that the correlation structure @xmath107 conditional on @xmath72 is uniquely defined by @xmath25 . hence , we prove @xmath114 , and , consequently , the identifiability of the ml estimator is established . we also note that when @xmath90 and no covariates are available , the repeated binary counts could be collapsed into a binomial distribution . the problem then reduces to the zero - inflated binomial model , which is clearly identifiable . in the presence of the random effects , collapsing the binary counts leads to an over - dispersed binomial distribution . @xcite discussed the zero - inflated beta - binomial model , where the over - dispersion is controlled by a beta distributed random intercept . our model assumes that the over - dispersion comes from a normal distributed random intercept . another way to view the proposed model is a mixture of random effect distributions . recall that @xmath26 follows a bernoulli distribution with probability @xmath115 , given by @xmath116 instead of introducing the latent class @xmath4 , we assume that @xmath20 is a mixture of normal distribution and a point mass at @xmath117 : @xmath118 when @xmath119 , the probability @xmath115 is always 0 for @xmath120 , so @xmath5 is the structural zero . it is easy to show that the likelihood is exactly the same as the proposed model . in practice , one may wish to test for the existence of zero inflation , which can be performed under the parametric model framework . the likelihood ratio statistic is given by @xmath121 where @xmath122 is the maximized log - likelihood for the zero - inflated model , and @xmath123 is the maximized log - likelihood for the ordinary glmm . as the null hypothesis ( @xmath87 ) is on the boundary of the parameter space , the asymptotic null distribution of @xmath124 is a mixture of @xmath125 and point mass at 0 , with equal mixture probabilities [ @xcite ] . theoretically , we could also construct a score test statistic similar to the test proposed by @xcite for zero inflation in a poisson distribution . however , for our problem , the likelihood function involves intractable integrals , making the score and information matrix both difficult to evaluate . so in our application , we apply the likelihood ratio test . motivated by the next study , the data generation for the simulation studies mimics the real example . to evaluate the statistical properties of the above methods , simulation studies of the true model and a misspecified model were run with two different levels of within - cluster correlation . a sample size of @xmath126 with a cluster size of @xmath127 questions is considered . the simulations were repeated 5000 times to compare the performance of the naive estimator ( glmm , where the zero - inflation is ignored ) , the maximum - likelihood ( ml ) estimator and the five gee estimators ( gee - mi , gee - ci , gee - me , gee - ce and gee - un ) . we calculated the average ( mean ) and standard deviation ( sd ) of the estimated parameters , average of the estimated standard errors ( se ) and 95% ci coverage rates ( cover ) based on the wald intervals to evaluate the robustness and efficiency of the gee and the maximum - likelihood approaches . twenty gaussian hermite quadrature points were used for computing the glmm and ml estimators . we also tried 10 and 40 quadrature points as well as the adaptive quadrature with 250 simulated data sets . in our simulations , the results are very similar for differing number of quadrature points . our experience for generalized linear mixed models with the logit link function is that gaussian quadrature works very well , and in most situations agq is not needed . in terms of numerical efficiency , we found that the computation time for agq is about 1020 times longer than the fixed quadrature . @@lcd2.3d2.3ccc@ & * parameter * & & & & * se * & * cover * + ml & @xmath128 & 0.700 & 0.700 & 0.014 & 0.014 & 0.949 + & @xmath129 & 0.500 & 0.499 & 0.040 & 0.039 & 0.953 + & @xmath130 & 0.000 & 0.000 & 0.039 & 0.040 & 0.959 + & @xmath131 & 0.894 & 0.895 & 0.027 & 0.027 & 0.949 + & @xmath132 & -0.447 & -0.448 & 0.051 & 0.051 & 0.951 + & @xmath133 & -0.358 & -0.358 & 0.050 & 0.051 & 0.956 + & @xmath134 & 0.179 & 0.179 & 0.051 & 0.050 & 0.949 + & @xmath135 & 0.358 & 0.358 & 0.050 & 0.051 & 0.948 + glmm & @xmath129 & 0.500 & 1.352 & 0.047 & 0.046 & 0.000 + & @xmath130 & 0.000 & -0.444 & 0.029 & 0.030 & 0.000 + & @xmath131 & 0.894 & 0.570 & 0.028 & 0.025 & 0.000 + & @xmath132 & -0.447 & -0.301 & 0.034 & 0.034 & 0.013 + & @xmath133 & -0.358 & -0.239 & 0.033 & 0.034 & 0.060 + & @xmath134 & 0.179 & 0.114 & 0.032 & 0.032 & 0.489 + & @xmath135 & 0.358 & 0.224 & 0.031 & 0.032 & 0.015 + the estimated parameters for ml and glmm methods were marginalized , as we described in section [ sec2.3 ] . in the following sections , we evaluate the performance of the maximum likelihood and gee under a correctly specified and a misspecified model . additional simulation results are reported in the supplementary material [ @xcite ] , including ( a ) the performance of the proposed model with a smaller sample size ( @xmath136 ) ; ( b ) sensitivity of assuming a constant zero - inflation probability when the probability is affected by covariates ; ( c ) performance of zero - inflated beta - binomial model . we generated a continuous subject - level covariate @xmath137 from a standard normal distribution and categorical covariate @xmath138 to denote the questions for each subject . the zero - inflation indicator @xmath4 was generated from @xmath139 with @xmath140 . the outcome @xmath26 was generated from a probit random effects model : @xmath141 where @xmath142 is the indicator function and @xmath82 is the random intercept following a normal distribution @xmath143 . we fixed the regression parameters @xmath144 . the variance component @xmath145 was taken to be @xmath146 and @xmath147 , respectively , to reflect weak ( pearson correlation of about 0.1 ) and strong ( pearson correlation of about 0.45 ) within - cluster correlations . the simulation results are shown in tables [ tab1 ] and [ tab2 ] , where the true regression parameters are the marginal covariate effects given by @xmath148 . @@lcd2.3d2.3ccc@ & * parameter * & & & & * se * & * cover * + gee - mi & @xmath128 & 0.700 & 0.704 & 0.041 & 0.040 & 0.951 + & @xmath130 & 0.000 & -0.002 & 0.096 & 0.096 & 0.951 + & @xmath131 & 0.894 & 0.897 & 0.061 & 0.061 & 0.947 + & @xmath132 & -0.447 & -0.448 & 0.060 & 0.060 & 0.950 + & @xmath133 & -0.358 & -0.358 & 0.056 & 0.057 & 0.952 + & @xmath134 & 0.179 & 0.179 & 0.055 & 0.054 & 0.952 + & @xmath135 & 0.358 & 0.358 & 0.059 & 0.059 & 0.946 + gee - ci & @xmath128 & 0.700 & 0.701 & 0.029 & 0.029 & 0.948 + & @xmath130 & 0.000 & 0.001 & 0.069 & 0.070 & 0.955 + & @xmath131 & 0.894 & 0.897 & 0.044 & 0.043 & 0.953 + & @xmath132 & -0.447 & -0.449 & 0.055 & 0.056 & 0.953 + & @xmath133 & -0.358 & -0.359 & 0.053 & 0.054 & 0.958 + & @xmath134 & 0.179 & 0.179 & 0.052 & 0.052 & 0.951 + & @xmath135 & 0.358 & 0.360 & 0.056 & 0.056 & 0.946 + gee - me & @xmath128 & 0.700 & 0.701 & 0.032 & 0.032 & 0.953 + & @xmath130 & 0.000 & 0.001 & 0.077 & 0.078 & 0.953 + & @xmath131 & 0.894 & 0.898 & 0.049 & 0.049 & 0.949 + & @xmath132 & -0.447 & -0.449 & 0.058 & 0.058 & 0.955 + & @xmath133 & -0.358 & -0.359 & 0.055 & 0.056 & 0.956 + & @xmath134 & 0.179 & 0.180 & 0.054 & 0.054 & 0.952 + & @xmath135 & 0.358 & 0.359 & 0.057 & 0.058 & 0.950 + gee - ce & @xmath128 & 0.700 & 0.701 & 0.029 & 0.029 & 0.950 + & @xmath130 & 0.000 & 0.001 & 0.069 & 0.070 & 0.955 + & @xmath131 & 0.894 & 0.897 & 0.044 & 0.043 & 0.953 + & @xmath132 & -0.447 & -0.449 & 0.055 & 0.056 & 0.954 + & @xmath133 & -0.358 & -0.359 & 0.052 & 0.054 & 0.958 + & @xmath134 & 0.179 & 0.179 & 0.052 & 0.052 & 0.951 + & @xmath135 & 0.358 & 0.360 & 0.056 & 0.056 & 0.946 + gee - un & @xmath128 & 0.700 & 0.702 & 0.036 & 0.036 & 0.952 + & @xmath130 & 0.000 & 0.000 & 0.085 & 0.085 & 0.952 + & @xmath131 & 0.894 & 0.897 & 0.053 & 0.053 & 0.948 + & @xmath132 & -0.447 & -0.448 & 0.058 & 0.059 & 0.954 + & @xmath133 & -0.358 & -0.358 & 0.055 & 0.056 & 0.954 + & @xmath134 & 0.179 & 0.179 & 0.055 & 0.054 & 0.952 + & @xmath135 & 0.358 & 0.359 & 0.058 & 0.058 & 0.948 + @@lcd2.3d2.3ccc@ & * parameter * & & & & * se * & * cover * + ml & @xmath128 & 0.700 & 0.701 & 0.024 & 0.024 & 0.951 + & @xmath129 & 1.500 & 1.503 & 0.088 & 0.089 & 0.955 + & @xmath130 & 0.000 & -0.001 & 0.053 & 0.053 & 0.950 + & @xmath131 & 0.555 & 0.555 & 0.033 & 0.033 & 0.949 + & @xmath132 & -0.277 & -0.278 & 0.037 & 0.038 & 0.952 + & @xmath133 & -0.222 & -0.222 & 0.037 & 0.037 & 0.949 + & @xmath134 & 0.111 & 0.111 & 0.036 & 0.036 & 0.954 + & @xmath135 & 0.222 & 0.222 & 0.037 & 0.037 & 0.951 + glmm & @xmath129 & 1.500 & 2.248 & 0.071 & 0.073 & 0.000 + & @xmath130 & 0.000 & -0.419 & 0.032 & 0.029 & 0.000 + & @xmath131 & 0.555 & 0.384 & 0.029 & 0.026 & 0.000 + & @xmath132 & -0.277 & -0.204 & 0.027 & 0.027 & 0.230 + & @xmath133 & -0.222 & -0.163 & 0.027 & 0.027 & 0.387 + & @xmath134 & 0.111 & 0.079 & 0.025 & 0.026 & 0.763 + & @xmath135 & 0.222 & 0.155 & 0.025 & 0.026 & 0.271 + gee - mi & @xmath128 & 0.700 & 0.712 & 0.083 & 0.088 & 0.949 + & @xmath130 & 0.000 & -0.002 & 0.161 & 0.170 & 0.966 + & @xmath131 & 0.555 & 0.560 & 0.072 & 0.075 & 0.962 + & @xmath132 & -0.277 & -0.279 & 0.050 & 0.050 & 0.949 + & @xmath133 & -0.222 & -0.223 & 0.046 & 0.046 & 0.950 + & @xmath134 & 0.111 & 0.111 & 0.040 & 0.040 & 0.954 + & @xmath135 & 0.222 & 0.223 & 0.047 & 0.049 & 0.953 + gee - ci & @xmath128 & 0.700 & 0.709 & 0.064 & 0.064 & 0.954 + & @xmath130 & 0.000 & -0.006 & 0.121 & 0.123 & 0.969 + & @xmath131 & 0.555 & 0.556 & 0.056 & 0.057 & 0.958 + & @xmath132 & -0.277 & -0.278 & 0.045 & 0.045 & 0.951 + & @xmath133 & -0.222 & -0.222 & 0.042 & 0.042 & 0.946 + & @xmath134 & 0.111 & 0.111 & 0.038 & 0.039 & 0.952 + & @xmath135 & 0.222 & 0.223 & 0.044 & 0.045 & 0.951 + gee - me & @xmath128 & 0.700 & 0.706 & 0.058 & 0.057 & 0.947 + & @xmath130 & 0.000 & -0.001 & 0.115 & 0.114 & 0.954 + & @xmath131 & 0.555 & 0.557 & 0.053 & 0.053 & 0.952 + & @xmath132 & -0.277 & -0.279 & 0.045 & 0.044 & 0.948 + & @xmath133 & -0.222 & -0.223 & 0.042 & 0.042 & 0.947 + & @xmath134 & 0.111 & 0.111 & 0.039 & 0.039 & 0.955 + & @xmath135 & 0.222 & 0.223 & 0.043 & 0.044 & 0.953 + @@lcd2.3d2.3ccc@ & * parameter * & & & & * se * & * cover * + gee - ce & @xmath128 & 0.700 & 0.707 & 0.056 & 0.056 & 0.951 + & @xmath130 & 0.000 & -0.003 & 0.111 & 0.111 & 0.956 + & @xmath131 & 0.555 & 0.556 & 0.052 & 0.051 & 0.951 + & @xmath132 & -0.277 & -0.278 & 0.043 & 0.043 & 0.952 + & @xmath133 & -0.222 & -0.222 & 0.041 & 0.041 & 0.950 + & @xmath134 & 0.111 & 0.111 & 0.038 & 0.038 & 0.955 + & @xmath135 & 0.222 & 0.223 & 0.043 & 0.043 & 0.954 + gee - un & @xmath128 & 0.700 & 0.709 & 0.068 & 0.068 & 0.950 + & @xmath130 & 0.000 & -0.003 & 0.132 & 0.134 & 0.959 + & @xmath131 & 0.555 & 0.558 & 0.060 & 0.060 & 0.952 + & @xmath132 & -0.277 & -0.278 & 0.046 & 0.046 & 0.948 + & @xmath133 & -0.222 & -0.223 & 0.044 & 0.043 & 0.948 + & @xmath134 & 0.111 & 0.111 & 0.039 & 0.039 & 0.951 + & @xmath135 & 0.222 & 0.223 & 0.045 & 0.046 & 0.952 + both the ml and the five gee methods perform reasonably well , in terms of small bias and good ci coverage rate . glmm is seriously biased with poor ci coverage . it can be seen that the ml method is the most efficient , as it makes use of the full distributional assumption of the observed data . on the contrary , gee only relies on the first moments of the outcome . in estimating @xmath16 , the zero - inflated probability , the ses of the gee approaches are more than twice as large as the se of the ml method . the ses for other parameters are also significantly smaller for the ml method . of the five gee methods , we found that gee - ce is the most efficient with the smallest se , while gee - mi is the least efficient . by exploiting the correlation structure induced by the zero - inflation process , the conditional independence and exchangeable working correlation both gain a substantial amount of efficiency , compared to their marginal counterparts . this result is consistent with the simulation results in @xcite . the ses for gee - ce and gee - ci are quite close , implying that adding working dependence to the outcome given @xmath9 would not help much as long as the dependence due to zero inflation is accounted for . we did observe a bigger improvement of gee - ce versus gee - ci for the strong correlation case . but the improvement of gee - ci versus gee - mi is even larger . therefore , we recommend that it is more important to make use of the zero - inflation structure in the gee estimators . although gee - un has the most flexible form of working correlation , it is not as efficient as gee - ci or gee - ce , probably due to estimating a larger number of nuisance parameters . we found that gees may occasionally not have a solution or have a boundary solution ( @xmath149 ) in about 12% of the simulations with @xmath150 . our experience is that nonconvergence or boundary solutions occur more often when the covariate effects are weaker , the within - cluster correlation is stronger , the true zero - inflation probability is closer to 1 , or the model is more severely misspecified . we consider a misspecified model where only the first three questions are correlated and the last two are independent . the data generation for @xmath26 with @xmath151 was the same as in section 3.1 , but @xmath26 for @xmath152 was generated as follows : @xmath153 the random intercept @xmath82 was not added to the last two questions , but a factor of @xmath154 was divided to the coefficients to keep the marginalized regression coefficients the same . in this case , the ml estimator is from a misspecified model since the random intercept model imposes correlation among all the questions . for gee , only the working correlation is misspecified , while the first moment of @xmath26 is still correct . the simulation results are presented in tables [ tab3 ] and [ tab4 ] . @@lcd2.3d2.3ccc@ & * parameter * & & & & * se * & * cover * + ml & @xmath128 & 0.700 & 0.706 & 0.013 & 0.013 & 0.925 + & @xmath129 & 0.500 & 0.273 & 0.047 & 0.047 & 0.000 + & @xmath130 & 0.000 & -0.008 & 0.038 & 0.039 & 0.953 + & @xmath131 & 0.894 & 0.899 & 0.024 & 0.024 & 0.948 + & @xmath132 & -0.447 & -0.446 & 0.051 & 0.053 & 0.963 + & @xmath133 & -0.358 & -0.357 & 0.049 & 0.053 & 0.964 + & @xmath134 & 0.179 & 0.177 & 0.053 & 0.052 & 0.949 + & @xmath135 & 0.358 & 0.355 & 0.053 & 0.053 & 0.949 + glmm & @xmath129 & 0.500 & 1.176 & 0.042 & 0.040 & 0.000 + & @xmath130 & 0.000 & -0.441 & 0.029 & 0.030 & 0.000 + & @xmath131 & 0.894 & 0.570 & 0.028 & 0.024 & 0.000 + & @xmath132 & -0.447 & -0.303 & 0.035 & 0.036 & 0.019 + & @xmath133 & -0.358 & -0.241 & 0.034 & 0.035 & 0.080 + & @xmath134 & 0.179 & 0.114 & 0.034 & 0.034 & 0.519 + & @xmath135 & 0.358 & 0.223 & 0.033 & 0.034 & 0.021 + gee - mi & @xmath128 & 0.700 & 0.704 & 0.041 & 0.040 & 0.951 + & @xmath130 & 0.000 & -0.002 & 0.095 & 0.095 & 0.951 + & @xmath131 & 0.894 & 0.896 & 0.059 & 0.060 & 0.949 + & @xmath132 & -0.447 & -0.447 & 0.059 & 0.060 & 0.953 + & @xmath133 & -0.358 & -0.357 & 0.056 & 0.057 & 0.953 + & @xmath134 & 0.179 & 0.178 & 0.057 & 0.057 & 0.950 + & @xmath135 & 0.358 & 0.358 & 0.061 & 0.062 & 0.948 + @@lcd2.3d2.3ccc@ & * parameter * & & & & * se * & * cover * + gee - ci & @xmath128 & 0.700 & 0.701 & 0.030 & 0.030 & 0.950 + & @xmath130 & 0.000 & 0.000 & 0.070 & 0.071 & 0.955 + & @xmath131 & 0.894 & 0.897 & 0.044 & 0.044 & 0.951 + & @xmath132 & -0.447 & -0.449 & 0.055 & 0.056 & 0.956 + & @xmath133 & -0.358 & -0.358 & 0.053 & 0.054 & 0.957 + & @xmath134 & 0.179 & 0.179 & 0.055 & 0.055 & 0.949 + & @xmath135 & 0.358 & 0.360 & 0.059 & 0.059 & 0.951 + gee - me & @xmath128 & 0.700 & 0.702 & 0.034 & 0.033 & 0.951 + & @xmath130 & 0.000 & 0.000 & 0.079 & 0.080 & 0.952 + & @xmath131 & 0.894 & 0.897 & 0.049 & 0.050 & 0.949 + & @xmath132 & -0.447 & -0.448 & 0.058 & 0.058 & 0.955 + & @xmath133 & -0.358 & -0.358 & 0.055 & 0.056 & 0.954 + & @xmath134 & 0.179 & 0.179 & 0.057 & 0.057 & 0.952 + & @xmath135 & 0.358 & 0.359 & 0.061 & 0.061 & 0.951 + gee - ce & @xmath128 & 0.700 & 0.701 & 0.030 & 0.030 & 0.951 + & @xmath130 & 0.000 & 0.000 & 0.070 & 0.071 & 0.955 + & @xmath131 & 0.894 & 0.897 & 0.044 & 0.044 & 0.950 + & @xmath132 & -0.447 & -0.448 & 0.055 & 0.056 & 0.955 + & @xmath133 & -0.358 & -0.358 & 0.053 & 0.054 & 0.956 + & @xmath134 & 0.179 & 0.179 & 0.055 & 0.055 & 0.950 + & @xmath135 & 0.358 & 0.359 & 0.059 & 0.059 & 0.952 + gee - un & @xmath128 & 0.700 & 0.703 & 0.037 & 0.036 & 0.952 + & @xmath130 & 0.000 & -0.001 & 0.086 & 0.087 & 0.951 + & @xmath131 & 0.894 & 0.897 & 0.053 & 0.054 & 0.948 + & @xmath132 & -0.447 & -0.448 & 0.058 & 0.059 & 0.953 + & @xmath133 & -0.358 & -0.358 & 0.055 & 0.056 & 0.954 + & @xmath134 & 0.179 & 0.179 & 0.057 & 0.057 & 0.950 + & @xmath135 & 0.358 & 0.359 & 0.061 & 0.062 & 0.950 + @@lcd2.3d2.3ccc@ & * parameter * & & & & * se * & * cover * + ml & @xmath128 & 0.700 & 0.730 & 0.015 & 0.015 & 0.472 + & @xmath129 & 1.500 & 0.624 & 0.037 & 0.038 & 0.000 + & @xmath130 & 0.000 & -0.046 & 0.039 & 0.039 & 0.776 + & @xmath131 & 0.555 & 0.561 & 0.024 & 0.024 & 0.941 + & @xmath132 & -0.277 & -0.273 & 0.036 & 0.046 & 0.986 + & @xmath133 & -0.222 & -0.218 & 0.036 & 0.045 & 0.987 + & @xmath134 & 0.111 & 0.103 & 0.047 & 0.044 & 0.933 + & @xmath135 & 0.222 & 0.208 & 0.046 & 0.045 & 0.934 + glmm & @xmath129 & 1.500 & 1.201 & 0.038 & 0.039 & 0.000 + & @xmath130 & 0.000 & -0.416 & 0.029 & 0.030 & 0.000 + & @xmath131 & 0.555 & 0.378 & 0.024 & 0.022 & 0.000 + & @xmath132 & -0.277 & -0.206 & 0.027 & 0.034 & 0.425 + & @xmath133 & -0.222 & -0.164 & 0.027 & 0.034 & 0.613 + & @xmath134 & 0.111 & 0.079 & 0.034 & 0.033 & 0.816 + & @xmath135 & 0.222 & 0.153 & 0.033 & 0.032 & 0.432 + gee - mi & @xmath128 & 0.700 & 0.712 & 0.080 & 0.082 & 0.948 + & @xmath130 & 0.000 & -0.005 & 0.153 & 0.159 & 0.963 + & @xmath131 & 0.555 & 0.558 & 0.066 & 0.067 & 0.954 + & @xmath132 & -0.277 & -0.278 & 0.048 & 0.049 & 0.946 + & @xmath133 & -0.222 & -0.222 & 0.045 & 0.045 & 0.949 + & @xmath134 & 0.111 & 0.111 & 0.054 & 0.054 & 0.951 + & @xmath135 & 0.222 & 0.223 & 0.058 & 0.06 & 0.954 + gee - ci & @xmath128 & 0.700 & 0.707 & 0.066 & 0.066 & 0.944 + & @xmath130 & 0.000 & -0.001 & 0.129 & 0.130 & 0.966 + & @xmath131 & 0.555 & 0.558 & 0.058 & 0.058 & 0.960 + & @xmath132 & -0.277 & -0.279 & 0.046 & 0.046 & 0.948 + & @xmath133 & -0.222 & -0.223 & 0.043 & 0.043 & 0.950 + & @xmath134 & 0.111 & 0.112 & 0.053 & 0.052 & 0.950 + & @xmath135 & 0.222 & 0.224 & 0.057 & 0.058 & 0.951 + gee - me & @xmath128 & 0.700 & 0.709 & 0.069 & 0.068 & 0.948 + & @xmath130 & 0.000 & -0.004 & 0.132 & 0.133 & 0.951 + & @xmath131 & 0.555 & 0.557 & 0.060 & 0.060 & 0.952 + & @xmath132 & -0.277 & -0.278 & 0.046 & 0.046 & 0.946 + & @xmath133 & -0.222 & -0.222 & 0.043 & 0.043 & 0.946 + & @xmath134 & 0.111 & 0.112 & 0.054 & 0.054 & 0.950 + & @xmath135 & 0.222 & 0.224 & 0.058 & 0.059 & 0.953 + @@lcd2.3d2.3ccc@ & * parameter * & & & & * se * & * cover * + gee - ce & @xmath128 & 0.700 & 0.709 & 0.067 & 0.066 & 0.946 + & @xmath130 & 0.000 & -0.004 & 0.129 & 0.130 & 0.959 + & @xmath131 & 0.555 & 0.556 & 0.058 & 0.058 & 0.954 + & @xmath132 & -0.277 & -0.279 & 0.045 & 0.046 & 0.948 + & @xmath133 & -0.222 & -0.223 & 0.043 & 0.043 & 0.948 + & @xmath134 & 0.111 & 0.111 & 0.053 & 0.052 & 0.948 + & @xmath135 & 0.222 & 0.224 & 0.057 & 0.058 & 0.951 + gee - un & @xmath128 & 0.700 & 0.711 & 0.073 & 0.073 & 0.951 + & @xmath130 & 0.000 & -0.005 & 0.140 & 0.142 & 0.954 + & @xmath131 & 0.555 & 0.557 & 0.061 & 0.061 & 0.952 + & @xmath132 & -0.277 & -0.278 & 0.047 & 0.047 & 0.947 + & @xmath133 & -0.222 & -0.222 & 0.044 & 0.044 & 0.948 + & @xmath134 & 0.111 & 0.111 & 0.053 & 0.053 & 0.950 + & @xmath135 & 0.222 & 0.223 & 0.058 & 0.059 & 0.953 + with @xmath155 , the ml approach is almost unbiased for estimating @xmath16 as well as the regression coefficients . when @xmath129 increases to 1.5 , the ml estimator becomes slightly biased with poor ci coverage , especially for @xmath16 and @xmath130 . the estimation of other parameters appears to be robust to the model misspecification , except that the ses for @xmath132 and @xmath133 overestimate the true variability . on the other hand , the five gee methods all perform quite well , in terms of little bias and close - to - nominal coverage rates . similar to the previous simulation study , we observed that about 1% of the gee simulations did not converge for @xmath150 . although the maximum - likelihood approach is biased , its standard error is much smaller than the gee approaches . for example , the ml estimator for @xmath16 in table [ tab4 ] has a se only a quarter as large as that of the gee - ci and gee - ce estimators . as a result of the variance - bias trade - off , the mean squared error for the ml estimator is still smaller than gee . if the interest is in estimation , one can still argue that the ml performs better ; but if the interest is in hypothesis testing , gee methods are preferred , as they are more robust and preserve the correct type i error rate . from the above two sets of simulation studies , we would generally recommend the ml estimator in practice because of its high efficiency . the correlation structure of the outcome is critical in identifying the zero - inflation process . therefore , a full parametric assumption for the correlation can lead to good efficiency in the estimation . however , if this parametric assumption does not hold , the ml estimator could have poor ci coverage rates . in order to perform hypothesis testing , we would prefer the gee approaches , which only rely on the correct mean model and are not sensitive to the working correlation assumption . among the five gee approaches , the gee - ci and gee - ce are the most favorable , because they are more efficient by exploiting the dependence structure induced by the zero - inflation process . in practice , the gee - ci and gee - ce estimators may be computed in conjunction with the ml estimator as a sensitivity analysis . in this section we fit our proposed ml and gee models to the dating violence example , together with the naive glmm model . a total of 2787 students were enrolled in the study , among which 664 left all the dating violence questions blank . these 664 subjects are either not in a relationship , and thus skipped these questions , or they did not respond at all to the whole survey . in the remaining 2123 subjects , 39 were excluded because they only answered part of the dating violence questions , and 61 were excluded because they have missing data in the covariates . the final analysis sample was @xmath156 . the clustered outcomes of interest ( @xmath26 ) are the ten questions of dating violence , including five victimization and five perpetration questions . we can see from figure [ fig1 ] that the frequency histogram of `` yes '' responses shows a huge spike at 0 . it seems likely that some students who answered all the questions with `` no '' were not in a relationship , that is , a zero inflation of the outcome . define the latent variable @xmath9 if the subject is in a relationship and 0 otherwise . we included gender ( @xmath157 ) , depressive symptoms ( @xmath158 ) , family relationship ( @xmath159 ) and family influence ( @xmath160 ) as the predictors of @xmath26 given @xmath9 . the ds score comes from the questionnaire of depressive symptoms and is on the continuous scale ranging from 1 to 5 , with the larger score indicating worse depressive symptoms . the fr ( ranging from 0 to 10 ) measures the participant s satisfaction with the relationship in his / her family , with 10 being a very good relationship . the fi score ( ranging from 1 to 7 ) is the family influence on the participants not verbally or physically abusing their romantic partner , with a higher score being greater influence . we adjust for question number as a factor and question type ( victimization vs. perpetration ) , in order to account for different prevalence of yes responses . the interactions between question type and other covariates ( @xmath161 , @xmath162 , @xmath163 and @xmath164 ) are also included . the summary statistics of these variables are described in table [ tab5 ] . @@ld2.6@ * variable * & + gender : female & 56.6% + ds score & 2.0 ( 1.0 ) + family relationship & 7.4 ( 2.3 ) + family influence & 5.7 ( 1.8 ) + question 1v insult you & 18.5% + question 1p insult your boyfriend / girlfriend & 16.8% + question 2v swear at you & 31.3% + question 2p swear at your boyfriend / girlfriend & 26.1% + question 3v threaten you & 7.2% + question 3p threaten your boyfriend / girlfriend & 5.6% + question 4v push you & 13.5% + question 4p push your boyfriend / girlfriend & 9.9% + question 5v throw object at you & 4.5% + question 5p throw object at your boyfriend / girlfriend & 3.8% + denote @xmath27 to be the design matrix including all the covariates and interaction terms mentioned above . we fit the probit random effects model @xmath165 where @xmath166 is the random intercept . this model has the same marginal mean as the marginal probit regression model : @xmath167 as @xmath168 for @xmath169 . for comparative purposes , we report the marginal regression coefficients @xmath40 for all the analyses . = @@ld2.10d2.10d2.10d2.10d2.10d2.10d2.10@ * parameter * & & & & & & & + ( intercept ) & -0.192 ( 0.176 ) & -0.746 ( 0.177 ) & -0.055 ( 0.254 ) & -0.159 ( 0.225 ) & -0.497 ( 0.222 ) & -0.341 ( 0.221 ) & -0.201 ( 0.232 ) + question 2 & 0.528 ( 0.037 ) & 0.374 ( 0.024 ) & 0.578 ( 0.071 ) & 0.508 ( 0.057 ) & 0.475 ( 0.058 ) & 0.488 ( 0.057 ) & 0.529 ( 0.061 ) + question 3 & -0.754 ( 0.043 ) & -0.577 ( 0.031 ) & -0.814 ( 0.069 ) & -0.750 ( 0.061 ) & -0.716 ( 0.063 ) & -0.729 ( 0.061 ) & -0.768 ( 0.063 ) + question 4 & -0.332 ( 0.035 ) & -0.250 ( 0.026 ) & -0.360 ( 0.053 ) & -0.335 ( 0.048 ) & -0.315 ( 0.047 ) & -0.324 ( 0.047 ) & -0.339 ( 0.049 ) + question 5 & -1.003 ( 0.050 ) & -0.773 ( 0.036 ) & -1.069 ( 0.083 ) & -0.996 ( 0.073 ) & -0.949 ( 0.077 ) & -0.969 ( 0.074 ) & -1.013 ( 0.075 ) + question type & 0.222 ( 0.122 ) & 0.158 ( 0.089 ) & 0.264 ( 0.173 ) & 0.234 ( 0.150 ) & 0.215 ( 0.146 ) & 0.223 ( 0.145 ) & 0.215 ( 0.155 ) + ds score & 0.192 ( 0.036 ) & 0.178 ( 0.035 ) & 0.250 ( 0.043 ) & 0.220 ( 0.038 ) & 0.226 ( 0.037 ) & 0.219 ( 0.037 ) & 0.230 ( 0.039 ) + gender & 0.132 ( 0.067 ) & 0.153 ( 0.054 ) & 0.171 ( 0.076 ) & 0.118 ( 0.068 ) & 0.186 ( 0.067 ) & 0.160 ( 0.066 ) & 0.164 ( 0.070 ) + family relationship & -0.037 ( 0.014 ) & -0.035 ( 0.012 ) & -0.042 ( 0.019 ) & -0.039 ( 0.017 ) & -0.041 ( 0.015 ) & -0.041 ( 0.016 ) & -0.041 ( 0.017 ) + family influence & -0.114 ( 0.017 ) & -0.087 ( 0.015 ) & -0.137 ( 0.021 ) & -0.135 ( 0.018 ) & -0.095 ( 0.017 ) & -0.112 ( 0.017 ) & -0.123 ( 0.018 ) + question type@xmath170ds score & 0.037 ( 0.025 ) & 0.026 ( 0.018 ) & 0.034 ( 0.032 ) & 0.037 ( 0.029 ) & 0.028 ( 0.026 ) & 0.034 ( 0.028 ) & 0.035 ( 0.029 ) + question type@xmath170gender & -0.409 ( 0.051 ) & -0.300 ( 0.037 ) & -0.446 ( 0.066 ) & -0.392 ( 0.057 ) & -0.373 ( 0.056 ) & -0.379 ( 0.056 ) & -0.399 ( 0.059 ) + question type@xmath170family & -0.003 ( 0.010 ) & -0.002 ( 0.008 ) & -0.009 ( 0.014 ) & -0.007 ( 0.012 ) & -0.006 ( 0.012 ) & -0.006 ( 0.012 ) & -0.005 ( 0.013 ) + relationship + question type@xmath170family & 0.025 ( 0.012 ) & 0.019 ( 0.009 ) & 0.031 ( 0.015 ) & 0.025 ( 0.013 ) & 0.026 ( 0.013 ) & 0.024 ( 0.013 ) & 0.028 ( 0.013 ) + influence + @xmath171 & 0.571 ( 0.031 ) & 1 & 0.523 ( 0.062 ) & 0.626 ( 0.075 ) & 0.690 ( 0.104 ) & 0.663 ( 0.089 ) & 0.587 ( 0.070 ) + @xmath129 & 1.033 ( 0.065 ) & 1.627 ( 0.055 ) & & & & & + the results of the ml , glmm and gee estimations are listed in table [ tab6 ] . from the ml estimation , we can see that all four subject - level covariates are significant : the probability of dating violence perpetration was higher for females , those who are more depressed , those who have a worse relationship in their family , and those who are less influenced by their guardians . the interaction terms between question type and gender , and question type and family influence were both significant , suggesting ( a ) boys are more likely to be the victims of dating violence , and ( b ) the family influence has a slightly higher impact on dating violence perpetration than victimization . the finding regarding greater male dating violence victimization in this age is in line with previous studies [ @xcite ; @xcite ] . the impacts of depression score and family relationship are similar regardless of question type . the directions of association are expected and consistent with some of the findings in @xcite . the zero - inflation probability is estimated to be 0.571 , that is , we expect about 43% of the sample ( about 860 subjects ) to be structural zeros . we suspect that a majority of them were not in a relationship , but answered all the dating violence questions with `` no . '' however , there may be alternative reasons for the zero inflation , for example , some kids may give socially desirable answers in the survey and hence underreport dating violence . however , we believe that this only accounts for a small fraction of the structural zeros . the likelihood ratio test statistic for zero inflation is @xmath172 ( @[email protected] ) . the parameter estimations by gee are generally close to ml , but the standard errors are larger . the naive glmm method estimated smaller covariate effects , which could be biased due to ignoring the zero - inflated nature of the data . as pointed out by a referee , the zero - inflation problem could be avoided by including a filter question of asking whether the subject had a relationship or not . the filter question was not included because the study investigators felt that it was an unreliable question to ask . relationships between teenagers today can not easily be characterized , and the investigators felt that explicitly asking this question may limit important responses about violence [ @xcite ] . there are other cases where the susceptible population can not be ascertained accurately . for example , in drug abuse studies , specifically asking whether individuals are drug abusers might not be a question that would result in a reliable response . but we may be interested in whether the abusers seek particular types of treatment . additionally , this approach may be relevant to questions regarding immigration status , where there may be legal implications ( perceived or real ) in answering , and in questions on mental health status , where the respondent may have a reduced ability to accurately report their status . in public health research , excessive zero responses may occur if the population which is susceptible to respond is not carefully screened or is unknown . the resulting zero inflation may have an effect on the results obtained by conventional methods of analysis . in the next generation health study , investigators were interested in identifying predictors of dating violence in teenagers . examining these regression relationships are of interest for those individuals who are in a relationship ( i.e. , the susceptible condition ) . many more individuals completed this study component than investigators thought would be in a relationship at this age . this led to what appears to be zero - inflated clustered binary data . we developed both ml and gee approaches for analyzing such data . through simulations and analysis of the real data example , we found that the ml approach is substantially more efficient than the gee approaches . however , under moderate model misspecification , the ml approach may result in biased inference . it is recommended that , as a sensitivity analysis , both ml and gee approaches be applied in applications . in the gee approach , we treat the regression parameters and nuisance working correlation as orthogonal , that is , a gee1 approach with the parameters in the working correlation estimated by a moment estimator . potential efficiency gain could be achieved using an improved version of gee1 [ @xcite ] or gee2 [ prentice and zhao ( @xcite ) ; liang , zeger and qaqish ( @xcite ) ] , by establishing another set of estimating equations on the second moment . however , gee2 requires a correct variance structure with working third and fourth moment model , which is hard to verify with the presence of zero inflation , and results in bias under second , third and fourth moment misspecification . the previous work of @xcite adopted prentice and zhao s gee2 that maintains the parameter orthogonality in their second moment estimating equations , and they argued that only making a first moment assumption may lead to parameter nonidentifiability . this is true in their case , where the zero - inflation probability is at the observation level , so @xmath174 and @xmath62 might be confounded in @xmath175 . however , in the case of subject - level zero inflation , we have shown the model identifiability of the gee1 estimators . in principle , zero - inflated models can not be identified nonparametrically ; parametric assumptions for the nonzero part play a fundamental role in model estimation . for example , zero inflation in poisson and binomial data can be determined by the lack of fit in the zero cells of these respective distributions . in this paper , we assume that the nonzero distribution is given by a generalized linear mixed model with normal random effects . the ml approach exploits the correlation structure in order to distinguish structural zeros and random zeros . intrinsically , gee only uses the mean structure of the binary data in order to estimate regression parameters and , unlike ml , does not use the entire distribution for estimation . in our application , there were very few missing data . however , in many studies with sensitive psychological or behavioral questions , there may be informative missingness . an advantage of the ml approach is that it can more easily be extended to account for informative missing responses [ see @xcite , e.g. ] . the proposed methodology implicitly assumes that the subjects answer the questions truthfully . if this assumption does not hold , the parameter estimation is likely to be biased . we could formulate a likelihood approach if we had good prior information about the distribution of false negative occurrence across the questions . this would require a verification subsample corresponding to the survey questions ( maybe obtained through interviews of parents and friends ) on a fraction of teenagers . the zero - inflation probability in our model is assumed to be constant , but it is straightforward to include covariates in both ml and gee approaches . in addition , our focus is on a cross - sectional inference , that is , analyzing the dating violence data at one point in time ( 11th grade ) . understanding behavior change from adolescence over time is interesting but also challenging . in the longitudinal setting , the zero - inflation probability is time - dependent that can probably be modeled by a latent markov process . we will leave it for future exploration . we greatly appreciate and enjoyed reading the very insightful comments and suggestions by the associate editor and two anonymous referees . their comments not only remarkably improved the quality of the paper , but also motivated our deeper thoughts of possible extensions of the methodology in the future .

the construction of joins and secant varieties is studied in the combinatorial context of monomial ideals . for ideals generated by quadratic monomials , the generators of the secant ideals are obstructions to graph colorings , and this leads to a commutative algebra version of the strong perfect graph theorem . given any projective variety and any term order , we explore whether the initial ideal of the secant ideal coincides with the secant ideal of the initial ideal . for toric varieties , this leads to the notion of delightful triangulations of convex polytopes .

math0506223

given two varieties @xmath0 and @xmath1 in a projective space , their _ join _ @xmath2 is the zariski closure of the union of all lines spanned by a point in @xmath0 and a point in @xmath3 . the join of a variety @xmath0 with itself is the _ secant variety _ of @xmath0 , and the @xmath4-fold join of @xmath0 with itself is the _ @xmath4-th secant variety _ of @xmath0 . it is denoted @xmath5 . the study of joins and secant varieties has a long tradition in algebraic geometry , and many authors have studied the dimension and degree of these varieties . recent references include @xcite . in the emerging field of _ algebraic statistics _ , the construction of joins and secant varieties corresponds to _ mixture models _ @xcite , and it is of considerable interest to compute the defining prime ideals of @xmath6 and @xmath7 from those of @xmath0 and @xmath3 . for recent successes along these lines see @xcite . in this paper , we present a combinatorial framework for the study of joins and secant varieties . the basic setup was already suggested by simis and ulrich @xcite , and our results are generalizations and extensions of theirs . our strategy is summarized by the following steps . first , we take secants and joins of arbitrary projective schemes , and , hence , of arbitrary homogeneous ideals in a polynomial ring . second , we develop the combinatorial study of secants and joins of monomial ideals , relating secants and joins to alexander duality , coloring properties of graphs , antichains in posets , and regular triangulations of polytopes . third , we use grbner degeneration as a tool to reduce questions about secants and joins of arbitrary projective schemes to secants and joins of monomial schemes . among the applications of our technique is a new perspective on classical determinantal ideals , yielding a short unified proof for the grbner basis property of minors and pfaffians . here is the outline for our paper . in section [ sec : joinmono ] we introduce secants and joins of arbitrary ideals and we study the secants of monomial ideals . we give an explicit formula ( theorem [ thm : alex1 ] ) , valid in characteristic zero , for computing the join of monomial ideals by multiplying their alexander duals . in section [ sec : graph ] we focus on the case of ideals generated by quadratic monomials . if the generators are squarefree , so the ideal is an _ edge ideal _ , then the secants reflect coloring properties of the graph . as a consequence , perfect graphs make a surprise appearance , and we get a commutative algebra version ( theorem [ caspgt ] ) of the celebrated _ strong perfect graph theorem _ @xcite . in section [ sec : initial ] we show that the secant of an initial ideal contains the initial ideal of the secant . this allows for the derivation of numerical invariants of the secants of an ideal from the secants of carefully chosen initial ideals . we also introduce the notion of a _ delightful term order _ for a variety @xmath0 . this is a term order where taking secants commutes with taking initial ideals . diagonal term orders for determinantal and pfaffian ideals are delightful . in section [ sec : triangle ] we apply our techniques to the study of secant varieties of toric varieties . we show how information about such secant varieties can be derived from regular triangulations of the corresponding polytopes . we are particularly interested in finding _ delightful triangulations _ which correspond to delightful term orders for toric varieties . the existence of delightful triangulations is explored for veronese varieties , segre varieties and scrolls . we close the introduction with an example which demonstrates how our approach can be used to derive equations defining secant varieties . let @xmath8 be the cubic veronese surface in its standard toric embedding . consider the grbner degeneration of @xmath0 into a union of nine coordinate planes corresponding to the triangulation depicted in figure 1 . * figure 1 * : a delightful triangulation for the cubic veronese surface . the initial ideal of the surface @xmath0 with respect to this term order is the edge ideal @xmath9 whose graph @xmath10 consists of all non - edges of this triangulation : @xmath11 consider the variety @xmath12 of secant planes . this is a hypersurface in @xmath13 and we wish to compute its defining polynomial @xmath14 . to do so , we apply theorem [ notcolorable ] below to see that the ideal of the combinatorial secant variety equals @xmath15 the generator is the unique clique of size four in @xmath10 . equivalently , it is the unique independent set of size four in the edge graph of the triangulation . the desired polynomial @xmath14 has degree at least four , and its leading term is a multiple of @xmath16 . consider the quartic invariant of ternary cubics @xcite @xmath17 this polynomial vanishes on @xmath12 and it has the correct leading term . this proves that the desired generator @xmath14 equals the quartic polynomial above . the idea of using grbner degenerations as a tool to study the dimensions of secant varieties was suggested to us by rick miranda . bernd sturmfels was supported by the nsf ( dms-0456960 ) . this work was carried out while he visited eth zrich in the summer of 2005 . seth sullivant was supported by an nsf graduate research fellowship . let @xmath18 be ideals in a polynomial ring @xmath19 = { \mathbb{k}}[x_1,\ldots , x_n]$ ] over a field @xmath20 . their _ join _ @xmath21 is a new ideal in @xmath19 $ ] which can be computed as follows . we introduce @xmath22 new unknowns , grouped in @xmath4 vectors @xmath23 , @xmath24 , and we consider the polynomial ring @xmath25 $ ] in all @xmath26 unknowns . let @xmath27 be the image of the ideal @xmath28 in @xmath25 $ ] under the map @xmath29 . then @xmath30 is the elimination ideal @xmath31.\ ] ] of particular interest is the case when all @xmath4 ideals are identical . we define the _ @xmath4th secant _ of an ideal @xmath32 $ ] to be the @xmath4-fold join of @xmath33 with itself : @xmath34 the join operation @xmath35 of ideals is commutative and associative . moreover , it satisfies the following distributive law with respect to intersection . [ distributive ] if @xmath36 and @xmath37 are ideals in @xmath19 $ ] then @xmath38 see proposition 1.2 ( i ) in @xcite . if the ideals @xmath39 are geometrically prime then their join @xmath40 is also geometrically prime . similarly , for the properties of being geometrically primary and for being radical , provided @xmath20 is a perfect field . see ( * ? ? ? * proposition 1.2 ) . thus for homogeneous prime ideals , the ideal - theoretic join and secant represent the prime ideals of the secant varieties and joins of irreducible projective varieties , the setting discussed in the introduction . for arbitrary ideals , the ideal - theoretic join corresponds to the minkowski sum of affine schemes . this section is concerned with another extreme case , namely , when the given ideals are monomial ideals . we start with the simplest example . [ onevariable ] let @xmath41 and consider the ideals @xmath42 and @xmath43 . then @xmath44 where @xmath45 is the smallest integer such that the characteristic of @xmath20 divides @xmath46 for all @xmath47 . in particular , @xmath48 this example generalizes to _ irreducible _ monomial ideals in @xmath49 variables . such an ideal is represented by an integer vector @xmath50 as follows : @xmath51 [ joinirred ] the join of two irreducible monomial ideals @xmath52 and @xmath53 is an irreducible monomial ideal @xmath54 . here @xmath55 if @xmath56 or @xmath57 . otherwise @xmath58 is the smallest integer such that the characteristic of @xmath20 divides @xmath59 for all @xmath60 with @xmath61 , and if @xmath62 then @xmath63 . a polynomial @xmath64 lies in the join @xmath65 if and only if @xmath66 lies in the monomial ideal @xmath67 this happens if and only if every term @xmath68 lies in ( [ mmonoid ] ) . hence @xmath65 is a monomial ideal . now , @xmath69 lies in @xmath65 if and only if every term of @xmath70 lies in @xmath71 if and only if @xmath72 for some @xmath73 with @xmath74 if and only if @xmath69 lies in @xmath75 . therefore , @xmath76 . we shall prove that the join of monomial ideals is always a monomial ideal . recall that the _ standard monomials _ of a monomial ideal @xmath77 are the monomials in @xmath19 \backslash j$ ] , so @xmath77 is characterized by its set of standard monomials . [ prodofstandard ] let @xmath78 be monomial ideals in @xmath19 $ ] . then @xmath40 is a monomial ideal . if @xmath79 then the standard monomials of @xmath40 are precisely the products @xmath80 where @xmath81 is standard for @xmath28 . if @xmath82 are squarefree , the monomial generating set of @xmath83 is independent of @xmath84 . it suffices to consider the case @xmath85 ; the general statement follows by induction on @xmath4 . if @xmath86 and @xmath87 are irreducible ideals then both statements follow from lemma [ joinirred ] . otherwise , we decompose @xmath88 and @xmath89 as intersections of irreducible monomials ideals ( see @xcite ) . using lemma [ distributive ] , we then write @xmath90 is an intersection of joins @xmath65 . hence @xmath90 is a monomial ideal , and its set of standard monomials is the union of the sets of standard monomials of its irreducible components @xmath65 . if @xmath91 and @xmath92 are irreducible and squarefree the formula for @xmath90 of lemma [ joinirred ] does not depend on @xmath93 . since every monomial ideal in the irreducible decomposition of squarefree monomial ideals is squarefree , we deduce that the monomial generators of @xmath90 are independent of the characteristic of the field @xmath20 . a statement equivalent to proposition [ prodofstandard ] appears in ( * ? ? ? * proposition 3.1 ) . [ cor : standard ] the @xmath4th secant @xmath94 of a monomial ideal @xmath33 is a monomial ideal . every standard monomial of @xmath94 is a product of @xmath4 standard monomials of @xmath33 . if @xmath79 then every such product is standard for @xmath94 . if @xmath33 is squarefree , the generating set of @xmath94 is independent of @xmath84 these results show that the operations of taking joins and secants are very natural from the point of view of alexander duality . namely , forming joins is alexander dual to taking products of monomial ideals , and forming secants is alexander dual to taking powers of monomial ideals . we make this statement precise using the @xmath95}$ ] notation . see ( * ? ? ? * chapter 5 ) for the relevant definitions and basic facts on alexander duality of monomial ideals . [ thm : alex1 ] let @xmath33 and @xmath77 be monomial ideals in @xmath19 $ ] , @xmath79 , and let @xmath96 be a vector in @xmath97 whose coordinates are sufficiently large . then @xmath98 } \cdot j^{[{\mathbf{a}}]}\bigr)^{[2{\mathbf{a } } ] } \quad \,\ , \hbox{\rm modulo } \quad { \bf m}^{{\mathbf{a}}+ { \bf 1 } } .\ ] ] here @xmath99 , and the operation modulo @xmath100 removes all the monomial generators that are divisible by @xmath101 for some @xmath73 . first assume that the given ideals are irreducible , say , @xmath102 and @xmath103 . then @xmath95}$ ] is the principal ideal generated by @xmath104 , and @xmath105}$ ] is generated by @xmath106 . their product is the principal ideal @xmath107 } \cdot j^{[{\mathbf{a } } ] } \ , \ , = \ , \ , \biggl < \ , \prod_{i : u_i>0,v_i>0 } \!\!\ ! x_i^{2a_i + 2-u_i - v_i } \,\,\cdot \!\!\ ! \prod_{i : u_i>0,v_i=0 } \!\!\ ! x_i^{a_i+1-u_i } \,\,\cdot \!\!\ ! \prod_{i : u_i=0,v_i>0 } \!\!\ ! x_i^{a_i+1-v_i } \,\ , \biggr>.\ ] ] taking the alexander dual again , we see that @xmath108 } \cdot j^{[{\mathbf{a}}]})^{[2{\mathbf{a } } ] } $ ] is an irreducible ideal which is generated by three groups of monomials . the first group is @xmath109 the second group of generators of @xmath108 } \cdot j^{[{\mathbf{a}}})^{[2{\mathbf{a } } ] } $ ] is @xmath110 and the third group of generators is @xmath111 reduction modulo @xmath112 removes the second and third group of generators . the remaining first group generates the irreducible ideal @xmath35 , by lemma [ joinirred ] . this proves theorem [ thm : alex1 ] for irreducible monomial ideals . for the general case , we decompose the two given monomial ideals into their irreducible components : @xmath113 and @xmath114 . alexander duality switches intersections of monomial ideals with sum of monomial ideals , so we get @xmath115 } = \sum_\nu ( i_\nu)^{[{\mathbf{a}}]}\,$ ] and @xmath116 } = \sum_\mu ( j_\mu)^{[{\mathbf{a}}]}$ ] . this implies @xmath107 } \cdot j^{[{\mathbf{a } } ] } \,=\ , \sum_{\nu,\mu } ( i_\nu)^{[{\mathbf{a } } ] } \cdot ( j_\mu)^{[{\mathbf{a}}]},\ ] ] and therefore @xmath117 } \cdot j^{[{\mathbf{a}}]})^{[2{\mathbf{a } } ] } \quad = \quad \bigcap_{\nu,\mu } \bigl ( ( i_\nu)^{[{\mathbf{a } } ] } \cdot ( j_\mu)^{[{\mathbf{a } } ] } \bigr)^{[2{\mathbf{a}}]}.\ ] ] using lemma [ distributive ] , and using the result for irreducible ideals , we find @xmath118 } \cdot j^{[{\mathbf{a}}]})^{[2{\mathbf{a } } ] } \quad { \rm modulo } \,\,\,\ , { \bf m}^{{\mathbf{a}}+ { \bf 1 } } .\ ] ] this completes the proof of theorem [ thm : alex1 ] . [ cor : alex2 ] let @xmath33 be a monomial ideal in @xmath19 $ ] , suppose that @xmath79 , and let @xmath119 be a vector in @xmath97 whose coordinates are sufficiently large . @xmath120 } ) ^r \right)^{[r { \mathbf{a } } ] } \quad \mbox { modulo } \quad { \bf m}^{{\mathbf{a}}+ { \bf 1}}.\ ] ] theorem [ thm : alex1 ] and corollary [ cor : alex2 ] can be used for the efficient computation of joins and secants of monomial ideals in characteristic zero . [ m2 ] we present some code in the computer algebra program macaulay 2 @xcite for computing the first secant of a monomial ideal . in our example , the input is the ideal @xmath121 in @xmath122 $ ] . .... r = qq[x , y , z ] ; a = 7 ; i = monomialideal ( x^3 , x^2*y^2 , x*z^3 , y^4 , y^2*z ) ; ma = monomialideal(apply(gens r , u - > u^ ( a+1 ) ) ) ; m2a = monomialideal(apply(gens r , u - > u^(2*a+1 ) ) ) ; ia = monomialideal ( ( gens ( ma : i ) ) % ma ) ; -- alexander dualize ia2 = ia*ia ; -- take square of the result ia22a = monomialideal((gens(m2a : ia2))%m2a);-- alexander dualize monomialideal ( ( gens ia22a ) % ma ) -- reduce modulo m^{a+1 } .... the output of these commands is the join of @xmath33 with itself : @xmath123 note that we compute the alexander dual in macaulay 2 using the formula @xmath124 } \quad = \quad ( { \bf m}^{{\mathbf{a}}+{\bf 1 } } : i ) \,\,{\rm modulo } \,\ , { \bf m}^{{\mathbf{a}}+{\bf 1}}.\ ] ] -.6 cm the proof of theorem [ thm : alex1 ] shows that the smallest possible choice for @xmath96 has @xmath125 where @xmath126 is the largest power of @xmath127 appearing in any minimal generator of @xmath33 or @xmath77 . this guarantees that none of the generators of the form @xmath128 are removed when reducing modulo @xmath129 . for the secant ideal @xmath94 , the smallest possible choice for @xmath96 has @xmath130 where @xmath126 is the largest power of @xmath127 appearing in any minimal generator of @xmath33 . in particular , if @xmath33 and @xmath77 are squarefree monomial ideals we may choose @xmath131 . note that , for @xmath33 squarefree , the ideal @xmath132 } $ ] coincides with the squarefree alexander dual @xmath133 , which is familiar from the study of stanley - reisner ideals @xmath33 . the code above was used with @xmath134 for many examples of squarefree ideals @xmath33 which we computed for the research presented in the next sections . [ complexr ] let @xmath135 be the simplicial complex of @xmath33 and @xmath136 the simplicial complex of @xmath94 . the simplices in @xmath136 are the unions of @xmath4 simplices in @xmath135 . this follows from corollary [ cor : standard ] . see also ( * ? ? ? * corollary 3.3 ) . let @xmath10 be an undirected graph with vertex set @xmath137 = \{1,2,\ldots , n\}$ ] . to @xmath10 we associate the _ edge ideal _ @xmath9 which is generated by the squarefree quadratic monomials @xmath138 corresponding to the edges @xmath139 of @xmath10 . for example , if @xmath10 is the five - cycle with edges @xmath140 then @xmath141 the results below ( or remark [ complexr ] ) show that the secants of this ideal are @xmath142 edge ideals have been much studied in combinatorial commutative algebra . the main emphasis has been on expressing homological invariants of the ideal @xmath9 in terms of the graph @xmath10 . in this section we relate coloring properties of the graph @xmath10 to algebraic properties of the secant ideals @xmath143 . recall that the _ chromatic number _ @xmath144 of a graph @xmath10 is the smallest number of colors which can be used to give a coloring of the vertices of @xmath10 such that no two adjacent vertices have the same color . to the subset @xmath145 $ ] we associate the monomial @xmath146 . a basic first result is : [ prop : chromatic ] the chromatic number @xmath144 of a graph @xmath10 is the smallest integer @xmath147 such that the @xmath4th secant ideal @xmath143 is the zero ideal @xmath148 . the monomial @xmath146 is a standard monomial of @xmath9 if and only if @xmath149 is an independent subset of the vertices of @xmath10 . an @xmath4-coloring is a partition @xmath150 of the vertices of @xmath10 such that each @xmath151 is an independent subset of vertices of @xmath10 . an @xmath4-coloring exists if and only if @xmath152 is a standard monomial of @xmath143 if and only if @xmath153 , since @xmath143 is radical . the proof of proposition [ prop : chromatic ] leads to a combinatorial description of the minimal generators of the secant ideals @xmath143 . given a subset @xmath145 $ ] , we write @xmath154 for subgraph of @xmath10 which is induced on the set of vertices @xmath149 . [ notcolorable ] the @xmath4th secant @xmath143 of an edge ideal @xmath9 is generated by the squarefree monomials @xmath155 whose subgraph @xmath154 is not @xmath4-colorable : @xmath156 the minimal generators of @xmath143 are those monomials @xmath155 such that @xmath154 is not @xmath4 colorable but @xmath157 is @xmath4-colorable for every proper subset @xmath158 . the minimal graphs that are not @xmath159-colorable are the cycles of odd length . this explains the computation for the five - cycle in ( [ fivecycle ] ) . the special case @xmath85 of theorem [ notcolorable ] was already obtained in ( * ? ? ? * proposition 5.1 ) : [ oddcycle ] the secant @xmath160 is minimally generated by the monomials @xmath155 whose corresponding induced subgraph @xmath154 is a cycle of odd length . this implies that even for a monomial ideal @xmath33 , there is no bound on the degrees of minimal generators of @xmath161 in terms of the degrees of the generators of @xmath33 alone . furthermore , if @xmath33 is generated by squarefree quadratic monomials then @xmath161 can not have any minimal generators of even degree . since every graph on @xmath162 vertices can be colored by @xmath4 colors , the minimal generators of @xmath143 have degree at least @xmath163 . this suggests the problem of characterizing the graphs @xmath10 that have the property that the secants @xmath143 are generated in degree @xmath163 . recall that a graph @xmath10 is _ perfect _ if the chromatic number @xmath164 equals the clique number @xmath165 for every subset @xmath145 $ ] . the _ clique number _ is the size of the largest complete subgraph . [ prop : perfect ] a graph @xmath10 is perfect if and only if every non - zero secant ideal @xmath143 is generated in degree @xmath163 . suppose @xmath10 is perfect and let @xmath166 be a minimal generator of @xmath143 . then @xmath154 is not @xmath4-colorable , i.e. @xmath167 . since @xmath10 is perfect , we have @xmath168 and hence @xmath169 . this means there exists a subset @xmath170 such that @xmath157 is a complete subgraph @xmath171 . since @xmath171 is not @xmath4-colorable , the monomial @xmath172 is in @xmath173 . since @xmath155 is a minimal generator , we conclude that @xmath174 . hence @xmath175 and @xmath155 has degree @xmath163 . conversely , if @xmath10 is not perfect then we pick a subset @xmath145 $ ] such that @xmath176 . we may assume that @xmath149 is minimal with this property . setting @xmath177 we have @xmath178 . the monomial @xmath166 is in @xmath143 , whilst @xmath179 for any proper subset @xmath158 . hence @xmath155 is a minimal generator of @xmath143 which has degree larger than @xmath163 . let @xmath180 be the complement of a cycle @xmath181 of length @xmath182 ; that is , the edges of @xmath10 are the edges not appearing in @xmath181 . the edge ideal @xmath9 is a combinatorial model for the elliptic normal curve in @xmath183 , since the variety of @xmath9 has degree @xmath49 and geometric genus 1 . we claim that the secant ideal @xmath143 is the stanley - reisner ideal of the boundary complex of the @xmath184-dimensional cyclic polytope with @xmath49 vertices . to see this , we must analyze the structure of the maximal simplices of the secant complex @xmath185 . if @xmath186 , each of these maximal simplices consists of @xmath184 points . a set @xmath187 $ ] of cardinality @xmath184 is a maximal face of @xmath185 if and only if @xmath188 is a union of @xmath4 pairwise disjoint pairs of the form @xmath189 or @xmath190 . this condition on @xmath188 is equivalent to saying that , for every pair @xmath191 , with @xmath192 , the cardinality of @xmath193 is even . thus , by _ gale s evenness condition _ ( e.g. theorem 0.7 in @xcite ) , the facets of @xmath185 are precisely the facets of the cyclic polytope . note that @xmath143 is generated in degree @xmath163 , unless @xmath194 , in which case there is a single generator in degree @xmath195 , since @xmath10 is perfect if @xmath49 is even and minimally imperfect if @xmath49 is odd . this derivation of the cyclic polytope from the `` stick elliptic normal curve '' was suggested to us by c. athanasiadis and f. santos . the most important development in graph theory in the past few years has been the proof , announced in 2002 , of berge s strong perfect graph conjecture by chudnovsky , robertson , seymour and thomas @xcite . their theorem characterizes perfect graphs in terms of excluded induced subgraphs . [ thm : strong ] the minimal imperfect graphs are precisely the odd holes and the complements of the odd holes . a _ hole _ is a cycle of length greater than 3 . the secants to the edge ideal @xmath9 detect the minimal imperfections in the graph @xmath10 . the strong perfect graph theorem implies the following strong result on the degrees of the minimal generators of the secant ideals @xmath143 . [ cor : perfectgens ] let @xmath10 be an imperfect graph . then either 1 . @xmath160 has a minimal generator of odd degree bigger than three , or 2 . for some @xmath196 , @xmath143 has its minimal generators in degrees @xmath163 and @xmath197 only , and @xmath198 is generated in degree @xmath199 for @xmath200 . let @xmath10 be an imperfect graph . if @xmath10 contains an odd cycle of length @xmath201 then @xmath160 has a minimal generator of degree @xmath202 by corollary [ oddcycle ] . if @xmath10 contains no such odd cycle then , by theorem [ thm : strong ] , the graph @xmath10 contains the complement of an odd hole . let @xmath195 be the minimal length of such a hole . that subgraph has chromatic number @xmath163 , so it contributes a minimal generator of degree @xmath195 to the ideal @xmath143 . theorem [ thm : strong ] also ensures that @xmath143 has no generators of degree other than @xmath163 or @xmath195 . corollary [ cor : perfectgens ] is , in fact , equivalent to the strong perfect graph theorem . if @xmath203 has a minimal generator @xmath155 of odd degree greater than @xmath204 , the induced subgraph @xmath154 must be an odd hole . if @xmath143 has a minimal generator @xmath155 of degree @xmath197 and each of @xmath198 is generated in degree @xmath199 for @xmath200 , then the induced subgraph @xmath154 must be minimally imperfect , with clique number @xmath205 and @xmath206 vertices . a theorem of lovsz @xcite implies that the complementary graph @xmath207 is also minimally imperfect with clique number @xmath208 . thus , @xmath154 must be the complement of an odd hole . one family of perfect graphs are the incomparability graphs of posets @xcite . if @xmath209 is a poset on the set @xmath137 $ ] then the edge ideal of its incomparability graph is the _ stanley - reisner ideal _ of @xmath209 which is defined as follows : @xmath210 in words , the ideal @xmath211 is generated by the @xmath159-element antichains of @xmath209 . that the incomparability graph is perfect follows from _ dilworth s theorem _ which states that the size of the largest antichain of any poset @xmath209 equals the minimal number of chains needed to partition @xmath209 . proposition [ prop : perfect ] implies [ cor : antichain ] let @xmath209 be a poset . then any non - zero secant ideal @xmath212 of the stanley - reisner ideal @xmath211 is generated in degree @xmath163 . more precisely , @xmath213 the secant ideals of graph ideals have other important connections to geometric constructions in the theory of graph coloring . given any projective scheme @xmath0 , there is a natural rational map @xmath214 from the @xmath4-fold free join of @xmath0 to the secant variety @xmath7 . in the statistics literature , the ( nonnegative real ) preimage of a point @xmath215 on ( the nonnegative real part of ) @xmath7 is known as the _ space of explanations _ for the point @xmath215 . see @xcite . in the situation where @xmath216 is the simplicial complex of independent sets in a graph @xmath10 , the space of explanations has a very nice combinatorial interpretation . namely , if @xmath217 and @xmath215 is generic then the space of explanations is a geometric realization of @xmath218 , the polyhedral cell complex of graph homomorphisms from @xmath10 to @xmath219 . see @xcite and @xcite . the graph - theoretic interpretation of secant ideals extends to arbitrary squarefree monomial ideals , by thinking of these as facet ideals as in @xcite . let @xmath220}$ ] be a collection of subsets of @xmath137 $ ] with the property that for every @xmath221 , neither @xmath222 nor @xmath223 . the collection @xmath224 is the set of hyperedges of a hypergraph , or the maximal faces of a simplicial complex . the _ facet ideal _ of the hypergraph @xmath224 is the squarefree monomial ideal @xmath225 a coloring of the vertices of the hypergraph is an assignment of colors with the property that no hyperedge has all its vertices the same color . the chromatic number of @xmath224 is the smallest number @xmath226 such that @xmath224 has a coloring using @xmath226 colors . proposition [ prop : chromatic ] easily generalizes to this setting : the chromatic number of a hypergraph @xmath224 is the smallest positive integer @xmath227 such that @xmath228 . in the next section we shall apply these results to the study of secant varieties of certain irreducible projective varieties . it might make sense to use graph - theoretic language even at that level of generality . we could say that a projective scheme @xmath0 is _ perfect _ if all its secant schemes are cut out by equations of minimal degree , and the smallest secant scheme of @xmath0 which fills the ambient projective space would determine the _ chromatic number _ of @xmath0 . we are inclined to speculate that some version of the strong perfect graph theorem generalizes to arbitrary projective schemes defined by quadrics . one piece of evidence is that the result about the generating degrees of secants in corollary [ cor : perfectgens ] generalizes to arbitrary quadratic monomial ideals . [ caspgt ] let @xmath33 be an ideal generated by quadratic ( not necessarily squarefree ) monomials whose projective scheme is not perfect and let @xmath79 . then either 1 . @xmath161 has a minimal generator of odd degree bigger than three , or 2 . for some @xmath196 , the ideal @xmath94 has its minimal generators in degrees @xmath163 and @xmath197 only , and @xmath229 is generated in degree @xmath199 for @xmath200 . fix a large integer @xmath230 and introduce a graph @xmath231 as follows . there is one vertex @xmath232 for each index @xmath73 such that @xmath233 , and there are @xmath234 vertices @xmath235 for each index @xmath73 such that @xmath236 . two distinct vertices @xmath237 and @xmath238 are connected by an edge in @xmath239 if and only if @xmath240 . we consider the edge ideal @xmath241 in the polynomial ring with variables @xmath237 . we claim that for every integer @xmath242 , the ideal @xmath94 is obtained from the squarefree ideal @xmath243 by replacing @xmath237 by @xmath127 . since theorem [ caspgt ] holds for @xmath241 , we conclude that it holds for the given ideal @xmath33 as well . to prove the claim , note that @xmath244 divides a minimal generator of @xmath94 if and only if @xmath244 is a minimal generator of @xmath94 if and only if @xmath236 . this chain of implications follows because @xmath33 is generated by quadrics and either @xmath245 , in which case @xmath94 is squarefree in @xmath127 , or @xmath236 in which case @xmath246 . since @xmath247 contains a generator that maps onto @xmath244 if and only if @xmath236 , it suffices to show that the replacement procedure sends @xmath247 to @xmath94 when restricted to those monomials not divisible by @xmath244 for any @xmath33 . let @xmath248 be such a monomial and suppose that @xmath249 is not in @xmath94 and that @xmath249 is squarefree in each variable @xmath127 such that @xmath245 . this condition holds if and only if @xmath249 admits a factorization @xmath250 into a product of @xmath4 monomials such that each @xmath251 is standard for @xmath33 . by our assumption on @xmath249 , each @xmath251 is squarefree . since the squarefree standard monomials of @xmath241 map onto the set of squarefree standard monomials of @xmath33 , we deduce that @xmath252 if and only if every squarefree preimage of @xmath249 is standard for @xmath247 . the existence of such squarefree preimages is guaranteed because @xmath242 . we conclude that every minimal generator of @xmath94 arises from a squarefree minimal generator of @xmath247 and hence , by corollary [ cor : perfectgens ] , satisfies the specified requirements on its degree . needless to say , it would be fantastic to find a commutative algebra proof of theorem [ caspgt ] and hence of the strong perfect graph theorem . note , however , that the statement of theorem [ caspgt ] does not hold for all ideals generated by quadrics . in particular , for a non - monomial ideal @xmath33 generated by quadrics , the secant ideal @xmath161 need not have a generator of odd degree . let @xmath33 be generated by two generic homogeneous quadrics in @xmath253 $ ] . the variety of @xmath33 consists of four points in general position in @xmath254 . the secant variety is the reduced union of six lines . hence @xmath255 is a principal ideal generated by a homogeneous polynomial of degree six . in this section we consider the degeneration of secant ideals to their initial ideals . theorem [ thm : inclusion ] and its corollaries on initial degrees appear already in @xcite , but we include some proofs because ours are , perhaps , more elementary . we then examine determinantal and pfaffian ideals . in contrast to the exposition in @xcite , we offer direct new proofs for the grbner basis properties of determinants and pfaffians , using results from section 3 . to be precise , we replace the use of the knuth - robinson - schensted correspondence , first proposed in @xcite , by dilworth s theorem ( corollary [ cor : antichain ] ) . besides dilworth s theorem , which is a relatively easy combinatorial result , our derivation depends only on elementary linear algebra . the full proof for generic matrices is presented in theorem [ nomorekrs ] and corollary [ cor : diagonalgb ] below . let @xmath256 be arbitrary ideals in @xmath19 $ ] and @xmath257 any term order . then the initial ideal of a join is contained in the join of the initial ideals . [ thm : inclusion ] we have the following inclusions of monomial ideals : @xmath258 it suffices to consider the case of the join of two ideals @xmath35 ; the general result following by induction on @xmath4 . consider any polynomial @xmath259 . let @xmath260 be a weight vector which represents the term order @xmath257 in the sense that @xmath261 , @xmath262 and @xmath263 . we denote the latter monomial by @xmath264 . we consider the ideal @xmath265 in the polynomial ring @xmath25 $ ] . the @xmath266-initial ideal of this ideal equals @xmath267 this is seen by refining @xmath266 to a term order and using buchberger s first criterion ( the s - pairs of polynomials with relatively prime leading terms reduce to zero ) . now , since @xmath259 , the polynomial @xmath268 lies in @xmath265 . its @xmath266-leading form equals @xmath269 . this polynomial lies in ( [ threeideals ] ) and hence @xmath270 lies in @xmath271 , as desired . in the special case when all @xmath4 ideals are equal , this proposition implies [ cor : secin ] a secant of an initial ideal contains the initial ideal of the corresponding secant ideal . for any ideal @xmath33 , term order @xmath257 and integer @xmath272 , @xmath273 theorem [ thm : inclusion ] implies lower bounds on the degrees of generators for the ideals of joins and secants of arbitrary projective schemes . for a homogeneous ideal @xmath32 $ ] let @xmath274 denote the smallest degree of any minimal generator of @xmath33 . we omit the proofs which appear in ( * ? ? ? * theorem 4.4 ) . [ cor : joinbound ] let @xmath79 and let @xmath33 and @xmath77 be homogeneous ideals in @xmath19 $ ] . then either @xmath275 or @xmath276 . [ cor : secdegrees ] let @xmath79 and @xmath33 be a homogeneous ideal such that @xmath277 . then either @xmath278 or @xmath279 . the lower bound of corollary [ cor : secdegrees ] is best possible . this is illustrated by the family of determinantal ideals to be featured below . namely , if @xmath33 is the ideal of @xmath280-minors of an @xmath281-matrix of unknowns ( for @xmath282 ) then @xmath94 is the ideal of @xmath283-minors . corollaries [ cor : joinbound ] and [ cor : secdegrees ] do not hold if the field @xmath20 has positive characteristic . for instance , take @xmath284 and @xmath285 , and consider the ideal @xmath286 . we have @xmath287 . by example [ onevariable ] , the first secant ideal is @xmath288 while the bound in corollary [ cor : secdegrees ] says @xmath289 . corollary [ cor : secin ] shows that the secant of the initial ideal @xmath290 can provide useful bounds on numerical invariants of the ideal @xmath94 . an inclusion of monomial ideals leads to a coefficientwise inequality among the hilbert series and hence among values of the hilbert polynomials . this implies : [ cor : dims ] we have the following inequality for the krull dimension : @xmath291/ ( { \rm in}_\prec(i))^{\{r\ } } \,\,\,\leq \,\,\,\dim { \mathbb{k}}[{\mathbf{x}}]/i^{\{r\}}.\ ] ] if these two algebras have the same krull dimension then their degrees satisfy @xmath292/ ( { \rm in}_\prec(i))^{\{r\ } } \,\,\,\leq \,\,\,\deg { \mathbb{k}}[{\mathbf{x}}]/i^{\{r\}}.\ ] ] if equality holds in ( [ secantinclusion ] ) then we say that the term order @xmath257 is _ @xmath4-delightful _ for the ideal @xmath33 . we call @xmath257 _ delightful _ for @xmath33 if this holds for all integers @xmath293 . being delightful implies that equalities hold in corollary [ cor : dims ] . a classical result in combinatorial commutative algebra states that the @xmath294 minors of a generic matrix , the @xmath294 minors of a generic symmetric matrix , and the @xmath295 sub - pfaffians of a generic skew - symmetric matrix are all grbner bases for the ideals they generate . as a corollary , one deduces that these ideals are all prime ideals and one gets formulas for their hilbert series . our approach through secants of initial ideals provides a unified framework for proving these results , using the following strategy : 1 . solve the `` easy '' @xmath296 case by specifying a quadratic grbner basis for @xmath33 whose leading terms correspond to the incomparable pairs in a poset @xmath209 . ( usually , one here has an _ algebra with straightening law_. ) 2 . determine a combinatorial description of the antichains of size @xmath163 in @xmath209 . ( by corollary [ cor : antichain ] , these antichains generate @xmath297 ) . 3 . find a set @xmath298 whose initial terms are the above antichains . 4 . conclude that @xmath299 and @xmath300 is a grbner basis . the following theorem was first proved in @xcite using the knuth - robinson - schensted correspondence . in our new proof , knuth - robinson - schensted is replaced by dilworth s theorem ( imcomparability graphs are perfect ) . [ nomorekrs ] let @xmath33 be the ideal generated by the @xmath301 minors of a generic @xmath302 matrix , and let @xmath257 be any term order on @xmath303 $ ] which selects the diagonal leading term of each @xmath301 minor . then @xmath257 is delightful for @xmath33 . the poset @xmath209 is the product of an @xmath234-chain with an @xmath49-chain , indexed so that the incomparable pairs are @xmath304 with @xmath305 and @xmath306 . one easily checks that @xmath307 . by corollary [ cor : antichain ] , @xmath212 is generated by the monomials @xmath308 with @xmath309 and @xmath310 . each such monomial is the @xmath257-leading term of an @xmath311-minor of the @xmath302-matrix . the affine variety @xmath312 consists of all matrices of rank @xmath313 . since a matrix has rank @xmath162 if and only if it is a sum of @xmath4 matrices of rank @xmath313 , the affine variety @xmath314 consists of all matrices of rank @xmath162 . hence the @xmath315 minors vanish on @xmath314 . now , the ideal @xmath33 is easily seen to be prime over any field , and hence @xmath94 is geometrically prime . hence the @xmath315 minors lie in the ideal @xmath316 . this proves that the monomial ideal @xmath317 is equal to the monomial ideal @xmath318 for all @xmath272 . we conclude that the term order @xmath257 is delightful for the ideal @xmath33 of @xmath301-minors . [ cor : diagonalgb ] the secant ideal @xmath319 is generated by the @xmath294 minors of a generic matrix , and these minors are a grbner basis under any diagonal term order @xmath257 as above . in the proof of theorem [ nomorekrs ] we have argued that the @xmath294-minors lie in @xmath319 , and their leading terms generate the initial ideal @xmath320 . this implies that the @xmath294-minors form a grbner basis for the ideal @xmath319 , and , in particular , they generate that ideal . the ideal of @xmath321 minors of a generic matrix is prime . the ideal @xmath33 of @xmath301-minors is geometrically prime . the secant ideal @xmath319 of a geometrically prime ideal @xmath33 is prime . now use corollary [ cor : diagonalgb ] . the same argument works also for symmetric minors and pfaffians . [ ex : symmetric ] consider a generic @xmath281 symmetric matrix @xmath322 and let @xmath33 be its ideal of @xmath301-minors . let @xmath209 be the poset on the set of variables @xmath323 defined by @xmath324 whenever @xmath325 and @xmath326 . let @xmath257 be the reverse lexicographic term order on any linear extension of @xmath209 . it is easy to check that the @xmath301-minors are a grbner basis for @xmath33 with respect to @xmath257 , and the generators of @xmath327 are the incomparable pairs in @xmath209 . every antichain of size @xmath328 in @xmath209 is the leading term of a @xmath294-subdeterminant of @xmath322 . hence the term order @xmath257 is delightful for @xmath33 , and we conclude that the @xmath294-minors of @xmath322 form a grbner basis of @xmath319 with respect to @xmath257 , and their ideal is prime . [ ex : pfaff ] consider a generic @xmath281 skew - symmetric matrix @xmath322 , and let @xmath33 be the ideal generated by its @xmath329-pfaffians @xmath330 for @xmath331 . these are the _ three - term plcker relations _ , and @xmath33 is the defining ideal of the grassmannian of lines in projective @xmath332-space and , hence , is geometrically prime . let @xmath209 be the poset on the variables @xmath333 defined by @xmath324 whenever @xmath325 and @xmath334 . let @xmath257 be the reverse lexicographic term order on any linear extension of @xmath209 . the three - term plcker relations are a grbner basis for @xmath33 with respect to @xmath257 , and the generators of @xmath327 are the incomparable pairs in @xmath209 ( see , for example , ( * ? ? ? * chapter 14 ) ) . every antichain of size @xmath328 in @xmath209 is the leading term of a @xmath295 subpfaffian of @xmath322 and each @xmath295 subpfaffian lies in @xmath319 . hence the term order @xmath257 is delightful for @xmath33 . we conclude that the @xmath295 subpfaffians of @xmath322 form a grbner basis of @xmath319 with respect to @xmath257 , and their ideal is prime . we do not know whether the plcker ideals of the higher grassmannians , @xmath335 for @xmath336 , and their schubert subvarieties , admit delightful term orders . some computational explorations would be worthwhile . the arguments we have presented also work to show the grbner basis property for ladder determinantal ideals and ladder pfaffian ideals . these ladder ideals consist of ideals generated by the minors and pfaffians contained in staircase shaped regions of a generic matrix , symmetric matrix , or skew - symmetric matrix . each poset @xmath209 for these ideals is a sub - poset of the posets described in theorem [ nomorekrs ] and examples [ ex : symmetric ] and [ ex : pfaff ] . for studies of such ideals and their posets we refer to @xcite and @xcite . in this section we consider the case when @xmath33 is a homogeneous toric ideal , and we examine when there exist delightful initial ideals for these toric ideals . in the case where @xmath327 is generated by squarefree monomials , this corresponds to finding a special regular unimodular triangulation of the point configuration underlying @xmath33 . we begin by briefly reviewing the connection between toric initial ideals and regular triangulations @xcite . let @xmath337 and suppose there is a vector @xmath338 such that @xmath339 for all @xmath73 . the toric ideal @xmath340 $ ] is the kernel of the map @xmath341 \rightarrow { \mathbb{k}}[t_1^{\pm 1 } , \ldots , t_d^{\pm 1 } ] \,,\,\ , x_j \mapsto \prod_{i = 1}^d t_i^{a_{ij}}.\ ] ] let @xmath257 be any term order on @xmath19 $ ] and @xmath342 the initial ideal of @xmath343 . then the radical of @xmath342 is a squarefree monomial ideal whose corresponding simplicial complex @xmath344 is a regular triangulation of @xmath345 . conversely , every regular triangulation of @xmath345 has the form @xmath344 for some term order @xmath257 . a subset @xmath346 of @xmath345 is a simplex of the triangulation @xmath344 if and only if every power of @xmath347 is a standard monomial modulo @xmath342 . if @xmath344 is not full then some @xmath348 is not a vertex of @xmath344 . hence @xmath349 for some @xmath350 . by example [ onevariable ] , @xmath351 contains the monomial @xmath352 . thus @xmath353 for @xmath354 . consider the embedding of the toric surface @xmath356 in @xmath357 by the line bundle @xmath358 . here @xmath359 and the defining configuration is @xmath360 the toric ideal @xmath343 is generated by the @xmath301-minors of the symmetric matrix @xmath361 fix a term order @xmath257 which selects the main diagonal product as the leading term for each @xmath301-minor . the regular triangulation @xmath344 is displayed on the left in figure 2 . the diagram on the right of figure 2 shows a poset @xmath209 to which we can apply steps ( 1)(4 ) of section 4 . note that the maximal chains in @xmath209 are the triangles in @xmath344 . using macaulay 2 we can verify that the @xmath362-minors of @xmath249 form a grbner basis of @xmath363 . in particular , the variety of secant planes to our surface is the hypersurface @xmath364 . this proves that this triangulation of @xmath345 is delightful . note that this example is a specialization of the @xmath365 case in example [ ex : symmetric ] . the veronese example discussed in the introduction has @xmath366 we have seen that the standard triangulation of @xmath345 is @xmath204-delightful . however , it turns out that no full triangulation of @xmath345 is @xmath159-delightful . this can be proved by a brute - force enumeration of all full triangulations of @xmath345 , using cats @xcite or topcom @xcite , and by using the following counting argument . results in @xcite imply that @xmath367 has the expected dimension ( krull dimension @xmath368 ) and its degree equals @xmath369 . for each of the triangulations @xmath344 we count the number of six - tuples of vertices which form the vertices of two disjoint triangles . if there is no such six - tuple then @xmath370 has dimension less than six , so @xmath257 can not be delightful . otherwise , the number of such six - tuples equals the degree of @xmath370 . now , the maximum number arising from any triangulation of @xmath345 is @xmath371 which is less than @xmath369 . the quest for delightful triangulations is a worthwhile undertaking even if no such triangulation exists . namely , the same approach can be used for showing that certain secant varieties of toric varieties are nondefective and to compute a non - trivial lower bound on their degree . recall that a @xmath372 dimensional subvariety @xmath0 of @xmath183 is called @xmath4-_defective _ if the secant variety @xmath7 has dimension less than @xmath373 , which is the _ expected dimension_. if all secant varieties @xmath7 have the expected dimension then @xmath0 is called _ nondefective_. regular triangulations of @xmath345 can be used to prove that the toric variety @xmath374 is nondefective . this was the original problem suggested to us by rick miranda . to make the idea precise , we introduce the following terminology . let @xmath135 be a full triangulation of a configuration @xmath345 of maximal rank @xmath202 . a subset @xmath181 of @xmath345 is called _ @xmath4-partitionable _ if @xmath181 is the disjoint union of @xmath4 maximal simplices in @xmath135 . naturally , if @xmath181 is @xmath4-partitionable then @xmath375 . we write @xmath374 for the projective toric variety in @xmath183 defined by @xmath343 . [ partitionable ] let @xmath135 be a regular triangulation of @xmath345 which has at least one @xmath4-partitionable set . then @xmath376 has the expected dimension , and the degree of @xmath376 is bounded below by the number of @xmath4-partitionable sets in @xmath135 . the @xmath4-partitionable sets are the @xmath377-dimensional simplices of @xmath136 by remark [ complexr ] . the number of @xmath4-partitionable sets is positive if and only if @xmath136 has the expected dimension @xmath378 . in this case , that number is the number of maximal - dimensional simplices in @xmath136 , which is the degree of @xmath136 when regarded as a reduced union of coordinate subspaces in @xmath183 . pick a term order @xmath257 such that @xmath379 . then we have @xmath380 the first equation holds because the degree is preserved under grbner degenerations , the middle inequality holds by corollary [ cor : dims ] , and the last inequality holds because @xmath136 is the reduced scheme defined by the ( possibly non - radical ) ideal @xmath381 . this proves the asserted lower bound . let @xmath382 be a vector of positive integers and fix the configuration @xmath383 represents a product of simplices , and the corresponding toric variety is the product @xmath384 in the standard segre embedding . consider a lexicographic term order @xmath257 such that @xmath385 is higher than all other elements of @xmath386 . since the polytope @xmath387 is smooth , the resulting _ lexicographic triangulation _ @xmath388 has exactly one maximal simplex which contains the vertex @xmath385 . this simplex is denoted @xmath389 , and it is formed by the vertices that are neighbors of @xmath385 . in other words , the simplex @xmath389 contains all @xmath390 such that the hamming distance between the vectors @xmath391 and @xmath392 is at most one . now consider a set of indices @xmath393 with the property that the hamming distance between @xmath394 and @xmath395 is greater than two for all @xmath396 . let @xmath135 be any lexicographic triangulation of @xmath386 which puts the elements in @xmath397 lexicographically larger than all elements of @xmath398 . by our assumption on the hamming distance between elements of @xmath33 , each simplex @xmath399 appears in the triangulation @xmath135 , and these simplices are disjoint . thus , if such an index set of cardinality @xmath400 exists , the secant varieties @xmath401 for @xmath402 will all have the expected dimension by theorem [ partitionable ] . this combinatorial technique for proving that secant varieties to certain segre varieties have the expected dimension was introduced by catalisano , geramita and gimigliano in @xcite . as pointed out in @xcite , finding an @xmath400-element index set @xmath33 with pairwise hamming distance greater than @xmath159 is equivalent to finding a placement of @xmath400 rooks on a @xmath403 chessboard with the property that no two rooks attack each other or attack the same square on the board . our approach via triangulations can be used to get information about further invariants ( beyond dimension ) of such secant varieties . to conclude this section , we explore the existence of delightful triangulations for the class of rational normal scrolls . while all the secant ideals in question are known to have nice determinantal presentations , not every scroll has a delightful term order . this is somewhat surprising , considering our results on delightful term orders for minors and pfaffians in section [ sec : initial ] . a determinantal presentation for the secant ideals @xmath409 is well known . namely , for each @xmath73 and @xmath4 with @xmath410 let @xmath411 denote the @xmath412 hankel matrix @xmath413 if @xmath414 then @xmath411 denotes the empty @xmath415 matrix . the concatenation @xmath416 is a matrix with @xmath163 rows and @xmath417 columns . we first reduce to the two dimensional case . this reduction is possible because a delightful triangulation of a polytope is delightful for any face , and the quadrangle @xmath423 appears as a face of the polytope @xmath424 . to analyze the two dimensional case , we first must understand the full triangulations of the sets @xmath425 . each of these triangulations is lexicographic . the full triangulations of @xmath426 correspond to certain bipartite graphs . namely , aside from the edges like @xmath427 and @xmath428 , the remaining edges form a bipartite planar spanning tree in the complete bipartite graph @xmath429 . planar means that there is no pair of edges @xmath430 @xmath431 with @xmath305 and @xmath432 . an example of such a triangulation and the associated bipartite planar spanning tree appear in figure 3 . if a triangulation of a scroll is delightful , then the associated bipartite planar graph can not possess certain induced subgraphs . we claim that , up to symmetry , these forbidden induced subgraphs are the following two : thus ( 1 ) is a claw @xmath441 which is not adjacent to a vertical boundary , and ( 2 ) is a claw @xmath442 which is adjacent to a vertical boundary . note that the triangulation in figure 3 contains both of these forbidden subgraphs . if the graph of a triangulation contains the @xmath441 in case ( 1 ) then the ideal @xmath443 contains the monomial @xmath444 . however , by virtue of the fact that the full triangulation can be chosen to be lexicographic with @xmath445 , @xmath446 and @xmath447 smaller than any other @xmath448 , and appealing to proposition [ prop : scrolldet ] , we see that this monomial can not be the leading monomial of any polynomial in @xmath449 . a similar argument rules out the subgraph @xmath442 in case ( 2 ) . to finish the proof , note that if @xmath450 and @xmath451 , then the induced graph of any full triangulation of @xmath452 must contain one of the two forbidden subgraphs ( or a subgraph symmetrically equivalent ) . the edge graph of every full triangulation of a configuration @xmath458 is a chordal graph . this can be proved by induction on @xmath459 . let @xmath460 be the complementary graph to that chordal graph . the initial ideal @xmath461 equals the edge ideal @xmath462 . since chordal graphs are perfect , and the complements of perfect graphs are perfect , it follows that @xmath460 is a perfect graph . for the particular lexicographic term order we have chosen , the edges in the graph @xmath460 are the pairs of the form @xmath463 such that @xmath464 or @xmath465 and @xmath305 . it is the simplicity of the graph @xmath460 which depends on @xmath466 for all @xmath73 and @xmath467 . to show that @xmath257 is delightful , we must show that for each clique of size @xmath4 in @xmath460 there is a polynomial in @xmath468 which has the clique as a leading term . let @xmath469 be such clique . we may suppose that @xmath470 . consider the @xmath471 matrix by construction , the polynomial @xmath473 belongs to @xmath468 since it is one of the minors appearing in proposition [ prop : scrolldet ] . furthermore , @xmath14 is not identically zero . the structure of @xmath249 implies that @xmath14 could be identically zero only if there were two identical columns in @xmath249 . this implies that there are indices @xmath400 and @xmath474 with @xmath475 such that @xmath476 and @xmath477 . but the conditions on the edges of @xmath460 make it impossible for there to be a lexicographically ordered clique in @xmath460 with these properties . furthermore , each indeterminate appearing in the matrix @xmath249 is a valid indeterminate in our polynomial ring . this follows because all the second indices , @xmath478 , lie between @xmath479 and @xmath480 . finally , the term order @xmath257 selects the main diagonal as the leading term . hence @xmath257 is delightful for @xmath408 . the ideal @xmath457 can be realized as the elimination ideal of @xmath482 where @xmath483 for all @xmath73 . the lexicographic ordering from theorem [ lemma : scrollsame ] realizes this elimination and thus the delightful property passes to this elimination ideal . we will prove the case @xmath486 . the case of @xmath487 follows by the elimination argument used in the proof of corollary [ cor : same1 ] , and the other two cases are proved in theorem [ lemma : scrollsame ] and corollary [ cor : same1 ] . now introduce the lexicographic term order @xmath257 , given by the rule @xmath488 @xmath489 we claim that this lexicographic term order is delightful for @xmath485 . to see this , let @xmath490 be an independent set in the triangulation corresponding to this term order , arranged in decreasing lexicographic order . since every full triangulation of @xmath491 is chordal and hence perfect , we must show that this independent set yields the initial term of a polynomial in @xmath492 . for a general pair of sequential elements in our independent set @xmath493 , this happens if and only if either : @xmath494 and @xmath495 ; or @xmath496 , and @xmath497 ; or @xmath498 , and @xmath499 . the only exceptions to these rules come at the ends : we can not have the pairs @xmath500 or @xmath501 in an independent set . now construct the matrix @xmath249 as in the proof of theorem [ lemma : scrollsame ] . our conditions on the sequence @xmath490 guarantee that all the entries in @xmath249 are valid indeterminates in our polynomial ring . furthermore , @xmath473 is not identically zero , and @xmath14 has leading term equal to @xmath502 . thus , the lexicographic term order @xmath257 is delightful . d. kozlov . chromatic numbers , morphism complexes , and stiefel - whitney characteristic classes , to appear in _ geometric combinatorics _ ( eds . e. miller , v. reiner , b. sturmfels ) , park city series in mathematics , ams , math.at/0505563 .

we study properties of _ quantum strategies _ , which are complete specifications of a given party s actions in any multiple - round interaction involving the exchange of quantum information with one or more other parties . in particular , we focus on a representation of quantum strategies that generalizes the choi - jamiokowski representation of quantum operations . this new representation associates with each strategy a positive semidefinite operator acting only on the tensor product of its input and output spaces . various facts about such representations are established , and two applications are discussed : the first is a new and conceptually simple proof of kitaev s lower bound for strong coin - flipping , and the second is a proof of the exact characterization of the class of problems having quantum refereed games .

quant-ph0611234

the theory of games provides a general structure within which both cooperation and competition among independent entities may be modeled , and provides powerful tools for analyzing these models . applications of this theory have fundamental importance in many areas of science . this paper considers games in which the players may exchange and process quantum information . we focus on competitive games , and within this context the types of games we consider are very general . for instance , they allow multiple rounds of interaction among the players involved , and place no restrictions on players strategies beyond those imposed by the theory of quantum information . while classical games can be viewed as a special case of quantum games , it is important to stress that there are fundamental differences between general quantum games and classical games . for example , the two most standard representations of classical games , namely the _ normal form _ and _ extensive form _ representations , are not directly applicable to general quantum games . this is due to the nature of quantum information , which admits a continuum of pure ( meaning extremal ) strategies , imposes bounds on players knowledge due to the uncertainty principle , and precludes the representation of general computational processes as trees . in light of such issues , it is necessary to give special consideration to the incorporation of quantum information into the theory of games . a general theory of quantum games has the potential to be useful in many situations that arise in quantum cryptography , computational complexity , communication complexity , and distributed computation . this potential is the primary motivation for the work presented in this paper , which we view as a first step in the development of a general theory of quantum games . the following facts are among those proved in this paper : * every multiple round quantum strategy can be faithfully represented by a single positive semidefinite operator acting only on the tensor product of the input and output spaces of the given player . this representation is a generalization of the choi - jamiokowski representation of super - operators . the set of all operators that arise in this way is precisely characterized by the set of positive semidefinite operators that satisfy a simple collection of linear constraints . * if a multiple round quantum strategy calls for one or more measurements then its representation consists of one operator for each of the possible measurement outcomes . the probability of any given pair of measurement outcomes for two interacting strategies is given by the inner product of their associated operators . * the maximum probability with which a given strategy can be forced to output a particular result is the _ minimum _ value of @xmath1 for which the positive semidefinite operator corresponding to the given measurement result is bounded above ( with respect to the lwner partial order ) by the representation of a valid strategy multiplied by @xmath1 . we give the following applications of these facts : * a new and conceptually simple proof of kitaev s bound for strong coin - flipping , which states that every quantum strong coin - flipping protocol allows a bias of at least @xmath2 . * the exact characterization @xmath0 of the class of problems having quantum refereed games ( i.e. , quantum interactive proof systems with two competing provers ) . this establishes that quantum and classical refereed games are equivalent in terms of expressive power : @xmath3 . it is appropriate for us to comment on the relationship between the present paper and a fairly large collection of papers written on a topic that has been called _ quantum game theory_. meyer s _ pq penny flip _ game @xcite is a well - known example of a game in the category these papers consider . the work of eisert , _ et al . _ @xcite is also commonly cited in this area . some controversy exists over the interpretations drawn in some quantum these papers see , for instance , refs . @xcite . a key difference between our work and previous work on quantum game theory is that our focus is on multiple - round interactions . understanding the actions available to players that have quantum memory is therefore critical to our work , and to our knowledge has not been previously considered in the context of quantum game theory . a second major difference is that , in most of the previous quantum game theory papers we are aware of , the focus is on rather specific examples of classical games and on identifying differences that arise when so - called quantum variants of these games are considered . as a possible consequence , it may arguably be said that none of the results proved in these papers has had sufficient generality to be applicable to any other studies in quantum information . in contrast , our interest is not on specific examples of games , but rather on the development of a general theory that holds for all games . it remains to be seen to what extent our work will be applied , but the applications that we provide suggest that it may have interesting uses in other areas of quantum information and computation . a different context in which games arise in quantum information theory is that of _ nonlocal games _ @xcite , which include _ pseudo - telepathy games _ @xcite as a special case . these are cooperative games of incomplete information that model situations that arise in the study of multiple - prover interactive proof systems , and provide a framework for studying bell inequalities and the notion of nonlocality that arises in quantum physics . while such games can be described within the general setting we consider , we have not yet found an application of the methods of the present paper to this type of game . possibly there is some potential for further development of our work to shed light on some of the difficult questions in this area . this section gives a brief overview of various quantum information - theoretic notions that will be needed for the remainder of the paper . we assume the reader has familiarity with quantum information theory , and intend only that this overview will serve to establish our notation and highlight the main concepts that we will need . readers not familiar with quantum information are referred to the books of nielsen and chuang @xcite and kitaev , shen and vyalyi @xcite . when we speak of the vector space associated with a given quantum system , we are referring to some complex euclidean space ( by which we mean a finite - dimensional inner product space over the complex numbers ) . such spaces will be denoted by capital script letters such as @xmath4 , @xmath5 , and @xmath6 . we always assume that an orthonormal _ standard basis _ of any such space has been chosen , and with respect to this basis elements of these spaces are associated with column vectors , and linear mappings from one space to another are associated with matrices in the usual way . we will often be concerned with finite sequences @xmath7 of complex euclidean spaces . we then define @xmath8 for nonnegative integers @xmath9 , and define @xmath10 for @xmath11 . it is convenient to define various sets of linear mappings between given complex euclidean spaces @xmath4 and @xmath5 as follows . let @xmath12 denote the space of all linear mappings ( or _ operators _ ) from @xmath4 to @xmath5 , and write @xmath13 as shorthand for @xmath14 . we write @xmath15 to denote the set of hermitian operators acting on @xmath4 , @xmath16 to denote the set of all positive semidefinite operators acting on @xmath4 , and @xmath17 to denote the set of all density operators on @xmath4 ( meaning positive semidefinite operators having trace equal to 1 . ) an operator @xmath18 is a _ linear isometry _ if @xmath19 . the existence of a linear isometry in @xmath12 of course requires that @xmath20 , and if @xmath21 then any linear isometry @xmath18 is unitary . we let @xmath22 denote the set of all linear isometries from @xmath4 to @xmath5 . the operator @xmath23 denotes the identity operator on @xmath4 . transposition of operators is always taken with respect to standard bases . the hilbert - schmidt inner product on @xmath13 is defined by @xmath24 for all @xmath25 . for given operators @xmath26 , the notation @xmath27 means that @xmath28 . this relation is sometimes called the _ lwner partial order _ on @xmath15 . when we refer to _ measurements _ , we mean povm - type measurements . formally , a measurement on a complex euclidean space @xmath4 is described by a collection of positive semidefinite operators @xmath29 satisfying the constraint @xmath30 here @xmath31 is a finite , non - empty set of _ measurement outcomes_. if a state represented by the density operator @xmath32 is measured with respect to such a measurement , each outcome @xmath33 results with probability @xmath34 . a _ super - operator _ is a linear mapping of the form @xmath35 where @xmath4 and @xmath5 are complex euclidean spaces . a super - operator of this form is said to be _ positive _ if @xmath36 for every choice of @xmath37 , and is _ completely positive _ if @xmath38 is positive for every choice of a complex euclidean space @xmath6 . the super - operator @xmath39 is said to be _ admissible _ if it is completely positive and preserves trace : @xmath40 for all @xmath41 . admissible super - operators represent discrete - time changes in quantum systems that can , in an idealized sense , be physically realized . the choi - jamiokowski representation @xcite of super - operators is as follows . suppose that @xmath42 be a given super - operator and let @xmath43 be the standard basis of @xmath4 . then the _ choi - jamiokowski representation _ of @xmath39 is the operator @xmath44 it holds that @xmath39 is completely positive if and only if @xmath45 is positive semidefinite , and that @xmath39 is trace - preserving if and only if @xmath46 . for two complex euclidean spaces @xmath4 and @xmath5 , we define a linear mapping @xmath47 by extending by linearity the action @xmath48 on standard basis states . we make extensive use of this mapping in some of our proofs , as it is very convenient in a variety of situations . let us now state some identities involving the @xmath49 mapping , each of which can be verified by a straightforward calculation . [ prop : identities ] the following hold : 5 mm for any choice of @xmath50 , @xmath51 , and @xmath52 for which the product @xmath53 makes sense we have @xmath54 [ item : trace ] for any choice of @xmath55 we have @xmath56 for any choice of @xmath57 we have @xmath58 [ item : choi ] let @xmath59 and suppose @xmath60 is given by @xmath61 for all @xmath41 . then @xmath62 for any non - empty set @xmath63 of hermitian operators , the _ polar _ of @xmath64 is defined as @xmath65 and the _ support _ and _ gauge _ functions of @xmath64 are defined as follows : @xmath66 these functions are partial functions in general , but it is typical to view them as total functions from @xmath15 to @xmath67 in the natural way . for any set @xmath64 of positive semidefinite operators , we denote @xmath68 [ prop : polar ] let @xmath4 be a complex euclidean space and let @xmath64 and @xmath69 be non - empty subsets of @xmath15 . then the following facts hold : 1 . [ item : polar - contain ] if @xmath70 then @xmath71 . [ item : polar - pos ] if @xmath72 for each @xmath37 then @xmath73 . [ item : polar - polar ] if @xmath64 is closed , convex , and contains the origin , then the same is true of @xmath74 . in this case we have @xmath75 , @xmath76 the first two items in the above proposition are elementary , and a proof of the third may be found in rockafellar @xcite . in this section we define the notions of a quantum _ strategy _ and the _ choi - jamiokowski representation _ of quantum strategy . the remainder of the paper is concerned with the study of these objects and their interactions . we begin with our definition for quantum strategies , which we will simply call _ strategies _ given that the focus of the paper is on the quantum setting . [ def : strategy ] let @xmath77 and let @xmath78 and @xmath79 be complex euclidean spaces . @xmath80-turn non - measuring strategy _ having input spaces @xmath78 and output spaces @xmath79 consists of : 5 mm complex euclidean spaces @xmath81 , which will be called _ memory spaces _ , and an @xmath80-tuple of admissible mappings @xmath82 having the form @xmath83 an _ @xmath80-turn measuring strategy _ consists of items 1 and 2 above , as well as : 5 mm a measurement @xmath84 on the last memory space @xmath85 . we will use the term _ @xmath80-turn strategy _ to refer to either a measuring or non - measuring @xmath80-turn strategy . figure [ fig : strategy ] illustrates an @xmath80-turn non - measuring strategy . ( 23734,2939)(-600,1700 ) ( 18612,4212)(19812,4212)(19812,3012)(18612,3012)(18612,4212 ) ( 1812,4212)(3012,4212)(3012,3012)(1812,3012)(1812,4212 ) ( 5412,4212)(6612,4212)(6612,3012)(5412,3012)(5412,4212 ) ( 9012,4212)(10212,4212)(10212,3012)(9012,3012)(9012,4212 ) ( 2412,3612)(0,0)@xmath86 ( 6012,3612)(0,0)@xmath87 ( 9612,3612)(0,0)@xmath88 ( 19212,3612)(0,0)@xmath89 ( 16512,3612)(18312,3612 ) ( 18072.000,3552.000)(18312.000,3612.000)(18072.000,3672.000 ) ( 18072.000,3552.000 ) ( 17712,1512)(18912,2712 ) ( 18784.721,2499.868)(18912.000,2712.000)(18699.868,2584.721 ) ( 18784.721,2499.868 ) ( 19512,2712)(20712,1512 ) ( 20499.868,1639.279)(20712.000,1512.000)(20584.721,1724.132 ) ( 20499.868,1639.279 ) ( 912,1512)(2112,2712 ) ( 1984.721,2499.868)(2112.000,2712.000)(1899.868,2584.721 ) ( 1984.721,2499.868 ) ( 2712,2712)(3912,1512 ) ( 3699.868,1639.279)(3912.000,1512.000)(3784.721,1724.132 ) ( 3699.868,1639.279 ) ( 4512,1512)(5712,2712 ) ( 5584.721,2499.868)(5712.000,2712.000)(5499.868,2584.721 ) ( 5584.721,2499.868 ) ( 6312,2712)(7512,1512 ) ( 7299.868,1639.279)(7512.000,1512.000)(7384.721,1724.132 ) ( 7299.868,1639.279 ) ( 8112,1512)(9312,2712 ) ( 9184.721,2499.868)(9312.000,2712.000)(9099.868,2584.721 ) ( 9184.721,2499.868 ) ( 9912,2712)(11112,1512 ) ( 10899.868,1639.279)(11112.000,1512.000)(10984.721,1724.132 ) ( 10899.868,1639.279 ) ( 3312,3612)(5112,3612 ) ( 4872.000,3552.000)(5112.000,3612.000)(4872.000,3672.000 ) ( 4872.000,3552.000 ) ( 6912,3612)(8712,3612 ) ( 8472.000,3552.000)(8712.000,3612.000)(8472.000,3672.000 ) ( 8472.000,3552.000 ) ( 10512,3612)(12312,3612 ) ( 12072.000,3552.000)(12312.000,3612.000)(12072.000,3672.000 ) ( 12072.000,3552.000 ) ( 20112,3612)(21912,3612 ) ( 21672.000,3552.000)(21912.000,3612.000)(21672.000,3672.000 ) ( 21672.000,3552.000 ) ( 15012,3612 ) ( 15012,3612 ) ( 13212,3612 ) ( 13212,3612 ) ( 14112,3612 ) ( 14112,3612 ) ( 15912,3612 ) ( 15912,3612 ) ( 1112,2112)(0,0)[b]@xmath90 ( 4712,2112)(0,0)[b]@xmath91 ( 8312,2112)(0,0)[b]@xmath92 ( 17912,2112)(0,0)[b]@xmath93 ( 3662,2112)(0,0)[b]@xmath94 ( 7262,2112)(0,0)[b]@xmath95 ( 10862,2112)(0,0)[b]@xmath96 ( 20612,2112)(0,0)[b]@xmath97 ( 4212,4012)(0,0)@xmath98 ( 7862,4012)(0,0)@xmath99 ( 11462,4012)(0,0)@xmath100 ( 17462,4012)(0,0)@xmath101 ( 21012,4012)(0,0)@xmath85 although there is no restriction on the dimension of the memory spaces in a quantum strategy , it is established in the proof of theorem [ theorem : characterization ] that every measuring strategy is equivalent to one in which @xmath102 for each @xmath103 . we also note that our definition of strategies allows the possibility that any of the input or output spaces is equal to @xmath104 , which corresponds to an empty message . one can therefore view simple actions such as the preparation of a quantum state or performing a measurement without producing a quantum output as special cases of strategies . when we say that an @xmath80-turn strategy is described by _ linear isometries _ @xmath105 , it is meant that the admissible super - operators @xmath106 defining the strategy are given by @xmath107 for @xmath108 . notice that , when it is convenient , there is no loss of generality in restricting ones attention to strategies described by linear isometries in this way . this is because every admissible super - operator can be expressed as a mapping @xmath109 for some linear isometry @xmath50 , followed by the partial trace over some `` garbage '' space that represents a tensor factor of the space to which @xmath50 maps . by including the necessary `` garbage '' spaces as tensor factors of the memory spaces , and therefore not tracing them out , there can be no change in the action of the strategy on the input and output spaces . along similar lines , there is no loss of generality in assuming that a given measuring strategy s measurement is projective . a given @xmath80-turn strategy expects to interact with something that provides the inputs corresponding to @xmath78 and accepts the strategy s outputs corresponding to @xmath79 . let us define an @xmath80-turn _ co - strategy _ to be the sort of object that a strategy interfaces with in the most natural way . let @xmath77 and let @xmath78 and @xmath79 be complex euclidean spaces . the spaces @xmath78 are viewed as the input spaces of some @xmath80-turn strategy while @xmath79 are to be viewed as its output spaces . @xmath80-turn non - measuring co - strategy _ to these spaces consists of : 5 mm complex euclidean _ memory _ spaces @xmath110 , a density operator @xmath111 , and an @xmath80-tuple of admissible mappings @xmath112 having the form @xmath113 an _ @xmath80-turn measuring co - strategy _ consists of items 1 , 2 and 3 above , as well as : 5 mm a measurement @xmath114 on the last memory space @xmath115 . as for strategies , we use the term _ @xmath80-turn co - strategy _ to refer to either a measuring or non - measuring @xmath80-turn co - strategy . figure [ fig : interaction ] represents the interaction between an @xmath80-turn strategy and co - strategy . ( 23734,4539)(-600,200 ) ( 20412,1212)(21612,1212)(21612,12)(20412,12)(20412,1212 ) ( 18612,4212)(19812,4212)(19812,3012)(18612,3012)(18612,4212 ) ( 1812,4212)(3012,4212)(3012,3012)(1812,3012)(1812,4212 ) ( 3612,1212)(4812,1212)(4812,12)(3612,12)(3612,1212 ) ( 5412,4212)(6612,4212)(6612,3012)(5412,3012)(5412,4212 ) ( 7212,1212)(8412,1212)(8412,12)(7212,12)(7212,1212 ) ( 9012,4212)(10212,4212)(10212,3012)(9012,3012)(9012,4212 ) ( 10812,1212)(12012,1212)(12012,12)(10812,12)(10812,1212 ) ( 12,1212)(1212,1212)(1212,12)(12,12)(12,1212 ) ( 612,612)(0,0)@xmath116 ( 4212,612)(0,0)@xmath117 ( 7812,612)(0,0)@xmath118 ( 11412,612)(0,0)@xmath119 ( 21012,612)(0,0)@xmath120 ( 2412,3612)(0,0)@xmath86 ( 6012,3612)(0,0)@xmath87 ( 9612,3612)(0,0)@xmath88 ( 19212,3612)(0,0)@xmath89 ( 16512,3612)(18312,3612 ) ( 18072.000,3552.000)(18312.000,3612.000)(18072.000,3672.000 ) ( 18072.000,3552.000 ) ( 18312,612)(20112,612 ) ( 19872.000,552.000)(20112.000,612.000)(19872.000,672.000 ) ( 19872.000,552.000 ) ( 17712,1512)(18912,2712 ) ( 18784.721,2499.868)(18912.000,2712.000)(18699.868,2584.721 ) ( 18784.721,2499.868 ) ( 19512,2712)(20712,1512 ) ( 20499.868,1639.279)(20712.000,1512.000)(20584.721,1724.132 ) ( 20499.868,1639.279 ) ( 912,1512)(2112,2712 ) ( 1984.721,2499.868)(2112.000,2712.000)(1899.868,2584.721 ) ( 1984.721,2499.868 ) ( 2712,2712)(3912,1512 ) ( 3699.868,1639.279)(3912.000,1512.000)(3784.721,1724.132 ) ( 3699.868,1639.279 ) ( 4512,1512)(5712,2712 ) ( 5584.721,2499.868)(5712.000,2712.000)(5499.868,2584.721 ) ( 5584.721,2499.868 ) ( 6312,2712)(7512,1512 ) ( 7299.868,1639.279)(7512.000,1512.000)(7384.721,1724.132 ) ( 7299.868,1639.279 ) ( 8112,1512)(9312,2712 ) ( 9184.721,2499.868)(9312.000,2712.000)(9099.868,2584.721 ) ( 9184.721,2499.868 ) ( 9912,2712)(11112,1512 ) ( 10899.868,1639.279)(11112.000,1512.000)(10984.721,1724.132 ) ( 10899.868,1639.279 ) ( 3312,3612)(5112,3612 ) ( 4872.000,3552.000)(5112.000,3612.000)(4872.000,3672.000 ) ( 4872.000,3552.000 ) ( 6912,3612)(8712,3612 ) ( 8472.000,3552.000)(8712.000,3612.000)(8472.000,3672.000 ) ( 8472.000,3552.000 ) ( 1512,612)(3312,612 ) ( 3072.000,552.000)(3312.000,612.000)(3072.000,672.000 ) ( 3072.000,552.000 ) ( 5112,612)(6912,612 ) ( 6672.000,552.000)(6912.000,612.000)(6672.000,672.000 ) ( 6672.000,552.000 ) ( 8712,612)(10512,612 ) ( 10272.000,552.000)(10512.000,612.000)(10272.000,672.000 ) ( 10272.000,552.000 ) ( 10512,3612)(12312,3612 ) ( 12072.000,3552.000)(12312.000,3612.000)(12072.000,3672.000 ) ( 12072.000,3552.000 ) ( 12312,612)(14112,612 ) ( 13872.000,552.000)(14112.000,612.000)(13872.000,672.000 ) ( 13872.000,552.000 ) ( 11712,1512)(12912,2712 ) ( 12784.721,2499.868)(12912.000,2712.000)(12699.868,2584.721 ) ( 12784.721,2499.868 ) ( 20112,3612)(21912,3612 ) ( 21672.000,3552.000)(21912.000,3612.000)(21672.000,3672.000 ) ( 21672.000,3552.000 ) ( 21912,612)(23712,612 ) ( 23472.000,552.000)(23712.000,612.000)(23472.000,672.000 ) ( 23472.000,552.000 ) ( 16812,612 ) ( 16812,612 ) ( 15012,3612 ) ( 15012,3612 ) ( 13212,3612 ) ( 13212,3612 ) ( 14112,3612 ) ( 14112,3612 ) ( 15912,3612 ) ( 15912,3612 ) ( 15012,612 ) ( 15012,612 ) ( 15912,612 ) ( 15912,612 ) ( 17712,612 ) ( 17712,612 ) ( 1112,2112)(0,0)[b]@xmath90 ( 4712,2112)(0,0)[b]@xmath91 ( 8312,2112)(0,0)[b]@xmath92 ( 11962,2112)(0,0)[b]@xmath121 ( 17912,2112)(0,0)[b]@xmath93 ( 3662,2112)(0,0)[b]@xmath94 ( 7262,2112)(0,0)[b]@xmath95 ( 10862,2112)(0,0)[b]@xmath96 ( 20612,2112)(0,0)[b]@xmath97 ( 4212,4012)(0,0)@xmath98 ( 7862,4012)(0,0)@xmath99 ( 11462,4012)(0,0)@xmath100 ( 17462,4012)(0,0)@xmath101 ( 2412,212)(0,0)@xmath122 ( 6012,212)(0,0)@xmath123 ( 9612,212)(0,0)@xmath124 ( 13212,212)(0,0)@xmath125 ( 19212,212)(0,0)@xmath126 ( 21012,4012)(0,0)@xmath85 ( 22662,212)(0,0)@xmath115 we have arbitrarily defined strategies and co - strategies in such a way that the co - strategy sends the first message . this accounts for the inevitable asymmetry in the definitions . while it is possible to view any @xmath80-turn co - strategy as being an @xmath127-turn strategy having input spaces @xmath128 and output spaces @xmath129 , it will be convenient for our purposes to view strategies and co - strategies as being distinct types of objects . similar to strategies , there will be no loss of generality in assuming that the initial state @xmath116 of a co - strategy is pure , that each of the admissible super - operators @xmath130 takes the form @xmath131 for some linear isometry @xmath132 , and , in the case of measuring co - strategies , that the measurement @xmath133 is a projective measurement . an @xmath80-turn strategy and co - strategy are _ compatible _ if they agree on the spaces @xmath78 and @xmath79 . by the _ output _ of a compatible strategy and co - strategy , assuming at least one of them is measuring , we mean the result of the measurement or measurements performed after the interaction between the strategies takes place . in particular , if both the strategy and co - strategy make measurements , then each output @xmath134 results with probability @xmath135 the definitions of strategies and co - strategies given above are natural from an operational point of view , in the sense that they clearly describe the actions of the players that they model . in some situations , however , representing a strategy ( or co - strategy ) in terms of a sequence of admissible super - operators is inconvenient . we now describe a different way to represent strategies that is based on the choi - jamiokowski representation of super - operators . let us first extend this representation to @xmath80-turn non - measuring strategies . to do this , we associate with the strategy described by @xmath82 a single admissible super - operator @xmath136 this is the super - operator that takes a given input @xmath137 and feeds the portions of this state corresponding to the input spaces @xmath78 into the network pictured in figure [ fig : strategy ] , one piece at a time . the memory space @xmath85 is then traced out , leaving some element @xmath138 . such a map is depicted in figure [ fig : choijam ] for the case @xmath139 . the choi - jamiokowski representation of the strategy described by @xmath82 is then simply defined as the choi - jamiokowski representation @xmath140 of the super - operator @xmath141 we have just defined . an alternate expression for the choi - jamiokowski representation of a strategy exists in the case that it is described by linear isometries @xmath142 . specifically , its representation is given by @xmath143 where @xmath144 is defined by the product @xmath145 ( 6999,2697)(0,-10 ) ( 1587,1137)(2262,1137)(2262,12)(1587,12)(1587,1137 ) ( 3162,1812)(3837,1812)(3837,687)(3162,687)(3162,1812 ) ( 4737,2487)(5412,2487)(5412,1362)(4737,1362)(4737,2487 ) ( 1925,575)(0,0)@xmath86 ( 3500,1250)(0,0)@xmath87 ( 5075,1925)(0,0)@xmath88 ( 6350,2262)(0,0)[l]traced out ( 687,1587)(0,0)[r ] @xmath146{0mm}{0mm}\right.$ ] ( 6312,912)(0,0)[l ] @xmath147{0mm}{0mm}\right\}\xi(\xi)$ ] ( 2362,912)(3062,912 ) ( 2942.000,882.000)(3062.000,912.000 ) ( 2942.000,942.000)(2942.000,882.000 ) ( 787,912)(1487,912 ) ( 1367.000,882.000)(1487.000,912.000 ) ( 1367.000,942.000)(1367.000,882.000 ) ( 3937,1587)(4637,1587 ) ( 4517.000,1557.000)(4637.000,1587.000 ) ( 4517.000,1617.000)(4517.000,1557.000 ) ( 5512,2262)(6212,2262 ) ( 6092.000,2232.000)(6212.000,2262.000 ) ( 6092.000,2292.000)(6092.000,2232.000 ) ( 5512,1587)(6212,1587 ) ( 6092.000,1557.000)(6212.000,1587.000 ) ( 6092.000,1617.000)(6092.000,1557.000 ) ( 2362,237)(6212,237 ) ( 6092.000,207.000)(6212.000,237.000 ) ( 6092.000,267.000)(6092.000,207.000 ) ( 787,1587)(3062,1587 ) ( 2942.000,1557.000)(3062.000,1587.000 ) ( 2942.000,1617.000)(2942.000,1557.000 ) ( 3937,912)(6212,912 ) ( 6092.000,882.000)(6212.000,912.000 ) ( 6092.000,942.000)(6092.000,882.000 ) ( 787,2262)(4637,2262 ) ( 4517.000,2232.000)(4637.000,2262.000 ) ( 4517.000,2292.000)(4517.000,2232.000 ) ( 5862,2487)(0,0)@xmath100 ( 5862,1812)(0,0)@xmath96 ( 5862,1137)(0,0)@xmath95 ( 5862,462)(0,0)@xmath94 ( 1137,2487)(0,0)@xmath92 ( 1137,1812)(0,0)@xmath91 ( 1137,1137)(0,0)@xmath90 ( 4287,1812)(0,0)@xmath99 ( 2712,1137)(0,0)@xmath98 next we consider measuring strategies . assume that an @xmath80-turn measuring strategy is given , where the measurement is described by @xmath148 for some finite , non - empty set @xmath31 of measurement outcomes . in this case we first associate with the strategy a collection of super - operators @xmath149 , each having the form @xmath150 the definition of each super - operator @xmath151 is precisely as in the non - measuring case , except that the partial trace over @xmath85 is replaced by the mapping @xmath152 notice that @xmath153 where @xmath141 is the mapping defined as in the non - measuring case . each super - operators @xmath151 is completely positive but generally is not trace - preserving . the choi - jamiokowski representation of the measuring strategy described by @xmath82 and @xmath154 is defined as @xmath155 . in the situation where a measuring strategy is described by linear isometries @xmath105 and a measurement @xmath154 , its choi - jamiokowski representation is given by @xmath156 for @xmath157 where @xmath158 for @xmath50 as defined above . it is not difficult to prove that for given input spaces @xmath78 and output spaces @xmath79 , a collection @xmath156 of operators is the choi - jamiokowski representation of some @xmath80-turn measuring strategy if and only if @xmath159 is the representation of an @xmath80-turn non - measuring strategy over the same spaces . finally , we define the choi - jamiokowski representation of measuring and non - measuring co - strategies in precisely the same way as for strategies , except that for technical reasons the resulting operators are transposed with respect to the standard basis . ( this essentially allows us to eliminate one transposition from almost every subsequent equation in this paper involving representations of co - strategies . ) specifically , a given @xmath80-turn co - strategy is viewed as an @xmath127-turn strategy by including an empty first and last message as discussed previously . this strategy s choi - jamiokowski representation is defined as above . the operators comprising this strategy representation are then transposed with respect to the standard basis to obtain the choi - jamiokowski representation of the co - strategy . as we work almost exclusively with the choi - jamiokowski representation of strategies and co - strategies hereafter , we will typically use the term _ representation _ rather than _ choi - jamiokowski representation _ for brevity . the applications of choi - jamiokowski representations of strategies given in this paper rely upon three key properties of these representations , stated below as theorems [ theorem : interaction - output - probability ] , [ theorem : characterization ] , and [ theorem : gauge ] . this section is devoted to establishing these properties . the first property provides a simple formula for the probability of a given output of an interaction between a strategy and a co - strategy . [ theorem : interaction - output - probability ] let @xmath156 be the representation of a strategy and let @xmath160 be the representation of a compatible co - strategy . for each pair @xmath161 of measurement outcomes , we have that the output of an interaction between the given strategy and co - strategy is @xmath162 with probability @xmath163 . let us fix a strategy and co - strategy whose representations are @xmath156 and @xmath164 , respectively . without loss of generality , the strategy is described by linear isometries @xmath105 and a projective measurement @xmath165 , while the co - strategy is described by a pure initial state @xmath166 , linear isometries @xmath167 and a projective measurement @xmath168 . for each output pair @xmath134 , define @xmath169 as follows : @xmath170 the probability of the outcome @xmath162 is @xmath171 for each pair @xmath162 . now , making use of the @xmath49 mapping , we may express @xmath172 in a different way : @xmath173 where @xmath174 the probability of outcome @xmath162 is therefore @xmath175 \op{vec}(i_{\y_{1\ldots n}\otimes\x_{1\ldots n}})\\ & = \tr\left [ \left(\tr_{\z_n } x_a x_a^{\ast}\right ) \left(\tr_{\w_n } y_b y_b^{\ast}\right)^{\t } \right ] \\ & = \ip{q_a}{r_b } \end{aligned}\ ] ] as required . let us denote by @xmath176 the set of all representations of @xmath80-turn strategies having input spaces @xmath78 and output spaces @xmath79 . we may abbreviate this set as @xmath177 or @xmath178 whenever the spaces or number of turns is clear from the context . similarly , we let @xmath179 denote the set of all representations of co - strategies for the same spaces . it will be convenient to define @xmath180 . the second property of strategy representations that we prove provides a characterization of @xmath181 in terms of linear constraints on @xmath182 . [ theorem : characterization ] let @xmath77 , let @xmath78 and @xmath79 be complex euclidean spaces , and let @xmath183 . then @xmath184 if and only if @xmath185 for @xmath186 . first assume that @xmath187 , which implies that there exist memory spaces @xmath81 and admissible super - operators @xmath188 that comprise a strategy whose representation is @xmath189 . let @xmath190 be the super - operator associated with this strategy as described in section [ sec : strategies ] , and let @xmath191 be the super - operator associated with the @xmath192-turn strategy obtained by terminating the strategy described by @xmath82 after @xmath193 turns . we have @xmath194 and so @xmath195 for @xmath196 as required . now assume that @xmath183 satisfies @xmath197 for some choice of @xmath186 . our goal is to prove that @xmath198 . this will be proved by induction on @xmath80 . in fact , it will be easier to prove a somewhat stronger statement , which is that the assumptions imply that there exists a strategy whose representation is @xmath189 that ( i ) is described by linear isometries @xmath105 , and ( ii ) satisfies @xmath199 . if @xmath200 , there is nothing new to prove : it is well - known that if @xmath201 satisfies @xmath202 , then @xmath203 for some admissible super - operator @xmath204 . any such super - operator can be expressed as @xmath205 for @xmath206 and some choice of a linear isometry @xmath207 . the 1-turn strategy we require is therefore the strategy described by @xmath208 . now assume that @xmath209 . by the induction hypothesis , there exist spaces @xmath210 and linear isometries @xmath211 with @xmath212 such that @xmath213 for @xmath214 defined as @xmath215 as required , we let @xmath85 be a complex euclidean space with dimension equal to the rank of @xmath189 , and let @xmath216 be any operator satisfying @xmath217 such a choice of @xmath51 must exist given that the dimension of @xmath85 is large enough to admit a purification of @xmath189 . note that @xmath218 next , let @xmath219 be a complex euclidean space with dimension equal to that of @xmath93 , and let @xmath220 be an arbitrary unitary operator . we have @xmath221 and therefore @xmath222 at this point we have identified two purifications of @xmath223 . we will use the isometric equivalence of purifications to define an isometry @xmath224 that will complete the proof . specifically , because @xmath225 has the minimal dimension required to admit a purification of @xmath223 , it follows that there must exist a linear isometry @xmath226 such that @xmath227 this equation may equivalently be written @xmath228 we now define @xmath229 . this is a linear isometry from @xmath230 to @xmath231 that satisfies @xmath232 this implies that the strategy described by @xmath105 has representation @xmath233 and therefore completes the proof . theorem [ theorem : characterization ] is equivalent to the following statement : an operator @xmath183 is the representation of some @xmath80-turn strategy with input spaces @xmath78 and output spaces @xmath79 if and only if there exist operators @xmath234 , where @xmath235 , such that the following linear constraints are satisfied : @xmath236 each operator @xmath237 in of course uniquely determined by the representation @xmath189 , and represents the strategy obtained by terminating any strategy represented by @xmath189 after @xmath238 turns . theorem [ theorem : characterization ] also gives a characterization of @xmath80-turn co - strategies , as stated in the following corollary . [ cor : characterization ] let @xmath77 , let @xmath78 and @xmath79 be complex euclidean spaces , and let @xmath183 . then @xmath239 if and only if @xmath240 for @xmath241 satisfying @xmath242 the fact that @xmath181 and @xmath243 are bounded and characterized by the positive semidefinite constraint together with finite collections of linear constraints yields the following corollary . let @xmath77 , let @xmath78 and @xmath79 be complex euclidean spaces . then the sets @xmath181 and @xmath243 are compact and convex . just as @xmath181 consists of all representations of non - measuring strategies , the set @xmath244 consists of all elements of representations of measuring strategies . in other words , for any @xmath80-turn measuring strategy representation @xmath245 , it holds that @xmath246 for each @xmath33 . moreover , for each operator @xmath189 there exists an @xmath80-turn measuring strategy @xmath156 of which @xmath189 is an element if and only if @xmath247 . the set @xmath248 has similar analogous properties . the final property of strategy representations that we will prove concerns the maximum probability with which some interacting co - strategy can force a given measuring strategy to output a given outcome . [ theorem : gauge ] let @xmath77 , let @xmath78 and @xmath79 be complex euclidean spaces , and let @xmath249 represent an @xmath80-turn measuring strategy with input spaces @xmath78 and output spaces @xmath79 . then for each @xmath33 , the maximum probability with which this strategy can be forced to output @xmath250 , maximized over all choices of compatible co - strategies , is given by @xmath251\,:\ , q_a \leq p r\;\ ; \text{for some}\;r\in{\mathcal{s}}_n(\x_{1\ldots n},\y_{1\ldots n})\}.\ ] ] an analogous result holds when @xmath156 is a measuring co - strategy . the remainder of the present section is devoted to a proof of this theorem . [ lemma : polar - technical ] let @xmath219 and @xmath252 be complex euclidean spaces and let @xmath253 be any closed , convex set that contains the origin . then for @xmath254 we have @xmath255 the assumption @xmath256 implies that @xmath257 for every @xmath258 , and therefore @xmath259 . consider any choice of @xmath260 , and note that @xmath261 for every choice of @xmath262 . if it is the case that @xmath263 then we have @xmath264 for all @xmath265 , and therefore @xmath266 . on the other hand , if @xmath266 then @xmath264 for all @xmath265 . it follows from the fact that @xmath189 is positive semidefinite that @xmath267 for all @xmath268 , and therefore @xmath269 . [ lemma : switch - polar ] let @xmath219 be a complex euclidean space , let @xmath270 be non - empty closed and convex sets , and suppose @xmath271 then @xmath272 let @xmath273 as @xmath274 for every @xmath275 , it follows that @xmath276 . clearly @xmath277 , and therefore @xmath278 . thus , @xmath279 on the other hand , we have that every @xmath280 is contained in @xmath281 , implying that @xmath282 for all @xmath283 . as @xmath284 , this implies that @xmath267 for @xmath285 . consequently , @xmath286 . thus @xmath287 , and so @xmath288 as required . [ lemma : polar ] let @xmath77 and let @xmath78 and @xmath79 be complex euclidean spaces . then for all @xmath289 we have 5 mm @xmath290 if and only if @xmath291 for some choice of @xmath292 . @xmath293 if and only if @xmath291 for some choice of @xmath187 . the proof is by induction on @xmath80 . as @xmath294 $ ] and @xmath295 $ ] , the lemma holds for the case @xmath296 . now suppose that @xmath77 . the two items in the statement of the lemma are equivalent by lemma [ lemma : switch - polar ] , so it will suffice to prove the first . define @xmath297 as @xmath298 by lemma [ lemma : polar - technical ] we have @xmath299 also define @xmath300 as @xmath301 again applying lemma [ lemma : polar - technical ] , we obtain @xmath302 now , by theorem [ theorem : characterization ] we have @xmath303 , and so @xmath304 . by the induction hypothesis we have @xmath305 and therefore @xmath306 by corollary [ cor : characterization ] we have that @xmath307 which completes the proof . let @xmath308 $ ] denote the maximum probability with which @xmath156 can be forced to output @xmath250 in an interaction with some compatible co - strategy . it follows from theorem [ theorem : interaction - output - probability ] that @xmath309 . using lemma [ lemma : polar ] , along with the fact that @xmath310 is positive semidefinite , we have @xmath311 which completes the proof . quantum coin - flipping protocols aim to solve the following problem : two parties , alice and bob , at separate locations , want to flip a coin but do not trust one another . a quantum coin - flipping protocol with bias @xmath312 is an interaction between two ( honest ) strategies @xmath50 and @xmath51 , both having output sets @xmath313 , that satisfies two properties : 1 . the interaction between the honest parties @xmath50 and @xmath51 produces the same outcome @xmath314 for both players , with probability @xmath315 for each outcome . ( neither player outputs `` abort '' when both are honest . ) 2 . if one of the parties does not follow the protocol but the other does , neither of the outcomes @xmath314 is output by the honest player with probability greater than @xmath316 . protocols that satisfy these conditions are generally referred to as _ strong _ coin - flipping protocols , because they require that a cheater can not bias an honest player s outcome toward either result 0 or 1 . ( in contrast , _ weak _ protocols assume that alice desires outcome 0 and bob desires outcome 1 , and only require that cheaters can not bias the outcome toward their desired outcome . ) kitaev @xcite proved that no strong quantum coin - flipping protocol can have bias smaller than @xmath2 , meaning that one cheating party can always force a given outcome on an honest party with probability at least @xmath317 . kitaev did not publish this proof , but it appears in refs . @xcite . here we give a different proof based on the results of the previous section . suppose @xmath318 is the representation of honest - alice s strategy and @xmath319 is the representation of honest - bob s co - strategy in some coin - flipping protocol . these strategies may involve any fixed number of rounds of interaction . the first condition above implies @xmath320 now , fix @xmath314 , and let @xmath1 be the maximum probability that a cheating bob can force honest - alice to output @xmath321 . obviously we have @xmath322 . theorem [ theorem : gauge ] implies that there must exist a strategy @xmath189 for alice such that @xmath323 . if a cheating alice plays this strategy @xmath189 , then honest - bob outputs @xmath321 with probability @xmath324 given that @xmath325 for all @xmath326 , we have that either honest - alice or honest - bob can be convinced to output @xmath321 with probability at least @xmath317 . this proof makes clear the limitations of strong coin - flipping protocols : the inability of bob to force alice to output @xmath321 directly implies that alice can herself bias the outcome toward @xmath321 . weak coin - flipping does not directly face this same limitation . currently the best bound known on the bias of weak quantum coin - flipping protocols , due to ambainis @xcite , is that @xmath327 rounds of communication are necessary to achieve a bias of @xmath312 . the best weak quantum coin - flipping protocol currently known , which is due to mochon @xcite , achieves bias approximately 0.192 ( which surpasses the barrier @xmath328 on strong quantum coin - flipping ) . next , we define quantum refereed games . it will be noted that von neumann s min - max theorem for zero - sum quantum refereed games follows from the facts we have proved about representations of strategies together with well - known generalizations of the classical min - max theorem . although it is completely expected that the min - max theorem should hold for quantum games , it has not been previously noted in the general case with which we are concerned . ( lee and johnson @xcite proved this fact in the one - round case . ) this discussion will also be helpful for the application to interactive proof systems with competing provers that follows . let us first define specifically what is meant by a zero - sum quantum refereed game . such a game is played between two players , alice and bob , and is arbitrated by a referee . the referee s output after interacting with alice and bob for some fixed number of rounds determines their pay - offs . [ def : referee ] an _ @xmath80-turn referee _ is an @xmath80-turn measuring co - strategy @xmath329 whose input spaces @xmath78 and output spaces @xmath79 take the form @xmath330 for complex euclidean spaces @xmath331 , @xmath332 , @xmath333 and @xmath334 , for @xmath335 . an _ @xmath80-turn quantum refereed game _ consists of an @xmath80-turn referee along with functions @xmath336 defined on the referee s set of measurement outcomes , representing alice s payoff and bob s payoff for each output @xmath33 . such a game is a _ zero - sum _ quantum refereed game if @xmath337 for all @xmath33 . the referee s actions in a quantum refereed game are completely determined by its representation @xmath329 . during each turn , the referee simultaneously sends a message to alice and a message bob , and a response is expected from each player . the spaces @xmath331 and @xmath332 correspond to the messages sent by the referee during turn number @xmath238 , while @xmath333 and @xmath334 correspond to their responses . after @xmath80 turns , the referee produces an output @xmath338 . a refereed quantum game does not in itself place any restrictions on the strategies available to alice and bob . for example , alice and bob might utilize a strategy that allows quantum communication , they might share entanglement but be forbidden from communicating , or might even be forbidden to share entanglement . specific characteristics of a given game , such as its nash equilibria , obviously depend on such restrictions in general . the focus of the remainder of the paper is on the comparatively simple setting of zero - sum quantum refereed games . in this case , we assume that alice and bob do not communicate or share entanglement before the interaction takes place . more precisely , we assume that alice and bob play independent strategies represented by @xmath339 and @xmath340 , respectively . the combined actions of alice and bob are therefore together described by the operator @xmath341 . it is a completely natural assumption that alice and bob play independent strategies in a zero - sum quantum refereed game , given that it can not simultaneously be to both players advantage to communicate directly with one another or to initially share an entangled state . this should not be confused with the possibility that entanglement among the players and referee might exist at various points in the game , or that the referee might choose to pass information from one player to the other . these possibilities are not disallowed when alice and bob s joint strategy is represented by @xmath342 . now , assume that a zero - sum quantum refereed game is given , and that alice and bob play independent strategies @xmath50 and @xmath51 as just discussed . let us write @xmath343 and define @xmath344 alice s expected pay - off is then given by @xmath345 while bob s expected pay - off is @xmath346 . now , @xmath347 is a real - valued bilinear function in @xmath50 and @xmath51 . because the operators @xmath50 and @xmath51 are drawn from compact , convex sets @xmath348 and @xmath349 respectively , we have that @xmath350 this is the min - max theorem for zero - sum quantum games . note that the above expression does not immediately follow from von neumann s original min - max theorem , but follows from an early generalization due to j. ville @xcite and several subsequent generalizations such as the well - known min - max theorem of ky fan @xcite . the real number represented by the two sides of this equation is called the _ value _ of the given game . classical interactive proof systems with competing provers have been studied by several authors , including feige , shamir , and tennenholts @xcite , feige and shamir @xcite , feigenbaum , koller , and shor @xcite , and feige and kilian @xcite . a quantum variant of this model is defined by allowing the verifier to exchange quantum information with the provers @xcite . in both cases these interactive proof systems are generalizations of single prover interactive proof systems @xcite . interactive proof systems with two competing provers are naturally modeled by zero - sum refereed games . to highlight this connection we will refer to the verifier as the referee and the two provers as alice and bob . the referee is assumed to be computationally bounded while alice and bob are computationally unrestricted . alice , bob , and the referee receive a common input string @xmath351 , and an interaction follows . after the interaction takes place , the referee decides that either alice wins or bob wins . a language or promise problem @xmath352 is said to have a classical refereed game if there exists a referee , described by a polynomial - time randomized computation , such that : ( i ) for every input @xmath353 , there is a strategy for alice that wins with probability at least @xmath354 against every strategy of bob , and ( ii ) for every input @xmath355 , there is a strategy for bob that wins with probability at least @xmath354 against every strategy of alice . the class of promise problems having classical refereed games is denoted @xmath356 . it is known that @xmath356 is equal to @xmath357 . the work of koller and megiddo @xcite implies @xmath358 , while feige and kilian @xcite proved the reverse containment . let us note that zero - sum classical refereed games , and therefore the class @xmath356 , are unaffected by the assumption that alice and bob may play quantum strategies , assuming the referee remains classical . this assumes of course that the classical referee is modeled properly within the setting of quantum information , which requires that any quantum information that it touches immediately loses coherence . equivalently , the referee effectively measures all messages sent to it by alice and bob with respect to the standard basis before any further processing takes place . as there also can not be a mutual advantage to alice and bob to correlate their strategies using shared entanglement , there is no advantage to alice or bob to play a quantum strategy against a classical referee . this is not the case in the cooperative setting , because there alice and bob might use a shared entangled state to their advantage @xcite . quantum interactive proof systems with competing provers are defined in a similar way to the classical case , except that the referee s actions are described by polynomial - time generated quantum circuits and the referee may exchange quantum information with alice and bob . the complexity class of all promise problems having quantum refereed games is denoted @xmath359 . the containment @xmath360 follows from @xmath361 . it was previously known that @xmath362 @xcite , and we will improve this to @xmath363 . this establishes the characterization @xmath0 , and implies that quantum and classical refereed games are equivalent with respect to their expressive power . in the refereed game associated with a competing prover quantum interactive proof system , the referee declares either alice or bob to be the winner . specifically , the referee outputs one of two possible values @xmath364 , with @xmath250 meaning that alice wins and @xmath321 meaning that bob wins . by setting @xmath365 and @xmath366 , we obtain a quantum refereed game whose value is the maximum probability with which alice can win . we will show that this optimal winning probability for alice can be efficiently approximated : it is the value of a semidefinite programming problem whose size is polynomial in the total dimension of the input and output spaces of the referee . it follows from the polynomial - time solvability of semidefinite programming problems @xcite that @xmath363 . we will use similar notation to the previous subsection : for a fixed input @xmath367 , the referee is represented by operators @xmath368 , and assuming @xmath80 is the number of turns for this referee we let the input and output spaces to alice be denoted by @xmath369 while @xmath370 denote the input and output spaces to bob . the assumption that the referee is described by polynomial - time generated quantum circuits implies that the entries in the matrix representations of @xmath371 and @xmath372 with respect to the standard basis can be approximated to very high precision in exponential time . now , given any strategy @xmath339 for alice , let us define @xmath373 the functions @xmath374 and @xmath375 are linear and extend to uniquely defined super - operators . under the assumption that alice plays the strategy represented by @xmath339 and bob plays the strategy represented by @xmath340 , we have that the referee outputs @xmath250 with probability @xmath376 and outputs @xmath321 with probability @xmath377 . one may think of @xmath378 as being the co - strategy that results from `` hard - wiring '' alice s strategy represented by @xmath50 into the referee . now , alice s goal is to minimize the maximum probability with which bob can win . for a given strategy @xmath50 for alice , the maximum probability with which bob can win is @xmath379 which , by theorem [ theorem : gauge ] , is given by @xmath380 the following optimization problem therefore determines the maximum probability @xmath1 for bob to win , minimized over all strategies for alice : @xmath381 this optimization problem can be expressed in terms of linear and semidefinite constraints as follows : @xmath382 & \tr_{\c_k}(a_k ) = a_{k-1 } \otimes i_{\a_k } \quad & ( 2\leq k\leq n),\\ & \tr_{\c_1}(a_1 ) = i_{\a_1},\\[2 mm ] & q_k = p_k \otimes i_{\d_k } \quad & ( 1\leq k\leq n),\\ & \tr_{\b_k}(p_k ) = q_{k-1 } \quad & ( 2\leq k\leq n),\\[2 mm ] & a_k \in \pos{\c_{1\ldots k}\otimes\a_{1\ldots k } } \quad & ( 1\leq k\leq n ) , \\ & q_k \in \pos{\d_{1\ldots k}\otimes\b_{1\ldots k } } \quad & ( 1\leq k\leq n ) , \\ & p_k \in \pos{\d_{1\ldots k-1}\otimes\b_{1\ldots k}}\quad & ( 1\leq k\leq n).\end{aligned}\ ] ] quantum interactive proofs with competing provers . in _ proceedings of the 22nd symposium on theoretical aspects of computer science _ ( 2005 ) , vol . 3404 of _ lecture notes in computer science _ , springer , pp .

the lhc higgs search data just released indicates the existence of a scalar resonance with mass . we examine the implications of the data reported by atlas , cms and the tevatron collaborations on understanding the properties of this scalar by performing joint fits on its couplings to other standard model ( sm ) particles . we discuss and characterize to what degree this resonance has the properties of the sm higgs , and consider what implications can be extracted for new physics in a ( mostly ) model - independent fashion . we find that , if the higgs couplings to fermions and weak vector bosons are allowed to differ from their standard values , the sm is from the best fit point to the current data . fitting to a possible invisible decay branching ratio , we find . we also discuss and develop some ways of using the data in order to bound or rule out models which modify significantly the properties of this scalar resonance , and apply these techniques to the current global dataset .

1207.1717

particle physics entered a new era with the announcement of the discovery of a new boson @xcite based on excess events in several higgs search channels using @xmath4 @xmath5 lhc data collected in 2011 - 2012 . in light of this discovery , it has become obvious that the question of central importance to address now is what are the properties of the scalar field responsible for the observed excesses ? the answer to this question determines if this field corresponds to the standard model ( sm ) higgs , with the specific sm mechanism of elegantly breaking electroweak ( ew ) symmetry , or whether a more complicated mechanism is involved in ew symmetry breaking . we study this question in detail in this paper , characterizing to what degree a sm higgs is consistent with the current global dataset and presenting several results on the properties of the scalar field . besides updating and expanding our past results @xcite , we also present new analyses that emphasize the power of the growing dataset to bound and rule out alternative models or to give hints of new physics ( np ) . it is worth emphasizing that it is very important to specify ( and justify ) a coherent theoretical framework in which to study the emerging evidence for the scalar field . however , without knowing the ultraviolet ( uv ) origin of this field , we do not know what effective field theory ( eft ) , or complete model , should be used to fit the data . we emphasize that at this time , the existing experimental evidence is not sufficiently strong to directly assume that the scalar resonance is the sm higgs boson , ascribing any deviations in the measured properties of the scalar field directly to the effects of np interactions expressed through higher dimensional operators . although this is certainly one possible interpretation of the data ( and we will examine the implications of higgs data for np in this framework ) , we emphasize that , in general , one should not assume what one wishes to prove . nevertheless , in formulating a theoretical framework for this study , a wealth of other experimental results that are also sensitive to the properties of scalar fields at the weak scale can be distilled into simple physical guiding principles . these are , namely , approximate minimal flavour violation ( mfv ) @xcite ; respecting the soft higgs theorems of refs . @xcite ( i.e. the scalar couples to the sm fields in proportion to their masses ) ; and an effective breaking of custodial symmetry , @xmath6 , @xcite approximately as in the sm . directly incorporating these principles in the formulation of the effective lagrangian allows us to restrict our attention to a few simple cases . in order to establish experimentally the properties of the scalar resonance in a model - independent way , one can utilize the effective field theory of the chiral ew lagrangian coupled to a scalar field that was emphasized in refs . @xcite to study recent higgs signal - strength data . depending on the assumptions of the uv origin of such a lagrangian , one is lead to various sets of free parameters to fit the data when studying the consistency of the sm higgs hypothesis with the current dataset . we discuss and utilize this framework extensively in this paper to examine the properties of the scalar field emerging from the data . we also emphasize that with the discovery of a new scalar resonance , one can also use the signal strength properties of the scalar field to bound and rule out models that provide @xmath7 signal events as well as models that provide too many signal events . further , one can also exclude allowed parameter space due to the degree of tension within the dataset that depends on the properties of the scalar field . as the dataset evolves , these techniques become complementary to direct @xmath8 fits on the signal strength dataset . these bounds can be quantified by excluding parameter space in the allowed couplings of the scalar using a gaussian probability density function approach . we develop and apply such an approach in this paper . the outline of this paper is as follows . in section [ frameworks ] we discuss the eft framework we employ , and the implicit assumptions about the uv origin of the scalar field that are adopted when fitting the data with various free parameters . in section [ data ] we review and discuss the manner in which we treat the scalar signal strength data and electroweak precision data ( ewpd ) , while in section [ results ] we present results , based on our fit method , of the status of the sm higgs hypothesis . in section [ bsm ] we discuss some of the implications of higgs signal strength parameters for beyond the sm ( bsm ) physics , expressed through model - independent free parameters . in section [ tension ] we discuss novel and complementary methods to identify the allowed parameter space , and in section [ concl ] we conclude . an effective chiral ew lagrangian with a nonlinear realization of the @xmath9 symmetry gives a minimal description of the ( non - scalar ) degrees of freedom of the sm consistent with the assumptions of sm - like @xmath6 violation and mfv . the goldstone bosons eaten by the @xmath10 are denoted by @xmath11 ( where @xmath12 ) , and are grouped as ( x ) = e^i _ a ^a / v , with @xmath13 the pauli matrices and @xmath14 . in this approach , the ew scale @xmath15 , which sets the mass of fermions and gauge bosons is introduced directly into the lagrangian . the @xmath16 field transforms linearly under @xmath17 as @xmath18 where @xmath19 indicate the transformation on the left and right under @xmath20 and @xmath21 , respectively , while @xmath6 is the diagonal subgroup of @xmath17 . adding a scalar field @xmath22 to this theory is trivial . one chooses @xmath22 to transform as a singlet under @xmath6 and a derivative expansion of such a theory is given by @xcite _ eff & = & ( _ h)^2 - v(h ) + ( d_^ d^ ) , + & & - ( |u_l^i |d_l^i ) ( c y_ij^u u_r^j + y_ij^d d_r^j ) + h.c . , [ leff ] with v(h ) & = & m_h^2 h^2 + ( ) h^3 + ( ) h^4 + . [ eq : efflag ] although we use the notation @xmath22 , we do not assume in principle that this scalar field is the higgs , or that the scale @xmath15 is somehow associated with the vacuum expectation value of this field as this is what we seek to establish from the data . as is well known , the @xmath23 and @xmath24 parameters control the couplings of @xmath22 to gauge bosons and fermions , respectively , and therefore , play a crucial role in the phenomenology of single @xmath22 production . note that in previous fits , and in the majority of this work , the assumption @xmath25 is used and no distinction is made between the rescaling of the @xmath22-coupling to the @xmath26 and @xmath27 quarks . we will relax this assumption later on . this lagrangian is common in the study of composite models and its relevance has been emphasized recently in refs . it is also appropriate to study a pseudo - goldstone boson ( pgb ) emerging out of an approximately conformal sector @xcite , or as the low - energy eft arising in many scenarios where the higgs is a composite pgb that emerges from the breaking of a larger chiral symmetry group @xcite . we emphasize that this eft setup can be matched to many uv frameworks and , being quite general , we do not confine ourselves to any particular uv scenario . constraints ) . note , however , that relaxing the assumption @xmath28 significantly , can lead to conflict with precision constraints sensitive to @xmath6 and flavour violation . ] we approach the data in this way to be as model - independent as possible . however , even specifying the free parameters that one will use to fit the data introduces implicit model dependence . one can nevertheless broadly characterize certain parameter choices . as this is an eft , @xmath29 is non - renormalizable . here ( and throughout this paper ) we take the cut - off scale to be @xmath30 . since we are concerned with the phenomenology of single scalar production , the higher - order derivative operators are suppressed by powers of @xmath31 . as such , we are justified in neglecting such sub - leading effects in this paper . non - derivative higher - dimensional operators will also exist in general . when the @xmath22 field is not assumed to have a uv origin such as the sm higgs , and is simply considered to be a singlet field [ that need not necessarily transform under the nonlinearly realized @xmath32 symmetry ] , the leading operators in the expansion in inverse powers of @xmath33 appear at dimension five and are given by ^5_hd & = & - h g^a _ g^a - h w^a _ w^a - h b _ b^ . here @xmath34 are the weak hypercharge , @xmath35 and @xmath36 gauge couplings , respectively , and the different tensor fields are the corresponding field strengths with their associated wilson coefficients @xmath37 . the scale @xmath33 corresponds to the mass scale of the lightest new state that is integrated out , which we assume is proximate to @xmath33 . we will neglect operators originating from cp - violating sources due to the lack of any clear evidence of beyond the sm cp violation in lower energy precision tests . note that the operators in @xmath38 can be further suppressed compared to the effects of @xmath39 on ( single ) scalar production when the scalar field has specific uv origins . this is the case for example when @xmath22 is a pgb , see ref . @xcite for a detailed discussion . when one assumes that @xmath22 is embedded into an @xmath20 doublet - @xmath40 - as in the sm , the operators in @xmath38 first appear at dimension six , and the coefficients are suppressed by an extra factor of @xmath41 when considering single scalar production . in this case , the dimension six operator basis is also extended by the operator ^6_hd & = & - h^ ^a h b _ w^a . for phenomenological purposes it is convenient to rotate to a basis for the operators given by [ phenolagrangian ] _ hd & = & - h g^a _ g^a - h f _ f^ , where @xmath42 is the electromagnetic field strength tensor and @xmath43 in the case of an @xmath20 singlet field , and @xmath44 if @xmath22 is embedded into an @xmath20 doublet . in this manner , one can understand that the choices to retain the effects of higher dimensional operators ( or not ) in performing global fits introduces implicit uv dependence . in introducing higher dimensional operators , one is also explicitly assuming the existence of new states charged under at least a subgroup of the sm group . alternatively , if np is uncharged under the sm group but couples to the @xmath45 operator , then it can impact @xmath22-phenomenology by inducing an invisible @xmath22-width ( when the scalar field takes a vacuum expectation value ) . this leads to the modification of the sm branching ratios for each decay into visible sm final states @xmath46 via ( h f ) ( 1 - br_inv ) _ sm(h f ) . [ eq : brinv ] we will update our recent fit of @xmath47 to higgs search data @xcite in a later section , and use this fit as a diagnostic tool to test aspects of our fit procedure . in general , the coefficients @xmath48 are arbitrary parameters subject to experimental constraints . the generic cases we consider are : * composite / pseudo - goldstone higgs / dilaton scalar theories . in this case , @xmath39 are free parameters in general , although they can be fixed in particular uv completions . if the scalar field is a pseudo - goldstone boson , it is also appropriate to neglect higher dimensional operators . we will fit to subsets of the parameters @xmath49 in what follows . as this is a more general framework than the sm higgs , we will use this eft in assessing to what degree the sm higgs hypothesis is consistent with the data or if deviations into this parameter space can give a substantially better fit . * the sm higgs as a low - energy eft . when the low - energy eft is just the sm , the field @xmath22 becomes part of a linear multiplet u = ( 1 + ) , reducing eq . ( [ leff ] ) to the sm higgs lagrangian . in this case @xmath50 and @xmath51 , and the only effect of np is through non - renormalizable higher - order operators . the naturalness problem of the sm higgs mass operator , and recent experimental hints of deviations in the observed properties of the ( assumed ) higgs , motivates moderately heavy np and the introduction of bsm parameters @xmath52 as free parameters . we study the constraints on these parameters in detail in this paper . we will not attempt to relate the constraints obtained on the various parameters to any particular underlying model , other than the sm , in this paper . this choice is motivated by the current lack of other clear experimental evidence of bsm states to guide coherent model - building . the classes of models discussed above can be considered as motivating examples . in this section we describe our method for globally fitting to the parameters discussed above , and incorporating the recently released @xmath53 tev data @xcite , updated @xmath54 tev results from atlas @xcite , the released 7 tev cms data @xcite , and the recently reported tevatron higgs search results @xcite . this work builds on our previous fits @xcite . we only summarize the main details of the fit procedure and method here . many subsidiary details of the fit procedure can be found in these reference works . we fit to the available higgs signal strength data , _ i = , [ mui ] for the production of a higgs that decays into the visible channels @xmath55 , where @xmath56 denotes the number of channels . the label @xmath57 in the cross section , @xmath58 , is due to the fact that some final states are defined to only be summed over a subset of higgs production processes @xmath57 . the reported best fit value of a signal strength we denote by @xmath59 , in terms of the observed numbers of events ( @xmath60 ) , the number of background events ( @xmath61 ) and the expected number of sm signal events ( @xmath62 ) . ] . the global @xmath8 we construct is defined via [ chi2 ] ^2(_i ) = _ i=1^n_ch . the covariance matrix has been taken to be diagonal with the square of the @xmath63 theory and experimental errors added in quadrature for each observable , giving the error @xmath64 in the equation above . correlation coefficients are neglected as they are not supplied by the experimental collaborations . for the experimental errors we use @xmath65 symmetric @xmath66 errors on the reported @xmath59 . for theory predictions of the @xmath58 and related errors , we use the numbers given on the webpage of the lhc higgs cross section working group @xcite . tev and we use the updated numbers . also note that @xmath67 is set to zero on this page but we use the latest version of hdecay @xcite to add in @xmath67 to the quoted results . this has a negligible impact on the reported numbers through the modification of the total width . ] the minimum ( @xmath68 ) is determined , and the @xmath69 best fit regions are plotted as @xmath70 , with the appropriate cumulative distribution function ( cdf ) defining @xmath71 . we assume , as in ref . @xcite , that the signal strength @xmath72 in a given channel @xmath73 follows a gaussian distribution with the probability density function ( pdf ) given by _ i(_i,_i,_i ) e^-(_i - _ i)^2/(2 _ i^2 ) , with one - sigma error @xmath64 , and best fit value @xmath59 . this is the case as long as the number of events is large , @xmath74 events , @xcite . we normalize these pdf s to 1 in the interval @xmath75 . in the framework of the sm , the predicted values of the @xmath72 are the same ( equal to 1 ) , and a universal signal strength modifier @xmath76 can be defined and applied to all channels . by multiplying together the individual channel pdf s ( or the pdf s of a single channel reported at two operating energies ) , we can also define a combined pdf for @xmath76 . this can be done for each separate experiment or for a global combination of all experiments . the combined pdf is also gaussian and has combined @xmath77 and @xmath78 values given approximately ) to include correlations is straightforward . these formulas are also easy to generalize in models with non - universal theory - predicted @xmath72 . ] by [ csigma ] = _ i^n_ch , = _ i^n_ch . we will use these relations to reconstruct the unreported @xmath0 data from the reported @xmath54 and @xmath79 data . armed with combined pdf(s ) , we can determine the 95% c.l . exclusion upper limits on the signal strength parameter @xmath76 ( @xmath80 ) @xcite . ( we shall explain momentarily how to introduce an overall signal strength parameter in models with non - universal theory - predicted @xmath72 . in the discussion that follows we are implicitly assuming that we are considering setting limits on combined signal strength parameters although we frame the discussion in terms of @xmath76 . a similar analysis can be carried through channel - by - channel instead of on the combined channels . ) such limits can be set on the combined signal strength parameter @xmath81 or for an individual channel s signal strength . with @xmath82 s settling around unity with @xmath64 errors getting smaller and smaller due to increasing integrated luminosities , we can already start to set also _ lower _ limits on @xmath76 . the condition for such lower bounds , say at 95@xmath83 c.l . , to be meaningful is that the symmetric interval @xmath84 containing 95% of the integrated probability has @xmath85 , in which case @xmath86 corresponds to the lower 95% c.l . bound on @xmath76 . in this same case , we will take @xmath87 as the upper limit @xmath88 . the conditions that define these limit are therefore ^__dwl pdf ( ) d= = 0.95/2 , and _ ^_upl pdf ( ) d= = 0.95/2 , where @xmath89 is the error function . however , when @xmath90 , we shift the 95% c.l . interval to the asymmetric one @xmath91 and revert to the upper limit definition @xcite , in which case there is no lower limit on @xmath76 and the upper limit is given by eq . ( [ usualuplim ] ) . ] [ usualuplim ] _ 0^_upl pdf ( ) d= = 0.95 . next consider models that depart from the sm , like @xmath92 with couplings of the higgs to fermions and gauge bosons modified by the @xmath39 factors as explained above . the predicted values of the @xmath72 s will deviate from 1 in a channel - dependent way . in order to keep the same expected value of the signal - strengths for all channels , it is convenient to normalize the @xmath72 s in eq . ( [ mui ] ) not to the sm signal expectation @xmath62 but to the expectation in the model considered @xmath93 . by doing this we can again use a universal signal - strength modifier @xmath76 , with expected value @xmath94 , corresponding to the model being tested . on the other hand , with this change in the @xmath72 definition , the observed @xmath95 values need to be rescaled by a factor @xmath96 . the individual pdf s are modified accordingly and can be combined in the usual way . to summarize , we can exclude ( at 95 % c.l . ) a given scenario not only if it predicts too many signal events which are not seen ( @xmath97 ) , but also if it predicts too few events , incompatible with the observed excesses associated with the reported discovery . the significance of such lower bounds will grow with more luminosity . significances above 5@xmath98 become possible ( by definition ) after discovery ( which excludes the background hypothesis @xmath99 ) . we will plot both of these bounds mapping the allowed @xmath100 into the relevant parameter space through the dependence of @xmath72 on the free parameters in our fit results . figure [ tenslim ] , in section [ tension ] , shows examples of such lower limits . we incorporate ewpd @xcite by adding it directly to the @xmath8 measure in eq . ( [ chi2 ] ) . when @xmath23 is considered a free parameter , the shifts of the oblique parameters @xmath101 and @xmath102 are given by @xcite s & & ( ) , t ( ) . the numerical coefficient is determined from the logarithmic large-@xmath103 dependence of @xmath104 given in ref . @xcite .. the degree to which @xmath33 properly captures the uv regularization of @xmath105 and @xmath106 is model - dependent . we assume that directly treating this cut - off scale as a proxy for a heavy mass scale integrated out is a good approximation , i.e. that further arbitrary parameters rescaling the cut - off scale terms need not be introduced . ] as for ewpd , recent updates to the measurement of @xmath107 at the tevatron @xcite have refined the world average @xcite , and have significantly reduced the quoted error . incorporating these new measurements we use @xcite s = 0.00 0.10 , t = 0.02 0.11 , u = 0.03 0.09 , while the matrix of correlation coefficients is given by c = ( ccc 1 & 0.89 & -0.55 + 0.89 & 1 & -0.80 + -0.55 & -0.80 & 1 + ) . here we have assumed @xmath108 , as corrections for shifting these results by a few gev are negligible . there is a strong preference for @xmath109 in the global fit when ewpd is used , and the constraints on the scalar field can be directly associated with ewpd bounds . note that the slight preference for @xmath110 in the best fit region when ewpd is taken into account is subject to uncertainties in cut - off scale effects . although the shift in the best fit point is of interest as a probe of possible new physics , we can not clearly disentangle such a hint from cut - off scale effects . the excess of events of @xmath111 significance reported by atlas and cms peaks , as a function of the scalar mass , at slightly different values : @xmath112 for atlas and @xmath113 for cms . this difference can be attributed at this stage of the search to statistical fluctuations in the data . monte - carlo studies @xcite indicate that such effects can shift the observed maximum signal strength compared to the true signal strength maximum by @xmath114 . due to this , a fit that combines data from different experiments at the same mass value might be biased and not necessarily better than using data at slightly different masses . as much more detailed data is available to us at the mass peaks , in our global fits we will use all available @xmath72 taken at @xmath115 @xmath116 for the cms and the tevatron ( which has a @xmath117 excess over a wider region of masses ) , and at @xmath118 for atlas ( see fig . [ fig.data ] in the appendix ) . for cms @xcite and at @xmath119 for atlas @xcite . however , as the amount of information relevant to the fits we will perform released to date is still greater at the mass scales @xmath120 for cms and @xmath121 for atlas we retain these mass choices in our fit . ] in order to assess the degree of consistency of the data with the sm hypothesis we first consider the effective lagrangian given by @xmath29 and assume that higher dimensional operators are sufficiently suppressed so that @xmath122 can be neglected . we then perform a two - parameter @xmath123 fit and examine the @xmath124 for the sm point @xmath125 compared with the best fit point . this defines a c.l . corresponding to the deviation of the sm hypothesis compared to the best fit point . the result is shown in figure [ bestfits ] ( left ) . we also show in fig . [ bestfits ] ( right ) the best fit regions when ewpd is added to the global @xmath8 measure . notice the dramatic reduction in the size of the best fit region along the @xmath23-parameter , which is forced to lie close to 1 . these results visually summarize the current experimental status of establishing the higgs hypothesis . when ewpd is not used , the sm higgs hypothesis of @xmath126 is @xmath127 ( c.l . of @xmath128 ) away from the best fit point , which sits at @xmath129 . note that here and in the following discussion we are choosing to round the c.l . and the best fit points . this is due to the preliminary nature of the @xmath130 data and is not limited directly to this accuracy due to the fit procedure we have adopted . the c.l . of the sm hypothesis in the combined data is consistent with our past results at @xmath54 @xmath5 @xcite . with the updated data released and incorporated in our fit since version one of this paper the parameter space that has the global minimum has changed from @xmath131 ( with initial ichep data ) to @xmath132 ( with post - ichep updates ) . is the primary reason we retain the use of @xmath118 for atlas data . ] the existence of the non negligible parameter space with @xmath131 is easy to understand . due to the interference term in the @xmath133 decay width which is @xmath134 , a negative @xmath135 allows a relatively larger excess in @xmath136 events due to constructive interference between the top and @xmath137 boson loops . when ewpd is used as in figure [ bestfits ] ( right ) we find that the sm is similarly residing at @xmath138 ( c.l . of @xmath139 ) away from the best fit point which is now @xmath140 and the best fit region where @xmath132 now has a ( significantly ) lower global minimum . the minima are no longer as degenerate with the addition of the most recent atlas data , @xmath141 . plane for all reported best fit values given by atlas and cms , left ( right ) without ewpd ( with ewpd ) . in both plots we take @xmath108 for the tevatron and cms7/8 and @xmath118 for atlas7/8 . the green , yellow , gray regions corresponds to the allowed @xmath142 spaces for a two parameter fit . the best fit point in each region is also labeled with a point . the thicker point indicates the one with the smaller @xmath68.,title="fig:",scaledwidth=45.0% ] plane for all reported best fit values given by atlas and cms , left ( right ) without ewpd ( with ewpd ) . in both plots we take @xmath108 for the tevatron and cms7/8 and @xmath118 for atlas7/8 . the green , yellow , gray regions corresponds to the allowed @xmath142 spaces for a two parameter fit . the best fit point in each region is also labeled with a point . the thicker point indicates the one with the smaller @xmath68.,title="fig:",scaledwidth=46.0% ] in view of the different masses of the signal - strength peaks in the various experiments ( which can be due to the statistical effects mentioned above ) and of the subtleties we have neglected in properly combining the results of these different experiments , it is also of interest to perform the fit in the @xmath143 space for each experiment individually . we show these results in figure [ individualexp ] . the cms experiment has the sm point residing about @xmath144 from the best fit point , with the c.l . of the sm case compared to the best fit point at @xmath145 . for atlas , the sm point is now at a c.l . of @xmath146 , within the @xmath147 region . the tevatron results have the sm point within the @xmath148 region with a c.l . of the sm case ( compared to the best fit point ) of @xmath149 . the allowed fit region for cms can be compared to the recently presented public results @xcite , which restrict the fit to the region @xmath132 ( physically different from the region @xmath131 ) . by a redefinition of the @xmath22 field . the sign of @xmath135 could in principle be changed also by rotating the fermion fields , but this would affect in the same way the fermion mass so that the sign of @xmath150 is in fact fixed and physically meaningful . ] as the absolute minimum of the @xmath8 lies in fact in the discarded region , the shape and size of the @xmath151 and @xmath152 c.l . regions presented by cms differ from the ones that we obtain in figure [ individualexp ] . if we also restrict our parameter space to positive @xmath135 we have checked that we get excellent agreement with the cms result . note however that there is no valid reason to discard a priori the negative @xmath135 region which offers in fact the interesting possibility of giving a good fit to the data by an enhancement of the @xmath153 coupling through constructive interference of the top and @xmath137 loops . plane for a fit to all reported signal - strength values given by atlas ( @xmath118 ) , cms ( @xmath108 ) and the tevatron ( @xmath108 ) collaborations individually . we plot the same best fit contours over the same domain of parameter space to allow a direct comparison amongst experimental results . the significant change in the atlas results from version one of this paper is due to the use of the atlas diphoton data broken into subcategories.,title="fig:",scaledwidth=32.0% ] plane for a fit to all reported signal - strength values given by atlas ( @xmath118 ) , cms ( @xmath108 ) and the tevatron ( @xmath108 ) collaborations individually . we plot the same best fit contours over the same domain of parameter space to allow a direct comparison amongst experimental results . the significant change in the atlas results from version one of this paper is due to the use of the atlas diphoton data broken into subcategories.,title="fig:",scaledwidth=32.0% ] plane for a fit to all reported signal - strength values given by atlas ( @xmath118 ) , cms ( @xmath108 ) and the tevatron ( @xmath108 ) collaborations individually . we plot the same best fit contours over the same domain of parameter space to allow a direct comparison amongst experimental results . the significant change in the atlas results from version one of this paper is due to the use of the atlas diphoton data broken into subcategories.,title="fig:",scaledwidth=32.0% ] the results of the last section can be interpreted as ( partial ) evidence in support of the sm higgs hypothesis . assuming then that the observed boson is the sm higgs , we can study possible deviations of its properties due to bsm effects . the simplest extension of the sm in terms of new parameters is perhaps the case where one only introduces a higgs invisible branching ratio , @xmath47 . such an extension is common in many bsm scenarios , _ e.g. _ when new physics couples through the higgs portal , and new states exist that are uncharged under the sm group . fitting to @xmath47 allows one to fit to the global combined signal strength values supplied by the experimental collaborations . the values we use to perform the fit are given in table [ table : combineddata ] . for the sm higgs using only cms data ( left figure ) for @xmath108 with two methods as a check of our fit procedure . blue curve - global signal strength based fit , red curve - individual channel fit . see text for further explanation . the middle figure shows the @xmath8 distribution developed from the combined best fit @xmath154 supplied by the four experiments , including the 7 and 8 tev lhc results for two mass values . in the right figure we show the discovery potential for @xmath47 , updating a result from the analysis in ref . @xcite with the new global signal strength data . _ , title="fig:",scaledwidth=32.0% ] for the sm higgs using only cms data ( left figure ) for @xmath108 with two methods as a check of our fit procedure . blue curve - global signal strength based fit , red curve - individual channel fit . see text for further explanation . the middle figure shows the @xmath8 distribution developed from the combined best fit @xmath154 supplied by the four experiments , including the 7 and 8 tev lhc results for two mass values . in the right figure we show the discovery potential for @xmath47 , updating a result from the analysis in ref . @xcite with the new global signal strength data . _ , title="fig:",scaledwidth=32.0% ] for the sm higgs using only cms data ( left figure ) for @xmath108 with two methods as a check of our fit procedure . blue curve - global signal strength based fit , red curve - individual channel fit . see text for further explanation . the middle figure shows the @xmath8 distribution developed from the combined best fit @xmath154 supplied by the four experiments , including the 7 and 8 tev lhc results for two mass values . in the right figure we show the discovery potential for @xmath47 , updating a result from the analysis in ref . @xcite with the new global signal strength data . _ , title="fig:",scaledwidth=32.0% ] .__combined signal strengths @xmath154 and errors @xmath78 from atlas , cms and the tevatron collaborations . here we quote @xmath65 symmetric @xmath66 errors . _ _ [ cols="^,^,^,^,^",options="header " , ] the left plot of fig.[fig.globalsm1 ] shows the result of extracting @xmath47 using our fit approach using two different methods . this allows an important cross check of our procedure and results . we have employed in our fit two approximations that require further justification . using the assumption of gaussian pdf s , we have extracted the @xmath0 data from the known @xmath155 data and the released @xmath79 data . in doing so we have neglected ( unsupplied ) correlations . further , in performing our fits we have neglected correlation coefficients for the @xmath72 . the left plot of fig.[fig.globalsm1 ] shows in blue the result of fitting directly to the combined cms data - which do take into account correlations . in red we show , for comparison , the fit results when we use our procedure to extract the @xmath0 data , and fit to the individual @xmath72 s . the best fit point and the @xmath156 c.l . regions are in good agreement , supporting the estimated accuracy of our results of @xmath157 . the middle plot of fig.[fig.globalsm1 ] shows instead the resulting @xmath8 distributions for @xmath47 for two different mass values @xmath158 when we combine the results from the three experiments and fit to the supplied signal strength parameters @xmath159 . this is the most accurate analysis on @xmath47 that we can perform with the released data ( more accurate than a fit on the individual channels ) . this is primarily due to the experimental correlation effects that are incorporated in the @xmath154 , and larger number of channels that are incorporated in the experimental likelihoods used to construct the combined signal strength parameter used . we find the @xmath156 c.l . regions @xmath160 for @xmath161 . these limits can be used to constrain many models that predict an invisible higgs width . for recent related work , see ref . @xcite . the right plot of fig.[fig.globalsm1 ] ( which updates a similar analysis in ref . @xcite ) shows the current status of the quantity @xmath162 ( with one sigma error band ) , interpreted as an invisible higgs branching ratio , as a function of the higgs mass . this result is also based on the combined signal strengths supplied by the experimental collaborations . with the current errors , no significant statement can be made about the possible nonzero central value of @xmath47 . the shaded areas indicate the 2@xmath98 range of possible fluctuations ( of the background in the upper region at large @xmath47 ; of the sm higgs signal in the lower region at small @xmath47 ) that could be miss - interpreted as an invisible higgs width . this plot shows that the current nonzero central values of @xmath47 at @xmath163 ( consistent with the fit results ) are perfectly compatible with downward fluctuations of the sm higgs signal . the main difference between the results presented in fig.[fig.globalsm1 ] with respect to the first version of this paper comes from the use of the atlas combined signal strength reported in ref . @xcite , that incorporates now the large 8 @xmath164 atlas ww @xcite signal strength @xmath165 . one finds the @xmath156 c.l . limits @xmath166 for @xmath161 when this signal strength is added to the global data set . one can also infer the current experimental bounds on the bsm parameters @xmath168 . we expect that these operators arise at the loop level , so we rescale the wilson coefficients as @xmath169 for @xmath170 . using the results of ref . @xcite , the effects of these operators are incorporated as rescaling factors used in the fit and given by [ higher-d.effect ] r_g & & | 1 - |^2 , + r _ & & | 1 + |^2 . here we have used @xmath171 , @xmath172 , @xmath108 and @xmath173 . when the higgs has sm renormalizable couplings , then @xmath174 . we have retained the two - loop qcd correction to the sm matching of the @xmath175 operator in the @xmath176 limit in these numerical coefficients . the operators in @xmath38 also affect @xmath177 , but the effects in our fit can be neglected . see ref . @xcite for further details and discussion on our fitting procedure to these higher dimensional operators . we show in figure [ higherdfigure ] the results of fitting to @xmath168 using the current dataset when one assumes the sm values @xmath178 ( left ) ; when one fits to @xmath179 and subsequently marginalizes over @xmath39 ( middle ) ; and finally , when one fits to @xmath180 and marginalizes over @xmath181 . these results summarize the preference in the current dataset for including higher dimensional operators when the scalar is not assumed to be the higgs . when the sm is assumed . note that in these results we have assumed that @xmath182 are real . ( middle ) results of fitting to @xmath179 and marginalizing over ( a , c ) subject to the constraint @xmath183 and @xmath184 . ( right ) results of fitting to @xmath180 and marginalizing over @xmath181.,title="fig:",scaledwidth=33.0% ] when the sm is assumed . note that in these results we have assumed that @xmath182 are real . ( middle ) results of fitting to @xmath179 and marginalizing over ( a , c ) subject to the constraint @xmath183 and @xmath184 . ( right ) results of fitting to @xmath180 and marginalizing over @xmath181.,title="fig:",scaledwidth=33.0% ] when the sm is assumed . note that in these results we have assumed that @xmath182 are real . ( middle ) results of fitting to @xmath179 and marginalizing over ( a , c ) subject to the constraint @xmath183 and @xmath184 . ( right ) results of fitting to @xmath180 and marginalizing over @xmath181.,title="fig:",scaledwidth=32.0% ] we also show in fig . [ moremarg ] ( left ) the fit to @xmath179 , marginalizing over the higher dimensional operators . this case shows the preference for @xmath39 even when more massive states are integrated out of a composite scalar theory ( for example ) contributing to @xmath168 . we see that the sm hypothesis is significantly improved in its consistency with the dataset in the context of np of this form . this is particularly relevant as it directly shows what the data tells about @xmath23 , the key parameter that probes the involvement of @xmath22 in electroweak symmetry breaking . a final plot in fig . [ moremarg ] ( right ) shows the allowed space when @xmath185 is varied independently from the remaining fermion couplings which are treated with a universal rescaling . and marginalizing over the higher dimensional operators . ( right ) allowed space when @xmath185 is varied independently . see text for more details.,title="fig:",scaledwidth=40.0% ] and marginalizing over the higher dimensional operators . ( right ) allowed space when @xmath185 is varied independently . see text for more details.,title="fig:",scaledwidth=40.0% ] the fit to higher dimensional operators shows a preference for a bsm contribution to @xmath186 , which deserves some comment . first , one can study the model independence of this preference for an enhancement of @xmath186 by constructing a one dimensional @xmath8 distribution , marginalizing over the unknown operator @xmath187 when the sm values @xmath178 are assumed , or over @xmath188 when one is considering non - sm scalar scenarios . performing this exercise we find that a preference for @xmath189 only exists in certain cases where implicit uv assumptions are adopted due to the parameters used to fit the data . one can make a number of observations regarding enhanced @xmath190 event rates . the coupling to @xmath42 can come about due to @xmath191 . if this is the case , one can bound the allowed enhancement of @xmath192 due to related ewpd constraints . using the relation @xcite _ wb = - 2 s , one finds the current experimental constraint @xmath193 for an @xmath194 deviation considered in the ewpd parameter @xmath101 , assuming @xmath108 . using @xmath195 , this leads to the bound @xmath196 for a one ( two ) sigma deviation in the @xmath101 parameter when @xmath197 . restricting the unknown higher dimensional operators to have a global @xmath8 measure in the @xmath198 allowed region ( @xmath199 ) , one can maximize the enhancement of @xmath200 considering the related constraints on @xmath191 . one finds that an enhancement of 1.4 is possible when a one sigma deviation in the @xmath101 parameter is also allowed . the @xmath201 wilson coefficient could also come about due to @xmath202 or a combination of the two . in this case , the ewpd constraint does not directly apply , and an enhancement of @xmath203 by a factor of @xmath204 is still allowed when maximizing the contribution subject to the constraint @xmath205 . this analysis has assumed no relationship between the wilson coefficients @xmath206 , which is fixed in any particular uv completion . it is interesting to note that in general when integrating out a single bsm field one expects a strong relationship between the wilson coefficients , with identical loop functions in many cases , the only differences in the matching onto the wilson coefficients is due to the relative charges of the new states under the sm subgroups @xmath207 and @xmath208 . parameter space derived from a ) 95% c.l . upper limits on the signal strength parameter @xmath209 ( gray shaded regions beyond the black lines towards the right ) , b ) 95% c.l . lower limits @xmath210 ( gray shaded regions enclosed by the black lines close to the origin ) and c ) `` tension limits '' ( purple dotted regions delimited by purple straight lines ) from the presence of search channels inconsistent with the rest of the dataset at 2 @xmath98 . the most powerful channels in setting such limits are listed in the text . for comparison , the best - fit regions discussed in previous sections are also shown . the fermiophobic scenario along @xmath211 is excluded . , title="fig:",scaledwidth=45.0% ] parameter space derived from a ) 95% c.l . upper limits on the signal strength parameter @xmath209 ( gray shaded regions beyond the black lines towards the right ) , b ) 95% c.l . lower limits @xmath210 ( gray shaded regions enclosed by the black lines close to the origin ) and c ) `` tension limits '' ( purple dotted regions delimited by purple straight lines ) from the presence of search channels inconsistent with the rest of the dataset at 2 @xmath98 . the most powerful channels in setting such limits are listed in the text . for comparison , the best - fit regions discussed in previous sections are also shown . the fermiophobic scenario along @xmath211 is excluded . , title="fig:",scaledwidth=45.0% ] parameter space derived from a ) 95% c.l . upper limits on the signal strength parameter @xmath209 ( gray shaded regions beyond the black lines towards the right ) , b ) 95% c.l . lower limits @xmath210 ( gray shaded regions enclosed by the black lines close to the origin ) and c ) `` tension limits '' ( purple dotted regions delimited by purple straight lines ) from the presence of search channels inconsistent with the rest of the dataset at 2 @xmath98 . the most powerful channels in setting such limits are listed in the text . for comparison , the best - fit regions discussed in previous sections are also shown . the fermiophobic scenario along @xmath211 is excluded . , title="fig:",scaledwidth=45.0% ] parameter space derived from a ) 95% c.l . upper limits on the signal strength parameter @xmath209 ( gray shaded regions beyond the black lines towards the right ) , b ) 95% c.l . lower limits @xmath210 ( gray shaded regions enclosed by the black lines close to the origin ) and c ) `` tension limits '' ( purple dotted regions delimited by purple straight lines ) from the presence of search channels inconsistent with the rest of the dataset at 2 @xmath98 . the most powerful channels in setting such limits are listed in the text . for comparison , the best - fit regions discussed in previous sections are also shown . the fermiophobic scenario along @xmath211 is excluded . , title="fig:",scaledwidth=45.0% ] the consistency between the search results ( @xmath212 ) from different channels can be quantified as follows . for each channel @xmath73 we construct its gaussian approximation to the pdf of the signal strength , @xmath213 , which we can contrast with the full combined pdf ( either in a given experiment or for the overall combination of all data ) , @xmath214 . for each single channel we calculate its @xmath215-value with respect to the global combined pdf , _ i.e. _ , @xmath216 is the @xmath215-value for @xmath82 assuming the full @xmath217 . we can also define the @xmath215-value of the global @xmath77 with respect to the individual @xmath218 , which we will denote by @xmath219 . in order to treat properly the @xmath215-values for cases with @xmath220 , we will normalize the pdf s in this section so that they give 1 when integrated over the whole interval @xmath221 . we then have , for @xmath222 , p_ic & & _ -^_ipdf()d= ( ) , + p_ci&&__c^pdf_i()d= ( ) , and , for @xmath223 , p_ic & & _ _ i^pdf()d= ( ) , + p_ci&&_-^_cpdf_i()d= ( ) , where @xmath224 is the complementary error function , @xmath225 . we will say channel @xmath73 is in tension with the rest of the data if @xmath216 and @xmath219 are both very small . for a given critical @xmath215-value @xmath226 , corresponding to @xmath227 standard deviations , channel @xmath73 is not consistent with the combined dataset at @xmath228 c.l . if @xmath229 and the model can be excluded based on that disagreement . the consistency condition reduces simply to discussion . when this test for consistency is applied to the ( @xmath212 ) dataset , interpreted as coming from a sm higgs signal , we find tension at this @xmath230 level for two atlas @xmath231 channels at 7 tev , those labelled urhpth and cchptt ( the two outliers easily identified in fig . [ fig.data ] ) . in extensions of the sm , like the two - parameter scenario @xmath92 we have discussed in previous sections , the rescaling of the different channels ( which affects @xmath212 as explained in section [ mudata ] ) can introduce very significant distortions in the pdfs and cause too large tensions for some other channels . such regions of parameter space can therefore be excluded on this basis . figure [ tenslim ] illustrates this by showing , on addition to the best fit regions and the 95% c.l . exclusion regions derived from upper and lower limits on the signal strength , the regions in @xmath143 space that would be excluded due to more than 2-@xmath98 tension in some search channel ( purple shaded regions delimited by straight lines ) . we show such limits both experiment by experiment and for the combined result . typically , in the excluded regions several channels at the same time cause the exclusion . we show in each case all the region that is excluded by at least one channel ( with the exception of the atlas case , see below ) . in each case , the channels that have bigger exclusion power are the following . for tevatron , the @xmath232 channel ; for cms , at 7 tev , @xmath233 and at 8 tev , @xmath234 and @xmath235 ; for atlas , the two @xmath231 channels mentioned above , urhpth and cchptt at 7 tev , cause @xmath230 tension in all the region of parameter space shown and we do not mark this area . besides these channels , there is @xmath236 tension also from @xmath237 at 7 tev and from @xmath238 at 8 tev . the fact that the tension limits are straight lines passing through the origin is due to the fact that any common rescaling of @xmath23 and @xmath135 ( that leaves @xmath239 invariant ) also leaves invariant the tension associated to any channel as the latter are functions of ratios @xmath240 , and such ratios are also functions of @xmath239 . we see from figure [ tenslim ] , that fermiophobic scenarios , along the axis @xmath211 , are excluded at this level of confidence . this channel - tension analysis is clearly related to the @xmath8 study we have also performed and the tension exclusion limits tend to exclude regions of parameter space that give a bad fit to the data . however , this approach seems to be more powerful in being able to exclude definitely some models . we also see that there are regions of parameter space that are not excluded by the conventional upper ( or lower ) limits imposed on the signal strength parameter , yet can be excluded by the tension exclusion limit . this is simply due to the fact that the dataset can contain two widely separate channels and still give a combined pdf that respects the upper limit , which is not able to probe in such cases the internal inconsistency of the individual channels . we conclude that this type of analysis offers a complementary tool in testing model performance . we have studied the evidence of a scalar field that has been discovered by the lhc collaborations using an effective field theory framework . we have discerned what can be inferred about the properties of the scalar field at this time using joint @xmath8 fits to available datasets . we have also developed and applied new techniques that allow one to exclude model classes that are in tension with the data through violating upper or lower c.l . bounds , or through introducing excessive tension into the signal strength parameter datasets . at this time , according to our fit method , and using publicly available data , the sm higgs hypothesis is consistent with the global dataset compared to the best fit point at the level of @xmath241 . our approach to the presented data is as follows . as the relative weights of the various contributions to the inclusive higgs production processes depend on the operating energy of the lhc , we need separate information on @xmath54 and @xmath0 data to perform our fit as the relative fraction of events at each operating energy depends on the unknown parameters @xmath242 . do not suffer this problem as this parameter affects universally all channels . ] most of the public results @xcite are now presented separately in @xmath243 signal strengths . when this information is available we use directly this data . for those cases in which only the combined @xmath244 tev signal strengths are available ( currently cms combined @xmath245 , @xmath246 used in the fit ) in addition to the @xmath54 tev results , we make use of eq . ( [ csigma ] ) to reconstruct the unreported @xmath0 data . note that this relies in the use of gaussian approximations to the pdf s to describe the data ( and signal strengths ) and should be increasingly accurate as the total number of events increases . this can be done without knowing directly the experimental likelihood function in the limit that correlations are neglected , which is an assumption we are already forced to adopt as this information is not supplied by the experimental collaborations . we show in fig . [ fig.data ] the resulting reconstructed data . an interesting check of our approach is to use a subset of the provided subclass signal strengths to reproduce a reported combined signal strength . this exercise can be carried out , for example , on the supplied vector boson fusion tagging @xmath247 signal strength and @xmath248 signal strength ( that uses @xmath249 ) and comparing to the presented combined @xmath249 signal strength . we have carried out this procedure for the presented cms data and find good agreement with the reported results , within our estimate of @xmath250 error introduced due to a lack of correlations . we have updated our fit from v1 of this paper to include the following information that has been released since the first version appeared on the archive . the full subclasses of @xmath136 events are now available from atlas and cms at @xmath243 . also , the production channel composition of the @xmath136 subclasses have been supplied @xcite . we have modified our fit to use this information consistently with our rescaling procedure . this procedure replaces our utilization of estimated @xmath251 contamination in the @xmath252 signal events . also note that despite the cuts of the tight " and loose " dijet channels of cms indicating they are not mutually exclusive event classes , the cms collaboration vetoes an event appearing in the loose " @xmath252 sample if it passes the tight " cuts @xcite . as such , these event classes can both be included in the global fit we perform . for the @xmath155 dijet tagged diphoton signal , @xmath253 for the @xmath0 tight " dijet tagged diphoton data and @xmath254 for the @xmath0 loose " dijet tagged diphoton data @xcite . here @xmath255 is the contamination of the @xmath252 signals due to @xmath251 higgs production events , when @xmath255 is defined such that the rate is given by @xmath256 . ] finally we note that the @xmath243 signal strengths for the @xmath257 and @xmath258 signal strengths have now been supplied and are incorporated in our fit . comparing our extracted @xmath258 @xmath0 result to the experimentally supplied number we find agreement within the estimated accuracy of our procedure . for the @xmath247 cms signal strength we note that the @xmath155 signal strength recently reported in ref . @xcite differs from the previously public @xmath155 signal strength . we thank a. djouadi , j. erler , b. feigl , r. gonalo , r. harlander , a. juste , m. martinez , m. schumacher , m. spira , w. fisher , j. huston , v. sanz , v. sharma , p. uwer , j. bendavid , a. delgado , k. tackmann and j. bergstrm for helpful communication on related theory and data . this work has been partly supported by the european commission under the contract erc advanced grant 226371 masstev , the contract pitn - ga-2009 - 237920 unilhc , and the contract mrtn - ct-2006 - 035863 forcesuniverse , as well as by the spanish consolider ingenio 2010 programme cpan ( csd2007 - 00042 ) and the spanish ministry micnn under contract fpa2010 - 17747 and fpa2008 - 01430 . mm is supported by the dfg sfb / tr9 computational particle physics . preprints : ka - tp-29 - 2012 , sfb / cpp-12 - 48 , cern - ph - th/2012 - 151 . 99 atlas and cms collaborations , we d july 4th , cern . atlas collaboration arxiv:1207.7214 ; cms collaboration arxiv:1207.7235 . j. r. espinosa , c. grojean , m. muhlleitner and m. trott , jhep * 1205 * ( 2012 ) 097 [ hep - ph/1202.3697 ] . r. s. chivukula and h. georgi , phys . b * 188 * ( 1987 ) 99 . g. dambrosio , g. f. giudice , g. isidori and a. strumia , nucl . b * 645 * ( 2002 ) 155 [ hep - ph/0207036 ] . a. j. buras , acta phys . b * 34 * , 5615 ( 2003 ) [ hep - ph/0310208 ] . v. cirigliano , b. grinstein , g. isidori and m. b. wise , nucl . b * 728 * ( 2005 ) 121 [ hep - ph/0507001 ] . a. azatov , r. contino and j. galloway , jhep * 1204 * ( 2012 ) 127 [ hep - ph/1202.3415 ] . d. carmi , a. falkowski , e. kuflik and t. volansky , [ hep - ph/1202.3144 ] . giardino , k. kannike , m. raidal and a. strumia , jhep * 1206 * ( 2012 ) 117 [ hep - ph/1203.4254 ] ; [ hep - ph/1207.1347 ] . j. ellis and t. you , jhep * 1206 * ( 2012 ) 140 [ hep - ph/1204.0464 ] . r. lafaye _ et al . _ jhep * 0908 * , 009 ( 2009 ) [ hep - ph/0904.3866 ] . c. englert _ et al . b * 707 * , 512 ( 2012 ) [ hep - ph/1112.3007 ] . m. klute _ et al . _ [ hep - ph/1205.2699 ] . t. corbett , o. j. p. eboli , j. gonzalez - fraile and m. c. gonzalez - garcia , [ hep - ph/1207.1344 ] . i. low , j. lykken and g. shaughnessy , [ hep - ph/1207.1093 ] . g. f. giudice , c. grojean , a. pomarol and r. rattazzi , jhep * 0706 * , 045 ( 2007 ) [ hep - ph/0703164 ] . r. contino _ et al _ jhep * 1005 * , 089 ( 2010 ) [ hep - ph/1002.1011 ] . r. grober and m. muhlleitner , jhep * 1106 * , 020 ( 2011 ) [ hep - ph/1012.1562 ] . m. bando , k. -i . matumoto and k. yamawaki , phys . b * 178 * , 308 ( 1986 ) . w. d. goldberger , b. grinstein and w. skiba , phys . lett . * 100 * , 111802 ( 2008 ) [ hep - ph/0708.1463 ] . d. b. kaplan and h. georgi , phys . lett . b * 136 * , 183 ( 1984 ) . d. b. kaplan , h. georgi and s. dimopoulos , phys . b * 136 * , 187 ( 1984 ) . k. agashe , r. contino and a. pomarol , nucl . b * 719 * , 165 ( 2005 ) [ hep - ph/0412089 ] . r. barbieri , b. bellazzini , v. s. rychkov and a. varagnolo , phys . d * 76 * , 115008 ( 2007 ) [ hep - ph/0706.0432 ] . s. chang , c. a. newby , n. raj and c. wanotayaroj , arxiv:1207.0493 [ hep - ph ] . j. chang , k. cheung , p. -y . tseng and t. -c . yuan , arxiv:1206.5853 [ hep - ph ] . m. farina , c. grojean and e. salvioni , [ hep - ph/1205.0011 ] . r. schabinger and j. d. wells , phys . d * 72 * , 093007 ( 2005 ) [ hep - ph/0509209 ] . b. patt and f. wilczek , hep - ph/0605188 . j. march - russell _ et al . _ jhep * 0807 * , 058 ( 2008 ) [ hep - ph/0801.3440 ] . j. r. espinosa and m. quiros , phys . d * 76 * , 076004 ( 2007 ) [ hep - ph/0701145 ] . m. pospelov and a. ritz , phys . d * 84 * , 113001 ( 2011 ) [ hep - ph/1109.4872 ] . j. r. espinosa , m. muhlleitner , c. grojean and m. trott , [ hep - ph/1205.6790 ] . g. aad _ et al . _ [ atlas collaboration ] , [ hep - ex/1207.0319 ] . b. holdom and j. terning , phys . b * 247 * ( 1990 ) 88 . m. e. peskin and t. takeuchi , phys . * 65 * ( 1990 ) 964 . m. golden and l. randall , nucl . b * 361 * ( 1991 ) 3 . m. e. peskin and t. takeuchi , phys . d * 46 * ( 1992 ) 381 . s. chatrchyan _ et al . _ [ the cms collaboration ] , arxiv:1207.7235 [ hep - ex ] . g. aad _ et al . _ [ the atlas collaboration ] , arxiv:1207.7214 [ hep - ex ] . atlas - conf-2012 - 098 m. frigerio , a. pomarol , f. riva and a. urbano , arxiv:1204.2808 [ hep - ph ] . a. v. manohar and m. b. wise , phys . b * 636 * , 107 ( 2006 ) [ hep - ph/0601212 ] . z. han and w. skiba , phys . d * 71 * , 075009 ( 2005 ) [ hep - ph/0412166 ] . b. grinstein and m. b. wise , phys . b * 265 * , 326 ( 1991 ) . espinosa , c. grojean , v. sanz and m.trott , arxiv:1207.7355 [ hep - ph ] . cms - pas - hig-12 - 015 .

lattices possess elegant mathematical properties which have been previously used in the literature to show that structured codes can be efficient in a variety of communication scenarios , including coding for the additive white gaussian noise ( awgn ) channel , dirty - paper channel , wyner - ziv coding , coding for relay networks and so forth . following the approach introduced by erez and zamir , we show that lattice codes are optimal for the family of gaussian broadcast channels where the source transmits a set of common messages to all receivers and each receiver has _ coded side information _ , i.e. , prior information in the form of linear combinations of the messages . this channel model , which is an instance of the gaussian version of index coding , is motivated by applications to multi - terminal networks where the nodes may have access to coded versions of the messages from previous signal hops or through orthogonal channels . the known results on the capacity of this channel are based on random gaussian codebooks . the structured coding scheme proposed in this paper utilizes construction a lattices designed over prime finite fields , and _ algebraic binning _ at the decoders to expurgate the channel code and obtain good lattice subcodes , for every possible set of linear combinations available as side information . as a corollary , we show that lattice codes based on construction a can achieve the capacity of single - user awgn channels with the size of the prime field growing as a function of the code length as , for any fixed , which is the slowest yet reported in the literature . additive white gaussian noise ( awgn ) , capacity , construction a , gaussian broadcast channel , index coding , lattice , side information , structured codes .

1509.01332

information - theoretic arguments often involve random coding techniques to prove the existence of good codes . usually , the codebook is constructed by randomly choosing the components of each codeword independently and identically from a judiciously chosen probability distribution . while this technique is powerful , the resulting codebooks do not exhibit any structure that may be of practical interest . one such desirable structure is linearity , which allows complexity reductions at the encoder and decoder by utilizing efficient algebraic processing techniques . further , in certain communication scenarios linear codes achieve higher rates than random code ensembles , as was shown by krner and marton @xcite for a distributed source coding problem . structured coding schemes have been widely studied in the literature , especially for communications in the presence of side information and in multi - terminal networks . for an overview of structured coding schemes we refer the reader to @xcite and the references therein . for communication in the wireless domain , structured codes can be obtained by choosing finite subsets of points from lattices @xcite . a lattice is an infinite discrete set of points in the euclidean space that are regularly arranged and are closed under addition . codes based on lattices , known as _ ( nested ) lattice codes _ or _ voronoi codes _ , are the analogues of linear codes in wireless communications . efficient lattice based strategies are known for a variety of communication scenarios , such as for achieving the capacity of the point - to - point additive white gaussian noise ( awgn ) channel @xcite , for dirty - paper coding @xcite , the wyner ziv problem @xcite and communication in relay networks @xcite , to name only a few . in this paper we present capacity - achieving lattice strategies for communication in gaussian broadcast channels where receivers have prior side information about the messages being transmitted . in particular , we assume that the transmitter is broadcasting @xmath4 message symbols @xmath5 from a finite field @xmath6 , of prime size @xmath0 , to all the receivers , and each receiver may have _ coded side information _ about the messages : the prior knowledge of the values of ( possibly multiple ) @xmath6-linear combinations of @xmath5 . the number of linear combinations available as side information and the coefficients of these linear combinations can differ from one receiver to the next . the capacity of this channel follows from the results of tuncel @xcite , where the achievability part requires the use of random gaussian codebooks . the broadcast channel with coded side information is an instance of a noisy wireless version of _ index coding _ index coding considers a noiseless broadcast link where each receiver demands a subset of the source messages and knows the values of some other subset as side information . the modification of the ( noiseless ) index coding problem in which the receivers have access to linear combinations of messages was studied recently in @xcite . the case of gaussian broadcast channel with coded side information is motivated by applications to multi - terminal communication networks . it is known that signal interference in wireless channels can be harnessed by decoding linear combinations of transmit messages instead of either treating interference as noise or decoding interference along with the intended message @xcite . when such a technique is used in a mutli - hop communication protocol , one encounters receivers that have coded side information obtained from transmissions in the previous phases . similarly , in a network that consists of both wired and wireless channels , the symbols received from wired links can be utilized as side information for decoding the wireless signals . if a linear network code is used in the wired part of the network , then the side information is in the form of linear combinations of the source messages . [ ex : relay_network ] consider a wireless network with two base stations @xmath7 and @xmath8 , that hold message symbols @xmath9 and @xmath10 , respectively . the base stations are required to broadcast @xmath9 and @xmath10 to four user nodes @xmath11 through the relay node @xmath12 , see fig . [ fig : relay_network ] . in the first phase of the protocol , @xmath7 and @xmath8 encode the data symbols @xmath9 and @xmath10 , and transmit the resulting codewords simultaneously . by using the decoding technique of compute - and - forward @xcite , @xmath13 reliably decodes some linear combination @xmath14 , @xmath15 , from the received noisy superposition of the two transmit signals . on the other hand , @xmath12 has a higher signal - to - noise ratio and successfully decodes both @xmath9 and @xmath10 by behaving as a multiple - access receiver . further , there is no signal interference at @xmath16 and @xmath17 , and these two nodes reliably decode @xmath9 and @xmath10 , respectively . we observe that the second phase of the protocol is a broadcast channel with coded side information at the receivers : the relay needs to jointly broadcast @xmath18 to four user nodes , the first three users @xmath19 have prior knowledge of the linear combinations @xmath20 , @xmath21 and @xmath14 , respectively , while the fourth user has no such side information . [ ex : overlay ] encodes @xmath5 and broadcasts wirelessly to the destination nodes @xmath22 in order to supplement the communication through the wired network . each destination receives ( possibly multiple ) @xmath6-linear combinations of the source symbols through a linear ( wireline ) network code and uses this information to decode the wireless signal broadcast by @xmath23.,width=288 ] assume a network of noiseless wired links in the form of a directed acyclic graph , where the source node @xmath23 desires to multicast @xmath4 independent messages @xmath24 to a set of destination nodes @xmath25 . the wireline network employs a traditional _ ( scalar ) linear network code _ @xcite , i.e. , the symbol transmitted on each outgoing edge of a node is an element of @xmath6 generated as a linear combination of the symbols received on its incoming edges . at every destination node @xmath26 , the decoder attempts to recover the @xmath4 message symbols from their @xmath6-linear combinations received on its incoming edges . recovery is possible if and only if the number of linearly independent equations available at @xmath27 is @xmath4 . it is known that the maximum number of linearly - independent equations that can be made available at @xmath27 is @xmath28 , where @xmath29 is the maximum number of edge - disjoint paths from @xmath23 to @xmath27 , see @xcite . it follows that multicasting is possible if and only if @xmath30 for every @xmath26 . now suppose there exist destination nodes with @xmath31 less than @xmath4 , i.e. , the communication demands are beyond the wireline network s capacity . a solution to meet the demands is to broadcast a wireless signal from the source to fill the capacity deficiency of the wired network , see fig . [ fig : overlay ] . at each destination , the @xmath6-linear combinations obtained from the wireline network serve as side information to decode the wireless broadcast signal . a special case of coded side information is the gaussian broadcast channel where each receiver has prior knowledge of the values of some subset of the @xmath4 messages . the known capacity - achieving coding schemes for this special case are based on random coding using i.i.d . ( independent and identically distributed ) codewords @xcite . constructions of binary codes for this channel were proposed in @xcite , explicit codes based on lattices were proposed in @xcite and codes based on quadrature amplitude modulation were constructed in @xcite . the objective of this paper is to prove that lattice codes achieve the capacity of gaussian broadcast channels with coded side information . we use the information - theoretic framework set by erez and zamir @xcite to this end . the proposed coding scheme uses an ensemble of construction a lattices and the decoding scheme involves _ algebraic binning _ @xcite where the receiver side information is used to expurgate the channel code and obtain a lower rate subcode . the set of linear equations available as side information may differ from one receiver to another , and hence , each receiver must employ a different binning scheme for the same channel code . the coding scheme ensures that the binning performed at each receiver produces a good lattice subcode of the transmitted code . following expurgation , each receiver decodes the channel output by minimum mean square error ( mmse ) scaling and quantization to an infinite lattice . the algebraic structure of the coding scheme facilitates the performance analysis by decomposing the original channel into multiple independent point - to - point awgn channels one corresponding to each receiver where each of the point - to - point awgn channels uses a lattice code for communication . unlike @xcite , where achievability in a point - to - point awgn channel was proved using error exponent analysis , we provide a direct proof based only on simple counting arguments . as a corollary to the main result , we obtain an alternative proof of the goodness of lattice codes in achieving the capacity of the point - to - point awgn channel . previous proofs of this result presented in @xcite also use ensembles of lattices obtained by applying construction a to random linear codes over a prime field @xmath6 ; see also @xcite . while @xcite used primes @xmath0 that were exponential in the code length @xmath1 , @xcite and @xcite improved this result to let @xmath0 grow as @xmath32 and @xmath33 , respectively . the corollary presented in this paper further improves these results by enabling a rate of growth of @xmath2 , for any choice of @xmath3 . the organization of this paper is as follows . we introduce the channel model in section [ sec : channel_model ] and review the relevant background on lattices and lattice codes in section [ sec : lattice_prelim ] . in section [ sec:3 ] , we state the main theorem , and describe the lattice code ensemble and encoding and decoding procedures . we prove the main theorem and state a few corollaries in section [ sec : proof_of_main ] , and finally , we discuss some concluding remarks in section [ sec : conclusion ] . _ notation : _ matrices and column vectors are denoted by bold upper and lower case letters , respectively . the symbol @xmath34 denotes the euclidean norm of a vector , and @xmath35 is the transpose of a matrix or a vector . the kronecker product of two matrices @xmath36 and @xmath37 is @xmath38 , @xmath39 is the @xmath40 identity matrix , and @xmath41 is the all zero matrix of appropriate dimension . the symbol @xmath42 denotes logarithm to the base @xmath43 and @xmath44 denotes logarithm to the base @xmath45 . the expectation operator is denoted by @xmath46 . the symbol @xmath47 denotes the elements in the set @xmath48 that do not belong to the set @xmath49 . we consider a ( non - fading ) gaussian broadcast channel with a single transmitter and finitely many receivers , where all terminals are equipped with single antennas . the transmitter operates under an average power constraint and the receivers are affected by additive white gaussian noise with possibly different noise powers . there are @xmath4 independent messages @xmath5 at the transmitter that assume values with a uniform probability distribution from a prime finite field @xmath6 . each receiver desires to decode all the @xmath4 messages while having prior knowledge of the values of some @xmath6-linear combinations of the messages @xmath5 . consider a generic receiver that has access to the values @xmath50 , @xmath51 , of the following set of @xmath52 linear equations @xmath53 we will denote this side information configuration using the matrix @xmath54 \in { \mathbb{f}_p}^{m \times k}$ ] , where each row of @xmath55 represents one linear equation . any row of @xmath55 that is linearly dependent on the other rows represents redundant information and can be discarded with no loss to the receiver side information , and hence , with no loss to system performance . hence , without loss in generality , we will assume that the rows of @xmath55 are linearly independent over @xmath6 , i.e. , @xmath56 , and @xmath57 . note that the values of @xmath55 and @xmath52 can be different across the receivers . a receiver with no side information is represented with an empty matrix for @xmath55 ( with @xmath58 ) . a receiver in the broadcast channel is completely characterized by its _ ( coded ) side information matrix _ @xmath55 and the variance @xmath59 of the additive noise . if we assume that the average transmit power at the source is @xmath60 , then the signal - to - noise ratio at this receiver is @xmath61 . we will denote a receiver by the pair @xmath62 , where @xmath55 is any matrix over @xmath6 with @xmath4 columns and linearly independent rows , and @xmath63 . note that _ uncoded _ side information , i.e. , the prior knowledge of the values of a size @xmath52 subset of @xmath5 , is a special case , and hence , is contained within the definition of our channel model . consider a source transmitting @xmath64 symbols , @xmath65 , from the finite field @xmath66 . a receiver that has prior knowledge of the value of @xmath10 has side information matrix @xmath67 this corresponds to the equation @xmath68 , and the number of linearly independent equations at this receiver is @xmath69 . now consider another receiver that has the knowledge of the values of the following three equations : @xmath70 , @xmath71 and @xmath72 . in matrix form , this side information is represented by @xmath73 where the three rows represent the three equations , in that order . the first row of this matrix is equal to the sum ( over @xmath74 ) of the second and third rows , and hence , the side information from the first equation is redundant and can be discarded . since the remaining two rows are linearly independent , the side information at this receiver can be represented by the following matrix that consists of these two rows , @xmath75 the number of linearly independent equations at this receiver is @xmath76 . from elementary linear algebra we know that if the values @xmath50 of @xmath52 linearly independent combinations of the variables @xmath5 are given , then the set of all possible solutions of @xmath77 is a coset of a @xmath78 dimensional linear subspace of @xmath79 . since the apriori probability distribution of @xmath5 is uniform , we conclude that , given the side information values @xmath50 , @xmath51 , the probability distribution of @xmath77 is uniform over this coset . using the fact that the number of elements in the coset is @xmath80 , we observe that the conditional entropy of @xmath77 given the side information is @xmath81 suppose we want to transmit , on the average , one realization of @xmath77 in every @xmath82 uses of the broadcast channel . the transmission rate of each message is @xmath83 b / dim ( bits per real dimension or bits per channel use ) . the information - theoretic result of @xcite , which is based on the average performance of random gaussian codebooks , shows that a transmission rate of @xmath84 is achievable if and only if the following condition is satisfied at every receiver @xmath62 of the broadcast channel , @xmath85 or equivalently , @xmath86 where @xmath87 . for the simplicity of exposition , we consider only the symmetric case where all the @xmath4 messages are required to be transmitted at the same rate @xmath84 . the general scenario , where the messages are of different rates , can be reduced to the symmetric case through _ rate - splitting _ : if there are @xmath88 messages with transmission rates @xmath89 , respectively , then by splitting each of these original sources into multiple virtual sources , one can generate a set of @xmath4 sources ( @xmath90 ) such that their rates @xmath91 are as close to each other as required . we will assume that the encoding at the transmitter is performed on a block of @xmath92 independent realizations of the @xmath4 message symbols , i.e. , the source jointly encodes @xmath4 message vectors @xmath93 . the transmitter uses an @xmath1-dimensional channel code @xmath94 together with a function @xmath95 to jointly encode the @xmath4 message vectors . the number of codewords in @xmath96 is @xmath97 , and we will assume that the codebook @xmath96 satisfies the per - codeword power constraint @xmath98 the average number of channel uses to transmit each realization of @xmath77 is @xmath99 . the resulting rate of transmission of each of the @xmath4 messages is @xmath100 the sum rate of all the messages is @xmath101 b / dim . of the broadcast channel uses its own side information @xmath102 to expurgate the channel code @xmath96 and obtain a subcode @xmath103 . note that the resulting subcodes can be different across the receivers . in order to achieve the capacity of the broadcast channel , we require that each of these expurgated codes be good for channel coding over the point - to - point awgn channel . , width=480 ] the side information at @xmath62 over a block of @xmath92 realizations of the @xmath4 message symbols is of the form @xmath104 , @xmath51 , where @xmath87 and @xmath105 . this side information allows the receiver to conclude that the transmitted codeword must belong to the following subcode of @xmath96 , @xmath106 the optimal decoder at @xmath62 decodes the channel output vector to the nearest codeword @xmath107 of this subcode , and the error probability at this receiver is the probability that the estimated message tuple @xmath108 is not equal to the transmit message @xmath109 . in order to achieve the optimal performance at a given receiver @xmath62 , we thus require that the expurgated code @xmath103 be a good channel code for the point - to - point awgn channel . in a broadcast channel with multiple receivers , the side information matrix @xmath55 can vary from one receiver to the next , and hence , the expurgated codes can be different at each receiver , see fig . [ fig : expurg ] . hence , a capacity - achieving broadcast channel code @xmath96 is such that the resulting expurgated code at every receiver is a good channel code for the awgn channel . we now briefly review the necessary background on lattices and lattice codes , and establish our notation and terminology . the material presented in this section consists of standard ingredients used in the literature , and is mainly based on @xcite . a _ lattice _ @xmath110 in @xmath1 dimensions is a discrete additive subgroup of @xmath111 . we will only consider lattices that have full - rank , i.e. , @xmath1-dimensional lattices of the form @xmath112 , where @xmath113 is known as the _ generator matrix _ of the lattice @xmath110 , and @xmath37 is of rank @xmath1 . the _ lattice quantizer _ corresponding to @xmath110 is the function @xmath114 given by @xmath115 where ties between competing lattice points @xmath116 are broken in a systematic way . the _ voronoi region _ @xmath117 of the lattice @xmath110 is the set of all points in @xmath111 that are mapped to @xmath118 by the function @xmath119 . the volume of the voronoi region , i.e. , @xmath120 , is denoted by @xmath121 . for any @xmath122 , @xmath123 is the set of all points in @xmath111 that are mapped to @xmath116 under @xmath119 , and it has the same volume @xmath121 as @xmath117 . for any two distinct lattice points @xmath124 , the sets @xmath125 and @xmath126 are disjoint . the _ modulo_-@xmath110 operation on @xmath127 is defined as @xmath128 { { \rm~mod~}}{\lambda}=\pmb{x}-q_{{\lambda}}(\pmb{x})$ ] , and denotes the error when @xmath129 is quantized to the nearest point in @xmath110 . the following identities are relevant @xmath130 { { \rm~mod~}}{\lambda}&\in { \mathcal{v}}({\lambda } ) \text { for all } { \pmb}{x } \in { \mathbb{r}}^n , \nonumber \\ [ { \pmb}{x}_1 + { \pmb}{x}_2 ] { { \rm~mod~}}{\lambda}&= \big [ \ , [ { \pmb}{x}_1 ] { { \rm~mod~}}{\lambda}+ { \pmb}{x}_2 \big ] { { \rm~mod~}}{\lambda}\text { for all } { \pmb}{x}_1,{\pmb}{x}_2 \in { \mathbb{r}}^n , \text { and } \label{eq : mod_is_distributive } \\ [ { \pmb}{x } ] { { \rm~mod~}}{\lambda}&= { \pmb}{0 } \text { if and only if } { \pmb}{x } \in { \lambda}. \label{eq : when_is_mod_zero}\end{aligned}\ ] ] we will denote the @xmath1-dimensional ball of radius @xmath131 with center @xmath132 as @xmath133 , i.e. @xmath134 we will denote the volume of a unit - radius ball in @xmath1 dimensions by @xmath135 . it follows that the volume of @xmath133 equals @xmath136 . the _ covering radius _ @xmath137 of the lattice @xmath110 is the radius of the smallest @xmath1-dimensional ball that contains @xmath117 , and it satisfies @xmath138 the _ effective radius _ @xmath139 of @xmath110 is the radius of the @xmath1-dimensional ball whose volume equals @xmath121 , and is given by @xmath140 evidently , @xmath141 for any lattice . rogers @xcite showed that for every dimension @xmath1 there exists a lattice @xmath110 such that @xmath142 where @xmath143 is a constant . note that the right hand side of the above inequality converges to @xmath60 as @xmath144 . a sequence of lattices of increasing dimension @xmath1 is said to be _ rogers - good _ if @xmath145 . rogers result shows that such a sequence exists ( see also @xcite ) . let @xmath146 and @xmath110 be two lattices such that @xmath147 . then @xmath146 is a _ sub - lattice _ of @xmath110 and forms a additive subgroup of @xmath110 . the index of @xmath146 in @xmath110 , i.e. , the size of the quotient group @xmath148 is given by @xmath149 each coset of @xmath146 in @xmath110 is represented by its unique coset leader in @xmath150 . hence , the additive group @xmath148 can be identified with the set @xmath151 , where the addition operation is performed modulo-@xmath146 . let @xmath152 be a fixed vector . a _ ( nested ) lattice code _ or a _ voronoi code _ @xmath153 is the set @xmath154 obtained by applying the @xmath155 operation on the points of the lattice translate @xmath156 . the code consists of all the points in @xmath156 that lie within the voronoi region of @xmath146 , i.e. , @xmath157 . the lattice @xmath146 is called the _ coarse lattice _ or the _ shaping lattice _ , @xmath110 is called the _ fine lattice _ or the _ coding lattice _ , and @xmath158 is the _ dither _ vector . the cardinality of this code is @xmath159 , and every codeword point @xmath160 satisfies @xmath161 . note that @xmath148 is a lattice code with zero dither . in this subsection we briefly describe the method proposed in @xcite to construct a pair @xmath147 of nested lattices , and recall its relevant properties . this construction uses a coarse lattice @xmath146 and a linear code @xmath162 to generate a fine lattice @xmath110 such that @xmath163 . let @xmath164 denote the natural map that embeds @xmath165 into @xmath166 . when applied to vectors , @xmath164 acts independently on each component of a vector . let @xmath167 be a linear code of rank @xmath168 , @xmath169 , @xmath170 where @xmath171 is the @xmath172 generator matrix with full column rank , and @xmath173 is the message encoded to @xmath162 . the set @xmath174 obtained by tiling copies of @xmath175 at every vector of @xmath176 is a lattice in @xmath111 and is known as the _ construction a _ lattice of the linear code @xmath162 @xcite . note that the number of points in @xmath177 contained in the voronoi region of the lattice @xmath176 is @xmath178 . we obtain @xmath110 by scaling down the construction a lattice by @xmath179 and transforming it by the generator matrix @xmath180 of @xmath146 @xmath181 since @xmath162 contains the all zero codeword , it follows that @xmath182 . we observe that applying the transformation @xmath183 to the lattice @xmath176 ( instead of the lattice @xmath177 ) generates @xmath146 ( instead of @xmath110 ) . hence , @xmath148 has the same algebraic structure as that of @xmath184 , which in turn , is equivalent to the linear code @xmath162 . in particular , @xmath185 encoded by @xmath162 and the points in the lattice code @xmath148 . this result , which is originally from ( * ? ? ? * lemma 5 ) , is proved below for completeness . [ lem : w_to_t ] the map @xmath186 { { \rm~mod~}}{\lambda_{\rm c}}$ ] is a bijection between @xmath187 and @xmath148 . from , we know that @xmath188 . hence , it only remains to show that no two distinct messages @xmath189 and @xmath190 are mapped to the same point in @xmath148 . assuming the contrary , we have @xmath191 { { \rm~mod~}}{\lambda_{\rm c}}= \left[{{\pmb}{b}_{\rm c}}p^{-1 } g({\pmb}{gw}_b)\right ] { { \rm~mod~}}{\lambda_{\rm c}}$ ] . using and , we obtain @xmath192 multiplying both sides by @xmath193 , we obtain @xmath194 . reducing this result modulo-@xmath0 , we have @xmath195 over @xmath6 . since this implies @xmath196 over @xmath6 while @xmath197 and @xmath171 has full column rank , we have arrived at a contradiction . in order to prove capacity achievability , we will rely on random coding arguments to show the existence of a good choice of @xmath171 . as in @xcite , we will assume that @xmath171 is a random matrix chosen with uniform probability distribution on @xmath198 . the following result is useful in upper bounding the decoding error probability over the ensemble of random codes . [ lem : t_is_uniform ] let @xmath199 be a given non - zero vector , and let @xmath171 be uniformly distributed in @xmath198 . then @xmath200 \ ! { { \rm~mod~}}{\lambda_{\rm c}}$ ] is uniformly distributed over @xmath201 , i.e. , over the lattice code @xmath202 . we will assume that the number of messages @xmath4 and a design rate @xmath84 are given , and show that there exist good lattice codes of sufficiently large dimension @xmath1 that encode @xmath4 messages over an appropriately chosen prime field @xmath6 at rates close to @xmath84 b / dim . in order to rigorously state the main result , we consider a fixed non - zero tolerance @xmath203 , and a free parameter @xmath3 that determines the value of the prime @xmath0 as a function of @xmath1 . [ thm : main ] let the number of messages @xmath4 , design rate @xmath84 , parameter @xmath3 and tolerance @xmath203 be given . for every sufficiently large dimension @xmath1 and any prime @xmath0 in the interval @xmath204 $ ] , there exists an @xmath1-dimensional lattice code @xmath205 that encodes @xmath4 message vectors over @xmath6 such that the rate of transmission of each message is at least @xmath206 b / dim and the probability of error at any receiver @xmath62 is less than @xmath207 if @xmath208 to prove theorem [ thm : main ] , we utilize the lattice code ensemble introduced in @xcite ( see section [ sec : prelim : lattices_from_codes ] of this paper ) . in the rest of this section we describe the construction of random lattice codes , and the encoding and decoding operations . we provide the proof of the theorem [ thm : main ] in section [ sec : proof_of_main ] . we require that the prime @xmath0 increase with the code dimension @xmath1 as @xmath209 for some fixed @xmath3 . from bertrand s postulate @xcite we know that for every @xmath210 there exists a prime in the interval @xmath211 $ ] . we will choose @xmath0 to be any prime in this interval , i.e. , @xmath212 . once @xmath0 is fixed , we choose @xmath92 as the largest integer that satisfies @xmath213 the left - hand side in the above inequality is the actual rate at which the lattice code encodes each message , while @xmath84 is the desired rate . the difference between the two is at the most @xmath214 which converges to @xmath215 as @xmath144 . hence , for a given @xmath216 and for all sufficiently large @xmath1 , the code rate is at least @xmath206 b / dim . from in section [ sec : prelim : lattices ] , we know that for a given @xmath203 and for all sufficiently large @xmath1 , there exists an @xmath1-dimensional lattice @xmath146 such that @xmath217 we will choose such a rogers - good lattice as @xmath146 , and scale it so that @xmath218 it follows that @xmath219 . using the definition of the effective radius , we arrive at the following lower bound on the volume of the voronoi region of @xmath146 @xmath220 the fine lattice is obtained by the construction of @xcite described in section [ sec : prelim : lattices_from_codes ] . the length of the linear code @xmath162 is @xmath1 , and its rank @xmath221 is the number of message symbols to be encoded by the lattice code . note that this requires that @xmath222 be true . using and the property @xmath223 , we have @xmath224 rearranging the terms in the above inequality , we obtain @xmath225 where the right - hand side converges to @xmath215 as @xmath144 . hence , for all sufficiently large @xmath1 , @xmath222 . if @xmath226 is the generator matrix of @xmath162 , then @xmath227 . we will choose @xmath171 uniformly random over @xmath228 matrices , resulting in a random ensemble of fine lattices @xmath110 . we will rely on random coding arguments to prove the existence of a translate @xmath158 such that the code @xmath205 performs close to capacity . we will assume that @xmath158 is distributed uniformly in @xmath150 and is chosen independently of @xmath171 . this random dither @xmath158 is usually viewed as a common randomness available at the transmitter and the receivers @xcite . note that @xmath229 . to a point in @xmath205.,width=480 ] we will now describe the encoding operation @xmath230 at the transmitter that maps the message vectors @xmath231 to a codeword @xmath232 . the encoder first concatenates the @xmath4 messages into the vector @xmath233 , encodes @xmath173 to a codeword in the linear code @xmath162 , and maps it to a point @xmath234 using construction a as follows @xmath235 { { \rm~mod~}}{\lambda_{\rm c}}.\ ] ] from the discussion in section [ sec : sub : choice_of_parameters ] , we know that @xmath236 , and hence , @xmath237 . finally , the transmit codeword @xmath238 is generated by dithering @xmath239 , @xmath240 { { \rm~mod~}}{\lambda_{\rm c}}= \left [ { { \pmb}{b}_{\rm c}}\ , p^{-1 } g\left ( { \pmb}{gw } \right ) - { \pmb}{d } \right ] { { \rm~mod~}}{\lambda_{\rm c}}.\ ] ] this sequence of operations is illustrated in fig . [ fig : encoder ] . note that since @xmath241 , each codeword @xmath238 satisfies @xmath242 , and hence , the power constraint @xmath243 it is straightforward to show that the dithering operation is a one - to - one correspondence between @xmath237 and @xmath232 . further , from lemma [ lem : w_to_t ] we know that is a bijection between the message space @xmath244 and the undithered codewords @xmath148 if @xmath171 is full rank . hence , to ensure that no two messages are mapped to the same codeword , we only require that the random matrix @xmath171 be full rank . it can be shown that ( see @xcite ) @xmath245 we will only require a relaxation based on the above inequality . from , for all sufficiently large @xmath1 , @xmath246 . similarly , for all large enough @xmath1 , we have @xmath247 , and hence , @xmath248 the receiver employs a two stage decoder : in the first stage the receiver identifies the subcode of @xmath205 corresponding to the available side information , and in the second stage it decodes the channel output to a point in this subcode . the side information at @xmath62 over a block of @xmath92 realizations of the @xmath4 messages is of the form @xmath249 the receiver desires to identify the set of all possible values of the message vector @xmath233 that satisfy . using the notation @xmath250 , the side information can be rewritten compactly in terms of @xmath173 and @xmath251 as @xmath252 where @xmath253 denotes the kronecker product of matrices and @xmath254 is the @xmath40 identity matrix over @xmath6 . observe that is an under - determined system of linear equations , and the set of solutions is a coset of the null space of @xmath255 . let @xmath256 be a rank @xmath257 matrix such that @xmath258 , i.e. , the columns of @xmath259 form a basis of the null space of @xmath260 . then the set of all solutions to is @xmath261 where @xmath262 is the coset leader . from , we conclude that the undithered codeword must be of the form @xmath263 \ ! { { \rm~mod~}}{\lambda_{\rm c}}\text { for some } { \pmb}{\tilde{w } } \in { \mathbb{f}_p}^{(k - m)\ell}.\ ] ] we will now use the property of @xmath164 that for any @xmath264 , @xmath265 . therefore , @xmath266 for some @xmath267 . using this in , we obtain @xmath268 \ ! { { \rm~mod~}}{\lambda_{\rm c}}\nonumber \\ = & \left[\ , { { \pmb}{b}_{\rm c}}p^{-1 } g\left ( { \pmb}{gv } \right ) + \left[{{\pmb}{b}_{\rm c}}p^{-1 } g\left({\pmb}{ga_s\tilde{w } } \right)\right ] \ ! { { \rm~mod~}}{\lambda_{\rm c}}+ \left[\ , { { \pmb}{b}_{\rm c}}{\pmb}{c } \ , \right]\ ! \ ! { { \rm~mod~}}{\lambda_{\rm c}}\ , \right ] { { \rm~mod~}}{\lambda_{\rm c}}\nonumber \\ = & \left[\ , { { \pmb}{b}_{\rm c}}p^{-1 } g\left ( { \pmb}{gv } \right ) + \left[{{\pmb}{b}_{\rm c}}p^{-1 } g\left({\pmb}{ga_s\tilde{w } } \right)\right ] \ ! { { \rm~mod~}}{\lambda_{\rm c}}\ , \right ] { { \rm~mod~}}{\lambda_{\rm c } } , \label{eq : subcode_of_valid_codewords}\end{aligned}\ ] ] where we have used , and the fact that @xmath269 . since the receiver knows @xmath262 , the component of @xmath239 unavailable from the side information is @xmath270 \ ! { { \rm~mod~}}{\lambda_{\rm c}}.\ ] ] let @xmath271 be the subcode of @xmath162 with generator matrix @xmath272 , and @xmath273 be the lattice obtained by applying construction a to @xmath274 and transforming it by @xmath183 , i.e. , @xmath275 using @xmath272 instead of @xmath171 in lemma [ lem : w_to_t ] , we see that @xmath276 and that is a one - to - one correspondence between @xmath277 and @xmath276 as long as @xmath272 is full rank . together with , , and , we conclude that the transmit vector @xmath238 belongs to the following lattice subcode of @xmath205 , @xmath278 \ ! { { \rm~mod~}}{\lambda_{\rm c}}~\big\vert~ { \pmb}{\tilde{t } } \in { \lambda_{{\pmb}{s}}}/{\lambda_{\rm c}}\right\}.\ ] ] the decoding problem at the second stage is to estimate @xmath279 , or equivalently @xmath280 , from the channel output . . the vector @xmath262 and the lattice @xmath273 are determined using the side information available at the receiver.,width=528 ] let the channel output at the receiver @xmath62 be @xmath281 , where @xmath282 is a gaussian vector with zero mean and variance @xmath59 per dimension . the received vector is scaled by the coefficient @xmath283 , resulting in @xmath284 this mmse pre - processing improves the effective signal - to - noise ratio of the system beyond the channel signal - to - noise ratio @xmath285 and allows the lattice decoder to perform close to capacity @xcite . let @xmath286 be the effective noise term in . using the facts that @xmath238 and @xmath282 are independent , @xmath287 , and @xmath282 has zero mean , we have @xmath288 where @xmath46 is the expectation operator . the choice of @xmath289 minimizes this upper bound and yields @xmath290 which is less than the gaussian noise power @xmath291 . in the rest of the paper we will assume that @xmath292 and use the notation @xmath293 the lower bound on signal - to - noise ratio can be rewritten in terms of @xmath294 as @xmath295 from , we know that @xmath296 for some @xmath297 . after mmse scaling , the decoder removes the contributions of the dither @xmath158 and the offset @xmath298 from @xmath299 to obtain @xmath300 the decoder proceeds by quantizing @xmath301 to the lattice @xmath273 and reducing the result modulo @xmath146 . if the noise @xmath302 is sufficiently ` small ' , then this sequence of operations will yield @xmath303 { { \rm~mod~}}{\lambda_{\rm c}}= \left [ q_{{\lambda_{{\pmb}{s}}}}({\pmb}{\tilde{t } } + { { \pmb}{\lambda}_{\rm c}}+ { \pmb}{z } ) \right ] { { \rm~mod~}}{\lambda_{\rm c}}= \left [ { \pmb}{\tilde{t } } + { { \pmb}{\lambda}_{\rm c}}\right ] { { \rm~mod~}}{\lambda_{\rm c}}= { \pmb}{\tilde{t}}.\end{aligned}\ ] ] given @xmath279 , the receiver uses to obtain the undithered codeword @xmath239 , and hence the message vector @xmath304 , as follows @xmath305 { { \rm~mod~}}{\lambda_{\rm c}}.\ ] ] to conclude , the decoder obtains the estimate @xmath306 of the undithered codeword @xmath239 from the received vector @xmath307 as @xmath308 { { \rm~mod~}}{\lambda_{\rm c}}+ { { \pmb}{b}_{\rm c}}p^{-1 } g({\pmb}{gv } ) \ , \big ] { { \rm~mod~}}{\lambda_{\rm c}}\\ & = \left [ \ , q_{{\lambda_{{\pmb}{s}}}}({\pmb}{y ' } ) + { { \pmb}{b}_{\rm c}}p^{-1 } g({\pmb}{gv } ) \ , \right ] { { \rm~mod~}}{\lambda_{\rm c}}\\ & = \left [ \ , q_{{\lambda_{{\pmb}{s}}}}\left(\alpha { \pmb}{y } - { { \pmb}{b}_{\rm c}}p^{-1 } g({\pmb}{gv } ) + { \pmb}{d}\right ) + { { \pmb}{b}_{\rm c}}p^{-1 } g({\pmb}{gv } ) \ , \right ] { { \rm~mod~}}{\lambda_{\rm c}},\end{aligned}\ ] ] which shows that the @xmath155 operation arising from can be ignored . the steps involved in the decoding operation are illustrated in fig . [ fig : decoder ] . note that the effective information vector @xmath280 is not encoded in the point @xmath309 , but is encoded in the coset @xmath310 . the error event for this decoder is @xmath311 , i.e. , @xmath312 , which is equivalent to @xmath313 . hence , a decoding error occurs if and only if @xmath302 is closer to a point in @xmath314 than any vector in the coarse lattice @xmath146 , i.e. , if and only if @xmath315 in this section we first state and prove two technical lemmas ( section [ sec : technical_lemma ] ) , use these lemmas to show that the error probability at a given fixed receiver @xmath62 is small ( section [ sec : error_at_single_receiver ] ) , and then complete the proof of the main theorem by showing that the error probability at every receiver of the broadcast channel is simultaneously small ( section [ sec : complete_the_proof ] ) . finally , we state some important corollaries of the main theorem ( section [ sec : corollaries ] ) . the first result , which is a direct generalization of ( * ? * lemma 1 ) and ( * ? ? ? * lemma 2.3 ) , gives an upper bound on the number of lattice points lying inside a ball . [ lem : counting ] for any @xmath132 , @xmath316 and any @xmath1-dimensional lattice @xmath146 , @xmath317 where @xmath135 is the volume of a unit ball in @xmath111 . let @xmath318 be the set of all points in @xmath111 that are mapped to one of the points in @xmath319 by the lattice quantizer @xmath320 . since @xmath321 is a union of the pairwise disjoint sets @xmath322 , @xmath323 , and since each of these sets has volume @xmath324 , we have @xmath325 using the fact that @xmath326 , we have the following @xmath327 where the last step follows from the triangle inequality . consequently , we have an upper bound on the volume of @xmath321 , @xmath328 . using this result with proves the lemma . as in @xcite , we will rely on the fact that , with very high probability , the norm of the noise @xmath302 is not much larger than @xmath329 . but unlike these previous works , we need to accommodate multiple receivers in the broadcast channel , and hence , we will require a slightly stronger result : the probability that the effective noise is large is exponentially small in @xmath1 . the proof of this result is given below . [ lem : noise_is_small ] let @xmath238 be uniformly distributed in @xmath150 and @xmath330 be any positive number . then @xmath331 we will prove for every fixed realization of @xmath238 in @xmath150 , which shows that the statement of the lemma is true for any distribution of @xmath238 on @xmath150 . in the rest of the proof we will assume that @xmath332 is an arbitrary fixed vector and @xmath282 is gaussian distributed . using @xmath333 , we have @xmath334 hence , we have the following upper bound @xmath335 from the definition of @xmath294 , we have @xmath336 . hence , the above upper bound corresponds to the event @xmath337 the event occurs _ only if _ at least one of the following two events occur @xmath338 therefore , @xmath339 . we will now individually upper bound @xmath340 and @xmath341 , and thereby complete the proof . a rearrangement of terms in yields @xmath342 . this is the probability that a gaussian vector with unit variance per dimension lies outside the sphere of squared radius @xmath343 . the following is a well known upper bound on this probability ( see @xcite ) @xmath344 the event @xmath345 is equivalent to @xmath346 . using @xmath289 , we can show that this is same as @xmath347 . since @xmath348 is a zero mean gaussian random variable with variance @xmath349 , we have @xmath350 where @xmath351 is the gaussian tail function . using @xmath352 and the chernoff bound @xmath353 , we arrive at @xmath354 this completes the proof . in this subsection we derive an upper bound on the decoding error probability @xmath355 at a receiver @xmath62 when averaged over the ensemble of lattice codes generated by choosing @xmath171 uniformly over @xmath356 and @xmath158 uniformly over @xmath150 . the following result from @xcite , known as the _ crypto lemma _ , captures an important characteristic of random dithering . [ lem : crypto ] let @xmath357 be any random vector . if @xmath158 is independent of @xmath239 and is uniformly distributed over @xmath150 , then @xmath358 \ ! { { \rm~mod~}}{\lambda_{\rm c}}$ ] is independent of @xmath239 and uniformly distributed over @xmath150 . the property that the transmit vector @xmath238 is statistically independent of @xmath239 implies that the effective noise @xmath359 is independent of the transmit message . this facilitates the error probability analysis through the observation that the error event is statistically independent of @xmath279 . for distinct messages @xmath280 to be mapped to distinct points @xmath279 , we require that @xmath272 be full rank . since @xmath360 is full rank , this is same as requiring that @xmath171 be full rank . apart from the event @xmath361 , we assume that the decoder declares an error whenever the event @xmath362 occurs . hence , the error probability @xmath355 at the receiver @xmath62 satisfies @xmath363 from , we already know that @xmath364 is exponentially small in @xmath1 . using the given design tolerance @xmath207 , we set @xmath365 , which is positive if @xmath203 . let @xmath366 be the radius of the typical noise vector and @xmath367 . then , @xmath368 lemma [ lem : noise_is_small ] provides an exponential upper bound on @xmath369 . in the following theorem we show that @xmath370 is also exponentially small in @xmath1 . the proof of this result uses the technique of @xcite to bound the number of lattice points lying in an @xmath1-dimensional ball . let @xmath371 be the least noise standard deviation @xmath372 among all the receivers . the knowledge of @xmath373 enables us to derive an upper bound on error probability which is independent of the side information matrix @xmath55 and the noise standard deviation @xmath372 as long as the condition on the minimum required @xmath374 is satisfied . the following result , which follows immediately from , will be used in the proof of theorem [ thm : pe_one_receiver ] , @xmath375 [ thm : pe_one_receiver ] for any receiver @xmath62 with @xmath376 , and for all large enough @xmath1 , @xmath377 when averaged over the ensemble of random lattice codes . from , we note that the decoder is in error when @xmath302 is closer to some coset @xmath378 , with @xmath379 and @xmath380 , than any point in @xmath146 . the number of competing cosets is @xmath381 , and we index them using the non - zero vectors @xmath382 . to each @xmath383 , we associate the coset corresponding to the coset leader @xmath384 \ ! { { \rm~mod~}}{\lambda_{\rm c}}.\ ] ] since @xmath171 is random , the coset leader @xmath385 associated with a given @xmath383 is a random vector . given that @xmath386 and @xmath387 , the euclidean distance between @xmath302 and @xmath146 is at the most @xmath388 . hence , for an error event to occur , there must exist a coset @xmath378 at a distance less than @xmath389 from @xmath302 , i.e. , @xmath390 . indexing the cosets by @xmath383 , we have @xmath391 where the second inequality follows from union bound , and the last inequality follows from the observation @xmath392 where @xmath393 is the indicator function . note that the expectation operation in is with respect to the random vector @xmath385 as well as the effective noise @xmath302 . the matrix @xmath360 has full column rank , and hence , @xmath394 for every @xmath395 . using and applying lemma [ lem : t_is_uniform ] , we see that @xmath385 is uniformly distributed in @xmath396 . further , from lemma [ lem : crypto ] the distribution of @xmath385 is independent of @xmath302 . hence , the probability mass function of @xmath385 equals @xmath397 over every element of the set @xmath398 . using this result , we obtain @xmath399 where the last equality follows from the fact that the set of cosets @xmath400 form a partition of @xmath401 . since the number of competing @xmath383 is less than @xmath402 , and @xmath403 , @xmath404 using lemma [ lem : counting ] , we bound the number of lattice points inside the ball @xmath405 , and obtain @xmath406 using the bounds @xmath407 , from ; @xmath408 , from ; @xmath409 , from ; and the relations @xmath241 , @xmath366 , and @xmath410 , we obtain @xmath411 since @xmath412 , we have @xmath413 to complete the analysis , we use the bound @xmath223 and the property that @xmath414 from below as @xmath415 . hence , for all @xmath3 , @xmath416 we want an upper bound on @xmath370 which is exponentially small in @xmath1 , and hence , we require an upper bound on @xmath417 which is exponentially smaller than @xmath418 , say , for instance , @xmath419 . using , we observe that @xmath420 for @xmath3 , @xmath203 and for all sufficiently large @xmath1 . hence , @xmath421 , and consequently , @xmath422 for all large enough @xmath1 . we will now combine the result of theorem [ thm : pe_one_receiver ] with and , and upper bound the error probability @xmath355 at the receiver @xmath62 as @xmath423 using theorem [ thm : pe_one_receiver ] , lemma [ lem : noise_is_small ] and , we obtain @xmath424 for sufficiently large @xmath1 . note that @xmath425 , and @xmath426 as long as @xmath365 is positive . consequently , the parameter @xmath427 is positive , and the value of each of the terms on the right - hand side of is at the most @xmath428 . hence the error probability at the receiver @xmath62 can be upper bounded as @xmath429 for all sufficiently large @xmath1 . we remark that the minimum required value of @xmath1 for this upper bound to hold is a function of only @xmath207 , @xmath430 and @xmath373 , and is independent of the side information matrix @xmath55 . the bound shows that the error probability for a fixed side information matrix @xmath55 , averaged over the random code ensemble , tends to @xmath215 as the code dimension increases . hence , there exists a choice of lattice code ( which is chosen for the given side information matrix @xmath55 ) with a small error probability at this receiver . we want to prove a slightly stronger result , viz . , there exists a lattice code such that the decoding error probability for every possible side information matrix @xmath55 is small as long as the receiver @xmath374 is large enough . in order to prove this result , we consider a hypothetical broadcast network that consists of one receiver for each possible choice of the matrix @xmath55 . note that two distinct values of the matrix @xmath55 that have identical row space constitute equivalent receiver side information configurations . hence , it is enough to consider a broadcast channel that consists of one receiver corresponding to each possible subspace of @xmath79 , where the dimension of the subspace can be between @xmath215 and @xmath431 . a subspace of dimension @xmath52 , @xmath432 , can be mapped to a @xmath433 matrix whose rows form a basis of the subspace . this map embeds the set @xmath434 of all non - equivalent choices of side information matrix @xmath55 into @xmath435 , which is the set of all matrices over @xmath6 with @xmath4 columns and at the most @xmath431 rows . hence , the number of receivers @xmath436 can be upper bounded as latexmath:[\[\begin{aligned } \label{eq : set_s } note that the number of receivers in this hypothetical broadcast channel grows polynomially in @xmath1 , while the probability of error at each receiver decays exponentially . this observation lets us complete the proof of theorem [ thm : main ] by strengthening the result of the previous subsection . we assume that each receiver @xmath62 , @xmath438 , satisfies the lower bound on @xmath374 and outputs an estimated message vector @xmath439 using its own channel observation . we say that the broadcast network is in error if any of the receivers commits a decoding error . using a union bound argument and the upper bounds and , we see that the network error probability @xmath440 averaged over the random ensemble of lattice codes satisfies @xmath441 which tends to @xmath215 as @xmath1 becomes arbitrarily large . hence , for a sufficiently large @xmath1 , there exists a lattice code such that the network error probability is as small as desired . in particular , this implies that there exists a choice of lattice code such that the decoding error probability at every receiver @xmath62 , @xmath438 , is simultaneously small . this completes the proof of the main theorem . using standard arguments based on markov inequality @xcite , we show that almost all codes from the random lattice code ensemble yield a small error probability . in order to prove this , it is sufficient to show that for almost all lattice codes the network error probability is small over the hypothetical broadcast channel that consists of one receiver for each possible side information matrix . for a given dimension @xmath1 , all the lattice codes in the random code ensemble use the same coarse lattice @xmath146 , but differ in the choice of the fine lattice @xmath110 and/or the dither vector @xmath158 . let @xmath442 denote the network error probability for a given choice of @xmath443 in the hypothetical broadcast channel . if @xmath110 and @xmath158 are chosen randomly , then @xmath444 is a random variable . from , we know that the expected value of @xmath444 , which is equal to the average network error rate @xmath440 , is small . suppose we want a lower bound on the fraction of random codes with error probability at the most @xmath445 . using markov inequality , we have @xmath446 it follows that , asymptotically , for almost all choices of the fine lattice @xmath110 and dither vector @xmath158 , the resulting lattice code @xmath205 provides an exponentially small error probability in the broadcast channel , i.e. , @xmath447 our model of broadcast channel includes as a special case the receiver with no side information , i.e. , @xmath55 is the empty matrix and @xmath448 . the decoder for such a receiver uses the @xmath449 identity matrix for @xmath360 and the all zero vector for @xmath262 , see . specializing the main theorem for a single receiver with @xmath58 , we immediately deduce that the ensemble of random lattice codes achieves the capacity of the single - user awgn channel . it is well known that ( nested ) lattice codes , and lattice constellations in general , can achieve the capacity of the point - to - point awgn channel @xcite . our corollary to the main theorem provides an alternate proof of this result which is based only on simple counting arguments . the proof technique presented in this paper relies on lattices obtained by applying construction a to random linear codes over a large enough prime field @xmath6 . this technique was introduced by loeliger in @xcite and used in @xcite to prove the goodness of lattice codes in awgn channel . each of these results requires a different rate of growth for the prime @xmath0 and places different requirements on the characteristics of the coarse lattice @xmath146 . the following are some of the properties that have been used in the literature : * _ rogers - good : _ the ratio of covering radius @xmath450 to the effective radius @xmath451 of the lattice must be close to @xmath60 , see . such a lattice is also said to be _ good for covering_. * _ mse - good : _ the value of the lattice parameter @xmath452 , known as the _ normalized second moment _ , is close to @xmath453 , see @xcite . every rogers - good lattice is also mse - good , and hence , this is a weaker requirement . * _ poltyrev - good : _ such a lattice , when used as an infinite constellation , achieves the capacity of an awgn channel in which the transmitter has no power constraints @xcite . these lattices are resilient against additive gaussian noise . the achievability result of @xcite requires @xmath146 to be simultaneously rogers - good and poltyrev - good , and uses @xmath454 ( exponential in @xmath1 ) . the random code ensemble of @xcite uses an mmse - good lattice for @xmath146 , lets @xmath0 grow as @xmath32 , and can accommodate a wide class of channel noise statistics , including white gaussian noise . the code construction of @xcite requires @xmath0 to be at least @xmath33 , needs no dithering operation , i.e. , uses @xmath455 , but is known to achieve capacity only if @xmath456 . in comparison , our proof method allows @xmath0 to grow as @xmath2 , for any fixed @xmath3 , and holds for all @xmath457 , while requiring that @xmath146 be rogers - good . we have showed that lattice codes are optimal for common - message broadcast in gaussian channels where receivers have side information . we used random lattice ensembles obtained by applying construction a to linear codes over appropriately large prime fields @xmath6 . the growth of @xmath0 as @xmath2 does not necessarily pose a limitation in communication applications . for instance , in the relay network of example [ ex : relay_network ] , the first phase of the protocol , namely compute - and - forward @xcite , only requires that @xmath458 as @xmath144 , which can be met by our scheme by choosing @xmath459 . similarly , with example [ ex : overlay ] , where the broadcast signal supplements a wired multicast network , it is known that wireline network codes meeting the @xmath31 bound exist over every large enough finite field @xcite . hence , we can choose @xmath1 to be sufficiently large to simultaneously optimize both the wired and wireless parts of the hybrid network . on the other hand , designing lattice strategies for a fixed size of the finite field , especially sizes that are powers of two , may have greater practical significance . the capacity of the gaussian broadcast channel with receiver side information under general message demands , such as with private message requests , is known only for some special cases @xcite . these known achievability schemes utilize random gaussian codebooks together with dirty - paper and superposition coding . it will be interesting to examine if the lattice structure of the codes proposed in this paper can be exploited to derive new capacity results beyond the known cases . o. ordentlich and u. erez , `` a simple proof for the existence of `` good '' pairs of nested lattices , '' in _ electrical electronics engineers in israel ( ieeei ) , 2012 ieee 27th convention of _ , nov . 2012 , pp . 112 . n. di pietro , `` on infinite and finite lattice constellations for the additive white gaussian noise channel , '' theses , universit de bordeaux , jan . 2014 . [ online ] . available : https://tel.archives-ouvertes.fr/tel-01135575 m. wilson , k. narayanan , h. pfister , and a. sprintson , `` joint physical layer coding and network coding for bidirectional relaying , '' _ ieee trans . inf . theory _ , 56 , no . 11 , pp . 56415654 , nov . 2010 . s. el rouayheb , a. sprintson , and c. georghiades , `` on the index coding problem and its relation to network coding and matroid theory , '' _ ieee trans . inf . theory _ 56 , no . 7 , pp . 31873195 , jul . k. shum , m. dai , and c. w. sung , `` broadcasting with coded side information , '' in _ personal indoor and mobile radio communications ( pimrc ) , 2012 ieee 23rd international symposium on _ , sep . 2012 , pp . 8994 . y. ma , z. lin , h. chen , and b. vucetic , `` multiple interpretations for multi - source multi - destination wireless relay network coded systems , '' in _ proc . ieee 23rd int . personal indoor and mobile radio communications ( pimrc ) _ , sep . 2012 , pp . 22532258 . l. natarajan , y. hong , and e. viterbo , `` lattice index coding for the broadcast channel , '' in _ information theory workshop ( itw ) , 2015 ieee _ , apr . 2015 , pp . 15 , extended version available online , arxiv:1410.6569 . b. asadi , l. ong , and s. johnson , `` the capacity of three - receiver awgn broadcast channels with receiver message side information , '' in _ proc . information theory ( isit ) _ , jun . 2014 , pp . 28992903 . j. sima and w. chen , `` joint network and gelfand - pinsker coding for 3-receiver gaussian broadcast channels with receiver message side information , '' in _ proc . information theory ( isit ) _ , jun . 2014 , pp . 8185 .

garoufalidis and levine defined a filtration for 3-manifolds equipped with some degree 1 map (-homology equivalence ) to a fixed 3-manifold and showed that there is a natural surjection from a space of-decorated graphs to the graded quotient of the filtration over $ ] . in this paper , we show that in the case of the surjection of garoufalidis levine is actually an isomorphism over . for the proof , we construct a perturbative invariant by applying fukaya s morse homotopy theoretic construction to a local system of the quotient field of . the first invariant is an extension of the casson invariant to-homology equivalences to the 3-torus . the results of this paper suggest that there is a highly nontrivial equivariant quantum invariants for 3-manifolds with . ( 2016 )

1608.08462

in @xcite , we constructed a perturbative invariant of 3-manifolds with @xmath8 by modifying fukaya s morse homotopy theoretic invariant ( @xcite ) . in this paper , we extend the invariant of @xcite to 3-manifolds with higher first betti numbers and see that it is effectively utilized in the study of finite type invariants of 3-manifolds defined by garoufalidis and levine ( @xcite ) . a theory of finite type invariant of 3-manifolds was first developed by ohtsuki in @xcite for integral homology 3-spheres . ohtsuki s finite type invariant was defined by means of a filtration on the module spanned by diffeomorphism classes of integral homology 3-spheres . ohtsuki s filtration was defined with respect to surgeries on algebraically split framed links in @xmath9 . it has been shown that the graded quotient of his filtration is isomorphic over @xmath5 to certain space of trivalent graphs ( @xcite ) , with the help of the lmo invariant ( @xcite ) . garoufalidis and levine defined in @xcite a filtration for 3-manifolds equipped with some degree 1 map ( @xmath0-homology equivalence ) to a fixed 3-manifold . they gave an upper bound for the graded quotient of their filtration by the space of @xmath10-decorated graphs . to see that this correspondence is an isomorphism , it is important to find a perturbative invariant that is a counterpart to the lmo invariant . some perturbative invariants were found for 3-manifolds with @xmath8 in @xcite . we construct a perturbative invariant for @xmath0-homology equivalences to the 3-torus via morse homotopy theory and by utilizing it we show that when the target manifold is the 3-torus @xmath11 , the space of @xmath10-decorated graph is naturally isomorphic to the graded quotient in garoufalidis levine s filtration . in this paper , we mainly consider the filtration of @xcite defined by surgery on @xmath12-links ( or graph claspers , @xcite ) . we also give a formula relating the value of our first invariant @xmath13 for the connected sum of @xmath11 with an integral homology sphere @xmath14 to the casson invariant of @xmath14 . first , we recall the definitions of @xcite . we fix an oriented closed 3-manifold @xmath1 and let @xmath2 . a degree 1 map @xmath15 from another oriented closed 3-manifold @xmath16 is said to be a _ @xmath0-homology equivalence _ if the induced map @xmath17 is an isomorphism . here , @xmath18 are the @xmath10-coverings of @xmath19 . two @xmath0-homology equivalences @xmath15 and @xmath20 are _ diffeomorphically equivalent _ if there exists a diffeomorphism @xmath21 such that @xmath22 is homotopic to @xmath23 . let @xmath24 be the set of all diffeomorphism equivalence classes of @xmath0-homology equivalences @xmath15 . in the following , we assume for simplicity that @xmath25 . in @xcite , it was shown that when @xmath25 a @xmath0-homology equivalence @xmath15 can be represented uniquely by a nullhomotopic framed link in @xmath1 is determined . ] . moreover , in @xcite , a `` surgery obstruction map '' @xmath26 to certain semi - group @xmath27 was defined by assigning a matrix of @xmath0-valued linking numbers ( or @xmath10-equivariant linking number ) of the framed link . . an element of @xmath28 is given by a _ @xmath10-algebraically split framed link _ , namely , a framed link such that its @xmath0-linking matrix is a diagonal matrix all of whose diagonal entries are @xmath30 and that its components are nullhomotopic . let @xmath31 be the vector space over @xmath5 spanned by the set @xmath28 . in @xcite , it was shown that any element of @xmath28 can be obtained from @xmath1 by surgery on a @xmath12-link . we consider the filtration of @xmath31 given by surgery on @xmath12-links . for the terminology related to @xmath12-links ( or graph - claspers ) , we refer the reader to @xcite . let @xmath32 be a @xmath12-link consisting of disjoint @xmath12-graphs in @xmath16 with only nullhomotopic leaves and let @xmath33=\sum_{g'\subset g}(-1)^{|g'|}m^{g'},\ ] ] where @xmath34 denotes the map @xmath35 obtained by surgery on @xmath36 . namely , by surgery a 4-cobordism @xmath37 between @xmath34 and @xmath16 is obtained and the identity map on @xmath16 extends over it uniquely up to homotopy , and then compose with the @xmath0-homology equivalence @xmath15 . let @xmath38 ( @xmath39 ) be the subspace of @xmath31 spanned by @xmath40 $ ] for @xmath12-link @xmath41 with @xmath42 disjoint @xmath12-graphs . this defines a descending filtration on @xmath31 : @xmath43 . a linear map @xmath44 to a vector space @xmath45 is said to be a _ finite type invariant of type @xmath42 _ if @xmath46 . we call a pair @xmath47 of an abstract , vertex - oriented , edge - oriented trivalent graph @xmath48 and a map @xmath49 a _ @xmath10-decorated graph_. we will denote @xmath47 also by @xmath50 or @xmath51 ( @xmath52 ) . we define the _ degree _ of a @xmath10-decorated graph as the number of vertices . let @xmath53 be the vector space over @xmath5 spanned by @xmath50 of degree @xmath42 quotiented by the relations as , ihx , orientation reversal , linearity and holonomy ( figure [ fig : relations ] ) . when @xmath54 , we call @xmath47 a _ monomial _ @xmath10-decorated graph . a monomial @xmath10-decorated graph @xmath47 determines uniquely a homotopy class @xmath55 of a map @xmath56 . we choose a ribbon structure on @xmath48 that is compatible with the vertex - orientation , choose an embedding of the ribbon graph into @xmath1 that represents @xmath55 and replace the image with a set @xmath57 of @xmath42 disjoint @xmath12-graphs . by assigning @xmath58 $ ] to @xmath50 , a well - defined linear map @xmath59 is obtained . in @xcite , it was shown that this map is surjective$]-coefficients . ] . when @xmath42 is odd , @xmath60 and hence @xmath61 . the main theorem of this paper is the following . [ thm : main ] when @xmath4 , the map @xmath62 is an isomorphism . in the rest of this paper we prove theorem [ thm : main ] . in section [ s : perturbative ] , we define a perturbative invariant @xmath63 for @xmath0-homology equivalence . the way of construction of the invariant is almost the same as @xcite . in section [ s : z - fti ] , we show that @xmath63 is a finite type invariant and compute its values on @xmath64 . the main theorem follows as a corollary of this result and of the surjectivity of @xmath65 of @xcite . the outline of the proof is analogous to that of @xcite . instead of modifying chains , we modify gradients of morse functions to localize the counts of graphs in an alternating sum . in section [ s : example ] , we relate the value of @xmath66 for an integral homology sphere @xmath14 to the casson invariant of @xmath14 , as a straightforward example . throughout this paper , manifolds and maps between them are assumed to be smooth . for a gradient @xmath67 of a morse function , we denote by @xmath68 ( resp . @xmath69 ) the ascending manifold ( resp . descending manifold ) of a critical point @xmath70 of @xmath67 . we denote by @xmath71 ( resp . @xmath72 ) the usual compactification of @xmath68 ( resp . @xmath69 ) into a smooth manifold with corners ( e.g. , @xcite ) . we will often abbreviate @xmath73 etc . as @xmath74 etc . for simplicity . by a ribbon graph , we mean a compact surface equipped with a handle structure with only 0- and 1-handles . in the following , let @xmath4 , @xmath75 and @xmath76 $ ] . let @xmath77 be the field of fractions @xmath78 of @xmath79 . we fix connected oriented 2-submanifolds @xmath80 of @xmath1 so that they form a basis of @xmath81 . by means of these data , we define the holonomy of a path in @xmath1 as follows . let @xmath82 be one of intervals @xmath83 $ ] , @xmath84 , @xmath85 $ ] , @xmath86 ( @xmath87 ) and consider a piecewise smooth path @xmath88 such that @xmath89 intersects each @xmath90 transversally in finitely many points . in this case , let @xmath91 ( @xmath92 ) and define @xmath93 for a @xmath89 with disconnected domain , we define @xmath94 as the product of @xmath95 of each components . for a @xmath0-homology equivalence @xmath15 and a morse function @xmath96 on @xmath16 , a chain complex @xmath97 is defined as follows . take a riemann metric @xmath98 on @xmath16 and consider the gradient @xmath67 of @xmath99 . here , we assume without loss of generality that the critical values of the critical points of @xmath99 are all different . the gradient @xmath67 is morse smale for generic metric @xmath98 , namely , ascending manifolds and descending manifolds of @xmath67 are transversal . in this case , there are finitely many ( unparametrized ) flow - lines between a critical point of index @xmath100 and a critical point of index @xmath101 . for each such flow - line @xmath102 , we assign its weight @xmath103 , where the sign is the same one as given in @xcite . for a critical point @xmath70 of @xmath99 of index @xmath100 and a critical point @xmath104 of @xmath99 of index @xmath101 , we define @xmath105 as the sum of the weights of flow - lines between @xmath70 and @xmath104 . let @xmath106 be the set of critical points of @xmath99 of index @xmath100 and let @xmath107 . we define the boundary @xmath108 by @xmath109 @xmath110 forms a chain complex and its homology is identified with @xmath111 as a @xmath79-module . next , we put @xmath112 . since @xmath77 is a flat @xmath79-module ( e.g. , ( * * theorem 3.2.2 ) ) , we have @xmath113 for every @xmath79-module @xmath45 . hence by the universal coefficient theorem ( ( * ? ? ? * theorem vi.3.3 ) ) , we have @xmath114 since @xmath22 is a @xmath0-homology equivalence , we have @xmath115 . since the action of @xmath10 on this module is trivial , we have @xmath116 . namely , @xmath117 is acyclic . since the chain complex @xmath118 of based free @xmath77-modules is acyclic , it follows that @xmath119 is acyclic too , by knneth theorem ( ( * ? ? ? * theorem vi.3.1a ) ) . since @xmath120 is a cycle of degree 0 , there exists an element @xmath121 of degree 1 such that @xmath122 ( @xmath123 is called a _ combinatorial propagator _ ) . this part is almost a copy from @xcite . let @xmath125 be a sequence of morse functions and let @xmath126 be the gradient of @xmath127 . we consider an edge - oriented trivalent graph with its sets of vertices and edges labelled and with @xmath128 vertices and @xmath129 edges . by the labelling @xmath130 of @xmath48 , we identify edges with numbers . choose some of the edges and split each chosen edge into two arcs . we attach elements of @xmath131 on the two 1-valent vertices ( _ white - vertices _ ) that appear after the splitting of the @xmath100-th edge . we call such obtained graph a _ @xmath132-graph _ ( @xmath133 , see figure [ fig : cgraph ] ) . if a @xmath132-graph is obtained from a connected trivalent graph , then we call it a _ primitive _ @xmath132-graph . a @xmath132-graph has two kinds of `` edges '' : a compact edge , which is connected , and a separated edge , which consists of two arcs . we call ( 3-valent ) vertices that are not white vertices _ black vertices_. if @xmath134 ( resp . @xmath135 ) is the critical point attached on the input ( resp . output ) white vertex of a separated edge @xmath100 , we define the degree of @xmath100 by @xmath136 . we define the degree of a compact edge @xmath100 by @xmath137 . we define the degree of a @xmath132-graph by @xmath138 . we say that a continuous map @xmath139 from a @xmath132-graph @xmath48 to @xmath16 is a _ flow - graph _ if it satisfies the following conditions ( see figure [ fig : flow - graph ] ) . 1 . a critical point @xmath134 attached on the @xmath100-th edge is mapped by @xmath139 to @xmath134 in @xmath16 . the restriction of @xmath139 to each edge of @xmath48 is a smooth embedding and at each point @xmath140 of the @xmath100-th edge that is not on a white vertex , the tangent vector of @xmath139 at @xmath140 ( chosen along the edge orientation ) is a positive multiple of @xmath141 . for a @xmath132-graph @xmath48 , let @xmath142 ( @xmath143 ) be the set of all flow - graphs from @xmath48 to @xmath16 . by extracting black vertices , a natural map from @xmath142 to the configuration space @xmath144 of ordered tuples of @xmath145 points is defined . it follows from a property of the gradient that this map is injective . this induces a topology on the set @xmath142 . [ lem:0-mfd ] if @xmath146 and @xmath98 are generic , then for a @xmath132-graph @xmath48 with @xmath147 , the space @xmath142 is a compact 0-dimensional manifold . moreover , this property can be assumed for all @xmath132-graphs with @xmath145 black vertices simultaneously . when the assumption of lemma [ lem:0-mfd ] is satisfied , we may define an orientation of @xmath142 by the same way as @xcite . here , we count points of @xmath142 with weights in @xmath148 as follows . @xmath149 we take the sign @xmath150 as the one determined by the orientation of @xmath139 . we define an action of a monomial @xmath151 on a graph as follows . @xmath152 the right hand side is a @xmath10-decorated @xmath132-graph . by extending this by @xmath5-linearity , the formal linear combination @xmath153 of @xmath10-decorated @xmath132-graphs is defined . we denote by @xmath154 the space of @xmath77-decorated graphs , defined by replacing @xmath6 in the definition of @xmath10-decorated graph with @xmath77 . let @xmath155 be a sequence of combinatorial propagators for @xmath133 . then we define @xmath156 by @xmath157 here , the sum is taken over all primitive @xmath132-graphs @xmath48 with @xmath147 , and @xmath158 is defined as follows . for simplicity , we assume that the labels for the separated edges in a @xmath132-graph @xmath159 are @xmath160 . let @xmath161 be the critical points on the input and output of the @xmath100-th edge , respectively and let @xmath162 be the coefficient of @xmath134 in @xmath163 . then @xmath164 is the equivalence class in @xmath154 of a @xmath10-decorated graph obtained by identifying each pair of the two white vertices of the separated edges in @xmath165 the definition of @xmath166 can be generalized to graphs with other degrees in the same manner . [ lem : l ] @xmath167 does not depend on the choices of @xmath168 and of the hypersurfaces @xmath169 within their oriented bordism classes . we shall define @xmath170 by adding a correction term to @xmath167 . take a compact oriented 4-manifold @xmath37 with @xmath171 and with @xmath172 . by the condition @xmath172 , the outward normal vector field to @xmath16 in @xmath173 can be extended to a nonsingular vector field @xmath174 on @xmath37 . let @xmath175 be the orthogonal complement of the span of @xmath174 . then @xmath175 is a rank 3 subbundle of @xmath176 that extends @xmath177 . take a sequence @xmath178 of generic sections of @xmath175 so that @xmath179 is an extension of @xmath180 . we define @xmath181 \in \cala_{2n}(\wlambda).\ ] ] the sum is taken over all primitive @xmath132-graphs with @xmath145 vertices and with only compact edges . here , @xmath182 is the moduli space of affine graphs in the fibers of @xmath175 whose @xmath100-th edge is a positive scalar multiple of @xmath179 . the number @xmath183 is the count of the signs of the affine graphs that are determined by transversal intersections of some codimension 2 chains in a configuration space bundle over @xmath37 . see @xcite for detail . by a similar argument as in @xcite , it can be shown that @xmath184 , where @xmath185 is the constant given in @xcite , does not depend on the choices of @xmath37 , @xmath186 and the extension @xmath187 of @xmath188 . we define @xmath189 by the following formula . @xmath190 [ thm:1 ] @xmath191 is an invariant of the diffeomorphism equivalence class of @xmath0-homology equivalence @xmath15 and of oriented bordism classes of @xmath80 . proof of theorem [ thm:1 ] can be done by using cerf theory and is almost the same as ( * ? ? ? * theorem 3.3 ) . the invariance under a homotopy of @xmath99 follows by the graph relations . next , we extend the invariant to possibly disconnected graphs . for each @xmath192 , we fix a sequence rational numbers @xmath193 ( @xmath48 is a labelled edge - oriented connected trivalent graph of degree @xmath128 ) so that @xmath194=\mu_{2n}$ ] . let @xmath48 be a not necessarily primitive @xmath132-graph and let @xmath195 be the set of connected components of @xmath48 . for a sequence @xmath196 of gradient on @xmath16 , let @xmath197 be the subsequence of @xmath188 restricted to the edge - labels for @xmath198 . let @xmath199 be the restriction of @xmath187 . let @xmath200 be the number of black vertices of @xmath198 . we define @xmath201 by the following formula . @xmath202 then we have @xmath203 . by means of these symbols , we define @xmath204 where the sum is taken for all @xmath132-graphs @xmath48 ( not necessarily primitive ) of degree @xmath128 . ] @xmath205 is an invariant of the diffeomorphism equivalence class of @xmath0-homology equivalence @xmath15 and of oriented bordism classes of @xmath80 . since the bifurcations of @xmath206 are exactly of the same kinds as those of @xmath191 , the invariance of @xmath205 follows from the proof of invariance of @xmath191 . in the following , we abbreviate @xmath205 as @xmath207 . we remark that @xmath207 depends not only on the diffeomorphism type of @xmath16 but also of the choice of a degree 1 map @xmath15 . in this section , we prove the following theorem , which implies that @xmath63 induces the inverse of @xmath65 . [ thm : eval ] let @xmath192 . 1 . @xmath63 is a finite type invariant of type @xmath128 in the sense of garoufalidis levine s filtration @xmath208 . 2 . let @xmath50 be a monomial @xmath10-decorated graph of degree @xmath128 . the following identity holds . @xmath209))=[\gamma(\alpha)].\ ] ] theorem [ thm : main ] follows as a corollary to theorem [ thm : eval ] and a result of @xcite . by considering similarly as the main theorem of @xcite , it turns out that @xmath210 is universal among @xmath5-valued finite type invariants . for @xmath211 , let @xmath212 be a @xmath12-link consisting of @xmath12-graphs with only nullhomotopic leaves . let @xmath213 be closed regular neighborhoods of @xmath214 respectively that are compact manifolds with boundaries . take a morse function @xmath215 $ ] on @xmath216 satisfying the following conditions . * @xmath217 . * the set of critical points of @xmath218 consists of one critical point of index 0 and three critical points of index 1 . we extend the morse function @xmath219 on @xmath220 over whole of @xmath1 to obtain a morse function @xmath221 ( figure [ fig : h - adapted ] ) . here , we choose @xmath99 so that @xmath222=v_1\cup\cdots\cup v_m$ ] . moreover , let @xmath223 be a sequence of morse functions that are obtained from @xmath99 by small perturbations so that they satisfy the genericity of lemma [ lem:0-mfd ] , and let @xmath224 be their gradients . we may also assume that @xmath225 and @xmath226=v_1\cup\cdots\cup v_m$ ] . [ lem : sigma ] we may assume that @xmath80 are all disjoint from @xmath227 by modifying @xmath80 within their oriented bordism classes suitably . if @xmath90 intersects a @xmath12-graph @xmath228 , then we may arrange that @xmath90 does not intersect @xmath228 by a bordism modification ( figure [ fig : leaf - sigma ] ) . the bordism may be taken to be disjoint from other @xmath12-graphs . in the following , we assume the condition of lemma [ lem : sigma ] . although the result of the bordism modification of @xmath90 may have self - intersections , such a modification of @xmath90 does not change the value of @xmath63 and there is no problem . we see that after modifying gradients for @xmath229 in @xmath230 the types of the flow - graphs that contribute to the alternating sum will be restricted to special ones . as is well - known , a @xmath12-surgery in @xmath216 can be realized by a torelli surgery along @xmath231 . namely , there is a diffeomorphism @xmath232 that represents an element of the torelli group and the surgery is obtained by cutting @xmath1 along @xmath231 and gluing @xmath216 back after applying @xmath233 to @xmath231 . after this surgery , a morse function @xmath234 is induced from @xmath99 . a morse function @xmath235 on the result of surgery on a @xmath12-link @xmath236 is induced in a similar way . let @xmath237 denote the gradient for @xmath238 . the manifold @xmath239 can be considered as the one obtained from @xmath1 by replacing a cylindrical neighborhood of @xmath231 in @xmath216 with the mapping cylinder of @xmath233 and similarly @xmath240 is obtained by iterating such replacements . let @xmath241 denote a @xmath0-homology equivalence @xmath242 obtained by projecting the mapping cylinder suitably onto @xmath231 . let @xmath243 be the critical points of the morse function @xmath215 $ ] of index 1 . let @xmath244 be the part of @xmath239 corresponding to @xmath216 . we also denote by @xmath243 the critical points of the induced morse function @xmath245 $ ] of index 1 . let @xmath246 ( @xmath247 ) be the intersection of @xmath248 with @xmath216 and let @xmath249 ( @xmath247 ) be the intersection of @xmath250 with @xmath244 . @xmath251 @xmath252 and @xmath253 @xmath252 both forms a disjoint triple of simple closed curves . note that @xmath251 and @xmath254 may intersect and they are bordant in @xmath231 . let @xmath255 be a bordism between @xmath251 and @xmath253 in @xmath231 whose image is not the whole of @xmath231 . [ lem : ribbon - graph ] the gradients of @xmath229 and the bordism @xmath255 can be chosen so that they satisfy the following . 1 . there exists an embedded ribbon graph @xmath256 and @xmath255 is a smaller ribbon graph immersed in @xmath257 . 2 . let @xmath258 be a handle filtration associated with the ribbon structure of @xmath257 . then the image of @xmath255 in each 1-handle of @xmath257 consists of parallel copies of ( not necessarily disjoint ) bands . the 0- and 1-handles of @xmath257 can be taken as arbitrarily small tubular neighborhoods of their cores . namely , for every @xmath259 , we may arrange that the diameters of 0-handles are less than @xmath260 and that the thickness of 1-handles are less than @xmath261 . we say that such a sequence @xmath262 is _ coherent_. put @xmath263 ( @xmath247 ) . then @xmath264 is nullhomologous in @xmath231 . take a minimal morse function @xmath265 such that the maximal point of @xmath266 is disjoint from every @xmath264 . now we define a filtration @xmath258 by the morse smale filtration for the gradient of @xmath266 up to dimension 1 . we modify @xmath264 by applying the negative gradient flow @xmath267 for @xmath266 . for a sufficiently large @xmath268 , the curve @xmath269 concentrates on a small neighborhood of the 1-skeleton of the cellular decomposition of @xmath231 with respect to the gradient of @xmath266 . moreover , since @xmath264 is nullhomologous , the image of @xmath269 in 1-handles of @xmath257 consist of parallel copies of pairs of two arcs with opposite orientations . the intersections between curves and self - intersections are all squeezed into 0-handles . then it will be easy to extend each curve @xmath264 to a nullbordism in @xmath257 so that the extension is an immersed ribbon graph as desired . all the modifications above can be realized by isotopies of @xmath231 and the isotopies can be obtained by modifying the gradient of @xmath218 . we remark that the curves @xmath255 for different @xmath270 may intersect each other . let @xmath271 be the 2-cycle in the closed 3-manifold @xmath272 defined by @xmath273 we denote the class of @xmath271 in @xmath274 by @xmath275 . [ lem : aaa=1 ] @xmath276 . we assume that the mapping cylinder @xmath277 for the torelli surgery is included in a small neighborhood of @xmath231 in @xmath244 . let @xmath278 be a level such that * all the critical points of @xmath218 and of @xmath279 lie in @xmath280 and in @xmath281 respectively and * @xmath282 $ ] . let @xmath283 $ ] and @xmath284 $ ] . consider the closed 3-manifold @xmath285\cup_\partial -(\mu_i^{g_i})^{-1}[\kappa,0],\ ] ] where @xmath286 and @xmath287 and @xmath288 are glued together by the natural identification of @xmath289 $ ] and @xmath290 $ ] . then @xmath291 is the mapping torus of @xmath233 . the 2-chains @xmath292 represent 2-cycles in @xmath291 . since @xmath293 $ ] ( @xmath247 ) are mutually disjoint , we have @xmath294 according to ( * ? ? ? * second definition , p.170 ) , the value of the first johnson homomorphism for @xmath233 is @xmath295\wedge [ \gamma_2^{(i)}]\wedge [ \gamma_3^{(i)}]\in \textstyle\bigwedge^3 h_1(\partial v_i),\ ] ] where @xmath296 is a 1-cycle of @xmath231 that are homologous to @xmath297 . this must agree with @xmath298\wedge [ \gamma_2^{(i)}]\wedge [ \gamma_3^{(i)}]$ ] for the @xmath12-surgery . hence we have @xmath299 . here we make the following assumption . [ hyp : a3=1 ] the indices of @xmath300 are chosen so that it is compatible with the cyclic order of @xmath300 determined from the vertex - orientation of the @xmath12-graph @xmath301 . moreover , we orient the 2-cycles so that @xmath302 . [ lem : morse - complex - invariant ] if @xmath262 is coherent , then the map @xmath303 induced by the @xmath0-homology equivalence @xmath304 is a chain isomorphism . since the sets of critical points of @xmath99 and @xmath238 are the same , the underlying @xmath77-modules can be taken to be the same one . hence it suffices to check that the boundary operator is invariant under the surgery on @xmath301 . the flow - lines that may change the count of flow - lines are those from critical points outside @xmath216 to critical points inside @xmath216 . hence the change of the count under surgery is given by the intersection of the curves @xmath305 and @xmath306 . generically , @xmath305 intersects transversally with only edges of the ribbon graph of lemma [ lem : ribbon - graph ] . since it always intersects both sides of each band simultaneously , the intersection number cancels in pair . under the isomorphism of lemma [ lem : morse - complex - invariant ] , we may identify all the morse complexes for @xmath307 ( @xmath236 ) . accordingly , the combinatorial propagators may be taken to be a common one . [ lem : occupied ] if @xmath308 is coherent for each @xmath309 and if the parameter @xmath260 in lemma [ lem : ribbon - graph ] is sufficiently small , then we have the following . 1 . @xmath310)$ ] agrees with the alternating sum of @xmath311 for @xmath58 $ ] ( we denote this value by @xmath312)$ ] ) . the flow - graphs that may contribute to @xmath312)$ ] are such that for each @xmath100 there are exactly one black vertex in @xmath216 or @xmath244 . ( we say that such a flow - graph _ occupies _ @xmath41 . ) let @xmath37 be a compact 4-manifold with @xmath313 considered in the definition of the correction term for @xmath170 and let @xmath314 be a sequence of sections of @xmath175 that extends @xmath188 on @xmath315 . there exists a compact 4-manifold @xmath316 and a sequence of sections @xmath317 of @xmath318 such that @xmath319 , @xmath320 and that @xmath317 agrees with @xmath314 on @xmath321 . by the additivity of the signature , the signature is unchanged under attaching @xmath316 to @xmath37 along @xmath216 . moreover , by connected summing several @xmath322 ( @xmath323 is the kummer @xmath323 surface , @xmath324 ) and @xmath325 ( @xmath326 ) to @xmath316 , we may assume that @xmath327 without changing the signature . by the assumption of @xmath316 , the term @xmath328 in @xmath329 does not change under the surgery . hence the terms of the form @xmath328 can be factored out as a constant @xmath330 for some @xmath331 . then the problem is reduced to the case of less @xmath42 without terms of the form @xmath328 . in the rest of the proof , we consider the alternating sum @xmath332 of the part of @xmath333 that does not have terms of @xmath334 . the first 2 terms @xmath335 in @xmath336 can be considered as counting flow - graphs in @xmath1 or @xmath240 ( @xmath236 ) or in @xmath175 . if a flow - graph of @xmath48 does not touch neither @xmath216 nor @xmath244 nor @xmath318 , it is counted as the same value before and after the surgery , and it is cancelled in the alternating sum @xmath332 . thus , only the flow - graphs from @xmath48 that touch @xmath216 or @xmath244 or @xmath318 for each @xmath100 may contribute to the alternating sum @xmath332 . if a flow - graph @xmath139 with @xmath128 black vertices touches @xmath216 or @xmath244 but no black vertices of @xmath139 lie there , then there are white vertices of @xmath139 in @xmath216 or @xmath244 . let @xmath70 be one of such white vertices . if @xmath337 , then the flow - graph @xmath139 can be interpreted as a 0-chain in @xmath338 . this shows that the count of @xmath139 does not change under the surgery on @xmath301 , and that @xmath139 does not contribute to the alternating sum @xmath332 . if @xmath339 , then the flow - graph @xmath139 , viewed as a 0-chain in @xmath340 , can be interpreted as the transversal intersection of the ascending manifold of @xmath70 and a piecewise smooth 1-chain of @xmath341 . the change of the count of @xmath139 before and after the surgery on @xmath301 is then the transversal intersection of the locus of the gradient flow from the boundary of the ribbon graph @xmath255 with the 1-chain . since the locus from the ribbon graph can be made arbitrarily thin , the transversal intersections always occur in pairs and the intersection number at the pair is 0 . namely , the flow - graphs that does not have any black vertices in neither @xmath216 nor @xmath244 nor @xmath318 for each @xmath100 does not contribute to the alternating sum @xmath332 . therefore , the only flow - graphs that may contribute to the alternating sum @xmath332 are those having a black vertex in either @xmath216 , @xmath244 or @xmath318 for each @xmath100 . under the assumption @xmath211 , such a situation can be achieved only for the case where @xmath342 and where there are exactly one black vertex in @xmath216 or @xmath244 for each @xmath100 . if there is a black vertex in @xmath318 , then the whole of a connected component of @xmath48 is mapped to @xmath318 . hence we do not need to consider such a case . namely , only the case where @xmath342 and where there is exactly one black vertex in @xmath216 or @xmath244 for each @xmath100 is left . the contribution of such a case agrees with @xmath312)$ ] . this follows immediately from lemma [ lem : occupied ] . [ cor : y - disjoint ] if @xmath342 , then only the flow - graphs that is the disjoint union of @xmath128 @xmath12-shaped flow - graphs may contribute nontrivially to @xmath312)$ ] . by lemma [ lem : occupied ] , a flow - graph @xmath139 that may contribute to @xmath312)$ ] occupies @xmath41 . thus there is no compact edge in @xmath139 connecting two black vertices . namely , every edge of @xmath139 must be separated edges . here , we shall give a preliminary for the proof of theorem [ thm : eval ] ( 2 ) . let @xmath343 be the complex of piecewise smooth singular chains with coefficients in @xmath5 . in the following , we identify @xmath344 with the image of the natural embedding @xmath345 defined by the cw - structure of @xmath346 that is determined by the morse smale gradient . @xmath347 is a chain equivalence . let @xmath70 be a critical point of @xmath215 $ ] of index 1 and let @xmath348 . @xmath349 corresponds to a leaf of a @xmath12-graph @xmath301 . [ lem : p - bound ] for a combinatorial propagator @xmath123 of the acyclic complex @xmath350 , we have @xmath351 . by definition of combinatorial propagator , we have @xmath352 . we shall define the intersection form with holonomy between base pointed cells . for two morse smale gradients @xmath353 on @xmath16 , we consider the associated cw - structures . we choose a base point of each @xmath354-dimensional cell as the critical point of index @xmath354 on the cell . let @xmath355 be two cells of the cw - structures for @xmath353 respectively such that @xmath356 and that they intersect transversally at finitely many points . let @xmath357 denote the base points of @xmath355 respectively . for each point @xmath140 of @xmath358 , choose a path @xmath359 in @xmath360 that goes from @xmath361 to @xmath140 and a path @xmath362 in @xmath363 that goes from @xmath140 to @xmath364 , and we define the weight @xmath365 of the intersection point @xmath140 as @xmath366 . here the sign is the sign of the intersection . then we define @xmath367 this does not depend on the choices of @xmath368 whereas it depends on the choices of @xmath80 . next , we shall define the triple intersection form with holonomies . for three morse smale gradients @xmath369 on @xmath16 , we consider the associated cw - structures . let @xmath370 be three cells of the cw - structures for @xmath369 respectively such that the sum of their codimensions is 3 and that they intersect `` transversally '' , that is , the orthogonal complements of @xmath371 ( @xmath92 ) span the direct sum in @xmath177 . let @xmath372 be the base points of @xmath370 respectively . for each point @xmath140 of @xmath373 , let @xmath374 be a flow - line of @xmath375 that connects @xmath376 and @xmath140 , and we define the weight @xmath365 of @xmath140 as @xmath377 . here the sign is the sign of the intersection . then we define @xmath378 by extending all the above by @xmath79-linearity , the intersection forms with holonomies among two or three @xmath79-chains are defined . these forms are defined only for transversal chains . if the chains are not transversal , then we consider the intersection form of them as the one defined after perturbing the gradients slightly so that the transversality is satisfied , not affecting the arguments in other part . in the following , we take for simplicity a special combinatorial propagator @xmath123 obtained as follows . since the abelian cover @xmath379 is homotopy equivalent to a point , for the morse complex @xmath380 of @xmath381 , there exists a degree 1 endomorphism @xmath382 such that @xmath383 where @xmath384 is the projection onto the free @xmath79-submodule generated by the critical point @xmath385 of @xmath99 that is the unique minimum . we choose a @xmath77-linear map @xmath386 that almost extends @xmath387 . namely , since @xmath385 is a boundary in @xmath350 , there exists a 1-chain @xmath388 in @xmath389 such that @xmath390 . for each critical point @xmath70 of @xmath99 , we define @xmath123 as @xmath391 then @xmath123 is a combinatorial propagator for @xmath350 . for @xmath392 , @xmath393 is a @xmath0-chain . let @xmath394 be a critical point of @xmath395 $ ] ( @xmath396 ) of index 1 and let @xmath397 . take @xmath123 as above and put @xmath398 . since @xmath399 represents leaves of @xmath12-graphs in @xmath41 , these are nullhomotopic in @xmath381 by the assumption of @xmath12-graphs . for two component link @xmath400 , their equivariant linking number is defined by @xmath401 where @xmath402 are lifts of @xmath399 in @xmath381 , and @xmath403 is the usual linking number in @xmath379 . the indeterminacy of changing the lifts @xmath402 and @xmath80 satisfying the condition of lemma [ lem : sigma ] is multiplication by elements of @xmath10 . thus @xmath404 is well - defined modulo @xmath10 . [ lem : g(p)-lk ] put @xmath405 . then the following identity holds . @xmath406 by lemma [ lem : sigma ] , both @xmath399 are both disjoint from @xmath80 , and thus @xmath402 are disjoint from every lifts of @xmath80 . since @xmath407 is a @xmath0-chain , @xmath408 bounds a @xmath409-chain corresponding to @xmath407 in @xmath381 . noting that @xmath410 does not intersect every lifts of @xmath80 , the following identity follows immediately . @xmath411 since @xmath412 , the lemma follows . the value of the left hand side of the equation in lemma [ lem : g(p)-lk ] does not depend on the choice of the lifts @xmath402 and is determined uniquely with respect to @xmath80 . [ lem : g(c)-cycle ] the two chains of @xmath413 and @xmath414 that are induced from @xmath407 induces relative 2-cycles of @xmath415 and @xmath416 respectively , and they induce the same 1-cycle on @xmath417 . hence , by gluing the two relative cycles along the common boundary , a 2-cycle of @xmath418 is obtained . that @xmath407 induces the same 1-cycle on @xmath417 is because the gradient of @xmath99 above the level @xmath419 does not change under the surgery . by lemma [ lem : p - bound ] , @xmath407 has no boundary in both @xmath413 and @xmath414 , and hence induces relative 2-cycles of @xmath415 and @xmath416 respectively . [ lem : g(p)-class ] let @xmath420 be the critical points of @xmath395 $ ] of index 1 and let @xmath421 be their poincar lefschetz dual . then the relative homology class in @xmath415 induced from @xmath407 is given by the following formula . @xmath422+\wlk_\pi(c , c_2^{(\ell)})[p_2^{(\ell)*}]+\wlk_\pi(c , c_3^{(\ell)})[p_3^{(\ell)*}]\ ] ] this follows immediately from lemma [ lem : g(p)-lk ] . in @xmath423 , the duals @xmath421 are explicitly given by @xmath424 respectively . as a @xmath12-link @xmath425 , we take the @xmath12-link obtained from an embedding @xmath56 with ribbon structure that represents a monomial @xmath10-decorated graph @xmath50 of degree @xmath128 . then @xmath426)$ ] is given by the equivalence class of @xmath58 $ ] in @xmath427 . we shall compute the value of @xmath63 for this alternating sum . by lemma [ lem : occupied ] , it suffices to compute the value of @xmath311 for the alternating sum . let @xmath428 be an edge - oriented , vertex - oriented labelled trivalent graph . let @xmath429 be the undecorated @xmath132-graph of degree @xmath128 obtained from @xmath428 by replacing every edge with separated edges . then @xmath429 is a disjoint union of @xmath12-shaped components . let @xmath430 ( @xmath431 ) be the @xmath12-component of @xmath429 that includes the @xmath100-th black vertex . we consider each of the two parts @xmath432 ( half - edges , @xmath433 is on the output - side , @xmath434 is on the input - side ) in the @xmath270-th separated edge as a degree 1 object , and consider the exterior product @xmath435 . by permuting the half - edges in this product , we obtain an expression of the form @xmath436 ( @xmath257 is the wedge of the three half - edges incident to the @xmath100-th black vertex ) , and then @xmath257 determines a vertex - orientation of the @xmath100-th black vertex . if the half - edges @xmath437 around the @xmath100-th black vertex of @xmath429 gives @xmath257 in this order , and if the critical points that are attached to the white vertices of @xmath437 are @xmath438 respectively , then we write @xmath439 ( @xmath440 ) . also , if @xmath441 , then we write @xmath442 . the contribution in @xmath443 of the flow - graphs from graphs of the form @xmath429 is @xmath444 we denote this by @xmath445 . each term @xmath446 is the triple product with holonomies for some three 2-chains formed along the vertex - orientation . for example , if among the three white vertices in @xmath447 the one with @xmath134 attached is input and the other two with @xmath448 attached are outputs , then @xmath449 . the term @xmath450 is a linear combination of graphs with @xmath129 inputs and @xmath129 outputs . let @xmath451 be the critical points that are attached on the input white vertices of @xmath452 , and let @xmath453 be those on the corresponding outputs , so that @xmath454 and @xmath455 are attached on a single separated edge . put @xmath456 then @xmath445 is rewritten as @xmath457 . \end{split}\ ] ] here , @xmath458 is extended to chains by @xmath79-linearity and @xmath459 is the trivalent graph that is obtained from the @xmath132-graph @xmath429 by identifying the two white vertices in each separated edge . the terms in the sum in the right hand side of ( [ eq : xr ] ) are linear combinations of @xmath459 together with holonomy data . the holonomy data is given by a product of @xmath128 terms like @xmath460 , where the product is taken by multiplying the holonomies of the two chains for each separated edge so that the coefficient belongs to @xmath148 . by corollary [ cor : y - disjoint ] , for the coherent families @xmath308 ( @xmath461 ) of morse smale pairs , we have @xmath462)=\frac{1}{2^{3n}(2n)!(3n)!}\sum_h\sum_{g'\subset g}(-1)^{|g'|}i(\vec{\xi}^{g'},h).\ ] ] here , the first sum is taken for edge - oriented , vertex - oriented labelled trivalent graphs @xmath428 . let @xmath463 denote the part in the alternating sum + @xmath464 consisting of terms of flow - graphs such that for each @xmath100 the @xmath100-th black vertex lies in @xmath465 or @xmath466 . [ lem : s_h ] the following identity holds . @xmath467\ ] ] here , @xmath468 , and for @xmath469 or @xmath470 , @xmath471 the critical points @xmath472 are taken so that they correspond to @xmath473 . by ( [ eq : xr ] ) , @xmath474 is @xmath475.\ ] ] the right hand side of the formula of the lemma is obtained by rewriting this expression . for @xmath463 , similar formula will be obtained by replacing @xmath476 in the formula of lemma [ lem : s_h ] with @xmath477 . the indices of the critical points @xmath478 in the formula of lemma [ lem : s_h ] may be restricted as follows . [ lem : y - restrict ] after perturbing the gradients @xmath479 without affecting the previous assumptions , we may restrict the @xmath12-shaped flow - graphs in corollary [ cor : y - disjoint ] to those such that the critical points attached on input white vertices are all of index 2 and that the critical points attached on output white vertices are all of index 1 . let @xmath139 be a flow - graph in @xmath240 for some @xmath236 with @xmath128 black vertices consisting only of @xmath12-shaped components and suppose that @xmath139 occupies @xmath41 . we may assume that @xmath139 has no output white vertex on which a critical point of index 2 is attached . indeed , if @xmath139 had such a white vertex @xmath480 , then there must be a black vertex @xmath140 that is on the half - edge incident to @xmath480 . since @xmath140 is mapped to some @xmath216 , @xmath480 is mapped to @xmath216 , too . this is a contradiction because @xmath218 does not have critical point of index 2 . moreover , we may assume that @xmath139 has no input white vertex on which a critical point of index 3 is attached . indeed , if @xmath139 had such a white vertex , then the input vertex is paired with an output white vertex on which a critical point of index 2 is attached . by the assumption of @xmath139 , such a critical point of index 2 must be in some @xmath413 or @xmath414 , and this is a contradiction . we define the index of a @xmath12-shaped component @xmath82 in @xmath139 as @xmath481 if the indices of the critical points attached on the input white vertices of @xmath82 are @xmath482 and if the indices of the critical points attached on the output white vertices of @xmath82 are @xmath483 . by the condition that the dimension of the moduli space is 0 , the possibilities for the indices of @xmath12-components in @xmath139 are as follows . @xmath484 among these , @xmath485 satisfy the condition of the lemma . further , by the result of the previous paragraph , we do not need to consider the cases @xmath486 . hence it suffices to check that the remaining cases @xmath487 does not contribute to the alternating sum @xmath312)$ ] . let @xmath488 be a @xmath12-shaped component in @xmath139 of index @xmath489 . suppose that @xmath488 has a black vertex in @xmath216 or @xmath244 and that the two input half - edges with critical points of indices @xmath490 are labelled by @xmath491 respectively . by ( [ eq : xr ] ) , we may consider that on the input white vertex of @xmath488 of index 2 , the chain @xmath393 for some critical point @xmath70 of @xmath492 of index 1 with negative critical value is attached , extending @xmath139 to chains . let @xmath104 be the critical point of @xmath493 of index 1 that is attached on an input white vertex of @xmath488 . according to lemma [ lem : g(c)-cycle ] , @xmath393 induces a relative 2-cycle in @xmath494 , which we now denote by @xmath495 . then the intersection between @xmath495 and @xmath496 corresponds to @xmath488 . if @xmath104 lies in @xmath216 , then by lemma [ lem : g(p)-lk ] , we have @xmath497 . since this value does not change under the surgery on @xmath301 , the flow - graph @xmath139 does not contribute to the alternating sum @xmath312)$ ] in this case . if @xmath104 lies outside @xmath216 , then a leg of @xmath496 converges to a critical point @xmath498 of index 0 in @xmath216 or @xmath244 by assumption . we remark that since @xmath499 induces a relative 1-cycle of @xmath500 , the value of @xmath501 above will not change by an addition of the boundary of a 3-chain in @xmath216 that does not touch @xmath502 , to @xmath495 . now we assume that the gradients for morse functions @xmath229 are as follows . we assume that the support of the torelli surgery for the surgery on @xmath301 is included in a ribbon graph @xmath503 in @xmath231 . then we may assume that for a real number @xmath278 with small absolute value , both the manifolds @xmath230 and the gradients for @xmath229 agree on the complement of @xmath504 $ ] , respectively . here , the direct product structure @xmath505 $ ] is the one generated by the gradient @xmath218 , and we consider that @xmath244 is obtained from @xmath216 by surgery within @xmath504 $ ] . furthermore , we may assume that this property is satisfied for every pairs @xmath506 . by perturbing the gradient of @xmath507 in @xmath508 , we may assume that @xmath509 does not touch @xmath504 $ ] in both @xmath216 and @xmath244 . note that this perturbation is independent of the gradients in @xmath216 s and does not affect all the previous vanishing results . by lemma [ lem : g(c)-cycle ] again , @xmath393 induces a relative 2-cycle of @xmath510 , which we now denote by @xmath511 , whose boundary agrees with that of @xmath495 . we may assume that the relative cycles @xmath512 are both disjoint from @xmath513 . by modifying @xmath511 by adding the boundary of a 3-chain in @xmath244 that does not touch @xmath514 , we may assume that @xmath495 and @xmath511 agree on the complement of @xmath504 $ ] . see figure [ fig : modify - sigma - g ] . under all the above assumptions , @xmath501 is invariant under the surgery on @xmath301 , and hence the flow - graph @xmath139 does not contribute to the alternating sum @xmath312)$ ] in this case , too . next , let @xmath515 be a @xmath12-shaped component in @xmath139 of index @xmath516 . suppose that @xmath515 has a black vertex in @xmath216 or @xmath244 and that the input and output half - edges with critical points of index 1 are labelled by @xmath491 respectively . @xmath515 corresponds to the intersection in @xmath216 or @xmath244 of the ascending manifold of a critical point @xmath517 of @xmath493 of index 1 in @xmath216 or @xmath244 , with the descending manifold of a critical point @xmath104 of @xmath492 of index 1 . by the same argument as above , the difference of @xmath518 before and after the surgery on @xmath301 can be squeezed into @xmath504 $ ] for a ribbon graph @xmath503 . since @xmath496 can be taken so that it does not touch @xmath504 $ ] , the intersection @xmath519 does not change under the surgery , and the flow - graph @xmath139 does not contribute to the alternating sum @xmath312)$ ] in this case . the gradients may be perturbed so that every @xmath12-shaped components in any @xmath139 with indices @xmath489 and @xmath516 vanish simultaneously . recall that @xmath520 . let @xmath521 be the basis of @xmath522 given by the three 1-handles of @xmath216 . the following identity holds . @xmath523 we consider the difference @xmath524 of triple products such that the index of @xmath525 satisfies the condition of lemma [ lem : y - restrict ] . by the same reason as in the proof of lemma [ lem : occupied ] , this difference may not vanish only when there are flow - graphs from @xmath526 with the @xmath100-th vertex included in @xmath216 or @xmath244 . there exist 2-chains @xmath527 ( @xmath528 ) of the form @xmath529 or @xmath530 such that the difference @xmath524 can be written as follows . @xmath531 this is the count of triple intersections in @xmath532 with weights . if @xmath533 is a descending manifold , then the boundaries of @xmath534 and @xmath535 agree on @xmath231 , and then @xmath536 is a 2-cycle in @xmath532 ( lemma [ lem : g(c)-cycle ] ) . by lemma [ lem : g(p)-class ] , @xmath537 = \wlk_\pi(x_k , c_1^{(i)})\alpha_1^{(i ) } + \wlk_\pi(x_k , c_2^{(i)})\alpha_2^{(i ) } + \wlk_\pi(x_k , c_3^{(i)})\alpha_3^{(i)}.\ ] ] if @xmath533 is an ascending manifold , then by closing @xmath536 by a ribbon graph as lemma [ lem : ribbon - graph ] , we obtain a 2-cycle in @xmath532 that is homologous to @xmath275 . further , if the ribbon graphs that close the three chains are chosen sufficiently thin , then the intersections between the ribbon graphs in @xmath231 are those induced from transversal double points between edges . hence no triple intersection of 2-chains involving the ribbon graphs will be produced after perturbing the 2-chains slightly . this shows that the additions of the ribbon graphs to @xmath536 does not change the value of the triple product ( [ eq : triple - prod ] ) . therefore , the triple product ( [ eq : triple - prod ] ) is a linear combination of terms of the forms @xmath538 and its coefficients are given by products of equivariant linking numbers between leaves of @xmath12-graphs . let @xmath539 be a separated edge of @xmath429 . consider the following element of @xmath540 with ( untwisted ) @xmath79-coefficients @xmath541 this is represented by the moduli space of flow - graphs from a separated edge whose black vertices lie in @xmath542 and @xmath543 . let @xmath544 be the element of @xmath545 obtained from @xmath546 by replacing the @xmath547-th and the @xmath270-th factors with @xmath548 . by the argument above , the following identity holds . @xmath549 the product in the right hand side is the usual intersection form @xmath550 this product can be rewritten as the right hand side of the formula of the lemma . although the sign in the right hand side of ( [ eq : q - q ] ) depends on the graph orientation of @xmath429 , it is cancelled out by the graph orientation after multiplying ( [ eq : q - q ] ) to a graph . for the details , we refer the reader to @xcite . the following lemma is immediate . [ lem : lk - nontrivial ] @xmath551 iff there exists an automorphism @xmath552 that preserves the labels of vertices . for an abstract trivalent graph @xmath48 , let @xmath553 be the group of automorphisms of @xmath48 that fix all vertices . let @xmath554 be the group of permutations of vertices of @xmath48 that give automorphism of @xmath48 . we have @xmath555 . by hypothesis [ hyp : a3=1 ] , the value of the right hand side of ( [ eq : q - q ] ) agrees with @xmath556 , whose value is the sum of monomials that correspond to the decorations of @xmath48 obtained by composing edge automorphisms of @xmath48 and a map @xmath56 that represents the monomial @xmath10-decorated graph @xmath50 . each term becomes the same element in @xmath557 after multiplying it to @xmath429 . thus @xmath558 the value of @xmath463 is the same as this . among the permutations @xmath559 of the set of vertices , there are @xmath560 elements satisfying the condition of lemma [ lem : lk - nontrivial ] . for a trivalent graph , there are @xmath561 different ways of giving edge - orientations and labellings . therefore , @xmath562)&=\frac{1}{2^{3n}(2n)!(3n)!}\sum_{h}\sum_{\sigma\in\mathfrak{s}_{2n}}s_h(\sigma(1),\ldots\sigma(2n))\\ & = \frac{1}{2^{3n}(2n)!(3n)!}\frac{2^{3n}(2n)!(3n)!}{|\mathrm{aut}\,\gamma|}|\mathrm{aut}_v\gamma||\mathrm{aut}_e\gamma|[\gamma(\alpha)]=[\gamma(\alpha ) ] . \end{split}\ ] ] this completes the proof . let @xmath4 and let @xmath15 be a @xmath0-homology equivalence that represents an element of @xmath28 . if we put @xmath563 then by theorem [ thm:2 ] , @xmath564 is an invariant of @xmath0-homology equivalences over @xmath11 . if @xmath565 for an integral homology sphere @xmath14 , then @xmath16 can be represented by an algebraically split framed link in a small ball in @xmath1 , and it determines a @xmath0-homology equivalence @xmath15 uniquely up to homotopy . [ thm : lambda ] for @xmath4 and an integral homology sphere @xmath14 , the following identity holds . @xmath566,\ ] ] where @xmath567 is the casson invariant , and @xmath568 is the 1-decorated theta - graph with vertex - orientation given by blackboard ribbon structure and counter - clockwise order at each trivalent vertex . consider the exact sequence @xmath569 here , since @xmath570 , we have @xmath571 , which is one - dimensional and is generated by the class of @xmath572 . the space @xmath573 is generated by elements of the form @xmath574-[n]$ ] ( @xmath574\in\calk(n)$ ] ) . in particular , if @xmath565 , then by theorem [ thm : eval](2 ) , @xmath575 is a linear combination of 1-decorated graphs of degree 2 . if @xmath14 is represented by a framed link @xmath515 in @xmath9 , then this value agrees with the difference @xmath576 of the ( unframed ) kontsevich kuperberg thurston invariant ( @xcite ) , and by a result of @xcite , this value agrees with @xmath577 $ ] . by theorem [ thm : lambda ] , @xmath564 can be considered as an extension of the casson invariant up to a constant . this can be computed by modifying @xmath578 for a @xmath10-algebraically split framed link @xmath515 to a linear combination of the images of @xmath579 of 2-loop graphs in @xmath580 , by the method given in @xcite . t. t. q. le , _ an invariant of integral homology 3-sphere which is universal for all finite type invariants _ , in `` soliton , geometry and topology : on the crossroad '' , ams transl . ser 2 * 179 * ( 1997 ) , 75100 . c. lescop , _ invariants of knots and 3-manifolds derived from the equivariant linking pairing _ , chern simons gauge theory : 20 years after , ams / ip stud . adv . math . * 50 * ( 2011 ) , amer . soc . , providence , ri , 217242 .

this paper investigates arbitrage chains involving four currencies and four foreign exchange trader - arbitrageurs . in contrast with the three - currency case , we find that arbitrage operations when four currencies are present may appear periodic in nature , and not involve smooth convergence to a `` balanced '' ensemble of exchange rates in which the law of one price holds . the goal of this article is to understand some interesting features of sequences of arbitrage operations , features which might well be relevant in other contexts in finance and economics . limits to arbitrage , four currencies , recurrent sequences , asynchronous systems _ jel classification _ : c60 , f31 , d82

1112.5850

an arbitrage operation involves buying some good or asset for a lower price than that for which it can be sold , taking advantage of any imbalance in the quoted prices . the `` law of one price '' is a statement of a key implication of the absence of arbitrage opportunities . in turn arbitrage is often the process invoked to explain why goods or assets that are in some sense `` identical '' should have a common price . a study of commodity prices since 1273 concluded that `` despite the steady decline in transportation costs over the past 700 years , the repeated intrusion of wars and disease , and the changing fashions of commercial policy , the volatility and persistence of deviations in the law of one price have remained quite stable '' @xcite . the present paper investigates a relatively neglected complication regarding arbitrage operations , namely the order in which information about arbitrage opportunities is presented , illustrating this in relation to arbitrage chains involving four currencies . the key finding is that arbitrage operations can be periodic in nature , rather than involving a smooth convergence to a law of one price . the early literature on the law of one price is coeval with the purchasing power parity explanation of foreign exchange rates . the terminology was coined in @xcite , involving arbitrage between relatively homogeneous goods priced in different currencies @xcite . empirical tests suggest that arbitrage operations in goods do not exert a strong influence on exchange rates until the price index deviations involved exceed about 25% @xcite . innovations that were expected to reduce price dispersion , such as the european single market legislation coming into effect in 1992 , and the economic and monetary union project beginning in 1999 , have had little effect on price level disparities @xcite . the degree of price level dispersion between us cities has displayed no marked trend over time @xcite . a study of the prices charged for identical products in ikea stores in twenty - five countries revealed typical price divergences of 2050% , differences that could not be attributed to just country or location - specific factors @xcite . among the most cited reasons for deviations from the law of one price are transaction costs , taxes , transport costs , trade barriers , the costs of searching for price differences , nominal price rigidities , customer market pricing , nominal exchange rate rigidities and differences in market power @xcite . in relation to assets , an early application of the law of one price was to the interest rate parity theory of the forward exchange rate , whereby the ratio of the forward to spot exchange rate between two currencies is equal to the ratio of the interest rates in the two currencies over the forward period in question @xcite . an arbitrage opportunity in relation to assets can be defined as `` an investment strategy that guarantees a positive payoff in some contingency with no possibility of a negative payoff and with no net investment '' ( * ? ? ? * online ) . the absence of such arbitrage opportunities has been seen as the unifying concept underlying mainstream theories in finance , no - arbitrage principles being applied in the modigliani miller theorem of corporate capital structure , in the black scholes model of option pricing and in the arbitrage pricing model of asset prices @xcite . actual arbitrage operations in relation to assets often involve net investment and risk and/or uncertainty , in addition to the complications arising in relation to arbitrage in goods . notable deviations from the law of one price in financial markets have been documented in relation to comparable circumstances applying to closed - end country funds , american depository receipts , twin shares , dual share classes and corporate spin - offs @xcite . among the limits to arbitrage in financial markets are those arising from transactions costs @xcite , and those involving the capital requirements of conducting arbitrage operations @xcite . a spectacular illustration of the capital limits to arbitrage was provided by the demise of the long - term capital management ( ltcm ) hedge funds . the arbitrage discrepancies being exploited in ltcm s `` convergence trades '' widened in 1998 . ltcm attempted unsuccessfully to raise new capital to finance its arbitrage positions . to avoid a major financial collapse the new york federal reserve board organised a bail - out by creditors @xcite . in what follows we focus on the limits to arbitrage arising from the order in which information is disseminated to arbitrage traders . the illustration used is for a foreign exchange ( fx ) market with four fx traders and four currencies , see sections [ s-3currencies ] and [ s-4currencies ] . an arbiter , the metaphorical equivalent of an unpaid auctioneer in a walrasian system , knows all the actual exchange rates . the individual fx traders , however , initially know only the exchange rates involving their own , domestic currencies . justification for the assumptions used in our model is provided in section [ s - structfx ] . so the us fx trader knows the exchange rates for the dollar against the euro , sterling and yen , but not the cross exchange rates for the non - dollar currencies . there are no transactions costs , no net capital requirements and no risks involved in the arbitrage operations . instead we focus on the information dissemination problem , and show that the order in which information about cross exchange rate discrepancies , and hence arbitrage opportunities , is presented makes an important difference to the sequences of arbitrage operations conducted . a general discussion of arbitrage dynamics is given in section [ s - arbitrages ] . an unexpected feature of the processes considered in this paper is that , _ rather than there being a smooth convergence to an ensemble of exchange rates with no arbitrage opportunities , the arbitrage operations may display periodicity and no necessary convergence on a cross exchange rate law of one price_. see proposition [ 32 ] in section [ mrss ] for a rigorous explanation . a further unexpected feature is that , _ starting at an ensemble of exchange rates which is not balanced , and using special periodic sequences of arbitrages , the arbiter can achieve balanced ( satisfying the law of one price ) exchange rate ensemble_. see , in particular , theorem [ arbh ] in section [ mrss ] and theorem [ irratbc ] in section [ s - gencase ] . these counter - intuitive results are new , as far as we are aware . in line with the renowned `` impossibility theorem '' of @xcite these results suggest an `` arbitrage impossibility theorem '' . proofs are relegated to section [ s - proofs ] . the mathematical approach taken in this paper to the analysis of arbitrage operation chains may be understood as a typical example of the asynchronous interactions that are important in systems theory and in control theory , see the monographs @xcite and the surveys @xcite . the arbitrage chains are particularly relevant to desynchronised systems theory , see @xcite . presence of an asynchronous interaction often leads to a dramatic complication of the related mathematical problems . @xcite proved that many asynchronous problems can not be solved algorithmically , and also @xcite and @xcite demonstrated that , even in the cases when the problem is algorithmically solvable , it is typically as hard to solve numerically as the famous `` travelling salesman problem , '' see @xcite ( that is , in the mathematical language , the problem is np - hard which is an abbreviation for `` non - deterministic polynomial - time hard '' which means in the theory of algorithms that a problem is very hard , if possible , to solve , see @xcite ) . in this context the fact that the principal questions that arise in analysis of arbitrage operation chains admit straightforward combinatorial analysis came to the authors as a pleasant surprise . our construction uses a geometrical approach to visualisation of arbitrage chains presented in sections [ simpleass][s - proofs ] , which may be useful in relation to other problems in mathematical economics . the periodicity results in this paper have implications for several strands of literature . one is that dealing with the disequilibrium foundations of equilibrium economics . the stability analysis of fisher poses the question : `` can one expect to prove that an economy with rational agents conscious of disequilibrium and taking advantage of arbitrage opportunities is driven ( asymptotically ) to any equilibrium , walrasian or constrained ? '' fisher uses the assumption of `` no favorable surprise '' as a means of demonstrating that a cessation of exogenous shocks can lead to convergence to equilibrium . the results in this paper suggest that there can be endogenous reasons , arising from the cyclical response of arbitrage sequences to an exogenous shock that gives rise to an arbitrage opportunity , why convergence to equilibrium may not take place . another strand of literature to which our results relate is that on market segmentation and arbitrage networks . goods and assets are not traded on a single exchange . instead there are various trading posts , such as commodity and stock exchanges . other trades , including a sizeable proportion of foreign exchange deals , are conducted `` over the counter '' in direct transactions that bypass formal exchanges . `` as a result , various clienteles trade on different exchanges , and very few retail clients trade on more than one exchange , let alone on all of them simultaneously '' @xcite . a key aspect of segmentation in the foreign exchange `` market '' is that dealing rooms tend to specialise in domestic currency trades . this provides a rationale for the specification in this paper that foreign exchange dealers initially are aware of only the exchange rates involving their domestic currencies . we restrict our analysis to the case of 4 currencies , with 6 principal exchange rates . the financial times gives daily quotes for 52 currencies . the 1,326 principal exchange rates involved suggest richer potential opportunities for arbitrage than in the four currency case studied in the present paper . bank for international settlements ( bis ) data indicate that , in 2010 , transactions in these four currencies counted for 155.9% of global fx market turnover , the currency components being us dollars ( 84.9% ) , euros ( 39.1% ) , japanese yen ( 19.0% ) and pound sterling ( 12.9% ) . because two currencies are involved in each transaction , the % shares sum to 200% ( * ? ? ? * table b.4 ) . in a centralised market trade takes place at prices that are public information and traders face the same potential trading opportunities . in contrast the fx market is decentralised , with the end - user bank customers , banks , brokers and central banks involved facing several possible methods of executing transactions , and possibly different exchange rate quotes , some of which constitute private information . bis data for fx spot exchange rate transactions in 2010 ( * ? ? ? * table e.24 ) indicate the following breakdown in execution methods as a % of total global turnover : inter - dealer direct ( 14.9% ) , customer direct ( 21.6% ) , voice broker ( 8.6% ) , electronic broking system ( 26.0% ) , single - bank electronic proprietary trading platforms ( 14.3% ) and multi - bank dealing systems ( 14.5% ) . until the late 1980s fx transactions were conducted largely by telephone , with fx dealers phoning counterparties to get bid ( buy ) and offer ( sell ) quotes for specific transaction amounts , there also being indirect dealing via voice brokers who would search for matching interests between clients , see @xcite . the last two decades have seen a growth in electronic methods of execution , a distinction being between electronic broking systems such as reuters matching and the electronic broking system spot dealing system ( ebs ) , and single or multi - bank proprietary dealing platforms . a burgeoning literature investigates how this fragmented trading structure impacts on price determination in fx markets , see @xcite and @xcite for surveys . the key contrast is between the decentralised transactions conducted by fx dealers who quote bid and offer prices that are not public information , and the one - way bid or offer limit orders to buy and sell currencies at a specific price that are accumulated by fx brokers in the quasi - centralised segment of the market . this means that market information is fragmented , fx dealers having private information about the transactions forthcoming at their own quoted bid and offer prices , and having access to the public information regarding the order flows accumulated by the fx brokers . for analysis , discussion and evidence regarding how order flows impact on intra - day exchange rates see @xcite . the analysis in the present paper assumes that fx dealers initially know only the exchange rates for their own domestic currencies , the order in which they discover imbalances in the exchange rate ensemble , in the form of cross exchange rate discrepancies , playing a key role in the arbitrage sequences conducted . the fragmented nature of information in fx markets suggests that this strong assumption has a whiff of reality in that different fx dealers are likely to have disjoint information sets and can conduct trades at different prices . individual fx dealers conducting bi - lateral trades with end - users receive private information in the form of the orders forthcoming at their quoted bid and ask prices , and this can give rise to profitable arbitrage opportunities . for example , an fx dealer specialising in us dollars might simultaneously receive large buy orders for euros and large sell orders for japanese yen , and suspect that euros are under - priced relative to japanese yen . after checking out the euro japanese yen exchange rates quoted in the inter - dealer market , or in the brokered section of the market where information is public , the dealer might discover that this is indeed the case , and exploit this arbitrage opportunity regarding which other fx traders are initially unaware . the bis data on the geographical distribution of fx market turnover is informative in relation to the assumption in the present paper that fx dealers initially are aware of only the exchange rates involving their own domestic currency . banks located in the uk account for 37% of global fx turnover , followed by the us ( 18% ) , japan ( 6% ) , singapore ( 5% ) , switzerland ( 5% ) , hong kong ( 5% ) and australia ( 4% ) see ( * ? ? ? * graph b.7 ) . although cross - border transactions account for nearly two - thirds of fx market turnover , this still leaves 35% of the turnover being local in nature ( * ? ? ? * table 3.2 ) , suggesting that the tendency of fx dealers initially to focus on the exchange rates involving their own domestic currencies assumed in the present paper is evident in a significant section of the fx market . traders could be better informed about exchange rate developments involving their own domestic currencies for a variety of reasons . this could be simply because their core end - users have the domestic currency as a unit of account , and means of payment , so the domestically - based fx dealers have a `` home bias '' when it comes to the exchange rates that they consider first . the psychology literature indicates that here are quite tight limits to the pieces of information that the working memory can take into account when decisions are made @xcite , suggesting that there could well be advantages to fx traders if they focus , at least initially , on a limited number of exchange rates . alternatively the `` home bias '' could be due to the existence of different time zones . so , for example , japanese fx traders may be more able to react to new information relevant to the japanese yen during the asian trading hours in which north american and european markets are closed . the evidence is that most fx trades initiated in japan and australia occur during asian trading hours ; most trades initiated in the us and canada occur during north american hours ; while uk - initiated trades tend to be bunched in the overlapping asia europe and europe north america time zones ( * ? ? ? * table 2 ) . a further reason for `` home bias '' is that the localised or institutionalised links that fx traders have with domestic clients gives them order flow information about the likely course of the exchange rates involving the domestic currency before the price impact of this information becomes publicly available , via the effects on inter - dealer trades , to fx traders operating in foreign locations . in @xcite the authors pose the question `` does tokyo know more about the yen ? '' . prior to december 22 , 1994 the japanese fx market closed for lunch from 1200 to 1330 hours , tokyo time . on the basis that local order flow conveys informational advantages , the authors postulate that the trades of informed tokyo traders would be bunched before the lunch - time fx market closure , an effect that would disappear once the lunch - time closure was abolished . they found a significant tendency of foreign quotes on the japanese yen us dollar market to lag behind the tokyo quotes in this pre - lunch period , suggesting either that tokyo - based traders were better informed about the japanese yen us dollar exchange rate than fx traders based in foreign locations , or that foreign - based traders believed this to be the case . further evidence for a `` home bias '' in the fx market was found in a study of the canadian dollar us dollar and australian dollar us dollar markets @xcite . the author calculates the impulse response functions of the exchange rates to trades , measured by the order flows , initiated in different locations . trades initiated in canada had a larger long - run impact on the canadian dollar us dollar exchange rate than those initiated the us during north american trading hours , and than australian and japanese trades initiated during asian trading hours . uk - initiated trades had a slightly larger long - run effect during european trading hours , but this effect was much larger before the start of north american trading hours . somewhat similarly , trades initiated in australia had a larger long - run impact on australian dollar us dollar exchange rate than trades initiated in the us and elsewhere . the conclusion is that `` dealers operating both at the same time and in the same geographic region as fundamentally driven customers have a natural informational advantage '' @xcite . a major challenge to theories based on the idea that macroeconomic `` fundamentals '' drive exchange rates was presented by the meese - rogoff results that such models did not forecast any better than the `` naive '' postulate that the exchange rate rate would remain unchanged @xcite . engel and west @xcite showed that exchange rates would display something close to the random walk implied by the naive forecast if the fundamentals followed an @xmath0 process and the factor for discounting future fundamentals was close to one . the microstructure literature has shown a way out of this impasse , showing that micro - based information regarding order flows , information which is not necessarily publicly available , can explain a significant component of exchange rate variation . so , for example , evans and lyons @xcite show that end - user order flow data can explain around 16% of the variance in the monthly spot rate between the us dollar and the euro , outperforming both standard macro fundamentals models and the random walk specification . the microstructure literature has also focussed attention onto high frequency data sets . osler @xcite , for example , analyses minute - by - minute quotes for the us dollar spot exchange rates with the deutschmark , japanese yen and pound sterling , discovering significant effects from stop - loss order flows , where the stop - loss order is one that instructs fx dealers to buy ( sell ) a certain amount of a currency at the `` market '' rate once the exchange rate has risen ( fallen ) to a pre - specified level . a full survey of the theoretical and empirical literature on fx exchange rate determination has been beyond the scope of the present paper ( see @xcite for such a survey ) . what we would argue is that the foregoing selective review of the literature provides some justification for the assumptions used in the analysis of the arbitrage sequences that follows . the assumption that fx dealers initially know only the exchange rates for their domestic currency finds some support in the `` home bias '' evidence cited above . the evidence on the fragmented nature of the fx market lends support to the assumption that fx traders can have privileged access to initially private information , stemming from order flows from end - user clients , that would allow them to identify imbalances in cross exchange rates , and hence identify arbitrage opportunities before fx traders based in other locations can identify such opportunities . there is also evidence that there are arbitrage opportunities to be exploited . in @xcite the authors use binding quote and transactions data from the electronic broking system , ebs , for the us dollar , euros , japanese yen , the pound sterling and the swiss franc . triangular arbitrage opportunities are identified within two - minute time horizons , and can be exploited by three trades on the ebs trading screen . each identified arbitrage opportunity involved the us dollar and the euro , the third currency being the japanese yen , pound sterling or swiss franc . the estimated mean arbitrage profits , net of bid - offer spreads and 0.2 basis point trade fees , ranged from 2.8 to 3.0 basis points @xcite . so here is evidence of , albeit small , profits to be had from arbitrage operations on a quasi - centralised , electronic broking trading platform . once the decentralised sections of the fx market are considered , the existence of initially private information is likely extend the range and size of profitable arbitrage arbitrage opportunities available . consider a foreign exchange ( fx ) market that involves only three currencies : dollars ( $ ) , euros ( ) and sterling ( ) . this fx market involves three pair - wise exchange operations : @xmath1 the currencies are measured in natural currency units , and the corresponding ( strictly positive ) exchange rates , @xmath2 , @xmath3 , @xmath4 , are well defined . for instance , one dollar can be exchanged for @xmath2 euros . the rates related to the inverted arrows are reciprocal : @xmath5 we treat the triplet @xmath6 as the ensemble of _ principal exchange rates . _ we suppose that , prior to a reference time moment @xmath7 , each fx trader knows only the exchange rates involving his domestic currency . so the dollar trader does not know the value of @xmath8 , the euro trader is unaware of @xmath3 , and the sterling trader is unaware of @xmath9 . we are interested in the case where the initial rates are unbalanced in the following sense . by assumption , the dollar trader can exchange one dollar for @xmath9 euros . let us suppose that unbeknownst to him the exchange rate between sterling and euro is such that the the dollar trader could make a profit by first exchanging a dollar for @xmath3 units of sterling and then exchanging these for euros . the inequality which guarantees that dollar trader can take advantage of this arbitrage opportunity is that the product @xmath10 is greater than @xmath9 : @xmath11 let us consider the situation where the inequality holds , and , after the reference time moment @xmath7 , one of the three traders becomes aware of the third exchange rate . the evolution of this fx market depends on _ which trader is the first to discover the information concerning the third exchange rate_. the following three cases are relevant . the dollar trader becomes aware of the value of the rate @xmath8 . therefore , the dollar trader contacts the euro trader and makes a request to increase the rate @xmath9 to the new fairer value @xmath12 the reciprocal exchange rate @xmath13 is also to be adjusted to the new level : @xmath14 the result is that the principal exchange rates become balanced at the levels : @xmath15 the euro trader is the first to discover the third exchange rate @xmath3 . by , inequality may be rewritten as @xmath16 which is , in turn , equivalent to @xmath17 . in this case the euro trader could do better by first exchanging euros for dollars , and then by exchanging the dollars for sterling . therefore , the euro trader requests adjustment of the rate @xmath8 to the value @xmath18 in terms of the principal exchange rates the outcome is that the fx market adjusts to the following balanced rates : @xmath19 the sterling trader is the first to discover the third exchange rate @xmath9 . the inequality may be rewritten as @xmath20 thus , the sterling trader requests adjustment of the rate @xmath21 to @xmath22 . in this case the principal exchange rates become balanced at the levels : @xmath23 _ after the adjustment of the principal exchange rates , following the new information being revealed , the exchange rates become balanced , and this is the end of the arbitrage evolution of an fx market with three currencies . having established the reasonably straightforward application of arbitrage to three currencies , we now turn to investigation what happens when the fx market contains four currencies and four currency traders . _ consider an fx market $ that involves four currencies : dollars ( $ ) , euros ( ) , sterling ( ) and yen ( ) . this fx market involves six exchange relationships : @xmath24 the exchange rates are : @xmath25 the rates relating to the inverted arrows are reciprocal : @xmath26 our market may be described by the ensemble of six principal exchange rates @xmath27 together with the reciprocal exchange rates . the following characterisation of balanced , no - arbitrage , exchange rates , that is the ensembles of exchange rates such that no trader could do better by trading indirectly , is convenient . [ balp ] ensemble of the principal exchange rates is balanced if and only if the following relationships hold : @xmath28 this assertion can be proved by inspection . let us suppose that initially each trader is aware only of the three exchange rates involving his domestic currency . for instance , the dollar trader knows only the rates @xmath9 , @xmath3 , @xmath29 . _ we are interested in the case where the rates @xmath9 , @xmath3 , @xmath29 , @xmath8 , @xmath30 , @xmath31 are unbalanced_. for instance , let us suppose that can make a profit by first exchanging one dollar for @xmath3 units of sterling , and then by exchanging this sterling for euros . this means that the product @xmath32 is greater than @xmath9 : @xmath33 suppose that becomes aware of the rate @xmath8 , and , therefore , about the inequality . the dollar trader then asks to increase the exchange rate @xmath9 to the new fairer value @xmath34 along with the adjustment of the exchange rate @xmath9 the reciprocal rate @xmath13 would be adjusted to @xmath35 we call this procedure $ -_arbitrage _ , and we use the notation @xmath36 to represent it . we denote by @xmath37 the ensemble of the new principal exchange rates : @xmath38 we also use the notation @xmath37 in the case where the inequality does not hold . in this case , of course , @xmath39 , and we say that arbitrage @xmath36 is _ not active _ in this case . this particular arbitrage is an example of the 24 possible arbitrages listed in table [ tab1 ] . we will also use , where convenient , the notation @xmath40 for the arbitrage number @xmath41 from this table : for instance , @xmath42 . llll number&arbitrage&activation condition&actions + + 1 & @xmath36 & @xmath43 & @xmath44 + 2 & @xmath45 & @xmath46 & @xmath47 + 3 & @xmath48 & @xmath49 & @xmath50 + 4 & @xmath51 & @xmath52 & @xmath53 + 5 & @xmath54 & @xmath55 & @xmath56 + 6 & @xmath57 & @xmath58 & @xmath59 + 7 & @xmath60 & @xmath61 & @xmath62 + 8 & @xmath63 & @xmath64 & @xmath47 + 9 & @xmath65 & @xmath66 & @xmath67 + 10 & @xmath68 & @xmath69 & @xmath70 + 11 & @xmath71 & @xmath72 & @xmath73 + 12 & @xmath74 & @xmath75 & @xmath76 + 13 & @xmath77 & @xmath78 & @xmath50 + 14 & @xmath79 & @xmath80 & @xmath53 + 15 & @xmath81 & @xmath82 & @xmath67 + 16 & @xmath83 & @xmath84 & @xmath70 + 17 & @xmath85 & @xmath86 & @xmath87 + 18 & @xmath88 & @xmath89 & @xmath90 + 19 & @xmath91 & @xmath92 & @xmath56 + 20 & @xmath93 & @xmath94 & @xmath59 + 21 & @xmath95 & @xmath96 & @xmath73 + 22 & @xmath97 & @xmath98 & @xmath76 + 23 & @xmath99 & @xmath100 & @xmath87 + 24 & @xmath101 & @xmath102 & @xmath90 + _ the principal distinction of the fx market with four currencies from that with only three currencies is that applying a single arbitrage operation does not bring the fx market to a balance in which no arbitrage opportunities exist , and in which the law of one price holds . _ one can apply arbitrages from table [ tab1 ] sequentially in any order and to any initial exchange rates @xmath103 . the situation that we have in mind is the following . suppose that there exists an _ arbiter _ who knows current ensemble @xmath103 of exchange rates . this _ arbiter _ could provide information to the fx traders in any order he wants , thus activating the _ chain _ ( or _ superposition _ ) of corresponding arbitrages . the principal question is : [ que1 ] how powerful is the _ arbiter _ ? the short answer is : the _ arbiter is surprisingly powerful . _ let us explain at a more formal level what we mean . for a finite chain of arbitrages @xmath104 , and for a given ensemble @xmath103 of initial exchange rates , we denote the resulting ensemble of principal exchange rates as @xmath105 if @xmath103 is balanced , then @xmath106 for any individual arbitrage , and therefore @xmath107 for any chain . if , on the contrary , @xmath103 is not balanced , then different arbitrage chains could result in different balanced or unbalanced ensembles of principal exchange rates . denote by @xmath108 the collection of the sets @xmath109 related to all possible chains . denote also by @xmath110 the subset of @xmath108 , that includes only balanced exchange rates ensembles . our principal observation is the following : _ for a typical unbalanced exchange rate ensemble @xmath103 , the set @xmath110 is unexpectedly rich ; therefore the arbiter , who prescribes a particular sequence of arbitrages , is an unexpectedly powerful figure . _ to avoid cumbersome notation and technical details when providing a rigorous formulation of this observation , we concentrate on the simplest initial ensemble . let us consider the ensemble @xmath111 where @xmath112 and @xmath113 is a given balanced ensemble of principal exchange rates . the ensemble is not balanced . the ensemble may have emerged as follows . let us suppose that the underlying balanced rates @xmath114 had been in operation up to a certain reference time moment @xmath7 . at this moment has decided to increase his price for euros by a factor @xmath115 . a natural respecification of question [ que1 ] is the following : [ que2 ] to which balanced exchange rates can the arbiter now bring the foreign exchange market ? the possible general structure of elements from the corresponding sets @xmath116 and @xmath117 is easy to describe . to this end we denote by @xmath118 the collection of all sextuples of the form @xmath119 where @xmath120 are integer numbers ( positive , negative or zero ) . we also denote by @xmath121 the subset of elements of @xmath122 , which satisfy the relationships @xmath123 [ rep1p ] the following inclusions hold : @xmath124 the ensemble belongs to @xmath125 . to verify we show that the set @xmath122 is invariant with respect to each arbitrage @xmath126 from table [ tab1 ] . this statement can be checked by inspection . let us , for instance , apply to a sextuple the first arbitrage @xmath36 . then , by definition , either this arbitrage is inactive , or it changes the first component @xmath127 of to the new value @xmath128 however , the ensemble @xmath129 is balanced , and , by the first equation , @xmath130 . therefore , implies that the ensemble @xmath131 also may be represented in the form . we have proved the first part of the proposition , related to the set @xmath116 . the inclusion follows now from proposition [ balp ] . proposition [ rep1p ] in no way answers question [ que2 ] . this proposition , however , allows us to reformulate this question in a more constructive form : [ que3 ] how big is the set @xmath117 , compared with the collection @xmath132 of all elements that satisfy the restrictions imposed by proposition [ rep1p ] ? the naive expectation would be that the set @xmath117 is finite and , at least for values of @xmath133 close to 1 , that all elements of @xmath117 are close to @xmath113 . however , the following statement , describing an unexpected feature of the power of the _ arbiter _ , is true . [ arbh ] the set @xmath117 coincides with @xmath132 : @xmath134 moreover each balanced ensemble may be achieved via a chain of arbitrage operations no longer than @xmath135 loosely speaking , this theorem means that the _ arbiter _ is extremely powerful . an assertion similar to theorem [ arbh ] was formulated as a hypothesis in @xcite . we describe the algorithms corresponding to this theorem in the next section . the following assertion certifies that the estimate from theorem [ arbh ] is pretty close to the optimal . [ victorp ] the inequalities @xmath136 hold for any @xmath137 . here @xmath138 are the integers from representation of @xmath103 . this assertion is a special case of lemma [ specialc ] which will be considered below . note that the set @xmath116 is , in contrast to , much smaller than the totality @xmath118 of all ensembles of the form . in particular , the following assertion holds : [ boundp ] let @xmath139 denote a chain of arbitrages of length @xmath140 , and @xmath141 . then @xmath142 , where @xmath143 are the integers from the representation of @xmath103 . let us consider an infinite arbitrage chain : @xmath144 this chain is periodic with minimal period @xmath145 if @xmath146 for @xmath147 , and @xmath145 is the minimal positive integer with this property . various periodic chains of arbitrage play a special role in context of this article , and we summairise below some interesting features of such periodic arbitrage chains . for a periodic chain and for an initial ( unbalanced ) exchange rate ensemble @xmath148 we consider the sequence @xmath149 defined by @xmath150 , @xmath151 . [ perpro ] either ( i ) the sequence is periodic for @xmath152 or ( ii ) this sequence is diverging : at least one of the following six relationships hold : @xmath153 moreover , in case ( i ) the minimal period of the sequence is a divisor of @xmath154 ; in case ( ii ) there exist a divisor @xmath155 of @xmath154 and factors @xmath156 such that the relationships @xmath157 hold for @xmath158 . this statement follows from lemmas [ comp ] and [ trans ] . to conclude this discussion , we note one more unexpected feature of periodic chains of arbitrage . a chain is _ regular for the initial ensemble @xmath159 _ if this chain includes all 24 arbitrages , and each arbitrage is active infinitely many times while generating the sequence . by analogy with typical results from the desynchronised systems theory , one could expect a regular chain of arbitrage elements of the corresponding sequence should be balanced for sufficiently large @xmath41 . however , this is not the case : the sequences may be both periodic ( after some transient period ) or diverging . as an instructive example consider the 24-periodic chain @xmath160 which is defined by the following equations : @xmath161 [ 32 ] for the initial ensemble @xmath162 the corresponding sequence is periodic with minimal period 24 , and all arbitrages from @xmath160 are active . by inspection . this proposition demonstrates that arbitrage operation chains may display periodicity and no necessary convergence on a cross exchange rate law of one price . see figs . [ graphleha1_24 ] , [ graphleha3 ] and formula below for an explanation of the geometrical meaning of the arbitrage chain @xmath160 . introduce the following chains of arbitrages of length @xmath163 : @xmath164 it is convenient to define the mapping @xmath165 which corresponds to a non - negative integer @xmath41 by the symbol `` @xmath166 '' , and by the symbol `` @xmath167 '' for a negative integer . [ algp ] the chain @xmath168 satisfies theorem [ arbh ] : the ensemble @xmath169 coincides with @xmath170 and the length @xmath140 of the chain satisfies @xmath171 the legitimacy of this algorithm may be verified by induction . however a simple geometric proof is much more instructive . this proof will be given later on . this chain is not always the shortest : for instance , in the case @xmath172 the shortest chain @xmath139 is of length one : @xmath173 . we begin with the following comment . the ensemble is the first item in the list @xmath174 a natural `` relabelling '' procedure confirms that the main results described in section [ mrss ] hold without any changes for first initial ensemble from the list . in particular , theorem [ arbh ] implies [ arbah ] the equality @xmath175 holds for @xmath176 . moreover each balanced ensemble may be achieved via a chain of arbitrage operations no longer than @xmath177 , where @xmath178 to describe the corresponding algorithms we introduce the auxiliary chains @xmath179 the equation can be modified to the form @xmath180 for @xmath181 , and to the form @xmath182 for @xmath183 . let us turn to the initial ensembles @xmath184 , @xmath185 . [ arba2h ] the equality @xmath186 holds for @xmath185 . moreover each balanced ensemble may be achieved via a chain of arbitrage no longer than @xmath177 , where @xmath187 the corresponding chains @xmath188 , @xmath185 , may be defined by the following equations : @xmath189 @xmath190 @xmath191 this assertion may be proved analogously to theorem [ arbh ] . to formulate further generalisations we need an additional notion . to each ensemble @xmath192 we attach an _ arbitrage discrepancies ensemble , _ using the relationships for balanced principal exchange rates given in above @xmath193 as follows : @xmath194 for instance @xmath195 where @xmath196 . [ disc0p ] the ensemble @xmath103 is balanced , if and only if @xmath197 . follows from proposition [ balp ] and equations . the case where two of the discrepancies are equal to zero was implicitly considered in section [ directss ] : see the second line in and proposition [ arba2h ] . consider now the case when one of the discrepancies in is equal to zero , while two others are not . we will be particularly interested in the situation where two nonzero discrepancies are different . this situation may have emerged , for instance , as follows . let us suppose that the underlying balanced rates had been in operation up to a certain reference time moment @xmath7 . at this moment the euro trader has decided to change two of three his rates , namely @xmath8 and @xmath30 , by different factors @xmath133 and @xmath198 . then at this moment the two discrepancies would acquire different non - zero values , while the third discrepancy remains equal to zero . suppose , for example that @xmath199 , while @xmath200 . we introduce the ratio @xmath201 [ irratbc ] let the number be irrational . then set @xmath110 is dense in the totality @xmath202 of all possible balanced ensembles . a proof of this assertion will be given later on . consider also the case where @xmath203 is a rational number : @xmath204 with co - prime integers @xmath205 ( including the possibilities @xmath206 or @xmath207 ) . denote also @xmath208 the following assertion is a straightforward analog of proposition [ rep1p ] . [ repbp ] the inclusions @xmath209 and @xmath210 hold . the following is an analog of theorem [ arbh ] : [ arbbh ] the equality @xmath211 holds . a proof of this assertion will be given later on . note that the expressions like ( [ estn ] ) are not valid in general . similar expressions may be established , however , for the cases @xmath206 or @xmath207 . note also that the case when the discrepancy triplet is of one the forms @xmath212 or @xmath213 or @xmath214 , @xmath215 , was implicitly considered in section [ directss ] : see the first line in and proposition [ arba2h ] . consider the case where all three arbitrage discrepancies are not equal to zero . [ irratcc ] let at least one of the ratios @xmath216 be irrational . then the set @xmath110 is dense in the totality @xmath202 of all possible balanced ensembles . suppose now that both ratios are rational : @xmath217 denote by @xmath218 the least common multiple of the corresponding denominators . denote @xmath219 [ rep1cp ] the relationships @xmath220 and @xmath210 hold . [ arbch ] let @xmath221 then @xmath222 . this assertion as well as formulated below corollary [ arbc2h ] follows from proposition [ arbbh ] together with lemma [ colp ] . consider finally the case when the ratios @xmath223 and @xmath224 are rational , but does not hold . in this case we introduce the number @xmath225 such that @xmath226 where the numbers @xmath227 are integers and their greatest common divisor , @xmath228 , is equal to @xmath229 . consider also the following six numbers : @xmath230 a_{3}&=\gcd(k_{2},k_{3 } ) , & a_{4}&=\gcd(k_{1},k_{2}-k_{3}),\\[1 mm ] a_{5}&=\gcd(k_{2},k_{1}+k_{3}),\quad & a_{6}&=\gcd(k_{3},k_{1}-k_{2 } ) . \end{alignedat}\ ] ] introduce also the numbers @xmath231 , @xmath232 . [ arbc2h ] the equation @xmath233 holds . note that all six numbers in may indeed be greater than one . for instance , consider : @xmath234 , @xmath235 , @xmath236 . by inspection , @xmath237 , and @xmath238 from this point onward we discuss the proofs of the theorems formulated above . this part of the paper is organised as follows . in section [ s - sarb ] we introduce , as a useful auxiliary tool , stronger arbitrage procedures . using strong arbitrages , we `` linearise the problem '' , reducing it to investigation of all possible products of 12 explicitly written @xmath239-matrices . afterwards , in section [ s - scs ] we separate a family of 12 @xmath240-matrices @xmath241 such that the products of these matrices completely describe the dynamics of the discrepancy triplets . the properties of such products appear to be of key importance , and these are investigated in section [ s - struct ] . the results are applied in section [ s - discrep ] . sections [ s - incdyn ] and [ s - prooft ] are dedicated to finalising the proof of theorem [ arbh ] . finally , in sections [ knotsss][fp2ss ] we provide proofs for theorem [ irratbc ] and proposition [ arbbh ] . we use , as an auxiliary tool , stronger arbitrage procedures . let us begin with an example . consider the currencies triplet @xmath242 . for a given @xmath103 we define the strong arbitrage @xmath243 as @xmath36 if the inequality holds , and as @xmath60 , otherwise . note that in both cases the result in terms of principal exchange rates is the same : the rate @xmath9 is changed to @xmath244 . the strong arbitrage @xmath245 is the second entry in table [ starbt ] of the possible 12 strong arbitrages . the meaning of a strong arbitrage is simple . this is an arbitrage balancing a sub - fx market such as @xmath246 by changing the exchange rate for a pair such as @xmath247 . we will use , where convenient , the notation @xmath248 for the arbitrage number @xmath41 from this table . llll number & strong arbitrage & action & numbers of arbitrages + + 1 & @xmath249 & @xmath250 & 1 , 7 + 2 & @xmath245 & @xmath251 & 2 , 8 + 3 & @xmath252 & @xmath253 & 3 , 13 + 4 & @xmath254 & @xmath255 & 4 , 14 + 5 & @xmath256 & @xmath257 & 5 , 19 + 6 & @xmath258 & @xmath259 & 6 , 20 + 7 & @xmath260 & @xmath261 & 9 , 15 + 8 & @xmath262 & @xmath263 & 10 , 16 + 9 & @xmath264 & @xmath265 & 11 , 21 + 10 & @xmath266 & @xmath267 & 12 , 22 + 11 & @xmath268 & @xmath269 & 17 , 23 + 12 & @xmath270 & @xmath271 & 18 , 24 for any arbitrage chain , and any initial exchange rates @xmath103 , there exists a chain @xmath272 of strong arbitrages such that @xmath273 . conversely , for any chain @xmath274 of strong arbitrages , and any initial exchange rates @xmath103 , there exists a chain of arbitrages such that @xmath273 . this proposition reduces investigation of the questions from the previous section to investigation of analogous questions related to chains of strong arbitrages . now we relate each strong arbitrage to a @xmath239 matrix @xmath275 as follows : @xmath276 @xmath277 @xmath278 @xmath279 @xmath280 @xmath281 for any ensemble @xmath282 we denote @xmath283 [ oldprop ] the equation @xmath284 holds for @xmath285 . follows from definitions . in the six - dimensional real coordinate space @xmath286 we introduce the vectors @xmath287 by definition for any ensemble @xmath103 @xmath288 where @xmath289 denotes the usual inner product in @xmath286 . propositions [ balp ] and [ oldprop ] together imply the three - dimensional subspace @xmath290 is invariant with respect to each linear operator @xmath291 , @xmath292 . we introduce in @xmath286 the new basis @xmath293 here @xmath294 , @xmath295 , @xmath296 . by the last corollary in this basis the matrices of the linear operators @xmath297 have the block - triangular form : @xmath298 here @xmath299 and @xmath300 are some @xmath240-matrices . denote @xmath301 [ blockc ] the equality @xmath302 holds for @xmath292 . follows from lemma [ qlem ] and proposition [ oldprop ] . the matrices @xmath303 may be written explicitly as @xmath304 where @xmath305 [ qlem ] the following equations are valid : @xmath306 @xmath307 @xmath308 @xmath309 and @xmath310 @xmath311 @xmath312 follows by inspection from , . [ disc1p ] the discrepancy ensemble @xmath313 depends only on @xmath314 and @xmath315 , and may be written as follows : @xmath316 . here @xmath317 is the number of a strong arbitrage as listed in table [ starbt ] . follows from proposition [ blockc ] . by the last proposition a discrepancy ensemble @xmath318 related to an arbitrage chain @xmath319 may be written as @xmath320 therefore the set @xmath321 of all possible products of the matrices @xmath241 is of interest . the following assertion is the key observation of our paper : [ gfini ] the set @xmath321 consists of 229 elements . by inspection denote by @xmath322 the totality of all finite chains of strong arbitrages . [ dfini ] for a given @xmath103 the set @xmath323 consists of less than 230 elements . let us discuss briefly the structure of the set @xmath321 . a subset @xmath324 of @xmath321 is called a connected component , if for any @xmath325 there exists @xmath326 satisfying @xmath327 . by the definition different connected components do not intersect . [ comp ] the set @xmath321 is partitioned into 14 connected components @xmath328 . each of the first six connected components includes 24 matrices of range 2 ; each of the connected components @xmath329 includes 12 matrices of range one ; the last component contains a single zero matrix . the sets @xmath330 may be characterised by the following inclusions : @xmath331 to identify the connected components @xmath329 we list below the smallest lexicographical matrices from these components @xmath332 @xmath333 @xmath334 one can move from one connected component @xmath335 to another component @xmath336 applying a matrix @xmath241 , @xmath292 . let us describe the set of possible transitions . we will use the notation @xmath337 if such a transition is possible . [ trans ] the following relationships hold : @xmath338 also @xmath339 , @xmath340 . by inspection . for any @xmath341 either @xmath342 or @xmath343 or @xmath344 is a projector . by inspection . the structure of the set @xmath321 explained above induces structuring of the set of discrepancies , which we discuss below . we say that a set @xmath345 of discrepancies is a connected component if for any @xmath346 there exists an arbitrage chain @xmath139 satisfying @xmath347 . for a given reals @xmath348 we denote by @xmath349 the set of different triplets from the collection @xmath350 each set @xmath349 is a connected component , and each connected component coincides with a certain set @xmath349 . this statement may be proved by inspection . let us discuss in brief the structure of the sets @xmath349 for different values @xmath348 . clearly , @xmath351 consists of the single zero triplet @xmath352 . the connected components @xmath353 , @xmath354 , @xmath355 , @xmath356 coincide and include the following 12 elements : @xmath357 we use notation @xmath358 for this set . geometrically the set @xmath358 represents vertices of a partly distorted truncated cuboctahedron , or triangular orthobicupola , shown in fig . the structure of this component will be explained in more detail in section [ s - prooft ] . the set @xmath359 , @xmath360 , @xmath361 , also consists of 12 elements . geometrically these sets @xmath358 represent vertices of a distorted truncated tetrahedron , shown in fig . [ tt ] . otherwise , a set @xmath362 , consists of 24 elements , and represents vertices of a distorted truncated octahedron , shown in fig . [ to ] . the structure of this component will be explained in more detail in section [ knotsss ] . , @xmath363 ; right : the same polyhedron transparent.[tc],title="fig:",scaledwidth=25.0% ] , @xmath363 ; right : the same polyhedron transparent.[tc],title="fig:",scaledwidth=25.0% ] , @xmath364 , or @xmath365 , @xmath363 ; right : the same polyhedron transparent.[tt],title="fig:",scaledwidth=25.0% ] , @xmath364 , or @xmath365 , @xmath363 ; right : the same polyhedron transparent.[tt],title="fig:",scaledwidth=25.0% ] ; right : the same polyhedron transparent.[to],title="fig:",scaledwidth=25.0% ] ; right : the same polyhedron transparent.[to],title="fig:",scaledwidth=25.0% ] we formulate also a corollary of proposition [ trans ] . for a set @xmath345 of discrepancies we denote by @xmath366 the collection of elements of the form @xmath367 , @xmath368 , @xmath369 . [ dfini2 ] the equality @xmath370 holds for @xmath371 . also @xmath372 . some discrepancy triplets do not belong to any connected component ; however any element of the form @xmath367 must belong to a connected component . more precisely : [ colp ] the following inclusions hold : @xmath373 this assertion may be proved by inspection . for a given sextuple @xmath103 we denote by @xmath374 the triplet of the first three components of @xmath103 : @xmath375 . denote further @xmath376 where @xmath315 is a strong arbitrage . [ incrp ] @xmath377 depends only on @xmath315 and @xmath314 and may be described as follows : @xmath378 also the equalities @xmath379 hold for @xmath380 . follows from corollary [ blockc ] . this proceeds by graphing the detailed dynamics of the arbitrage discrepancies . in this section we use the shorthand notation @xmath381 instead of @xmath382 . [ specialc ] for any initial exchange rate ensemble belonging to the list , and for any arbitrage chain , the corresponding sequence of discrepancies includes only elements from the union @xmath383 , @xmath384 , see . the possible transition paths , arising from the strong arbitrages listed in table [ tab1 ] , are given in table [ tab3 ] . figure [ graphleha1 ] plots the corresponding graph . figure [ graphleha1_24 ] plots a similar graph , where the numbers of the arbitrages from table [ tab1 ] are included , instead of the numbers of strong arbitrages . by inspection follows from proposition [ disc1p ] . ignoring the zero vertex @xmath385 , the edges that lead to this vertex and directions of the edges , another , polyhedral , representation of the graph plotted in fig . [ graphleha1 ] is given in fig . [ graphleha2 ] . the corresponding polyhedron is a distorted triangular orthobicupola , shown in fig . the incidence matrix @xmath386 of the graph plotted in fig . [ graphleha2 ] is as follows : @xmath387 l|llllllllllll & @xmath388 & @xmath389 & @xmath390 & @xmath391 & @xmath392 & @xmath393 & @xmath394 & @xmath395 & @xmath396 & @xmath397 & @xmath398 & @xmath399 + + @xmath400 & @xmath388 & @xmath401 & @xmath398 & @xmath402 & @xmath392 & @xmath393 & @xmath394 & @xmath393 & @xmath392 & @xmath402 & @xmath398 & @xmath399 + @xmath403 & @xmath388 & @xmath389 & @xmath390 & @xmath402 & @xmath396 & @xmath395 & @xmath394 & @xmath395 & @xmath396 & @xmath402 & @xmath390 & @xmath404 + @xmath405 & @xmath388 & @xmath402 & @xmath394 & @xmath393 & @xmath392 & @xmath393 & @xmath394 & @xmath402 & @xmath388 & @xmath401 & @xmath398 & @xmath399 + @xmath406 & @xmath388 & @xmath402 & @xmath390 & @xmath391 & @xmath392 & @xmath391 & @xmath390 & @xmath402 & @xmath396 & @xmath397 & @xmath398 & @xmath407 + @xmath408 & @xmath388 & @xmath389 & @xmath390 & @xmath389 & @xmath388 & @xmath402 & @xmath394 & @xmath395 & @xmath396 & @xmath395 & @xmath394 & @xmath409 + @xmath410 & @xmath392 & @xmath391 & @xmath390 & @xmath391 & @xmath392 & @xmath402 & @xmath398 & @xmath397 & @xmath396 & @xmath397 & @xmath398 & @xmath409 + @xmath411 & @xmath388 & @xmath388 & @xmath402 & @xmath392 & @xmath392 & @xmath393 & @xmath394 & @xmath394 & @xmath402 & @xmath398 & @xmath398 & @xmath399 + @xmath412 & @xmath389 & @xmath389 & @xmath402 & @xmath391 & @xmath391 & @xmath393 & @xmath395 & @xmath395 & @xmath402 & @xmath397 & @xmath397 & @xmath399 + @xmath413 & @xmath388 & @xmath389 & @xmath390 & @xmath390 & @xmath402 & @xmath394 & @xmath394 & @xmath395 & @xmath396 & @xmath396 & @xmath402 & @xmath414 + @xmath415 & @xmath401 & @xmath389 & @xmath390 & @xmath390 & @xmath402 & @xmath393 & @xmath393 & @xmath395 & @xmath396 & @xmath397 & @xmath402 & @xmath399 + @xmath416 & @xmath402 & @xmath390 & @xmath390 & @xmath391 & @xmath392 & @xmath392 & @xmath402 & @xmath396 & @xmath396 & @xmath397 & @xmath398 & @xmath407 + @xmath417 & @xmath402 & @xmath389 & @xmath389 & @xmath391 & @xmath393 & @xmath393 & @xmath402 & @xmath418 & @xmath395 & @xmath397 & @xmath401 & @xmath419 ] ] ] now let us deal with the coupled discrepancies and the incremental dynamics . [ specialic ] for any arbitrage chain the corresponding sequence of increments includes only the zero triplet @xmath420 or one of the following six triplets : @xmath421 the dynamics of the increments @xmath422 is conveniently visualised in fig . [ graphleha3 ] . ] the correctness of this description of the dynamics of the increments follows immediately from corollary [ specialc ] and proposition [ incrp ] . the legitimacy of the algorithms relevant to theorem [ arbh ] and , therefore , proof of theorem [ arbh ] and proposition [ arba2h ] follows from figs . [ graphleha1_24 ] and [ graphleha3 ] . we note also that the 24-periodic chain of arbitrage from proposition [ 32 ] was also found looking at fig . [ graphleha1_24 ] and [ graphleha3 ] . the corresponding route is quite natural from this perspective , and is given by @xmath423 now we move to a proof of theorem [ irratbc ] and proposition [ arbbh ] . the case @xmath424 has been considered in section [ directss ] . thus we can assume that @xmath425 . the focus is again on the dynamics of the exchange rate discrepancies . the set of all discrepancies that may be achievable from @xmath426 contains altogether 61 different elements , see corollary [ dfini2 ] . the corresponding connected component @xmath427 , which contains @xmath426 , see , contains 24 elements listed in . to describe the detailed structure of this set we will introduce a new notation . the set @xmath427 , contains six elements that have all three components that are non - zero , and we re - denote these elements by @xmath428 we call these ensembles _ commuters _ by way of analogy with passenger travel . we call an element with two non - zero components a _ terminal _ , if @xmath429 . there are altogether 18 terminals in @xmath349 . to each commuter @xmath430 , @xmath431 , we relate three _ terminals _ @xmath432 , @xmath433 , as follows : @xmath434 [ commul ] the equalities @xmath435 hold for @xmath436 . also the following equalities hold : @xmath437 , for @xmath431 , @xmath433 , @xmath438 . we group the commuters and terminals in six knots , @xmath439 as follows : @xmath440 figure [ commuterleha ] illustrates behaviour at a knot . ] departing from a particular terminal , and applying some arbitrages with numbers @xmath441 , one can travel to another terminal belonging to a different knot , simultaneously `` loading some cargo '' upon the corresponded triplet @xmath374 . details are given in the following proposition . [ cargop ] the following groups of equalities hold : @xmath442\ ] ] @xmath443\ ] ] @xmath444\ ] ] @xmath445\ ] ] @xmath446\ ] ] @xmath447 we introduce the `` travel between knots '' directed graph @xmath448 , shown in fig . [ graphlehacargo ] , as follows . this graph has @xmath449 vertices that correspond to the knots @xmath450 . a knot @xmath451 is connected by an arrow with another knot @xmath452 if one of terminals belonging to @xmath453 figures in the rows belonging to the @xmath317-th subset of equalities from proposition [ cargop ] . for instance , the knot @xmath454 is connected with @xmath455 . moreover each arrow corresponds to the three dimensional `` cargo vector(s ) '' : these vectors are related in a natural way to the increment vectors in the equalities above . for instance , we attach the cargo - vectors @xmath456 and @xmath457 to the @xmath458 arrow . the incidence matrix of this graph is written below . @xmath459 [ graphlehacargo ] ] if the single transition @xmath460 is possible we use @xmath461 for the corresponding cargo ; we will use @xmath462 if two transitions are possible . in the latter case @xmath463 refers to the upper vector indicated at graph @xmath448 . for instance , @xmath464 , @xmath465 , @xmath466 , etc . [ cyclesp ] for any positive integers @xmath467 there exists a chain @xmath468 of strong arbitrages such that @xmath469 has the form @xmath470 where @xmath471 are some positive integer numbers , since the moves from one terminal to another , within a particular knot , are always possible and do not change @xmath374 ( see lemma [ commul ] ) , any route allowed by the graph @xmath448 can be performed , and any combination of corresponding cargo can be loaded . for the cycle @xmath472 we have @xmath473 for the cycle @xmath474 we have @xmath475 for the cycle @xmath476 we have @xmath477 [ abc ] for any non - negative integers @xmath478 and @xmath479 there exists a chain @xmath468 of strong arbitrages such that @xmath469 has the form @xmath480 . from the lemma above it follows that we can achieve the state @xmath481 then moving to the terminal @xmath482 and applying arbitrage @xmath483 we arrive at @xmath484 however , from this state we can , by proposition [ arbah ] , adjust the numbers @xmath485 to the targets @xmath486 . theorem [ irratbc ] and proposition [ arbbh ] follow immediately from this last corollary . the key contribution of this paper is to ask what happens to arbitrage sequences when the number of goods or assets under consideration is four , rather than the two , or occasionally three , usually considered . the model is illustrated with regard to a foreign exchange market with four currencies and traders , so there are @xmath487 principal exchange rates . despite abstracting from various complications such as transaction costs , capital requirements and risk that are often invoked to explain the limits to arbitrage , we find that the arbitrage operations conducted by the fx traders can generate periodicity or more complicated behaviour in the ensemble of exchange rates , rather than smooth convergence to a `` balanced '' ensemble where the law of one price holds . we use the fiction of an arbiter , who knows all the actual exchange rates and what a balanced ensemble would be , to bring out the information problem . fx traders tend to specialise in particular currencies , so the assumption that the fx traders are initially aware only of the exchange rates for their own `` domestic '' currencies is not entirely implausible . we show that the order in which the arbiter reveals information to individual traders regarding discrepancies in exchange rate ensembles makes a key difference to the arbitrage sequences that will be pursued . the sequences are periodic in nature , and show no clear signs of convergence on a balanced ensemble of exchange rates . the arbiter might know the law of one price exchange rate ensemble , but the traders have little chance of stumbling onto such an ensemble by way of their arbitrage operations . the analysis in the present paper raises several issues to pursue in future research . an obvious extension is to allow for a larger number of currencies and ask what happens to the arbitrage sequences as this number becomes large . one interesting modification of the analysis would allow the fx traders to learn that arbitrage sequences tend to be periodic and modify their arbitrage strategies to take the periodicity into account . another modification would allow some arbitrage operations to be pursued simultaneously , and ask what happens as the limiting case where all arbitrage operations are exploited simultaneously is approached . an alternative reformulation of the analysis would be as a markov process where the states are sextuples of exchange rates between the four currencies and the passages between the states reflect the effects of arbitrage operations pursued . it would be interesting to see if this could be done without compromising the relative simplicity of the present formulation . finally , but by no means exhaustively , it would be interesting to work with high frequency data sets to test for the existence of the types of arbitrage sequences postulated in the present paper . applegate , d. l. , bixby , r. e. , chvtal , v. , cook , w. j. , 2006 . the traveling salesman problem : a computational study . princeton series in applied mathematics . princeton university press , princeton , nj . covrig , v. , melvin , m. , august 2002 . asymmetric information and price discovery in the fx market : does tokyo know more about the yen ? journal of empirical finance 9 ( 3 ) , 271285 . http://ideas.repec.org/a/eee/empfin/v9y2002i3p271-285.html dsouza , c. , 2008 . price discovery across geographic locations in the foreign exchange market . bank of canada review spring 2008 , 1927 . http://ideas.repec.org/a/bca/bcarev/v2008y2008ispring08p19-27.html evans , m. d. d. , lyons , r. k. , may 2005 . meese - rogoff redux : micro - based exchange - rate forecasting . american economic review 95 ( 2 ) , 405414 . froot , k. a. , kim , m. , rogoff , k. , nov . the law of one price over 700 years . imf working paper wp/01/174 , international monetary fund , washington dc . http://www.imf.org/external/pubs/ft/wp/2001/wp01174.pdf kozyakin , v. , 2004 . a short introduction to asynchronous systems . in : aulbach , b. , elaydi , s. , ladas , g. ( eds . ) , proceedings of the sixth international conference on difference equations . crc , boca raton , fl , pp . 153165 . kozyakin , v. s. , 1990 . algebraic unsolvability of a problem on the absolute stability of desynchronized systems . i telemekh . ( 6 ) , 4147 , in russian , translation in automat . remote control 51 ( 1990 ) , no . 6 , part 1 , 754759 . kozyakin , v. s. , 2003 . indefinability in o - minimal structures of finite sets of matrices whose infinite products converge and are bounded or unbounded . i telemekh . ( 9 ) , 2441 , in russian , translation in autom . remote control 64 ( 2003 ) , no . 9 , 13861400 . kozyakin , v. s. , may 2003 . asynchronous systems : a short survey and problems . preprint 13/2003 , boole centre for research in informatics , university college cork national university of ireland , cork . marshall , b. r. , treepongkaruna , s. , young , m. , 2007 . exploitable arbitrage opportunities exist in the foreign exchange market . discussion paper , 10 september , massey university , palmerston north , new zealand . http://wwwdocs.fce.unsw.edu.au/banking/seminar/2007/exploitablearbitrage_marshall_sept13.pdf meese , r. a. , rogoff , k. , february 1983 . empirical exchange rate models of the seventies : do they fit out of sample ? journal of international economics 14 ( 1 - 2 ) , 324 . http://ideas.repec.org/a/eee/inecon/v14y1983i1-2p3-24.html osler , c. l. , march 2005 . stop - loss orders and price cascades in currency markets . journal of international money and finance 24 ( 2 ) , 219241 . http://ideas.repec.org/a/eee/jimfin/v24y2005i2p219-241.html tsitsiklis , j. n. , blondel , v. d. , 1997 . lyapunov exponents of pairs of matrices . a correction : `` the lyapunov exponent and joint spectral radius of pairs of matrices are hard when not impossible to compute and to approximate '' . control signals systems 10 ( 4 ) , 381 . http://dx.doi.org/10.1007/bf01211553

we propose a model to account for the bolometric light curve , quasi - continuum and the ca ii emission features of the peculiar type ia supernova ( sn ) 2002ic , which exploded in a dense circumstellar envelope . the model suggests that the sn ia had the maximum possible kinetic energy and that the ejecta expand in an approximately spherically symmetric ( possibly clumpy ) circumstellar environment . the ca ii and quasi - continuum are emitted by shocked sn ejecta that underwent deformation and fragmentation in the intershock layer . modeling of the ca ii triplet implies that the contribution of the o i 8446 line is about 25% of the 8500 feature on day 234 , which permits us to recover the flux in the ca ii 8579 triplet from the flux of 8500 blend reported by deng et al . ( 2004 ) . we use the ca ii doublet and triplet fluxes on day 234 to derive the electron temperature ( k ) in the ca ii line - emitting zone and the ratio of the total area of dense fragments to the area of the shell , . we argue that ca ii bands and quasi - continuum originate from different zones of the shocked ejecta that reflect the abundance stratification of the supernova . stars : mass - loss supernovae : general supernovae : individual ( sn 2002ic )

astro-ph0408508

recently , strong evidence has been presented that a sub - class of type iin supernovae ( sn ) ( ` n ' stands for narrow h@xmath2 ) includes rare sn ia events exploding in a dense circumstellar medium ( csm ) ( hamuy et al . 2003 ) . to date , the new variety comprises three events : sn 2002ic , sn 1997cy and sn 1999e ( cf . deng et al . statistical arguments indicate that the new family constitutes less than one percent of all sn ia ( chugai & yungelson 2004 ) . all these supernovae remain very luminous after the light curve maximum , which is related to the cs interaction . a light curve model of sn 1997cy suggests that the cs envelope within @xmath3 cm contains several solar masses ( chugai & yungelson 2004 ) . two options for the origin of the csm have been proposed ( hamuy et al . 2003 ) : mass - loss by a supergiant in a binary scenario of sn ia ( whelan & iben 1973 ) or the mass - loss by a single supergiant in the sn1.5 scenario ( iben & renzini 1983 ) . the similarity of sn 2002ic events seems to favour the sn1.5 scenario , with the progenitor mass around @xmath4 ( chugai & yungelson 2004 ) . the possible scenario of a c / o white dwarf ( wd ) merging with the co - core of a supergiant companion ( livio & riess 2003 ) faces a serious problem in explaining the short time lag between the ejection of the common envelope and the explosion ( chugai & yungelson 2004 ) . the study of the cs interaction and modeling of the spectra should elucidate the late evolution and the origin of these interesting events . an important recent conclusion is that sn 2002ic interacts with an equatorially concentrated csm ( deng et al . this statement is based upon the interpretation that line profiles in sn 2002ic are too broad compared with the predictions of a spherical model . also , spectropolarimetric observations of sn 2002ic reveal the presence of polarization which is likely to be related to the csm ( wang et al . the claimed asymmetry of the csm may have an important implication for the problem of the evolutionary scenario , because the single and binary scenarios are likely to possess different degrees of symmetry . the aspherical model predicts variations of the emission line profiles of sn 2002ic - like supernovae viewed from different angles . however , all three known objects ( see deng et al . 2004 , their fig . 1 ) show similar shapes and widths of spectral features . this fact implies that , despite the detected polarization of the emission from sn 2002ic , the interaction possesses a high degree of sphericity . in view of this , we investigate a spherical model here . we attempt to explain the light curve , the ca ii line profiles and the quasi - continuum of sn 2002ic - like events on the basis of a circumstellar interaction model . we rely on the spherical approximation , although we allow for local deviations from sphericity ( e.g. clumpy structure ) . we start with the simulation of the cs interaction dynamics and the bolometric light curve of sn 2002ic using the thin shell approximation ( section 2 ) . although the model is the same as before ( chugai & yungelson 2004 ) , we include more realistic ejecta abundances , which affects the light curve . we then analyse the line formation in the inhomogeneous intershock layer and propose a simple treatment of this problem ( section 3 ) . using this approximation we simulate the ca ii line profiles and the quasi - continuum of sn 2002ic with the kinematic parameters of the interaction model . we demonstrate that the spherical model successfully accounts for the basic observational properties of sn 2002ic and discuss implications . the interaction of a sn ia with a dense spherical cs envelope and the light curve formation is calculated using the model applied earlier for the analysis of the cs interaction of sn 2002ic ( chugai & yungelson 2004 ) . we recapitulate the basic assumptions of the model and emphasize some modifications . the density distribution in a freely expanding sn ia ejecta ( @xmath5 ) is approximated by the exponential law @xmath6 with @xmath7 determined by the the mass @xmath8 and kinetic energy @xmath9 . an exponential law better represents the density distribution in sn ia ejecta than a power law ( dwarkadas & chevalier 1998 ) . in an attempt to provide the maximum expansion velocity of the interaction shell we consider a sn ia model with the maximum kinetic energy , corresponding to the detonation model det1 of khokhlov et al . ( 1993 ) : @xmath10 erg with the @xmath11ni mass of @xmath12 . we do not claim that the det1 detonation regime actually occurs in sn 2002ic . this model strongly underproduces intermediate mass elements ( e.g. ca ) ( khokhlov et al . 1993 ) , which is at odds with the strong ca ii lines observed in sn 2002ic ( see below sec . . parameters of the det1 model are adopted because this case illustrates what happens if all of a chandrasekhar mass c / o wd is burned . the important point here is that the complete incineration of c and o is not required : if the burning in the outer layers is terminated at the intermediate elements , the released nuclear energy is practically the same as in the case of complete incineration . the cs density distribution is described by power laws ( @xmath13 ) with different power law indices for each of three suggested radial zones : @xmath14 in the range @xmath15 , @xmath16 for @xmath17 , and @xmath18 for @xmath19 . the variation of the power law index is suggested by two observations : ( i ) the luminosity decline after about day 400 ( deng et al . 2004 ) suggests a cs density drop beyond some radius @xmath19 ; ( ii ) the rising early visual luminosity of sn 2002ic ( wood - vasey et al . 2002 ; hamuy et al . 2003 ) suggests a flat ( or rising outwards ) cs density distribution in the inner part ( @xmath15 ) of the cs envelope ( chugai & yungelson 2004 ) . the hydrodynamic interaction is calculated in the thin shell approximation , in which the layer between the forward and reverse shocks is replaced by a geometrically thin shell ( chevalier 1982b ) . the kinetic luminosities of the forward and reverse shock waves are converted into x - ray luminosities using transformation coefficients dependent on cooling rates of the postshock gas and the expansion time ( chugai 1992 ; chevalier & fransson 1994 ) . for the cs density range of interest , the reverse shock is always radiative , which leads to the formation of a cool dense shell ( cds ) composed of the shocked sn ejecta . the forward shock can be also radiative and contribute to the cds formation . the x - ray radiation of both shocks absorbed by the cool material ( sn ejecta , cds , and csm ) is identified with the optical bolometric luminosity powered by the cs interaction . the additional luminosity powered by the radioactive decay @xmath11ni @xmath11co @xmath11fe is calculated using an analytical theory for the sn light curve ( arnett 1980 ) . the resulting light curve is the sum of the interaction and radioactive luminosities . an important modification in the present version is the inclusion of a ` realistic ' chemical composition of the sn ia ejecta instead of the previously assumed solar composition ( chugai & yungelson 2004 ) . the sn ia composition is now approximated by a mixture of fe and si , where the amount of fe is equal to the ejected @xmath11ni mass . the new composition results in more efficient cooling in the reverse shock and more efficient absorption of the x - rays in the sn ejecta . the adopted absorption coefficient for x - rays in the sn material is @xmath20 @xmath21 g@xmath22 , i.e. , 50 times larger than that with a solar composition . owing to the efficient absorption of x - rays in the sn ia ejecta with the ` realistic ' composition , the rate of emission of optical photons per unit of kinetic luminosity is higher in the present model compared with the earlier version . our modeling of the bolometric optical light curve is compared with the sn 2002ic data from deng et al . ( 2004 ) , modified here for a hubble constant of @xmath23 km s@xmath22 mpc@xmath22 . two versions of the light curve computations along with the corresponding shell radius and velocity are presented in fig . [ f - blc ] . fig . [ f - blc]a shows the model with @xmath24 , whereas the second model ( panel _ b _ ) corresponds to @xmath25 . both cs density distributions are shown in fig . [ f - rho ] . the model with @xmath25 provides the largest expansion velocities of the thin shell ( @xmath26 ) and of the sn at the boundary ( @xmath27 ) . from the point of view of the interpretation of the observed broad line profiles , this model is preferred . henceforth , this case is referred to as the standard model . the flatter cs density distribution of the model compared with a steady - state wind implies that after the main mass loss episode , the mass loss rate gradually decreased while the presupernova evolved towards the explosion . the total mass of the cs envelope in the standard model is @xmath28 . this estimate should be considered a lower limit , because we neglect the possible wind at @xmath29 that may comprise a significant mass , despite its lower density . if the wind velocity out to @xmath30 cm is @xmath31 km s@xmath22 , then the age of the strong mass loss is @xmath32 yr and the mass loss rate is @xmath33 yr@xmath22 . we recalculated the light curve of sn 1997cy for the detonation model with a ` realistic ' composition and found that the mass of the cs envelope in this case is @xmath34 , a factor two lower than our previous estimate for sn 1997cy based upon a model with lower explosion energy and normal composition of the sn ejecta ( chugai & yungelson 2004 ) . most of the difference is related to the efficient conversion of the kinetic luminosity into the bolometric luminosity owing to the efficient absorption of x - rays by the ejecta . the model of the optical light curve predicts the luminosity and flux of the escaping x - rays , which may be used as an additional test of the model for sn 2002ic - like events . in fig . [ f - xlc ] , we plot the total luminosity of the escaping x - rays from the forward and reverse shocks , together with the temperatures of both shocks for the standard model . due to the high cs density and rather flat cs density distribution , the velocity jump at the inner shock is relatively high ( fig.[f - blc ] ) . this , combined with the higher molecular weight of the sn ia matter , results in a high temperature of the reverse shock that is comparable to the temperature of the forward shock after day 250 . in a sn ii interacting with a wind , the reverse shock is usually markedly slower and cooler than the forward shock ( chevalier 1982b ) . remarkably , the luminosity of the forward shock is large and provides a substantial contribution to the optical sn luminosity . this is due to the fact that the forward shock is radiative : it enters the fully radiative regime on day @xmath35 . the reverse shock is radiative through the time period we have modeled . the inflection in the forward shock luminosity on day @xmath36 reflects the corresponding inflection in the adopted cs density distribution at @xmath37 cm ( fig . [ f - rho ] ) . the forward shock becomes nonradiative soon after that point , at @xmath38 d. the photon spectrum of the emergent x - ray flux at different epochs between day 30 and day 600 is shown in fig . [ f - xfc ] together with the detection limit of the jem - x monitor of the integral satellite assuming a 3@xmath39 detection level with the integration time of @xmath40 s ( http://astro.estec.esa.nl/integral/ ) . the plot demonstrates that the detection of x - rays from sn 2002ic is not expected in our model for a reasonable integration time . on the other hand , a sn 2002ic - like event exploding at a distance @xmath41 mpc would be detectable at the favorable epoch of 200400 days by the jem - x monitor with an integration time @xmath42 s. chandra and xmm may be even more efficient in detecting the x - rays from distant sn 2002ic - like events . between days @xmath43 the predicted flux of sn 2002ic in the @xmath44 kev band is @xmath45 erg s@xmath22 @xmath46 . given the chandra detection limit @xmath47 erg s@xmath22 @xmath46 in the @xmath44 kev band , we conclude that x - rays from sn 2002ic could have been detected at an age @xmath43 days . at later epochs , the flux quickly decreases ( fig . [ f - xfc ] ) and on day 600 it becomes far below the chandra sensitivity . the above model for the escaping x - ray flux assumes a smooth csm . clumpiness may modify these results in two ways : a different temperature in the forward shock , and a different escape probability for photons through the clumpy csm . the late time subaru spectrum on day @xmath48 ( assuming the explosion date jd 2452585 ) consists of broad emission bumps ( we dub them quasi - continuum ) with superimposed emission lines of h@xmath2 , h@xmath49 , ca ii 3900 , and ca ii 8600 blended with o i 8446 ( deng et al . most of the bolometric luminosity in our dynamical model comes from the shocked ejecta . we , therefore , believe that , apart from h@xmath2 and h@xmath49 and other narrow lines , most of the optical spectrum on day 234 ( i.e. quasi - continuum and ca ii lines ) originates from shocked sn ejecta . this conjecture will be checked below . we only briefly address the important issue of h@xmath2 line formation . although the cs origin of the narrow h@xmath2 emission is certain , the nature of the broad h@xmath2 is unclear . multiple thomson scattering as a mechanism for the broad h@xmath2 , applied earlier to sn 1998s ( chugai 2001 ) , and discussed by wang et al . ( 2004 ) in connection with sn 2002ic , requires a large optical depth @xmath50 . this is in conflict with our dynamical model that predicts a thomson optical depth of the undisturbed cs ejecta of @xmath51 on day @xmath52 . in our opinion , the broad h@xmath2 comes from the shocked cool dense hydrogen . there are three possible origins for that line - emitting gas : ( i ) the outer layers of the sn ejecta may contain the remains of the hydrogen envelope of the presupernova ( in the sn 1.5 scenario ) , so the cool shocked ejecta enriched by hydrogen may be partially responsible for the broad h@xmath2 component ; ( ii ) the forward shock becomes radiative in the case of smooth cs hydrogen after about day 150 , so the shocked cs hydrogen cools at that epoch and becomes able to emit broad h@xmath2 ; ( iii ) in the case of a clumpy cs envelope radiative shocks in cs clouds and the shear flow around the shocked clouds may produce the broad velocity spectrum of the line - emitting gas that can be responsible for the intermediate and broad h@xmath2 line . in reality , all three sources of the cool shocked hydrogen may contribute to the broad h@xmath2 component . however , only in the third option can the observed intermediate h@xmath2component with a width of @xmath53 km s@xmath22 ( hamuy et al . 2003 ) be produced . henceforth , we concentrate on the study of the ca ii emission lines and the quasi - continuum . momentum and mass conservation imply a very small thickness of the cds , of the order of @xmath54 , where @xmath55 is the velocity of the reverse shock and @xmath56 is the ion sound speed in the cool gas . with @xmath57 km s@xmath22 at an age of 230 - 250 d ( see fig . 1 ) , adopting the temperature of the cool gas of @xmath58 k and the ion average atomic weight @xmath59 ( i.e. , @xmath60 km s@xmath22 ) , @xmath61 . for such a thin shell the velocity dispersion along the line of sight is negligible , so the line radiation transfer in this shell may be treated in the static approximation ( chevalier and fransson 1994 ) . the expansion effect appears only in the transformation of the shell emission to the observer frame . however , the picture of a spherical thin cds is an idealization ; in reality , as we will show below , the shell should be strongly distorted . a straightforward argument against the existence of a geometrically thin spherical shell is based upon the observed profiles of the ca ii features . it is well known that in the optically thin case the line profile from a geometrically thin expanding shell should be boxy . this might be a sensible approximation for the observed ca ii features given some unavoidable perturbation and smearing of the boxy profile . the problem , however , is that the ca ii lines are expected to be optically thick , not thin . to show that the dense shell is opaque in the ca ii lines , we consider the conditions on day 240 . the swept up sn mass in the reverse shock in the standard model is @xmath62 and the shell radius is @xmath63 cm . assuming that the swept up mass resides in the cds and that the ca abundance among fe - peak elements is solar , i.e. , @xmath64 by mass ( a sensible approximation for a rough estimate ) one gets the column density of ca ii in the cds of @xmath65 ( where @xmath66 is the ionization fraction of ca ii ) . the optical depth in the ca iidoublet lines is then @xmath67 where @xmath68 is the oscillator strength for the ca ii doublet and @xmath69 km s@xmath22 is the thermal velocity assuming @xmath70 k. on day 234 , the number density in the cds under pressure equilibrium is @xmath71 @xmath72 . assuming lte in the cool dense gas we obtain @xmath73 for @xmath74 k. the equation ( [ eq - tauca ] ) shows that the optical depth of the cds in the ca ii doublet is very large for the estimated ca ii ionization fraction . it is easy to check that , given the low excitation potential of the ca ii term @xmath75d , the ca ii triplet should be optically thick too . optically thick ca ii lines from a geometrically thin shell are expected to be double - horned ( m - shaped ) @xmath76 with peaks at the shell velocity ( @xmath77 ) ( gerasimovic 1933 ) . this type of profile is apparently inconsistent with the smooth , round - topped ( flat - topped ) ca ii doublet and triplet profiles observed in sn 2002ic ( see deng et al . 2004 ) . to resolve this controversy , we follow cid fernandes & terlevich ( 1994 ) in suggesting that the round - topped h@xmath2 line profile observed in sne ii interacting with a dense cs environment are produced by a thin shell corrugated due to the rayleigh - taylor ( rt ) instability . it is well known that the decelerated thin shell in the intershock layer is subject to the rayleigh - taylor ( rt ) instability ( chevalier 1982a ; nadyozhin 1985 ) . rt spikes grow and degrade via shearing and advective flows , thus producing a two - fluid mixture consisting of dense cool gas and hot postshock gas behind the forward shock wave . as a result , a mixing layer with an extent @xmath78 forms between the thin shell and the forward shock as demonstrated by 2d and 3d simulations ( chevalier & blondin 1995 ; blondin & ellison 2001 ) . the clumpy structure of the csm may cause additional perturbations of the mixing layer and lead to the corrugation of the forward and reverse shocks ( blondin 2001 , his fig . if both reverse and forward shocks are radiative , the shell instability becomes even more vigorous due to the nonlinear thin shell instability ( blondin & marks 1996 ; hueckstaedt 2003 ) . the above arguments based upon the profile shape and the interaction hydrodynamics suggest that a realistic model of the intershock line - emitting layer should include the inhomogeneous distribution of the dense shocked gas . to treat the line formation in the inhomogeneous intershock layer , we reduce it to a spherical ` shell ' of a thickness @xmath79 in the range of radii @xmath80 filled with similar , randomly oriented , dense planar fragments ( fig . [ f - cart ] ) . each fragment is a thin sheet with an area @xmath39 and total surface area @xmath81 . as a result of the deformation and stretching of dense fragments , their cumulative area @xmath82 should , generally , exceed the area of the unperturbed spherical dense shell @xmath83 by a factor @xmath84 . as we will see below , this parameter ( called the ` area ratio ' ) turns out to be crucial for the interpretation of the high luminosity of the ca ii features in sn 2002ic . moreover , the high luminosity of the ca ii features requires @xmath85 . this is the reason why we omit in our simple model ( fig . [ f - cart ] ) the thin spherical shell that is produced and destroyed continuously in the reverse shock . we consider two plausible velocity distributions of filaments in the shell @xmath86 : a constant velocity and a linear law @xmath87 . line photons emitted by randomly distributed dense fragments may escape or be absorbed by another fragment of the shell . to describe the line radiative transfer inside the inhomogeneous shell we apply the local escape probability language adapted for the ` gas ' of opaque macroscopic chunks of dense material . definitions of the local optical depth and emissivity must be modified accordingly . the total number of fragments in the shell is @xmath88 and the fragment number density is @xmath89 , where @xmath90 . the average cross - section of a fragment is @xmath91 with the ratio @xmath92 for plane sheets and @xmath93 for deformed sheets . for instance , for a cylindrical surface with a radius @xmath94 and a height @xmath95 this ratio is @xmath96 with the accuracy of @xmath97 . the radiation escapes from an optically thick fragment in the frequency band @xmath98 assuming a gaussian absorption coefficient @xmath99 $ ] , where @xmath100 and @xmath101 is defined by the condition @xmath102 , i.e. , @xmath103 . a fragment absorbs the incident radiation in the same frequency band , @xmath98 . the analog of the sobolev optical depth in our case is the local occultation optical depth , @xmath104 . to take into account the finite optical depth of a fragment this expression should be multiplied by the absorption probability of a fragment @xmath105 . for a shell with constant velocity , the velocity gradient along some direction is @xmath106 , where @xmath107 is the angle between the photon wave vector and the radius . the expression for @xmath108 then reads @xmath109 where @xmath92 is used . for the linear velocity case , @xmath110 . to estimate @xmath111 let us adopt @xmath112 km s@xmath22 , @xmath113 km s@xmath22 and @xmath114 . equation ( [ eq - tauoc ] ) gives @xmath115 . therefore , the shell becomes optically thick ( @xmath116 ) for an area ratio @xmath117 provided @xmath118 . the emissivity of the fragment ensemble is @xmath119 , where @xmath120 is the frequency integrated flux . in the isothermal approximation and assuming boltzman populations , the flux is @xmath121 , so the emissivity is @xmath122 to compute the luminosity , one needs to take into account self - absorption via the angle - averaged escape probability @xmath123/q\rangle$ ] . assuming pure absorption one gets the luminosity @xmath124 , where @xmath125 is the shell volume @xmath126 . with the line emissivity @xmath127 from equation ( [ eq - emiss ] ) , the luminosity of ca ii triplet is @xmath128 the luminosity increases linearly with the area ratio for @xmath129 ( i.e. , @xmath130 ) and saturates for large @xmath131 , when @xmath111 becomes also large . actually , for large @xmath111 one gets @xmath132 so @xmath133 becomes independent of @xmath131 . remarkably , the line luminosity according to equation ( [ eq - lumir ] ) exceeds the upper limit on the luminosity of a static shell by a factor @xmath134 that , for large area ratio , may be substantially greater than unity . this apparent paradox is related to the velocity gradient that permits the volume emission regime for resonance photons as opposed to the surface emission regime in the static case . a finite albedo @xmath135 of a fragment for incident photons requires a modification of the escape probability : @xmath49 in equation ( [ eq - lumir ] ) should be replaced by @xmath136 $ ] . however , we compute the line profiles on the assumption of zero albedo , which , as we checked , does not affect the profile significantly . in the limit of a small area ratio ( @xmath137 ) the luminosity of the ca ii triplet on day 234 for the probable temperature range @xmath138 k for @xmath139 is @xmath140 erg s@xmath22 . on the other hand , using the flux in the 8500 band , @xmath141 erg s@xmath22 cm@xmath22 on day 234 ( deng et al . 2004 ) , we find the luminosity @xmath143 erg s@xmath22 which significantly exceeds the upper limit found for @xmath144 . this means that either @xmath145 or the emission feature at 8500 is dominated by o i 8446 as deng at al . ( 2004 ) suggest . below we will see that the contribution of the o i 8446 line is small so the high luminosity of the ca ii triplet is primarily due to the large area ratio . to show the effect of @xmath111 we consider the transparent case @xmath146 and modestly opaque shell @xmath147 for @xmath148 . the calculated profiles for a constant velocity shell with @xmath112 km s@xmath22 and @xmath149 are shown in fig . [ f - single ] ( upper panel ) . the finite optical depth smooths the boxy profile as expected . in the limit of @xmath150 , the profile is parabolic with a weak asymmetry due to relativistic effects . the relativistic effects arising from the lorentz transformation of the frequency , wave vector and intensity are usually ignored in sn spectrum calculations . however as was demonstrated by jeffery ( 1993 ) these effects become noticeable even for expansion velocities of @xmath151 km s@xmath22 and generally result in the skewing of an emission component towards blue . although the contribution of the net emission from the unshocked sn ejecta is small for sn2002ic given the domination of the shock luminosity , the sn ejecta can affect the line profile via the scattering of line photon emitted by the cds . if the sn ejecta are optically thick in the same line , then a line photon emitted by the shell inwards can scatter on the unshocked ca ii ions at the resonance point with the distance from the emission point @xmath152 where @xmath153 . the resonance points reside on a sphere of radius @xmath154 that contains the emitting point ( @xmath155 for @xmath156 ) . to compute the effect of the non - local scattering of line photons in sn ejecta we adopt the boundary velocity of the sn ejecta @xmath157 km s@xmath22 ; this value and the shell velocity ( @xmath158 km s@xmath22 ) are close to velocities in the standard model around day 230 ( fig . [ f - blc ] ) . let @xmath159 be a cut - off radius ( in units of @xmath160 ) defined in such a way that the sobolev optical depth of the sn ejecta @xmath161 in the outer zone , @xmath162 , and @xmath163 in the inner zone , @xmath164 . in both models the effect of non - local resonance scattering is shown for two cases : @xmath165 and @xmath166 , assuming @xmath167 in the unshocked sn ( fig . [ f - single ] , lower panel ) . profiles demonstrate significant skewing towards the blue and the emergence of a symmetric broad base produced by scattered photons in the fast sn ejecta with velocity @xmath168 . the shape of the line profile depends on @xmath159 : the skewing is larger for larger @xmath159 . as we see below , the non - local scattering can account for the asymmetry of the ca ii doublet observed in sn 2002ic . we start with the modeling of the ca ii doublet . a monte carlo technique is used for the radiation transfer computations . the parametrized emission rate of the cds fragments in lines of the ca ii doublet is set assuming equipartition of the upper levels . the non - local interaction in the shell is treated as a pure absorption process , whereas the resonance interaction of line photons with the unshocked sn ejecta is considered as pure scattering . the quasi - continuum radiation , presumably emitted by the shell can be scattered by ca ii lines of sn ejecta , whereas the scattering in other lines ( e.g. , fe ii ) is neglected . the quasi - continuum intensity , the slope of the emergent spectrum , and the relative line intensity are free parameters . .model parameters [ cols="<,^,^,^,^,^,^,^ " , ] [ t - par ] the calculated ca ii doublet assuming a constant velocity and an occultation optical depth @xmath146 ( model dc0 ) and @xmath147 ( model dc1 ) , and assuming a linear velocity distribution with the same @xmath111 values ( models dl0 and dl1 , respectively ) are shown in fig . [ f - cad ] overplotted on the spectrum on day 244 ( wang et al . 2004 ) . the computed spectra are smoothed by a gaussian filter with fwhm=7 to take into account the finite spectral resolution . velocities @xmath26 and @xmath27 of the standard interaction model as well as other parameters are presented in table 1 . the table gives also @xmath111 , the optimal cut - off radius ( @xmath159 ) and the optical depth of the sn ejecta in the ca ii 3934 line ( @xmath169 ) . we found that efficient scattering in the sn ejecta is needed with @xmath170 in ca ii 3934 . the optimal cut - off radius for all the models , @xmath165 , coincides with the inner boundary of the line - emitting shell . all four models shown in fig . [ f - cad ] reproduce the major properties of the observed ca iidoublet , particularly , the width , shape and the skewing towards the blue . the deviation in the red wing may be related to the behavior of the underlying quasi - continuum in the observed spectrum . models with constant and linear velocity laws in the shell are practically indistinguishable . we will use below only the linear law model that presumably mimics the possible random component of the velocity field in the shell . using models dl0 and dl1 , we also simulated the ca ii triplet ( fig . [ f - cat ] ) . the model profiles are overplotted on the observed profile in the subaru spectrum on day 234 ( deng et al . 2004 ) . parameters of the corresponding models ( tl0 and tl1 ) including velocities of the standard model are given in table 1 together with the optimal contribution of the o i 8446 line to the 8500 feature @xmath171 . the overall fit of the model tl0 is satisfactory . the derived contribution of the o i 8446 line to the 8500 feature @xmath172 is constrained with an uncertainty of @xmath173 . however , model tl1 with @xmath147 shows a pronounced central absorption feature caused by the resonance non - local absorption in the shell . this feature is inconsistent with the observed profile . we conclude that the fragmented shell should be characterized by a low probability of non - local radiation coupling in the ca ii triplet . we will return to this issue below ( sec . the observed 8500 profile is slightly broader than the model feature , especially in the wings . however , one must be cautious about attributing the wings to the ca ii triplet because the quasi - continuum also contributes to the 8500 feature , as will be demonstrated below . given the uncertainties of the behavior of the quasi - continuum and the satisfactory fit of the ca ii doublet and triplet , we conclude that the ca ii features are consistent with a spherical model and expansion velocities predicted by the interaction model . we performed computations with velocities multiplied by a factor larger than unity . we found that slightly larger expansion velocities ( by @xmath174% ) produce better fits to the 8500 features . however , this conclusion is uncertain because of a possible contribution from the quasi - continuum . the contribution of the o i 8446 line to the 8500 feature found on day 234 ( @xmath175% ) , combined with the total flux @xmath176 erg s@xmath22 @xmath46 ( deng et al . 2004 ) , implies a flux in the o i 8446 line @xmath177 erg s@xmath22 @xmath46 , and @xmath178 erg s@xmath22 @xmath46 in ca ii triplet . given the ca ii doublet flux @xmath179 erg s@xmath22 @xmath46 in the same spectrum ( deng et al 2004 ) , one gets the triplet - to - doublet ratio @xmath180 . assuming that both ca ii excited terms ( @xmath75d and @xmath75p ) have the same excitation temperature and using the scaling @xmath181 from the ratio @xmath182 we obtain @xmath183 k. the absolute luminosities of the ca ii features can be used to estimate @xmath184 and the area ratio @xmath131 simultaneously ( see equation [ [ eq - lumir ] ] ) . to make the procedure reliable , we abandoned the assumption that the excitation temperature for both excited terms is the same . instead , the statistical equilibrium for a three level ca ii model is solved with collisional and radiative transitions taken into account . the radiation transfer in a fragment is treated in the static average escape probability approximation . the adopted ca abundance is 0.03 by mass , the same as used in section [ sec - shell ] . the ionization fraction of ca ii is calculated in the lte approximation , which implies that ca ii is the dominant stage for a wide range of temperature , @xmath185 k. the solution of statistical equilibrium then yields the excitation temperature in both ca ii transitions as a function of the electron temperature @xmath186 and @xmath131 , which permits us to calculate the luminosities of the ca ii features ( equation [ [ eq - lumir ] ] ) and to determine @xmath186 and @xmath131 from a comparison with the observed luminosities . the ca ii levels are found to be well thermalized , so to reproduce the observed luminosities of ca ii features we need only two parameters : @xmath186 and @xmath131 , assuming @xmath149 . figure [ f - lums ] shows the behavior of the luminosities of the ca ii features as a function of @xmath131 for the optimal temperature @xmath187 k. the plot clearly demonstrates that the observed luminosities of the ca ii features can only be reproduced for a large area ratio , @xmath1 . this result provides strong evidence that the dense material of the shocked ejecta experienced substantial deformation , fragmentation and mixing in the intershock layer . we studied the sensitivity of the luminosity of ca ii lines to the adopted mass locked in line - emitting fragments ( @xmath188 ) and found that in the range of @xmath189 the luminosity of both ca ii multiplets are well reproduced for @xmath190 and can not be reproduced for smaller mass , @xmath191 . this indicates that the amount of the ca - rich matter locked in the ca ii line - emitting gas is @xmath192 ( for the mass fraction of ca @xmath193 ) . the conclusion that the area ratio is as large as @xmath194 raises a serious problem . the local occultation optical depth of the shell with an area ratio @xmath1 is @xmath195 . this seems inconsistent with the claim ( sec . 3.2 ) that the non - local radiative coupling in the ca ii triplet is weak . for the spherical shell , this requirement implies @xmath129 . in an attempt to resolve the inconsistency , we considered a model of the ca ii feature formation with a finite albedo @xmath135 of dense fragments for the incident line photons . to determine @xmath135 we performed a monte - carlo simulation of the history of the incident ca ii doublet and triplet photons assuming a three level model of ca ii . the line luminosities for the finite albedo are shown in figure [ f - lums ] . the required value of the area ratio in the new model is only slightly lower compared with the previous model , so the problem of large @xmath111 remains unsettled . a way out might be in the assumption that the accepted model of the fragmented shell is too simple to predict the upper limit for the occultation optical depth from the ca ii triplet . for instance , the large scale corrugation of the shell and/or patchy structure of the line - emitting zones might result in voids lacking the absorbing dense gas at the resonance points . as a result , the large local value of @xmath111 may be reconciled with the low absorption probability for the non - local interaction in the ca ii triplet . we now check our conjecture that the quasi - continuum of sn 2002ic is produced by the emission of the fragmented cool shocked ejecta in numerous lines of fe - peak elements . to this end , we compute the emergent spectrum assuming a model of the fragmented shell width @xmath196 for two values of the area ratio @xmath197 and @xmath144 . the shell patchiness indicated by the ca iitriplet modelling will be described by the surface ( and volume ) filling factor @xmath198 of quasi - continuum emitters . it should be emphasized that the filling factor refers only to the volume fraction of mixture and not to the volume fraction of the dense material that is , of course , much lower . the radiation transfer in the shell will be computed assuming a spherical shell , with two corrections . first , the emitting volume will be multiplied by the factor @xmath198 , and , second , we take into account a finite probability of the photon escape through the holes of the opposite side of shell , @xmath199 . here @xmath200 , the cosine between the photon wave vector and the radius , describes approximately the angular dependence of @xmath201 . the unshocked ejecta will be considered as transparent and the net emission of this ejecta will be neglected . we also omit the presence of a shocked h - rich cool gas in the intershock layer . we consider the seven most abundant fe - peak elements , ti , v , cr , mn , fe , co , ni , assuming their relative abundance to be solar . the latter is a sensible approximation for the explosive nucleosynthesis in the full incineration regime . doublet and triplet lines of ca ii are also included in the line list . the initial line list is extracted from the vald data base ( kupka et al . 1999 ) assuming conditions in a red giant atmosphere with @xmath202 k and gravitational acceleration @xmath203 . the list contains about 19000 lines of neutral and singly ionized species with a line optical depth @xmath204 in the range of @xmath205 . using this list , we compile the input line list for a particular density , assuming saha ionization with a temperature @xmath206 and boltzmann excitation with a temperature @xmath207 . the line optical depth is restricted by the condition @xmath208 . for the relevant conditions , the most abundant species are singly ionized ( @xmath209 ) , while the ionization fraction of neutrals is @xmath210 and that of double ions is @xmath211 . the radiation transfer is calculated using a monte carlo technique . the photon interaction with a dense fragment is treated as a partial absorption with an albedo @xmath135 . we adopt @xmath212 for all the lines , although we found that the result is practically the same in the case of pure absorption ( @xmath213 ) . scattering is assumed to be isotropic for the back and forward scattering with a diffusion reflection probability @xmath214 . the computed spectrum is reddened using @xmath215 ( deng et al . 2004 ) . an extensive search in the parameter space ( @xmath206 , @xmath207 , @xmath198 ) for the model with @xmath197 led us to conclude that a best fit of the quasi - continuum in the spectrum on day 244 is attained for the parameters @xmath216 k , @xmath217 k , and @xmath218 ( fig . [ f - qcon]a ) . note , the result is not very much sensitive to the area ratio ; we adopt @xmath197 , not @xmath219 that follows from ca ii line analysis , because the latter value leads to a too large occultation optical depth ( see discussion in section [ sec - area ] ) . on the other hand , the optimal model with the small area ratio , @xmath144 and @xmath220 ( fig . [ f - qcon]b ) seems to be inconsistent with the observed shape of the quasi - continuum . this implies that the model with a large area ratio is preferred thus supporting the conclusion made from ca ii line analysis . the fit is good despite the fact that the quasi - continuum model is rudimentary ; the number of free parameters is small . the agreement suggests that we are on the right track in the interpretation of the quasi - continuum . this result makes the identification of [ o i ] , [ o ii ] , [ ca ii ] lines for the 6400 and 7300 bumps ( cf . deng et al . 2004 ) unneccessary . otherwise , there would be serious problems in the interpretation of the identified emission lines . another remarkable fact is that the quasi - continuum contributes to the 8500 emission feature thus alleviating the problem of the origin of broad wings in the ca ii/ o i 8500 feature . we checked the contribution of different ions to the quasi - continuum and found that , as expected , the fe ii lines are dominant . this elucidates the role of the ionization temperature as a fitting parameter that regulates the contribution of neutral species , fe i in particular . the quasi - continuum in the region @xmath221 is very reminiscent to that in early sn 1988z , which also has been identified with fe ii lines ( chugai & danziger 1994 ) . interestingly , the smooth quasi - continuum in the range @xmath222 (fig . [ f - qcon]a ) is the result of radiation saturation at the planck intensity due to the high density of lines in this band that results in the large optical depth of the shell , viz . @xmath223 , where @xmath224 is the number of lines in the characteristic wavelength interval @xmath225 with the average local occultation optical depth @xmath226 . the saturation effect is absent in a model with small area ratio @xmath137 ( fig . [ f - qcon]b ) . because the line density in the ultraviolet ( uv ) region is very high , we expect that the uv quasi - continuum should retain a smooth black - body appearance with @xmath227 k in the short wavelength region ( @xmath228 ) . as shown in fig . [ f - qcon]a , the quasi - continuum model is unable to produce strong ca ii emission features . this is related to the fact that the quasi - continuum is a superposition of a large number of weak lines affected additionally by the line overlapping . we propose , therefore , that the ca ii emission and quasi - continuum should form in _ different _ zones of the shocked ejecta . these zones , presumably , correspond to fe - poor ( ca - rich ) gas and fe - rich gas , respectively . the obvious physical reason for these distinctive zones is the abundance stratification of sne ia with the inner layers incinerated up to fe - peak elements and the outer layers having undergone incomplete burning . in this situation one expects that the outer fe - poor gas that has passed through the reverse shock first has enough time to spread over the large volume of the intershock layer , whereas the inner fe - rich material shocked recently occupies only a small fraction of the intershock layer . to illustrate the proposed scenario for the ca ii lines , we consider a combined model of the fragmented shell : apart from the quasi - continuum line - emitting zones suggested by the model with @xmath218 , we consider the ca ii line - emitting zones with a filling factor @xmath229 , where @xmath230 is a fitting parameter . we assume a ca abundance @xmath193 by mass ( the result only weakly sensitive to ca abundance ) and neglect the fe - peak elements in the ca ii line - emitting zone . for ca ii lines , the scattering in the unshocked sn ejecta on ca iiis taken into account as described in the previous section . the same area ratio @xmath197 and @xmath217 k are adopted in the ca ii and quasi - continuum zones . we omit here the contribution of o i 8446 . the result presented in fig . [ f - qcon]c is obtained for @xmath231 , which suggests that @xmath232% of the fragmented shell volume is occupied by fe - poor ca ii line - emitting material . the filling factor again refers only to the volume fraction of the mixture and not the dense material . although we do not intend to describe the ca ii profiles and intensities in detail , the combined model provides a satisfactory representation of the overall spectrum , except for h@xmath2 and h@xmath49 . the modelling thus confirms the proposed scenario based upon the expected abundance stratification in the outer layers of the sn ejecta . we now address the question of how low the abundance of fe - peak elements in the ca ii line - emitting matter can be and retain sufficient contrast between the quasi - continuum and the ca ii lines , especially the ca ii doublet that resides in the line - rich spectral range . the answer depends on the assumed mass locked in the ca ii line - emitting dense gas ( @xmath188 ) . we adopt @xmath233 and a ca abundance 0.03 , which is close to the admissible lower limit ( @xmath234 ) and is consistent with the mass and composition of the fe - poor ca - rich layer produced by the delayed detonation model ( iwamoto et al . 1999 ) . for these parameters and the values of @xmath206 and @xmath207 adopted above for the case of @xmath197 , we found that the mass fraction of fe - peak elements should not exceed @xmath235 or @xmath236 of solar abundance . this upper limit indicates a low primordial metallicity of the pre - sn and , possibly , of the host galaxy as well . to summarize , the major observational properties of the sn 2002ic spectrum at an age of @xmath237 days may be explained in the framework of a model of a spherical , locally inhomogeneous shell with an expansion velocity predicted by the standard hydrodynamic model . the aim of this paper was to develop a unified model for the light curve and spectrum of sn 2002ic assuming approximate spherical symmetry . our principal result is that the ca ii line profiles and the quasi - continuum are consistent with the interaction model provided that the sn ia has the maximum possible kinetic energy . the high explosion energy is in line with the high peak absolute magnitude of sn 2002ic . according to hamuy et al . ( 2003 ) , at light maximum sn 2002ic was as bright as @xmath238 ( @xmath239 km s@xmath22 mpc@xmath22 ) . given the contribution of the veiling continuum of about @xmath240 mag on dec . 3 in @xmath125 band ( @xmath241 ) ( hamuy et al . 2003 ) , we conclude that the peak absolute magnitude of the sn ia was @xmath242 , close to that of extremely bright sn ia , sn 1991 t ( @xmath243 , cf . fisher et al . 1999 ) . the magnitude at the maximum coincides also with that predicted by the detonation model det1 ( hflich & khokhlov 1996 ) . this coincidence is another justification for our choice of the high kinetic energy for the sn 2002ic event . yet , as we noted above , the strong ca ii emission lines indicate that the explosion should be different from the det1 model in the sense that the degree of incineration in the outer layers should be lower to produce enough ( @xmath244 ) ca - rich fe - poor material . modeling of the spectra of the fragmented cool dense shell formed in the intershock layer provides strong evidence that the quasi - continuum and ca ii lines are emitted by different zones . this presumably reflects the stratification of the composition of the sn ejecta in the sense that the outer layers responsible for the ca ii lines are fe - poor , while the inner layers responsible for the quasi - continuum are fe - rich . the fe - poor ejecta are shocked earlier and , therefore , spread throughout the intershock layer , while fe - rich ejecta are shocked later and thus have not yet mixed with the fe - poor material at @xmath245 d. this picture predicts that at a later epoch ( @xmath246 d ) both fe - poor and fe - rich shocked components should be mixed with each other . as a result , the ca ii doublet must disappear at a later epoch due to the mixing of the ca ii line - emitting gas with optically thick quasi - continuum emitters . sn 1999e showed a gradual disappearance of the ca ii 3945 doublet in spectra between days 139 and 361 ( rigon et al . 2003 ) , in accord with these expectations . the high luminosity of the ca ii doublet and triplet in sn 2002ic on day 234 indicates that the dense shocked ejecta form a structure with a large area ratio , @xmath247 . a high area ratio ( @xmath248 ) is also preferred by the quasi - continuum model . the area ratio , which is common in the treatment of the physics of mixing fluids ( e.g. , catrakis et al . 2002 ) , has not been used before in the interpretation of optically thick emission lines from an intershock layer . it was recognized earlier that the interpretation of the x - ray and radio emission of an interacting sn depends on the presence of a mixing layer , while the h@xmath2 emitted by the mixed dense material may reveal the fine structure related to the rt spikes ( chevalier & blondin 1995 ) . here we add to the list of effects of the mixing layer the crucial role of the large area ratio for the interpretation of the ca ii lines and quasi - continuum in sn 2002ic . the fact that the area ratio in sn 2002ic on day 234 is large , @xmath1 , suggests that we observe an advanced stage of the mixing of dense shocked ejecta in the intershock zone . experiments on the mixing of fluids show that the interface between fluids reveals fractal behavior . specifically , the cumulative area of the interface above the scale @xmath249 is @xmath250 , where @xmath251 ( sreenivasan et al . 1989 ) . if this law also holds in the case of the mixing of the dense shocked ejecta in the intershock layer of sn 2002ic , then the dense gas should produce structures with scales as small as @xmath252 . the minimum linear scale that corresponds to the thickness of the dense fragments should be even smaller , @xmath253 , where @xmath254 is the original thickness of the cds ( section [ sec - mech ] ) . it would be interesting to check the possibility of creating the mixing layer in sn 2002ic with @xmath1 in numerical experiments using high resolution 3d hydrodynamical simulations , although it seems to be beyond the capabilities of present - day computers . to estimate the area ratio from a numerical experiment , one needs to find the average number of intersections of dense peaks with the radius in the intershock layer @xmath255 . the area ratio is then @xmath256 . we do not yet fully understand all the mechanisms responsible for the major spectral features seen in the late emission from sn 2002ic . in addition to the h@xmath2 line , an important issue ignored in our study is the problem of the o i 8446 emission , although it is not a dominant constituent of the 8500 bump anymore . the oxygen excitation mechanism should predict the absence of strong o i 7774 in sn 2002ic . the fluorescence pumping of o i 8446 in the h@xmath2-emitting gas ( deng et al . 2004 ) meets this requirement . this mechanism predicts a comparable intensity of o i 11287 . the absence of this line in the spectrum of sn 1999e around @xmath257 d ( rigon et al . 2003 , their fig . 6 ) casts doubt on the fluorescence mechanism . an alternative possibility is that the o i 8446 line is emitted by the oxygen - rich matter in the outermost layers of the sn ejecta due to non - thermal collisional excitation by secondary electrons that are created by high energy photoelectrons . excitation by secondary electrons also yields a high o i 8446 / o i 7774 intensity ratio ( stolarski & green 1967 ) . an interesting possibility that ly@xmath2 pumping line of fe ii 8451 might contribute to 8500 band instead of o i 8446 , as was the case , possibly , for sn 1995n ( fransson et al . 2000 ) , seems unlikely here because of the absence of another strong ly@xmath2 fluorescent band of fe ii at @xmath258 . despite of a possible presence of a large amount of oxygen in the cool shocked sn gas ( up to @xmath259 ) , the strong [ o i ] line is not expected . adopting boltzmann population for @xmath260 k and oxygen mass of @xmath261 one gets the luminosity of this line of about @xmath262 erg s@xmath22 , i.e. , a factor of @xmath263 lower compared to the luminosity of the observed `` 6300 '' emission band on day 244 . in view of the possibility that our hydrodynamic model may underestimate the expansion velocity of the sn / csm interface by about @xmath264% , we discuss the implications for our model . the emergent x - ray luminosity from the forward shock in the standard model is comparable or even exceeds the optical bolometric luminosity after maximum light ( see fig . [ f - blc ] and fig . [ f - xlc ] ) . hence , there is an extra energy reservoir which might , in principal , be transformed into the optical band under the proper conditions . this possibility could be realized in a model of interaction with a clumpy csm . as a result of ( i ) softer x - rays from cloud shocks than for a smooth csm , and ( ii ) the larger geometrical probability of absorption of x - rays from clouds penetrating the sn ejecta , the optical output from the interaction should be larger . therefore , the same bolometric luminosity can be produced for a lower average cs density and higher expansion velocity of the sn / csm interface compared with a smooth csm . the spherical geometry of the cs interaction of sn 2002ic , advocated here , finds additional support from statistical arguments already mentioned in the introduction . we know three events of sne ia in a dense csm ( sn 1997cy , sn 1999e , sn 2002ic ) and they show similar widths and shapes of spectral features at a similar phase ( see deng et al . 2004 , their fig . this fact suggests that a bi - polar structure of the sn / csm interface with high aspect ratio is unlikely . at first glance the picture of spherical interaction is incompatible with the detection of intrinsic polarization of sn 2002ic radiation at a level of @xmath265 % ( wang et al . simple estimates show that the thomson scattering in a moderately aspherical cs envelope with @xmath51 is not able to produce such a high polarization . the issue may be resolved in a model in which the polarization is acquired in the scattering of the sn radiation on the dust in a distant ( @xmath266 cm ) aspherical csm . the resemblance of the light curves for all three sn 2002ic - like events ( see deng et al . 2004 ) suggests a similar regime of mass - loss from their presupernovae and a common origin of the progenitors . as mentioned before ( chugai & yungelson 2004 ) , the uniformity lends support for the sn1.5 evolutionary scenario suggested by iben & renzini ( 1983 ) . yet we do not rule out the binary scenario . in this case , the wd companion should be a supergiant with the mass @xmath267 ( the cs envelope , @xmath268 , plus the supergiant core , @xmath269 ) . in the binary scenario , there is no clear reason for the supergiant to have experienced strong mass loss just before the sn explosion . the synchronization of the vigorous mass loss and the sn event might be a more natural outcome in the single agb star scenario with the c / o core approaching the chandrasekhar limit . a general belief is that chandrasekhar mass white dwarfs can not be produced from intermediate - mass single stars because the growth of the c / o core during the thermal pulsation stage is prevented by the removal of the hydrogen envelope ( blcker 1995 , and references therein ) . in this regard , it is highly remarkable that at least two supernovae ( sn 2002ic and sn 1997cy ) exploded in dwarf galaxies , as already noted by chugai & yungelson ( 2004 ) . dwarf galaxies are characterized by low metallicity , which thus might explain why the co core of an asymptotic giant branch star is successful in attaining the chandrasekhar limit before the hydrogen envelope is lost ( chugai & yungelson 2004 ) . an evidence for the low metallicity of the presupernova of sn 2002ic comes from our modelling of the ca ii lines and quasi - continuum ( sec . [ sec - cont ] ) . the possibility of sn1.5 explosions in metal - poor galaxies was suggested also by zijlstra ( 2004 ) . it would be interesting to check whether all the host galaxies of sn 2002ic - like supernovae are actually metal deficient . if the suggested link between sn 2002ic - like events and low metallicity is correct , then their occurrence rate relative to normal sn ia must increase with the redshift for @xmath270 , when the average cosmic metallicity was lower than solar . if sn 2002ic - like events stem from sn 1.5 explosions , their low fraction of sn ia , @xmath271% , can be used to estimate the the fraction of white dwarfs created with a mass similar to the chandrasekhar mass . if only @xmath272% of all wds can become sn ia , this suggests that only @xmath273 of white dwarfs in our galaxy were formed with a mass close to the chandrasekhar mass . since these white dwarfs are likely to have formed in a metal poor environment , they are probably very old ( age @xmath274 yr ) and thus of low luminosity , @xmath275 ( prada moroni & straniero 2003 ) . the number of known cool white dwarfs is low , @xmath276 ( salim et al . 2004 ) , and will probably remain too low for the nearest future to test the idea whether high mass white dwarfs were produced during early epochs when the metallicity was low . the possible link between sn 2002ic - like supernovae and hypothetical sn1.5 events , combined with arguments in favour of a high energy sn ia event in this case , raises the question of whether conditions in a hot chandrasekhar mass c / o wd that just got rid of the remnants of the hydrogen envelope favour a wd explosion with a high kinetic energy of ejecta . the sn1.5 origin of sn 2002ic suggests that the initial mass at the main sequence is @xmath277 . with the chandrasekhar core and @xmath28 in the dense cs envelope we expect that about @xmath278 had to be lost at the previous agb stage . this means that outside the dense cs envelope ( @xmath279 cm ) there should be a dense red supergiant wind with a velocity of the order @xmath280 km s@xmath22 and a mass loss rate of @xmath281 yr@xmath22 . keeping in mind the analogy with sn 1979c , this wind may be revealed in the infrared ( ir ) as an ir echo ( dwek 1983 ) , and in the radio as a result of cs interaction ( chevalier 1982a , weiler et al . however , the radio emission from the distant sn 2002ic ( @xmath282 mpc ) may pose a difficult challenge to detect . if the radio luminosity is comparable to that of sn 1979c ( weiler et al . 1986 ) then the expected radio flux from sn 2002ic around 1000 day should be of the order of @xmath283 mjy in the @xmath284 cm band , which is close to the detection limit of vla . a more promising target for the detection of the late time interaction of sn 2002ic - like supernovae with the presumed agb wind is sn 1999e at a distance of @xmath285 mpc , or any closer sn 2002ic - like event that will be discovered in the future . note , that at the epoch of the interaction with the dense cs shell ( @xmath286 d ) , the radio emission is severely attenuated by free - free absorption at wavelengths @xmath287 cm . with the parameters used in our models , and adopting temperatures for the shell and wind in accordance with the models of lundqvist & fransson ( 1988 ) , we find that free - free absorption is a likely reason why the observations of berger & soderberg ( 2003 ) and stockdale et al . ( 2003 ) did not detect radio emission from sn 2002ic . a better understanding of the phenomenon of sn 2002ic - like events requires an accurate energy audit , for which the direct detection of the x - ray emission as well as ir measurements are important . sn 2002ic - like events at an age of @xmath43 days are excellent targets for chandra , xmm and integral . we are grateful to lifan wang for sharing his spectrum of sn 2002ic and tanya ryabchikova for the help in retrieving the line list from vald . nnc received partial support from rfbr grant 02 - 17255 and rac from nsf grant ast-0307366 . support from the royal swedish academy is also acknowledged . pl is further supported by the swedish research council . arnett w.d . , 1980 , apj , 237 , 541 blcker t. , 1995 , a&a , 297 , 727 berger t. , soderberg a.m. , 2003 , iauc no . 8157 blondin j.m . 2001 , in s.s . holt and u. hwang eds . young supernova remnants . eleventh astrophysics conf . , aip , p.59 blondin j.m , marks b.s . , 1996 , newa , 1 , 235 blondin j.m . , ellison d.c . 2001 , apj , 560 , 244 catrakis h.j . , agguirre r.c . , ruiz - plancarte j. 2002 , j. fluid . mech . , 462 , 245 chevalier r.a . , 1982a , apj , 258 , 790 chevalier r.a . , 1982b , apj , 259 , 302 chevalier r.a . , blondin j.m . , 1995 , apj , 444 , 312 chevalier , r.a . , fransson , c. , 1994 , apj , 420 , 268 chugai n.n . , 1992 , , 36 , 63 chugai n.n . , danziger i.j . , 1994 , mnras , 268 , 173 chugai n.n . 2001 , mnras , 326 , 1448 chugai n.n . , yungelson l.r . , 2004 , astron . lett . 30 , 65 cid fernandes r.c . , terlevich r. , 1994 , in tenorio - tagle g. ed . violent star formation . from s doradus to qso . cup , cambridge , p. 365 deng j. , kawabata k.s . , ohyama y. , nomoto k. , mazzali p.a . , wang l. , jeffery d.j . , iye m. , 2004 , apj , 605 , l37 dwarkadas v.v . , chevalier r.a . , 1998 , apj , 497 , 807 dwek e. , 1983 , apj , 274 , 175 fisher a. , branch d. , hatano k. , baron e. , 1999 , mnras , 304 , 67 fransson c. et al . , 2002 , apj , 572 , 350 gerasimovic b.p . , 1933 , zeitschrift astrophys . , 7 , 23 hamuy m. et al . , 2003 , nature , 424 , 651 hueckstaedt r.m . , 2003 , newa , 8 , 295 hflich p. , khokhlov a.m. , 1996 , apj , 457 , 500 iben i. jr . , renzini a. , 1983 , annual rev . 21 , 271 itikawa y. , ichimura a. , 1990 , j. phys . chem . ref . data , 19 , 637 iwamoto k. , brachwitz f. , nomoto k. , kishimoto n. , umeda h. , hix w.r . , thielemann f .- k . , 1999 , apjs , 125 , 439 jeffery d.j . , 1993 , apj , 415 , 734 khokhlov a.m. , mller e. , hflich p. , 1993 , a&a , 270 , 223 . kupka f. , piskunov n. , ryabchikova t.a . , stempels h.c . & weiss w.w . , 1999 , a&as , 138 , 119 livio m. , riess a. , 2003 , apjl , 594 , l93 lundqvist p. , fransson c. 1988 , a&a , 192 , 221 madej j. , naleyty n. , althaus l.g . , 2004 , a&a , 419 , l5 prada moroni p.g . , straniero o. , 2003 , mem.s.a.it . , 74 , 508 rigon l. et al . , 2003 , mnras , 340 , 191 salim s. , rich r.m . , hansen b.m . , koopmans l.v.e . , oppenheimer b.r . , blandford r.d . , 2004 , apj , 6001 , 1075 sreenivasan k.r . , ramshankar r. , meneveau c. , 1989 , proc . a 421 , 79 stockdale j. , sramek r.a . , weiler , k.w . , van dyk , s.d . , panagia , n. , 2003 , iauc no . 8157 stolarski r.s . , green a.e.s . , 1967 , jgr , 72 , 3967 turatto m. et al . , 2000 , apjl , 534 , l57 wang l. , baade d. , hflich p. , wheeler j.c . , kawabata k. , nomoto k. , 2004 , apj , 604 , l53 whelan j. , iben i. jr . , 1973 , apj , 186 , 1007 weiler k.w . , sramek r.a . , panagia n. , van der hulst j.m . , salvati m. , 1986 , apj , 301 , 790 wood - vasey , w. m. , et al . , 2002 , iauc no . 8019 , 2 zijlstra a.a . , 2004 , mnras , 348 , l23

we present a detailed , quantitative study of the competition between interaction- and disorder - induced effects in electronic systems . for this the anderson - hubbard model with diagonal disorder is investigated analytically and numerically in the limit of infinite spatial dimensions , i.e. within a dynamical mean - field theory , at half filling . numerical results are obtained for three different disorder distributions by employing quantum monte carlo techniques which provide an explicit finite - temperature solution of the model in this limit . the magnetic phase diagram is constructed from the zeros of the inverse averaged staggered susceptibility . we find that at low enough temperatures and sufficiently strong interaction there always exists a phase with antiferromagnetic long - range order . a novel strong coupling anomaly , i.e. an _ increase _ of the nel - temperature for increasing disorder , is discovered . an explicit explanation is given which shows that in the case of diagonal disorder this is a generic effect . the existence of metal - insulator transitions is studied by evaluating the averaged compressibility both in the paramagnetic and antiferromagnetic phase . a rich transition scenario , involving metal - insulator and magnetic transitions , is found and its dependence on the choice of the disorder distribution is discussed .

cond-mat9411034

the investigation of interacting electronic systems is one of the most intriguing , albeit difficult , subjects in condensed matter physics . the same is true for the study of disordered , non - interacting electrons . in view of the theoretical complexity of the two problems taken separately it is understandable that their combination , i.e. the _ simultaneous _ presence of randomness and interactions , as found in many real systems ( e.g. doped semiconductors near the metal - insulator transition , high-@xmath0 superconducting materials close to @xmath0 , etc . ) , leads to new , fundamental questions to which only few secured answers are known . this is all the more true when the interaction and/or the disorder is strong , since there exist hardly any tractable , and at the same time controlled , theoretical method of investigation in this limit . an important starting point for the investigation of interacting , disordered systems was the field - theoretic approach developed for the treatment of non - interacting electrons , i.e. the scaling theory of anderson localization @xcite . a generalization of this theory to finite interactions by finkelshtein @xcite provided essential new insight @xcite . however , the appearance of local magnetic moments in the renormalization group treatment discovered by him @xcite and castellani et al . @xcite turned out to be a fundamental obstacle for the study of the metal - insulator transition ( mit ) itself @xcite . a microscopic origin of this instability towards the formation of localized moments can already be traced within hartree fock theory for the disordered hubbard model ( `` anderson - hubbard model '' ) with off - diagonal disorder @xcite . the results indicate that the above renormalization - group approach , as well as is starting point , i. e. anderson - localization , are not appropriate when it comes to the investigation of three - dimensional disordered electrons with intermediate or strong interactions close to half - filling ( @xmath1 ) . after all , a mit occurs even without disorder in this case , and the lattice periodicity becomes essential . a non - perturbative and qualitatively quite different , but essentially uncontrolled , renormalization technique by which disorder and interactions can be treated on equal footing is the real space renormalization group approach . it was first applied by ma @xcite and recently used by yi et al . @xcite to investigate the anderson - hubbard model with diagonal disorder , i.e. random on - site energies , in dimensions @xmath2 at @xmath3 and in @xmath4 for @xmath5 , respectively . however , in these investigations the formation of antiferromagnetic long range order ( aflro ) which will always set in at large enough repulsion ( at least at @xmath3 ) was not considered at all . the strong coupling limit of this model was studied by zimanyi and abrahams @xcite using a slave - boson formulation of the corresponding @xmath6 model . the considerably simpler , but still highly non - trivial , case of disordered , _ spinless _ fermions in @xmath7 was also addressed recently @xcite . to obtain a global picture of the properties of interacting , disordered systems it is desirable to know the solution of a simple , microscopic model which is valid for all input parameters ( interaction , disorder , temperature , band filling ) . since exact solutions are not available in @xmath8 one would like to construct , at least , a thermodynamically consistent mean - field theory that is valid also at strong coupling . such a ( non - perturbative ) approximation is provided by the exact solution of a model in @xmath9 . it is now known that even in the limit @xmath10 @xcite the hubbard interaction remains dynamical @xcite and leads to a highly non - trivial single - site problem @xcite with infinitely many coupled quantum degrees of freedom . this problem is , in fact , equivalent with an anderson impurity model complemented by a self - consistency condition @xcite and is thus amenable to numerical investigations within a finite - temperature quantum monte - carlo approach @xcite . in the absence of disorder this technique was already used by several groups to investigate the magnetic phase diagram @xcite , the mott - hubbard transition @xcite , transport properties @xcite , and lately also superconductivity in a two - band version @xcite of the hubbard model in @xmath9 . these investigations were also extended to the periodic anderson model @xcite and the holstein model @xcite . the dynamical mean - field theory of interacting electrons obtained in the limit @xmath9 has many appealing features ; however , it also has its limitations which are due to the purely local nature of the theory . one effect which can not be described within the @xmath9 mean - field theory of disordered electrons is the phenomenon of anderson localization . that is , the electrons in infinite dimensions are delocalized for arbitrary strength of the disorder . the absence of localized states is caused by the above mentioned single - site dependence , i.e. by the fact that momentum dependence enters the physical quantities only through the dispersion relation whereby vertex corrections to the conductivity vanish identically @xcite . in fact , it was shown @xcite that for vanishing interaction , when the anderson - hubbard model reduces to the anderson disorder model , the well - known and much - used `` coherent potential approximation ( cpa ) '' @xcite , which does not describe anderson localization , yields the _ exact _ solution in @xmath9 . nevertheless it should be stressed that even in the absence of anderson localization the presence of disorder can have a significant , non - trivial effect on the properties of an interacting system . first results of a quantum monte - carlo study of the competition between disorder and interaction in the anderson - hubbard in @xmath9 with diagonal `` binary alloy '' disorder were recently reported by us @xcite . the importance of separable , off - diagonal disorder in the formation of local moments and the mott transition was investigated by dobrosavljevi and kotliar @xcite . a good impression of how complicated the phase diagram and the dynamical properties of interacting electrons with and without disorder are may be inferred from the results for spinless electron in @xmath9 @xcite . the fact that the interactions reduce to their hartree - contribution in this case @xcite , i.e. are no longer dynamical , does not at all imply that the properties of the model become trivial quite the contrary . many details of this model can be obtained analytically in @xmath9 , thereby providing valuable insight into the effects of interactions @xcite , the interplay between interaction and disorder @xcite , and the @xmath9 limit itself @xcite . the correlations caused by interactions between electrons and by the scattering of electrons in a disordered medium , respectively , can be very different ; indeed , they may lead to opposite effects . for example , even an arbitrarily weak repulsive hubbard interaction between electrons on a bipartite lattice will induce aflro when the band filling is sufficiently close to @xmath3 ( at least in dimensions @xmath11 ) . by contrast , it is the very nature of a random potential to oppose order . hence interactions and disorder are expected to compete with each other . this conclusion may appear self - evident and almost trivial but is really only based on a simple - minded superposition of the two individual physical effects ( which may , in fact , be quite correct for weak interactions and disorder ) . when the interaction and/or disorder is strong , however , this picture looses its basis . in general the mutual interplay of those two effects can be expected to lead to _ novel _ , non - perturbative , quantum - mechanical many - body phenomena , e.g. new phases , which have no analog in non - random , interacting or disordered , non - interacting systems . in view of the interplay between disorder and interaction there are a number of simple questions whose answers will be non - trivial , for example 1 . is there a _ finite _ disorder strength above which aflro ceases to exist ? ( after all , the fluctuations of the ensembles around the average value @xmath3 induced by the disorder might become so large that aflro is made impossible ) . 2 . does disorder always _ reduce _ the critical temperature for aflro , as one might expect intuitively ? do interactions always drive a system _ away _ from metallic behavior into an insulating state ? 4 . how important is the choice of disorder distribution on the results ? these are questions which need to be answered by means of controlled investigations of a well - defined microscopic model . it is our intention in this paper to provide explicit , numerically exact answers to the above questions within the dynamical mean - field theory obtained in the limit of high spatial dimensions . a preliminary account was presented in ref . the paper is structured as follows . in sect . 2 we introduce the model and discuss the form of the corresponding averaged free energy in @xmath9 . in sect . 3 we calculate correlation functions , i.e. the averaged compressibility and staggered susceptibility , to determine the thermodynamic stability of various phases . the details of the model ( input parameters , disorder distribution etc . ) are specified and the numerical procedure is discussed in sect . 4 . this leads to the construction of the magnetic phase diagram in sect . 5 . in sect . 6 the results on magnetic phase transitions are then complemented by those on metal - insulator transitions . a discussion of the combined phase transition scenario closes the presentation . the simplest microscopic hamiltonian for conduction electrons interacting via the local hubbard interaction in a disordered system is the so - called anderson - hubbard model , which can be written as @xmath12 here @xmath13 create ( annihilate ) a @xmath14-electron on site @xmath15 , with @xmath16 and @xmath17 is the chemical potential of the @xmath14-electrons . in general both the hopping amplitude @xmath18 and the atomic energies @xmath19 in ( [ gl1 ] ) depend on the configuration of the atoms distributed on the lattice . in this paper we restrict ourselves to the case of uncorrelated , diagonal disorder . hence we assume the hopping elements to be independent of the randomness ( with @xmath20 for nearest neighbor sites and @xmath21 otherwise ) while the atomic potentials @xmath19 are chosen as random , their distribution being purely local ; thus short - range order is neglected . the hamiltonian under investigation is then given by @xmath22 where @xmath23 is the usual hubbard model @xmath24 here the hopping matrix element was written as @xcite @xmath25 with @xmath26 as the number of nearest - neighbors on the lattice ( e.g. @xmath27 on hypercubic lattices in @xmath28 dimensions ) . this scaling ensures that ( [ gl3 ] ) remains non - trivial even in the limit @xmath29 . we can then write the averaged full energy ( grand potential ) as [ gl5 ] @xmath30 where @xmath31 . here the average over the disorder is defined as usual as @xmath32 x ( \epsilon_1 , \ldots , \epsilon_l ) \label{gl5b}\ ] ] where @xmath33 is the number of lattice sites , and @xmath34 is the distribution of the random potentials @xmath19 . the explicit form of the function @xmath34 will be specified later . in realistic dimensions @xmath35 it is impossible to perform the trace over all fermionic states in ( [ gl2 ] ) exactly . however , in the limit of infinite coordination number or spatial dimensions significant simplifications occur in the evaluation of thermodynamic properties . two of us recently showed @xcite that the averaged grand potential of the disordered hubbard model in @xmath9 can be reduced to an expression where the averaging is performed only on a single site . the explicit expression for the grand potential in the paramagnetic phase is given by @xmath36 \nonumber \\[10pt ] & + & \beta^{-1 } \sum_{\sigma n } \ln g_{\sigma n}^{-1 } - \beta^{-1 } \big\langle \ln { \cal z } \ { { \cal g}^{-1 } , \epsilon_i \ } \big\rangle_{av } \quad , \label{gl6}\end{aligned}\ ] ] where @xmath37 is the density of states of non - interacting electrons , @xmath38 are matsubara frequencies and the generalized atomic partition function @xmath39 is represented by a grassmann functional integral [ gl7 ] @xmath40 with the local action @xmath41 \label{gl7b}\ ] ] here @xmath42 are grassmann variables that depend on the imaginary time @xmath43 , and @xmath44 are their fourier transforms into @xmath45-space . only the random potentials @xmath19 are configuration dependent . furthermore , @xmath46 is an effective local propagator of the electrons . the grand potential ( [ gl6 ] ) is a functional of the complex quantities @xmath47 and @xmath48 which at this point enter as variational parameters . that is , the _ physical _ values of @xmath47 and @xmath48 ( namely , the local green function , @xmath49 , and the self - energy , @xmath50 , of the electrons , respectively ) correspond to those for which ( [ gl5 ] ) is stationary @xcite . each term of @xmath51 in ( [ gl6 ] ) has a clear mean - field interpretation , which allows us to outline a simple construction of @xmath51 on a mean - field level @xcite . we first introduce a homogeneous @xcite , effective , energy - dependent potential @xmath48 which is defined in such a way that in the thermodynamic limit the system of disordered , interacting electrons is equivalently described by a system of non - interacting electrons in the potential @xmath52 . the free energy @xmath53 is now constructed as follows @xcite : ( i ) we start with the free energy of non - interacting electrons in the potential @xmath48 ( first term in ( [ gl6 ] ) ) , ( ii ) then we remove the potential @xmath48 from site @xmath15 , i. e. subtract its energy contribution ( second term in ( [ gl6 ] ) ) , and ( iii ) replace it by the potential @xmath54 , and finally average over @xmath55 ( last term in ( [ gl6 ] ) ) @xcite . to determine the potential @xmath48 we demand that the free energy obtained by this construction be stationary under variation of @xmath48 , i. e. @xmath56 . this is a self - consistent equation for @xmath48 . the major advantage of the mean - field grand potential ( [ gl6 ] ) is the property that the variational parameters are not explicitly configuration dependent @xcite , i.e. they are determined only by _ averaged _ quantities which result from the averaging in ( [ gl6 ] ) . since the averaging is local , the disorder only leads to local correlations on the same site . different lattice sites are effectively decoupled and all the information about the surrounding sites is contained in the effective potential @xmath48 . indeed , @xmath48 is a generalization of the `` coherent potential '' known from the theory of random alloys @xcite . this may be seen as follows @xcite : if we neglect the hubbard interaction @xmath57 , the grassmann functional integral ( [ gl7a ] ) becomes gaussian and can be performed exactly , leading to @xmath58 \nonumber \\[10pt ] & + & \langle \ln ( 1 + g_{\sigma n } ( \sigma_{\sigma n } - \epsilon_i ) ) \rangle_{av } \big\ } \label{gl9}\end{aligned}\ ] ] this is precisely the grand potential of a random alloy with non - interacting electrons within cpa , with @xmath48 as the coherent potential expressed in matsubara frequencies @xcite . hence ( [ gl6 ] ) can be understood as a field - theoretic generalization of the cpa to random alloys with _ interacting _ electrons which becomes exact for @xmath59 . it should be noted that the averaged grand potential ( [ gl6 ] ) does not contain the complete information about the equilibrium physics of random systems @xcite . for example , averaged _ products _ of green functions contain new , non - trivial correlations and can not be derived from ( [ gl6 ] ) . the averaged grand potential does , however , carry the full information about the equilibrium _ thermodynamics _ of the model . we concentrate in our paper exclusively on the thermodynamics of the anderson - hubbard model . it is well - known that , for large enough @xmath57 , the hubbard model itself leads to aflro at ( or even close to ) half - filling . this type of long range order must therefore be considered even in the presence of disorder . for the sake of simplicity we investigate only aflro on bipartite lattices . in this case the breaking of symmetry is caused by a staggered magnetic field @xmath60 with ( @xmath61 on @xmath62- and ( @xmath63 ) on @xmath64-sites . to be able to study this type of order quantitatively we extend the averaged grand potential ( [ gl6 ] ) to the symmetry - broken phase . the averaged grand potential of a general antiferromagnetic solution has the form @xmath65 \nonumber\\[10pt ] & - & \sum\limits_{\sigma , n } \sum\limits_{\gamma = a , b } \ln g _ { \gamma \sigma n}^{-1 } + \sum\limits_{\gamma= a , b } \ ; \langle \ln { \cal z}_{\gamma } \rangle_{av } \big\ } \label{gl10}\end{aligned}\ ] ] where the index @xmath66 corresponds to sublattices @xmath67 respectively , and @xmath68 , with @xmath69 as a magnetic field . the local partition function @xmath70 in ( [ gl10 ] ) is given by ( [ gl7 ] ) with the replacements @xmath71 , @xmath72 etc . , and integration over all grassmann fields , i.e. @xmath73 with @xmath74 the full partition function is given by @xmath75 . as described above the physical values of @xmath76 and @xmath77 in ( [ gl10 ] ) are then found from the stationarity conditions @xmath78 in the following we will set @xmath79 in ( [ gl10 ] ) and exclude the existence of _ ferri_magnetism . in this case the average number of electrons on sublattices @xmath62 and @xmath64 are equal . then there appears a new symmetry : quantities with indices @xmath80 and @xmath81 coincide . hence the double index @xmath80 can be combined into a single one , @xmath82 with @xmath83 for @xmath84 and @xmath85 for @xmath86 . thereby the number of independent parameters is reduced . ( [ gl12 ] ) then yields two coupled sets of self - consistent equations for @xmath87 and @xmath88 [ gl13 ] @xmath89 g_{\alpha n } & = & - \int\limits_0^\beta d\tau \ ; e^{i \omega_n \tau } \langle \langle \psi_{\alpha } ( \tau ) \psi_{\alpha}^ * ( \tau^+ ) \rangle_t \rangle_{av } \label{gl13b}\end{aligned}\ ] ] where @xmath90 . here @xmath91 represents the statistical average with the local action as @xmath92 note that ( [ gl13a ] ) merely expresses the fact that @xmath93 is the local element of the full green function , i.e. is an explicit function of the self energy @xmath88 . ( [ gl13b ] ) describes the actual dynamics of the system , determining the self - energy @xmath88 as a function of the input parameters @xmath94 and the disorder distribution . ( [ gl13 ] ) fully determine the one - particle properties of the anderson - hubbard model in @xmath9 . to decide about the stability of a particular solution of ( [ gl13 ] ) we have to evaluate averaged _ two_-particle correlation functions or susceptibilities . two such quantities are of particular interest in the anderson - hubbard model : the averaged staggered magnetic susceptibility @xmath95 and the averaged compressibility @xmath96 . the former decides about the instability of the paramagnetic phase w.r.t . aflro , while the latter not only contains the information about a possible instability w.r.t . phase separation ( @xmath97 ) , but also provides a thermodynamic criterion for a solution to be insulating ( @xmath98 . quite generally susceptibilities can be obtained from the second derivative of the averaged grand potential @xmath51 w.r.t . to some variable @xmath99 as @xmath100 \kappa_{av } , & x = \mu . % \hspace{72 mm } & { \rm ( 15b ) } \end{array } \right.\ ] ] the first derivative of @xmath51 w.r.t . @xmath99 yields @xmath101 where @xmath102 and @xmath103 . differentiating once more we obtain the desired susceptibilities ( 15 ) . note that these susceptibilities are evaluated at zero staggered field @xmath60 , i.e. @xmath60 is put to zero once the second derivative of @xmath51 w.r.t . @xmath104 or @xmath60 is taken ( in this case @xmath105 ) . thus the susceptibilities are given by @xmath106 where @xmath107 is the local two - particle correlation function @xmath108 @xmath109 \label{gl18}\ ] ] the quantity @xmath110 in ( [ gl17 ] ) measures the response of the averaged medium to an infinitesimal change of the field @xmath99 . we will later see that @xmath111 decides about the ( in-)stability of a given phase . this dynamical response function is determined by an integral equation in frequency - space which does not explicitly depend on momentum . ( note that there are no convolutions in @xmath112-space in the @xmath9 limit as is typical for a mean - field theory ) . this property does not imply , however , that the response function @xmath111 is local , too . it only indicates that @xmath111 is _ diagonal _ in the momentum @xmath113 . a momentum dependence will enter by taking derivatives , e.g. of the local green function , w.r.t . an external field with a particular @xmath113-dependence ( @xmath114 in the case of the compressibility or the ferromagnetic susceptibility , and @xmath115 for the staggered susceptibility ) . infrared divergencies show up in the spectrum of @xmath116 . the integral equations for @xmath111 are derived as follows . using the effective propagator @xmath117 , such that @xmath118 we replace the self - energy in ( 13 ) by @xmath119 . then ( 13 ) reads [ gl20 ] @xmath120 where @xmath121 , with @xmath122 , and @xmath123 to obtain @xmath111 we differentiate ( [ gl20 ] ) w.r.t . this yields [ gl22 ] @xmath124 where @xmath125 and @xmath126 . eliminating @xmath127 from ( [ gl22 ] ) we obtain the following matrix equation for @xmath128 and @xmath129 : @xmath130 the quantities @xmath131 and @xmath132 are a matrix and a vector , respectively , w.r.t . the sublattice - spin index @xmath133 and may be written as @xmath134 where [ gl25 ] @xmath135 \frac{\eta_{-,n}}{g^2_{+,n } } & 1- \frac{\xi_{+,n}}{g^2_{+,n } } \end{array } \right ) \label{gl25c}\ ] ] @xmath136 \eta_{-,n } & \xi_{-,n } \end{array } \right ) \label{gl25a}\ ] ] @xmath137 f_-^x \end{array } \right ) . \label{gl25b}\ ] ] in the paramagnetic phase , where @xmath138 , eqs . ( 23)-(25 ) reduce to ( [ gl6 ] ) in ref.@xcite . to determine the boundary between the paramagnetic and the antiferromagnetic phase it is sufficient to calculate @xmath139 only in the paramagnetic phase . by contrast , @xmath140 has to be evaluated in _ both _ phases since in both phases a metal - insulator transition ( mit ) is possible , in principle . indeed , in the paramagnetic phase of a random alloy with discrete disorder spectrum a mit due to band - splitting may occur ( at @xmath141 this is _ known _ to be the case , at least within cpa @xcite ) , and even in the presence of aflro there are indications that both a metallic and an insulating phase exist @xcite . we will find that the most sensitive indicator for an incipient transition are the dynamic response functions @xmath142 especially their behavior close to the fermi energy ( i.e. their values for the lowest matsubara frequencies ) , rather than @xmath143 and @xmath96 itself . the latter quantities are much less sensitive to changes in the ground state since they represent _ sums _ over the matsubara frequencies . eqs . ( 15)- ( [ gl25 ] ) form the basis for the numerical evaluation of @xmath96 and @xmath144 . before they can be solved quantitatively we have to specify the model parameters we use . the numerical calculations were performed with a semi - elliptic density of states ( dos ) with total width @xmath145 , i.e. @xmath146 this dos is chosen because of its sharp algebraic band edges , resembling those typical for @xmath147 , and its simple analytic form ; it is exact for a bethe lattice in the limit @xmath59 . to study the influence of the disorder we investigate , and compare , three qualitatively different distributions of random potentials : + a ) [ gl27 ] @xmath148 the atomic potentials @xmath149 are chosen with equal probability to ensure an average band filling of @xmath3 ( only in this case , or @xmath1 , is aflro expected to occur at all ) . this distribution is important since it leads to a disorder - induced mit due to band - splitting in the noninteracting system ( an exact result in @xmath9 [ 38,39 ] ) which may compete with the interaction - induced mott - hubbard mit . + b ) @xmath150^{1/2 } \label{gl27b}\ ] ] this is a much softer type of disorder . we found that a continuous , _ constant _ disorder distribution @xmath151 , leads to essentially identical results . + c ) @xmath152 by using @xmath153 one may simulate doping in a non - random system with impurities which do not hybridize with conduction electrons : with probability @xmath99 an infinite energy barrier is created which prevents the electrons from visiting these sites . the three distributions allow us to test the universality of the magnetic behavior of the model obtained for different types of disorder . to be able to study the competition between magnetic order caused by the electronic interactions and the disordering effects caused by the random potential , respectively , we work with an average band filling @xmath154 . due to the symmetry of the distributions ( [ gl27a],b ) we can fix the chemical potential at @xmath155 . this can even be done with ( [ gl27c ] ) in which case the green function has to be weighted with an additional factor of @xmath156 . for the numerical evaluation of the functional integral ( [ gl14 ] ) we employ the algorithm of hirsch and fye @xcite . we discretize the time variable , i. e. @xmath157 , with @xmath158 , and then extrapolate the quantities under investigation to @xmath159 . to obtain a smooth imaginary - time green function @xmath160 even for discrete values of @xmath43 we use instead of @xmath161 ( the fourier transform of @xmath160 w.r.t . matsubara frequencies @xmath45 ) the function [ gl28 ] @xmath162 \label{gl28a } \\ g(\tau ) & = & \beta^{-1 } \sum_n e^{i \omega_n \tau } \overline{g}_n \label{gl 28b}\end{aligned}\ ] ] with @xmath157 . in the continuum limit @xmath163 the functions @xmath164 and @xmath161 coincide . the definition ( [ gl28 ] ) preserves the relation @xmath165 even in the case of discrete @xmath166 and suppresses unphysical and undesirable oscillations in @xmath160 due to the discretization ( see fig . 1 ) . in particular , by smoothing @xmath160 in this simple way large - scale fourier transformations ( namely for every iteration ) described in ref . @xcite become unnecessary . exact summations over spin variables in the discrete hubbard - stratonovich transformation were used whenever possible , i. e. for @xmath167 . for @xmath168 we used monte - carlo sampling . after typically 4 - 8 iterations an accuracy of @xmath169 and @xmath170 was reached in the exact summations and in the monte - carlo sampling with @xmath171 sweeps per iteration , respectively . close to the magnetic transition the convergence becomes significantly slower and the monte - carlo sampling less and less efficient . this implies that in the immediate vicinity of the transition this method can not be used to obtain accurate results . the integration over the continuous disorder distributions was performed using the gauss - legendre quadrature @xcite . a discretization of the random energies in steps of @xmath172 gave an accuracy of @xmath173 for all observables . the required cpu time increases proportional to @xmath174 and @xmath175 for the exact summation and the monte - carlo sampling , respectively @xcite . although the random energies break the particle - hole symmetry we never encountered a minus - sign problem . setting @xmath176 the only remaining physical dimension is that of an energy ( @xmath177 etc . ) or inverse energy ( @xmath178 ) . departing from our earlier convention we now choose the _ half - band width _ @xmath179 as our energy unit since it does not depend on the limiting process @xmath59 . ( this is in contrast to the scaling of the hopping amplitude @xmath180 , where @xmath133 may be chosen at will ) . this convention agrees with that used by kotliar and collaborators @xcite . to be able to compare the results presented in this paper with our earlier ones @xcite , all numerical values of quantities with dimension of energy ( inverse energy ) obtained in ref . @xcite must be divided ( multiplied ) by a factor of 2 . to compare with the results of jarrell s group @xcite and that of georges and krauth @xcite , their numbers have to be divided ( multiplied ) by a factor of @xmath181 . the competition between disorder and correlations in the anderson - hubbard model strongly affects its properties both in the ground state and at finite temperatures . concerning the thermodynamics the influence of different kinds of disorder on the stability of aflro near half filling is of particular interest . to determine the instability of the paramagnetic phase w.r.t . the formation of aflro we evaluate eqs . ( 15 ) - ( 25 ) in the paramagnetic limit for @xmath182 , in which case the dependence on the index @xmath133 drops out ; we set @xmath183 , etc . and @xmath184 . note that the matrix equation ( [ gl23 ] ) separates into two identical scalar equations in this case . as already mentioned below ( [ gl18 ] ) it is not the averaged susceptibility itself which is of primary interest in the investigation of the stability of the paramagnetic phase but the dynamic response function @xmath185 . the averaged susceptibility diverges ( becomes negative ) if and only if the real part of the response function @xmath185 diverges for at least one frequency . this is characteristic for the dynamics of a quantum system of interacting particles possessing infinitely many coupled internal degrees of freedom labelled by the matsubara frequencies . each frequency corresponds to one mode in the quantum mechanical system which is described by complex variables . note , however , that the matsubara frequencies do not index the actual independent modes , since @xmath186 is not diagonal in frequency . only the eigenvalues of the integral equation ( [ gl23 ] ) represent the independent ( eigen-)modes of the interacting quantum system . we can ascribe a critical ( nel ) temperature @xmath187 to each eigenmode @xmath188 whereby @xmath189 if @xmath190 . in fig . 2 we show the function @xmath191 , being a linear combination of eigenmodes for the lowest matsubara frequencies which lie very close to the diverging independent soft mode . the functions @xmath191 diverge due to the contribution of the eigenmode with the highest @xmath192 . the highest critical temperature @xmath193 is the thermodynamic critical temperature @xmath194 . once the function @xmath185 and the local vertex function @xmath195 are known one may calculate the averaged susceptibility . above @xmath194 this susceptiblity must obey the curie - weiss law , provided there is such a @xmath196 . in fig . 3 three characteristically different temperature dependencies of @xmath197 are shown for one value of ( binary alloy ) disorder , @xmath198 , at different interactions strengths . if @xmath199 is sufficiently large the curie - weiss law with a finite @xmath194 is obeyed as in the case without disorder . for values of @xmath200 the low - temperature behavior begins to be determined by the scattering off the frozen random configurations and deviations from the curie - weiss - law become apparent . for @xmath201 we observe a minimum in @xmath197 ( i.e. a maximum in @xmath139 itself ) which separates the temperature - dominated regime from the disorder - dominated regime . in the latter case the long - range correlations are continually suppressed due to impurity scattering . the phase diagram in the @xmath202-plane calculated from the zeros of @xmath197 is plotted in fig . 4a for different values of the binary alloy disorder . we can distinguish two different regimes , a ) the disorder gradually suppressed the long - range order and thus reduces the critical temperature . this is exactly what one expects from scattering off frozen random configurations ; b ) : here the situation is strikingly different since @xmath194-curves of constant disorder start to _ cross_. this means that a small amount of disorder _ supports _ the formation of aflro , i.e. the critical temperature _ increases _ with disorder as shown in fig . 5 . a maximal critical temperature is reached at some value @xmath203 beyond which a further increase of disorder causes the critical temperature to decrease monotonically to zero . this effect is particularly pronounced at strong coupling . it is also observed , but less pronounced , in the case of the continuous disorder distribution ( fig . 4b ) . in the limit @xmath204 the enhancement of @xmath194 may be explained as follows @xcite . the virtual hopping of an electron with spin @xmath14 from a given site @xmath62 with local energy @xmath205 to a neighboring site @xmath64 with energy @xmath206 occupied by a ( @xmath207 ) electron leads to an energy gain @xmath208 $ ] and to @xmath209 $ ] for the reverse process . the effective spin coupling @xmath210 is given by the sum of these energies , @xmath211 , which for a bounded disorder distribution , @xmath212 , with @xmath213 ( strong coupling ) implies @xmath214 \label{gl29}\ ] ] assuming that even in the presence of disorder the nel - temperature @xmath215 \mbox{$\propto < j ( \epsilon_a , \epsilon_b ) > _ { av}$}$ ] , we find @xmath216 \label{gl29a}\ ] ] for arbitrary disorder distribution @xmath217 . @xmath218 and @xmath219 are the first and second moment of @xmath217 , where @xmath220 . for a symmetric , bounded distribution @xmath221 one therefore finds @xmath222 with @xmath223 , i.e. the disorder is indeed found to _ increase _ @xmath194 irrespective of the type of disorder . this effect is the more pronounced the more structure @xmath217 has at @xmath224 , i.e. is larger for the binary alloy @xmath225 than for the semi - elliptic @xmath226 distribution . this quadratic increase of @xmath194 with @xmath227 is indeed found numerically at large @xmath57 ( fig . apparently the mechanism that enhances @xmath194 is effective already at @xmath57-values as small as @xmath228 . it is interesting to note that for given strength of disorder @xmath227 the @xmath57-value where aflro begins to set in , @xmath229 , given by @xmath230 , tracks almost perfectly with @xmath227 , i.e. @xmath231 . more generally , for binary alloy disorder the @xmath194-curves are found to obey the phenomenological scaling law @xmath232,0 \right ) \left [ 1- \frac{1}{2 } \left ( \frac{\delta}{u } \right)^2 \right ] \label{gl30}\ ] ] for _ arbitrary _ @xmath57 ( fig . 6 ) . for @xmath213 ( 31 ) reduces to ( 29b ) . it is well - known that aflro may be destroyed by doping the ( non - random ) system with holes , such that @xmath233 . this is a very interesting effect both from an experimental and theoretical point of view . the effect of adding holes may be approximated by introducing percolation - type disorder , ( [ gl27c ] ) , where the random site - energy is zero with probability @xmath156 and infinite with probability @xmath99 . since in @xmath9 the critical value for percolation is @xmath234 @xcite aflro will persist up to @xmath234 . the boundaries between the paramagnetic and the antiferromagnetic phases for this type of disorder are shown in fig . 7 . at weak coupling the critical temperature remains constant over a wide range of @xmath99 and then rapidly falls to zero . at strong coupling , however , the critical temperature _ monotonically _ decreases and approaches the linear behavior @xmath235 . the latter dependence of @xmath194 which is very different from the one observed in the previous two cases is an exact result in the case of the disordered falicov - kimball model @xcite . the magnetic correlations are mediated by the local magnetic moments . their static average , @xmath236 , is defined by @xmath237 . here @xmath238 is the average double occupancy of lattice sites . for the binary and the semi - elliptic disorder distributions discussed above @xmath239 is shown as a function of @xmath57 in fig . the disorder is seen to have two main effects which are independent of the specific disorder distribution : 1 ) at a fixed value of @xmath57 an increase of the disorder _ reduces _ the moments . 2 ) for fixed disorder strength @xmath227 an increase of @xmath57 leads to an increase of the moments , with saturation starting at @xmath240 . both features are easily explained in terms of the effect the disorder and the on - site repulsion , respectively , have on the average double occupancy @xmath241 . for @xmath242 the local repulsion is weak while the spatial fluctuations of the atomic potentials are strong , such that @xmath241 is at its maximum value ( @xmath243 and @xmath236 is small . as @xmath57 increases the particles are forced to separate and @xmath241 decreases , i.e. @xmath236 increases , too . the local magnetic moments do not show a critical behavior close to the transition temperature @xmath194 . in the case of the percolation - type disorder ( fig . 8b ) the density of local moments on the reduced lattice with @xmath244 sites increases with concentration @xmath99 . in other words , in the presence of disorder , the saturation of @xmath239 sets in at smaller values of @xmath57 than in the pure system . the influence of the percolation - type disorder can be explained by the reduction of the kinetic energy by a factor of @xmath156 which is due to the reduced average number of nearest neighbors . for large @xmath57 , in the heisenberg limit , the factor @xmath245 in the antiferromagnetic coupling , @xmath246 , has to be replaced by @xmath247 leading to a linear decrease of @xmath194 . for small values of @xmath57 , the influence of the repulsion increases due to the reduction of the kinetic energy . therefore , the double occupancy is suppressed , i.e. @xmath239 is enhanced by the percolation disorder . the enhancement of @xmath239 leads to the stabilization of aflro . this effect is definitely not described within the hartree - fock approximation ; the latter only leads to an exponential suppression of @xmath194 if @xmath57 or @xmath248 becomes small . in fig . 9 the temperature dependence of the order parameter ( the averaged staggered magnetization @xmath249 ) is shown . in the vicinity of the transition point , @xmath250 , it is well represented by a mean - field - type dependence @xmath251 indicated by the dashed line . we observe that the extrapolation of the monte - carlo data to @xmath252 using this law leads to the _ same _ critical temperature as that obtained from the divergence of @xmath144 , i.e. @xmath253 , assuming a curie - weiss law for @xmath254 . deviations from the square - root behavior of @xmath255 set in at lower temperatures . in the preceding section we found that at weak coupling the disorder suppresses the aflro . in the ground state the system is then a disordered paramagnet . whether it is metallic or not depends on the type and strength of the disorder . in @xmath9 , where anderson localization does not occur [ 38,39 ] a paramagnetic insulator only forms if the spectrum of the disorder distribution is multiple connected , such as in the case of the binary - alloy disorder . for this type of randomness a mit due to band splitting is expected to occur at some value @xmath256 . in the non - interacting case , @xmath141 , the exact result in @xmath9 ( obtained by cpa @xcite ) is @xmath257 . for @xmath258 there are two equivalent criteria to decide on whether the ground state in @xmath9 is metallic or insulating . they correspond to ( i ) a _ spectral _ definition of an insulator , based on the disappearance of the dos at the fermi level ( assuming that only extended electrons are present ) , and ( ii ) a _ thermodynamic _ definition , employing the disappearance of the compressibility @xcite . however , in interacting systems these two criteria need not coincide . hence we will investigate both . in fig . 10a the averaged compressibility @xmath96 of interacting electrons in the presence of binary - alloy disorder is shown as a function of @xmath57 for @xmath259 . hence @xmath96 is calculated across the line @xmath260 in fig . 4a , i.e. within the paramagnetic and the antiferromagnetic phase , respectively ( the transition points are indicated by arrows ) . for a given @xmath261 the curves display a common behavior as a function of @xmath57 : @xmath96 is ( exponentially ) small at @xmath141 , then increases and , at @xmath262 , approaches a maximum , beyond which it becomes ( exponentially ) small again for @xmath213 . this behavior has a clear physical interpretation : for @xmath242 , with @xmath261 , the double occupancy of lattice sites is at its maximum value ( @xmath263 ) , at the same time the kinetic energy can not delocalize these states . hence the system is insulating . at @xmath264 and @xmath141 this is an exact property for @xmath265 ( split - band limit ) in @xmath9 . as @xmath57 increases the on - site coulomb repulsion forces the particles to separate from each other . as a consequence the particles become less localized . we then expect that , at some critical value @xmath266 , a macroscopic fraction of the electrons becomes extended and the system starts to be metallic . this real - space picture has its analogy in @xmath112-space ( or w.r.t . energy ) : as the interaction increases the previously separated bands change their shapes , i.e. the interaction - induced energy exchange between particles leads to a transfer of states into the energy gap where the fermi energy is located . at the critical value @xmath266 the ( still algebraic ) band edges reach the fermi level , producing a finite dos there as well as a finite overall compressibility . ( we can not rule out that @xmath267 , i.e. that the dos at the fermi level is finite even at arbitrarily small @xmath57 ; however , this would require the bands to acquire exponential tails a feature which can not be observed numerically ) . hence , in contrast to an interacting system without disorder the coulomb interaction in a disordered system is able to _ improve _ the metallicity of the system and may even turn an insulator into a metal . apparently the interaction - induced energy exchange smooths the energy spectrum of the disordered system , thus leading to an easier transfer of energy . for @xmath268 the system is then expected to be a metal without fermi liquid properties ( since @xmath269 ) , at least in @xmath9 . as @xmath57 is further increased the dos and @xmath96 increase , reaching a maximum at @xmath262 . for even larger @xmath57 the effect of pushing states into the gap is reversed and @xmath96 decreases again . hence at some critical value @xmath270 a _ second _ transition occurs , back into an insulating state , where doubly occupied sites are almost completely suppressed , whereby the mobility is obstructed by the repulsive interaction . in fig . 10b @xmath96 for the continuous , semi - elliptic disorder distribution is shown . here @xmath96 behaves qualitatively as in the non - random case , i.e. @xmath96 decreases monotonically since the dos does not split at @xmath271 . this reduction of @xmath96 occurs in two steps : for @xmath272 @xmath96 decreases almost linearly with @xmath57 , corresponding to the linear suppression of doubly occupied sites @xmath241 , or @xmath273 ( see fig . 8a ) ; in this region the system is metallic but not a fermi liquid . then , for @xmath274 , when the number of doubly occupied sites is almost zero so that the local moments are almost saturated , @xmath96 approaches zero ( exponentially ) slowly . strictly speaking , at finite temperatures one has @xmath275 for @xmath276 . next we discuss the relation of @xmath266 and @xmath270 to the critical interaction strength @xmath277 where the paramagnet becomes unstable w.r.t . aflro . in fig . 10a , b the position of @xmath277 is indicated by an arrow . since the local moments vanish in the ( binary - alloy)-disorder - induced insulating phase , the insulator - to - metal transition at @xmath266 occurs _ before _ the magnetic order sets in , i.e. @xmath278 . this raises the question of what happens as one goes through the magnetic transition into the ordered phase : will the compressibility jump to zero discontinuously or will there be an _ antiferromagnetic metal _ , i.e. does the interaction - induced mit coincide with the magnetic transition or not ? to answer this question the behavior of @xmath96 near the transition point has to be investigated . in fig.11a the change of @xmath96 due to the onset of aflro in the case of binary - alloy disorder is shown . the results for the paramagnetic phase are compared with those for the ordered phase . at @xmath279 the aflro is seen to suppress @xmath96 drastically . by contrast , at finite disorder ( @xmath280 , this difference almost vanishes . apparently the disorder _ stabilizes _ the metallic state close to @xmath277 . a similar behavior is observed for the continuous disorder ( fig . these results suggest that in the vicinity of @xmath277 the system is an _ antiferromagnetic metal_. however , mc - techniques are not able to decide whether this is true even at @xmath264 . we attempted to extrapolate our finite - temperature data to @xmath264 but could not find a simple , accurate extrapolation law . in fig . 12 the averaged compressibility for the system with binary - alloy disorder , calculated at @xmath281 and @xmath282 , is shown . although at the lower temperature @xmath96 is lower and the slope is slightly steeper at the transition point , the critical point ( indicated by an arrow ) seems to be well inside the metallic phase . in particular , the tail behavior for @xmath283 does not show any significant change . to investigate whether , and how , an insulator evolves it is instructive to plot the dynamical response function @xmath284 , or the quantities @xmath285 with @xmath286 as a function of matsubara frequency @xmath287 . in fig . 13a the real part of @xmath288 is shown for a system without disorder , @xmath279 , for @xmath289 and @xmath290 ( the system is then well inside the antiferromagnetic region , see fig . the behavior of @xmath291 close to @xmath292 is seen to be very different for the antiferromagnetic and the ( hypothetical ) paramagnetic solution . for the paramagnetic solution , which is metallic , @xmath291 is a monotonically decreasing function of @xmath293 . by contrast , in the antiferromagnetic ( i.e. insulating ) phase this is only so for @xmath294 , while for @xmath45 close to zero @xmath291becomes strongly negative . in the former case the sum over @xmath288 , i.e. @xmath96 , is then clearly positive , while in the latter case @xmath96 becomes very small due to the negative contributions close to @xmath292 . this behavior at @xmath279 is now contrasted with that in the presence of binary - alloy disorder ( @xmath295 see fig . 13b ) at @xmath296 for two different temperatures : @xmath297 ( close to the magnetic transition ) and @xmath290 ( well inside the ordered phase ; see fig . 4a ) . at @xmath290 the range of @xmath45values for which @xmath298 is now even narrower the in the case without disorder . this shows that a definite answer to the question of whether the antiferromagnetic phase close to @xmath277 is insulating @xmath299 or not at @xmath264 can only be obtained from the behavior of @xmath300 at very small @xmath45 , i.e. from @xmath96 at very low temperatures @xmath301 . within the mc - approach used here such low temperatures are not attainable . the situation is similar in the case of @xmath302 , the imaginary part of the one - particle green function @xmath303 , which yields a _ condition for an insulator ( for @xmath304 and in the limit @xmath305 the function @xmath306 coincides with the dos at the fermi level ) . if @xmath307 for @xmath305 the ground state of the system is insulating . in fig . 14 the influence of the onset of aflro on @xmath308 is shown . although in the disordered system the difference between the magnetic and non - magnetic solution is now slightly stronger than in the case of @xmath96 ( see fig . 13a ) the magnetic transition occurs where @xmath309 is large , i.e. well inside the metallic phase . in fig . 15 the behavior of @xmath302 at the few lowest values of @xmath310 is plotted for the same values of @xmath57 and @xmath311 as for @xmath291 in fig . 13 . in the case without disorder ( fig . 15a ) this behavior is qualitatively similar to that of @xmath291 ( fig . 13a ) : the paramagnetic solution is clearly metallic while the antiferromagnetic phase is insulating since @xmath312 . in the presence of disorder ( fig . 15b ) , however , the downturn of @xmath313 in the phase with aflro ( @xmath314 sets in only very close to @xmath315 i.e. at very low temperatures . if close to the transition the antiferromagnetic phase were really insulating the decrease of @xmath302 will have to be very rapid . in fig . 16 we plotted our monte carlo data for @xmath316 in the case of binary - alloy disorder with @xmath317 as a function of temperature for several values of @xmath57 . both at very small @xmath57 ( i.e. @xmath318 , where the disorder - induced gap of the dos dominates the behavior ) and at large @xmath57 ( heisenberg - limit ) the temperature dependence may be clearly extrapolated to @xmath319 for @xmath320 , i.e. the ground state is insulating . however , for @xmath321 , with @xmath322 , these results can not be safely extrapolated . hence the question whether in the disordered system close to the magnetic transition the antiferromagnetic phase is metallic or not remains open . as in the case of @xmath96 one would have to go to temperature @xmath323 to be able to decide whether the ground state of the disordered system with aflro is metallic or insulating . in this paper we presented a detailed , quantitative study of the physical effects caused by the simultaneous presence of interactions and randomness in a system of lattice electrons . to this end we investigated the anderson - hubbard model with diagonal disorder at half filling in the limit of infinite spatial dimensions , i.e. within a dynamical mean - field theory , for three different disorder distributions . numerical results were obtained by employing quantum monte carlo techniques that provide an explicit finite - temperature solution of the model in @xmath9 . no further approximation were used . to construct the thermodynamic phase diagram we derived and evaluated the appropriate averaged two - particle correlation function , i.e. a dynamical response function , whose poles determine the magnetic instabilities of the disordered , interacting system . only this function and not the averaged staggered susceptibility @xmath144 itself , which is only a weighted sum of the response function over the ( matsubara ) frequencies fulfills a closed equation that determines the two - particle spectrum . the value of the response function at the lowest frequency is the most sensitive indicator for an instability of the system . in the temperature dependence of @xmath144 two distinct disorder regimes are observed : ( i ) for weak disorder the curie - weiss law holds , while ( ii ) at strong disorder @xmath144 acquires a maximum at a temperature below which a crossover from the temperature- to the disorder- dominated regime takes place . we demonstrated that at low temperatures and sufficiently strong interaction there always exists a phase with antiferromagnetic long - range order ( aflro ) . furthermore we discovered a new strong - coupling anomaly , namely that the nel - temperature @xmath194 is not always a monotonously decreasing function of disorder . indeed , at strong coupling and not too large disorder @xmath194 is always found to be an _ increasing _ function of disorder , i.e. disorder favors the formation of aflro in this regime . this implies the existence of an unexpected _ disorder - induced _ transition to a phase with aflro . under the assumption that @xmath194 is proportional to the effective exchange coupling between spins even in the disordered system we proved that for diagonal disorder the anomalous behavior is generic , i.e. is independent of the type of disorder distribution . it is a consequence of the fact that for diagonal disorder the difference between the local energies of neighboring sites becomes _ smaller _ on average , thus leading to stronger effective exchange coupling . we then studied the existence of metal - insulator transitions in the anderson - hubbard model . although in @xmath9 anderson localization does not take place the presence of disorder may well have other strong effects . in particular , binary - alloy disorder is able to cause band - splitting ( thereby resembling the effect of genuine interactions ) and hence may induce a metal - insulator transition all by itself . special attention was given to the question whether or not the disorder allows for the stabilization of an antiferromagnetic ( af ) _ metal_. to this end the average compressibility @xmath96 was evaluated both in the paramagnetic ( p ) and af phase . contrary to our expectation the presence of disorder was found to enhance the metallicity of the af - phase close to the p - af transition . this enhancement strongly suggests the existence of an af metal in the low - temperature phase of the anderson - hubbard mode . to investigate whether this af phase persists to be metallic down to @xmath320 , at least in @xmath9 , we studied the frequency components of the one - particle green function and of @xmath96 , respectively , down to @xmath290 , our lowest temperature . however , a reliable answer to this question can only be found at still lower temperatures , which at present are beyond the reach of the finite - temperature monte carlo techniques used here . in any case , the transition scenario involving metal - insulator and p - af transitions obtained for the anderson - hubbard model is remarkably rich . for alloy - type disorder with @xmath265 ( split - band limit for @xmath324 ) at @xmath320 an increase of the interaction @xmath57 from zero will probably first lead to a transition from a paramagnetic insulator to a paramagnetic metal at @xmath266 , then , at @xmath277 , to an antiferromagnetic metal and finally , at @xmath270 , to an antiferromagnetic insulator . no compelling evidence was found that @xmath277 and @xmath270 coincide . in the case of a continuous disorder distribution one has @xmath267 since band splitting never occurs . the above findings prove that the interplay between electronic interactions and scattering from disorder leads to interesting , and even novel , physical effects . in particular , the strong coupling anomaly discovered here call for an experimental verification . we are grateful to a. altland , p. van dongen , m. jarrell , h. mller - krumbhaar and especially e. mller - hartmann for very useful discussions and comments . this work was supported in part by the sonderforschungsbereich 341 of the deutsche forschungsgemeinschaft . in fact in their exact solution of the falicov - kimball model in @xmath9 brandt and mielsch [ z. phys . * b75 * , 365 , ( 1989 ) ; ibid . * b79 * , 295 ( 1990 ) ; ibid * b82 * , 37 ( 1991 ) ] were the first to show that the local self - energy may be used to map the lattice problem in @xmath9 onto its atomic counterpart with an additional `` kadanoff - baym field '' which has to be determined self - consistently ; for an application to other hubbard - like models see g. hlsenbeck and f. stephan ( preprint ) . x. y. zhang , m. j. rozenberg and g. kotliar , phys . lett . * 70 * , 1666 ( 1993 ) ; m. j. rozenberg , g. kotliar and x.y . zhang , phys . rev . * b49 * , 10181 ( 1994 ) ; q. si , m.j . rozenberg , g. kotliar , and a. e. ruckenstein , preprint . in principle the grand potential ( 6 ) may equally be viewed as a functional of the set of complex functions @xmath48 _ only _ in which case , however , one has to _ supplement _ ( 6 ) by the definition of @xmath47 as the local ( i.e. site - diagonal ) element of the full temperature green function ( see below ) . in the initial formulation of the theory it is advisable to treat _ both _ set of legendre - conjugate variables @xmath48 and @xmath47 as variational parameters , since only then does the expression ( 6 ) contain the full information about the thermodynamic equilibrium of the system @xcite . an equilibrium corresponds to a saddle point where the variations of @xmath51 w.r.t . @xmath48 and @xmath47 vanish , i.e. @xmath325 and @xmath326 . these variational conditions determine the physical values of @xmath47 and @xmath48 as the local green function and self - energy of the electrons , respectively . a theory generated in this way is guaranteed to be thermodynamically consistent and conserving in the spirit of the baym - kadanoff formalism [ g. baym , phys . rev . * 127 * , 836 ( 1962 ) ] . fig.1 : : local green function @xmath160 for @xmath324 and @xmath328 ( a ) , @xmath329 ( b ) obtained by usual fourier transformation ( dotted line ) and by redefinition acc . to eq.(13 ) ( solid line ) fig.2 : : dynamical antiferromagnetic response function @xmath185 vs. temperature @xmath330 . binary alloy with disorder strength @xmath331 and @xmath332 . here and in the following figures lines are usually guides for the eye , and error - bars are roughly of the size of the symbols unless shown explicitly . fig.3 : : inverse averaged antiferromagnetic susceptibility @xmath333 vs. @xmath330 for the binary alloy with @xmath334 and several values of @xmath57 . fig.4a : : @xmath202-phase diagram for the binary alloy with @xmath335 obtained from the zeroes of @xmath333 ( see fig . 4 ) . the af - phase is stable below the curves . the dotted lines at @xmath320 depict the regimes where the curie law would give negative transition temperatures . below the crosses @xmath333 has no zeroes but a minimum and an af - phase can no longer be expected . fig.4b : : @xmath202-phase diagram for the semi - elliptic distribution of the random energies with width @xmath336 . fig.5 : : @xmath330 vs. @xmath227 for the binary - alloy distribution at @xmath337 and 5.5 . dashed lines : quadratic increase of @xmath194 according to eq . ( 30b ) with @xmath338 ; dotted lines are guides to the eye only . fig.6 : : scaling plot of @xmath339 for the binary - alloy distribution according to eq . ( 30 ) ; symbols as in fig . 4a . fig.7 : : @xmath340-phase diagram for the percolation - type disorder for several values of @xmath57 . the dashed line depicts the hartree - fock ( hf ) result for the smallest value of @xmath341 . fig.8a : : averaged quadratic local moment @xmath239 vs. u at inverse temperature @xmath342 . without disorder ( full circle ) ; binary alloy with @xmath334 ( full square ) ; semi - elliptic distribution with @xmath334 ( open circle ) and @xmath343 ( open square ) . fig.8b : : averaged quadratic local moment @xmath344 ( normalized to the concentration of sites with random energy zero ) vs. @xmath99 for the percolation - type disorder at @xmath342 for several values of @xmath57 . fig.9 : : staggered magnetization @xmath249 and inverse averaged susceptibility @xmath333 vs. @xmath330 for the binary alloy with @xmath331 , @xmath332 , @xmath345 . dashed line : square - root fit of the last two points of @xmath249 below the transition ; dotted line : linear fit of @xmath333 . the arrow indicates the extrapolated nel temperature . fig.10 : : averaged compressibility @xmath96 vs. @xmath57 at @xmath342 . a ) binary alloy , @xmath335 ; b ) semi - elliptic distribution with @xmath346 and 4 . arrows indicate the transition to the antiferromagnetic state . fig.11 : : averaged compressibility @xmath96 vs. @xmath57 at @xmath342 in the paramagnetic phase ( dashed lines ) and in the antiferromagnetic phase ( dotted lines ) . a ) binary alloy , b ) semi - elliptic distribution . fig.12 : : averaged compressibility @xmath96 vs. @xmath57 for the binary alloy at @xmath342 and 40 . arrows indicate the transition to the antiferromagnetic state . fig.13 : : real part of @xmath347 vs. @xmath45 , a ) for @xmath348 , @xmath349 and @xmath350 in the paramagnetic phase ( p ) and the antiferromagnetic phase ( af ) ; b ) for the binary alloy with @xmath334 , @xmath351 in the af - phase at @xmath352 and 64 . fig.14 : : imaginary part of the one - particle green function @xmath161 at the lowest matsubara frequency @xmath353 vs. @xmath57 for a system without disorder ( @xmath279 ) and with binary - alloy disorder ( @xmath280 at @xmath342 ; p - phase ( dashed lines ) , af - phase ( dotted lines ) . fig.15 : : imaginary part of @xmath161 vs. @xmath45 in the p- and af - phase ; a ) no disorder , u=1.75 , @xmath350 ; b ) binary alloy with @xmath334 , @xmath351 , @xmath354 ( the two curves at @xmath355 in the middle are undistinguishable ) . fig.16 : : imaginary part of @xmath356 vs. @xmath330 for the binary alloy with @xmath334 for several values of @xmath57 .

let be a closed orientable surface of genus and let be the identity component of the group of area - preserving diffeomorphisms of . in this work we present the extension of gambaudo - ghys construction to the case of a closed hyperbolic surface , i.e. we show that every non - trivial homogeneous quasi - morphism on the braid group on strings of defines a non - trivial homogeneous quasi - morphism on the group . as a consequence we give another proof of the fact that the space of homogeneous quasi - morphisms on is infinite dimensional . let be the group of hamiltonian diffeomorphisms of . as an application of the above construction we construct two injective homomorphisms , which are bi - lipschitz with respect to the word metric on and the autonomous and fragmentation metrics on . in addition , we construct a new infinite family of calabi quasi - morphisms on .

1306.3350

let @xmath0 be a closed orientable surface of genus @xmath1 equipped with the area form , and let @xmath2 be the identity component of the group of area - preserving diffeomorphisms of @xmath0 . denote by @xmath7 and @xmath8 the unit @xmath9-disc in a plane and the standard @xmath9-sphere respectively . in @xcite gambaudo and ghys gave a construction of quasi - morphisms on @xmath10 and on @xmath11 from quasi - morphisms on the artin pure braid group @xmath12 and on the pure sphere braid group @xmath13 respectively . in this paper we are going to extend their construction to the case of @xmath2 , when @xmath14 , and to the case of @xmath4 , when @xmath15 . here @xmath4 is the group of hamiltonian diffeomorphisms of @xmath0 . more precisely , we are going to show that every non - trivial homogeneous quasi - morphism on the full surface braid group @xmath16 defines a non - trivial homogeneous quasi - morphism on the group @xmath2 . as a consequence of this construction , we construct two injective homomorphisms @xmath5 , which are bi - lipschitz with respect to the word metric on @xmath6 and the autonomous and fragmentation metrics on @xmath4 respectively . in particular , we generalize the results of the author and kedra proved for the case of @xmath10 , see @xcite . a quasi - morphism on @xmath4 is called calabi if it equals to the calabi homomorphism on every diffeomorphism supported in a displaceable disc . they have various applications in dynamics and symplectic geometry , see e.g. @xcite . calabi quasi - morphisms on @xmath4 were constructed by entov - polterovich @xcite ( in case when @xmath17 ) and by py @xcite ( in case when @xmath18 ) . in this paper we give a new topological construction of infinitely many linearly independent calabi quasi - morphisms on @xmath4 for every @xmath14 . let us start with a definition . a function @xmath19 from a group @xmath20 to the reals is called a _ quasi - morphism _ if there exists a real number @xmath21 such that @xmath22 for all @xmath23 . the infimum of such @xmath24 s is called the _ defect _ of @xmath25 and is denoted by @xmath26 . if @xmath27 for all @xmath28 and all @xmath29 then @xmath25 is called _ homogeneous_. any quasi - morphism @xmath25 can be homogenized by setting @xmath30 the vector space of homogeneous quasi - morphisms on @xmath20 is denoted by @xmath31 . the space of homogeneous quasi - morphisms on @xmath20 modulo the space of homomorphisms on @xmath20 is denoted by @xmath32 . for more information about quasi - morphisms and their connections to different brunches of mathematics , see @xcite . for every pair of points @xmath33 let us choose a geodesic path @xmath34\to { { \mathbf \sigma}}_g$ ] from @xmath35 to @xmath36 . let @xmath37 be an isotopy from the identity to @xmath38 and let @xmath39 be a basepoint . for @xmath40 we define a loop @xmath41\to{{\mathbf \sigma}}_g$ ] by @xmath42\\ f_{3t-1}(y ) & \text { for } t\in \left [ \frac13,\frac23\right ] \\ s_{f(y)z}(3t-2 ) & \text { for } t\in \left [ \frac23,1\right ] . \end{cases}\ ] ] let @xmath43 be the configuration space of all ordered @xmath3-tuples of pairwise distinct points in the surface @xmath0 . it s fundamental group @xmath44 is identified with the pure surface braid group @xmath45 . let point @xmath46 in @xmath43 be a base point . for almost each point @xmath47 the @xmath3-tuple of loops @xmath48 is a based loop in the configuration space @xmath43 . let @xmath14 . then the group @xmath2 is simply - connected ( see e.g. ( * ? ? ? * section 7.2 ) ) . hence the based homotopy class of this loop does not depend on the choice of the isotopy @xmath49 . let @xmath50 be an element represented by this loop . let @xmath15 . in this case the group @xmath51 is not simply - connected . however , the group @xmath52 is simply - connected ( see e.g. ( * ? ? ? * section 7.2 ) ) . hence , for every hamiltonian isotopy @xmath49 between @xmath53 and @xmath54 , the based homotopy class of the loop @xmath48 does not depend on the choice of the isotopy @xmath49 . let @xmath55 be a homogeneous quasi - morphism , where @xmath16 is a full braid group on @xmath3 strings on a surface @xmath0 . for @xmath14 we define the map @xmath56 by @xmath57 for @xmath15 the map @xmath58 is defined as in . [ t : gen - gg ] let @xmath14 . then the function @xmath59 is a well defined homogeneous quasi - morphism . let @xmath15 . then the function @xmath60 is a well defined homogeneous quasi - morphism . note that when @xmath61 the group @xmath62 . in this case theorem [ t : gen - gg ] was proved by l. polterovich in @xcite and by gambaudo - ghys in @xcite . we must add that when @xmath15 and @xmath61 the polterovich quasi - morphism @xmath63 becomes trivial on @xmath52 . by theorem [ t : gen - gg ] , in case when @xmath14 , the above construction defines linear map @xmath64 and in case when @xmath15 it defines linear map @xmath65 [ t : prop - gg - map ] let @xmath66 . the map @xmath67 is injective for each @xmath68 . in case when @xmath15 , one can similarly show that the map @xmath69 is injective . in case when @xmath17 , theorem [ t : prop - gg - map ] was proved by ishida @xcite . recall that @xmath0 is a compact orientable surface of genus @xmath1 . let @xmath70 be a smooth function with zero mean . there exists a unique vector field @xmath71 such that @xmath72 it is easy to see that @xmath71 is tangent to the level sets of @xmath73 . let @xmath74 be the time - one map of the flow @xmath75 generated by @xmath71 . the diffeomorphism @xmath74 is area - preserving and every diffeomorphism arising in this way is called _ autonomous_. such a diffeomorphism is relatively easy to understand in terms of its generating function . let @xmath4 be the group of hamiltonian diffeomorphisms of @xmath0 . since it is a simple group @xcite , every hamiltonian diffeomorphism of @xmath0 is a composition of finitely many autonomous diffeomorphisms . the set of such diffeomorphisms is invariant under conjugation . hence we define the conjugation - invariant _ autonomous norm _ on the group @xmath4 by @xmath76 where each @xmath77 is autonomous . the associated metric is defined by @xmath78 since the autonomous norm is conjugation - invariant , the autonomous metric is bi - invariant . [ t : main ] let @xmath14 . then for every natural number @xmath79 there exists an injective homomorphism @xmath80 which is bi - lipschitz with respect to the word metric on @xmath6 and the autonomous metric on @xmath4 . * remarks . * 1 . the study of bi - invariant metrics on groups of geometric origin was initiated by burago , ivanov and polterovich @xcite . the autonomous metric is a particular case of such metrics . moreover , it is the most natural word metric on the group of hamiltonian diffeomorphisms of a symplectic manifold . it is particulary interesting in the two - dimensional case due to the following observation . + let @xmath81 be a morse function and let @xmath74 be a hamiltonian diffeomorphism generated by @xmath73 . after cutting the surface @xmath0 along critical level sets we are left with finite number of regions , so that each one of them is diffeomorphic to annulus . by arnold - liouville theorem @xcite , there exist angle - action symplectic coordinates on each one of these annuli , so that @xmath74 rotates each point on a regular level curve with the same speed , i.e. the speed depends only on the level curve . it follows that a generic hamiltonian diffeomorphism of @xmath0 may be written as a finite composition of autonomous diffeomorphisms , such that each one of these diffeomorphisms is `` almost everywhere rotation '' in right coordinates , and hence relatively simple . of course this decomposition is nor unique , neither canonical . however , it is plausible that it might be useful in dynamical systems . 2 . let @xmath82 be the group of smooth compactly supported area - preserving diffeomorphisms of the open unit disc @xmath83 in the euclidean plane . in @xcite together with kedra we showed that for every positive number @xmath84 there exists an injective homomorphism @xmath85 which is bi - lipschitz with respect to the word metric on @xmath86 and the autonomous metric on @xmath82 . we should mention that gambaudo and ghys showed that the diameter of the group of area - preserving diffeomorphisms of the 2-sphere equipped with the autonomous metric is infinite as well , see ( * ? ? ? * section 6.3 ) . 3 . let @xmath87 be a closed symplectic manifold . it is interesting to know whether the autonomous metric is unbounded on the group of hamiltonian diffeomorphisms @xmath88 , and if yes then can one prove theorem [ t : main ] in the case when the dimension of @xmath89 is greater than two ? let @xmath90 be a closed symplectic manifold , and denote by @xmath88 the group of hamiltonian diffeomorphisms of @xmath89 . let @xmath91 be a collection of regions such that * for each @xmath92 the region @xmath93 is diffeomorphic to a ball of the same dimension as @xmath89 . * all elements in @xmath91 have the same volume @xmath94 . every region of volume @xmath94 in @xmath89 , which is diffeomorphic to a @xmath95-dimensional ball , lies in @xmath91 . since @xmath88 is a simple group @xcite , every diffeomorphism in @xmath88 is a composition of finitely many diffeomorphisms , such that each such diffeomorphism is supported in some ball @xmath96 . the set of such diffeomorphisms is invariant under conjugation . we define the conjugation - invariant _ fragmentation norm _ on the group @xmath88 with respect to the cover @xmath91 by @xmath97 where the support of @xmath77 lies in some @xmath98 . the associated metric is defined by @xmath99 . since the fragmentation norm is conjugation - invariant , the fragmentation metric is bi - invariant . recently , fragmentation metric received a considerable attention , since , in particular , it is related to the question about the simplicity of the group of compactly supported area - preserving homeomorphisms of an open 2-disc , see e.g. @xcite . [ t : frag - metric ] * 1 . * let @xmath0 be any closed surface of genus @xmath14 . then there exists an injective homomorphism @xmath100 which is bi - lipschitz with respect to the word metric on @xmath101 and the fragmentation metric on @xmath4 . let @xmath90 be a closed symplectic manifold of dimension greater then two . suppose that there exists an embedding of the free group on two generators @xmath102 such that the image of the induced map @xmath103 is infinite dimensional . then for every natural number @xmath79 there exists an injective homomorphism @xmath104 which is bi - lipschitz with respect to the word metric on @xmath6 and the fragmentation metric on @xmath88 . it is known that every non - elementary word - hyperbolic group @xmath20 contains @xmath105 such that the induced map @xmath106 is infinite dimensional , see @xcite . let @xmath90 be a closed negatively curved symplectic manifold of dimension greater then two . then @xmath107 contains bi - lipschitz embedded finitely generated free abelian group of an arbitrary rank . let @xmath90 be a symplectic manifold . recall that a homogeneous quasi - morphism @xmath108 on @xmath88 is called calabi if its restriction to a subgroup @xmath109 , where @xmath110 is a displaceable ball in @xmath89 , coincides with calabi homomorphism on @xmath109 , see e.g. @xcite . here we focus on the case of surfaces , but the works quoted above deal also with groups of hamiltonian diffeomorphisms of higher dimensional symplectic manifolds . in @xcite entov and polterovich constructed a calabi quasi - morphism on a group of hamiltonian diffeomorphisms of the two - sphere . their construction uses quantum homology . they asked whether one can construct a calabi quasi - morphism on @xmath4 , where @xmath111 . around 2006 pierre py gave a positive answer to this question , see @xcite . let @xmath14 . in this paper we give a different construction of an infinite family of calabi quasi - morphisms on @xmath4 . let @xmath112 be an open disc in the euclidean plane and denote by @xmath113 the group of compactly supported hamiltonian diffeomorphisms of @xmath112 . let @xmath114 be a homomorphism , which takes value @xmath115 on the generator of the braid group @xmath116 on two generators , which is cyclic . let @xmath117 be the induced , by gambaudo - ghys construction , homomorphism from @xmath113 to the reals . the calabi homomorphism @xmath118 may be defined as follows : @xmath119 for the proof of this fact see e.g. @xcite . it turns out that the generalized gambaudo - ghys construction allows us to construct an infinite family of linearly independent calabi quasi - morphisms , i.e. we prove the following [ t : calabi - qm ] let @xmath14 . the space of calabi quasi - morphisms on @xmath4 is infinite dimensional . we should note that the above theorem is not entirely new , because py construction @xcite implies a proof of this theorem as well . nevertheless , we think that this theorem is interesting , since our construction is entirely different from the one given by pierre py . it is enough to prove that the functions @xmath120 are well defined quasi - morphisms for each @xmath121 . let @xmath14 . in @xcite e. calabi constructed a homomorphism @xmath122 banyaga proved that the kernel of @xmath123 is a simple group , see @xcite . let us recall a useful definition of the calabi homomorphism . for each @xmath124 the fundamental group @xmath125 is generated by homotopy classes of @xmath126 simple closed curves such that @xmath127=1\rangle .\ ] ] let @xmath128 and @xmath129 be homomorphisms , where @xmath130 equals to the number of times @xmath131 appears in @xmath132 minus number of times @xmath133 appears in @xmath132 , and @xmath134 equals to the number of times @xmath135 appears in @xmath132 minus number of times @xmath136 appears in @xmath132 . then @xmath137 may be defined as follows : @xmath138 where @xmath139 is defined in , and @xmath140 is the @xmath141-th coordinate of @xmath142 . let @xmath38 . it follows that @xmath54 can be written as a product @xmath143 , where @xmath144 and each @xmath77 is a time - one map of the isotopy @xmath145 which is defined as follows : * for each @xmath146 and @xmath147 the diffeomorphism @xmath148 equals to the identity outside some tubular neighborhood of the curve @xmath131 , that is identified with @xmath149\times{{\mathbf s}}^1 $ ] , and the time - one map @xmath77 is equal to the identity on @xmath131 , which is identified with @xmath150 . * for each @xmath151 and @xmath147 the diffeomorphism @xmath148 equals to the identity outside some tubular neighborhood of the curve @xmath152 , that is identified with @xmath149\times{{\mathbf s}}^1 $ ] , and the time - one map @xmath77 is equal to the identity on @xmath152 , which is identified with @xmath150 . * each diffeomorphism @xmath148 preserves the foliation of @xmath149\times{{\mathbf s}}^1 $ ] by the circles @xmath153 . each diffeomorphism @xmath148 preserves the orientation for @xmath154 . for all @xmath155 we have @xmath156 . * in addition we require that @xmath157 . it was proved by banyaga that there exists @xmath158 and a family @xmath159 of diffeomorphisms in the group @xmath2 , such that each @xmath160 is supported in some disc @xmath161 and @xmath162 , see @xcite . for each @xmath163 pick an isotopy @xmath164 in @xmath2 between the identity and @xmath160 , such that the support of @xmath164 lies in the same disc @xmath165 . first we are going to show that @xmath166 . let @xmath167 be the configuration space of all ordered @xmath3-tuples of pairwise distinct points in the subsurface @xmath161 . since changing a basepoint in @xmath43 changes @xmath168 by a bounded value , we can assume that the basepoint @xmath169 lies in @xmath167 . theorems 2 and 4 in @xcite imply that the inclusion of @xmath165 into @xmath0 induces an inclusion of the artin braid group @xmath170 into the surface braid group @xmath16 . for @xmath171 the group @xmath172 may be viewed as a subgroup of @xmath170 by adding @xmath173 strings . since @xmath55 is invariant under conjugation and the support of the isotopy @xmath164 lies in @xmath165 we have @xmath174 since the integral @xmath175 is well defined for each @xmath176 ( * ? ? ? * lemma 4.1 ) , we have @xmath166 . denote by @xmath177 the tubular neighborhood of @xmath131 diffeomorphic to @xmath149\times{{\mathbf s}}^1 $ ] if @xmath146 and of @xmath152 if @xmath151 . by definition @xmath178 since @xmath77 rotates all circles of @xmath177 and is identity on @xmath179 , it is easy to compute @xmath180 . the computation is very similar to one done by gambaudo - ghys in ( * ? ? ? * section 5.2 ) , and hence we omit it . in particular , it follows that @xmath181 . let @xmath182 . then @xmath183 i.e. @xmath168 satisfies the quasi - morphism condition . it follows that @xmath184 hence @xmath168 is a well defined quasi - morphism . let @xmath15 and @xmath185 . since the group @xmath52 is simple , there exists @xmath158 and a family @xmath159 of diffeomorphisms in @xmath52 , such that each @xmath160 is supported in some disc @xmath186 and @xmath187 , see @xcite . for each @xmath163 we pick an isotopy @xmath164 in @xmath52 between the identity and @xmath160 , such that the support of @xmath164 lies in the same disc @xmath165 . now we proceed as in the case when @xmath14 . let @xmath14 . recall that the group @xmath188 is generated by homotopy classes of @xmath126 simple closed curves such that @xmath189=1\rangle .\ ] ] at this point we describe a particular generating set @xmath190 for the pure surface braid group @xmath45 . in @xcite bellingeri showed that @xmath45 is generated by the set @xmath191 we view the artin pure braid group @xmath12 as a subgroup of @xmath45 . each generator in the set @xmath192 is the corresponding band generator of @xmath12 , i.e. it twists only @xmath141-th and @xmath193-th strands . each generator in the set @xmath194 naturally corresponds to one of the generators of @xmath188 described above , see @xcite . let us start the proof . in case when @xmath17 this theorem was proved by ishida in @xcite . in what follows we adopt the proof of ishida to the general case of @xmath14 . let @xmath195 . it is sufficient to prove that @xmath196 . since @xmath197 is a non - trivial homogeneous quasi - morphism , there exists a braid @xmath198 such that @xmath199 . the braid @xmath200 is pure , hence it can be written as a composition of generators and their inverses from the set @xmath190 . let @xmath201 be a family of disjoint subsets of @xmath0 , such that @xmath202 and each @xmath203 is diffeomorphic to a disc . for any pair @xmath204 we take subsets @xmath205 such that @xmath206 and @xmath207 are diffeomorphic to a disc , @xmath208 and @xmath209 for @xmath210 . let @xmath211 be a path in @xmath2 which fixes the outside and a small neighborhood of @xmath207 and rotates @xmath206 once . such @xmath211 twists all @xmath203 s in the form of a generator @xmath212 . in addition , for each @xmath203 we take @xmath126 subsets @xmath213 such that both @xmath214 and @xmath215 are diffeomorphic to @xmath216 $ ] and @xmath217 for each @xmath218 . we identify @xmath214 with @xmath216 $ ] , and require that for each @xmath218 the curve @xmath219 in @xmath214 represents a unique generator from the presentation of @xmath188 shown in . let @xmath220 be a path in @xmath2 which fixes the outside and a small neighborhood of @xmath215 , rotates @xmath215 in the same direction , and rotates @xmath214 exactly once . such @xmath220 twists all @xmath203 s in the form of a generator @xmath221 . we compose all the paths @xmath211 s and @xmath220 s such that the time - one map of the composition flow @xmath222 twists all @xmath203 s in the form of pure braid @xmath200 . set @xmath223 , then @xmath54 is identity on each @xmath203 and @xmath224 for @xmath225 . if we denote by @xmath226 , then @xmath227 let us start with the second term in the above equation . we denote by @xmath228 to be the set of artin generators for the artin braid group @xmath170 . hence the group @xmath16 is generated by the set @xmath229 , see @xcite . for @xmath230 denote by @xmath231 the length of the word in generators from @xmath232 , which represent the braid @xmath233 and is given by @xmath234 concatenations of the flow @xmath235 . let @xmath236 . it follows from the construction of @xmath54 that there exists a constant @xmath237 , such that for each @xmath230 we have @xmath238 for each @xmath239 denote by @xmath240 the word length of @xmath241 with respect to the generating set @xmath232 . since @xmath242 is a homogenous quasi - morphism , we obtain @xmath243 it follows that for each @xmath230 we have @xmath244 it follows that @xmath245{}0\thinspace .\ ] ] now we turn to the first term in . there exists a constant @xmath246 such that in every neighborhood of @xmath247 there exists a number @xmath248 , where @xmath226 and @xmath203 s as before , such that @xmath249 the proof of this fact is identical to the proof of ( * ? ? ? * theorem 1.2 ) . this concludes the proof of the theorem . [ r : genuine - quasi - morphisms ] let @xmath242 be a genuine homogeneous quasi - morphism in @xmath250 , i.e. @xmath242 is not a homomorphism . then the induced quasi - morphism @xmath251 is not a homomorphism . the proof of this fact is very similar to the proof of theorem [ t : prop - gg - map ] , and is omitted . recall that for a group @xmath20 we denoted by @xmath32 the space of homogeneous quasi - morphisms on @xmath20 modulo the space of homomorphisms on @xmath20 . [ cor : inf - dim ] the vector space @xmath252 is infinite dimensional for @xmath14 . moreover , for @xmath14 , the map @xmath253 is injective and hence the vector space @xmath254 is also infinite dimensional . it is a well known fact that for each @xmath14 the braid group @xmath16 is a subgroup of the mapping class group of the surface @xmath0 with @xmath3 punctures , see @xcite . since the group @xmath16 is not virtually abelian , the space @xmath255 is infinite dimensional , see ( * ? ? ? * theorem 12 ) . it follows from theorem [ t : prop - gg - map ] and remark [ r : genuine - quasi - morphisms ] that the space @xmath252 is also infinite dimensional . it is known that , for @xmath14 , the group @xmath4 is simple and it coincides with the kernel of calabi homomorphism and with the commutator subgroup of @xmath2 , see @xcite . it follows that there are no non - trivial quasi - morphisms in @xmath252 which are homomorphisms when restricted on @xmath4 . hence the map @xmath256 is injective and its image in @xmath254 is infinite dimensional . the above proof gives another proof to the fact that the spaces @xmath252 and @xmath254 are infinite dimensional for @xmath14 . originally this fact was proved by gambaudo - ghys in @xcite . let @xmath257 be a compact orientable surface of genus @xmath1 with @xmath258 boundary components . then , by the proof above , the spaces @xmath259 are also infinite dimensional for @xmath14 . let @xmath260 and @xmath75 be an autonomous flow generated by a morse function @xmath261 . we pick a point @xmath47 which satisfies the following conditions : * @xmath262 is a regular point for each @xmath124 , * connected components of @xmath263 and @xmath264 , which are simple closed curves , are disjoint for each @xmath265 . such a set of points in @xmath43 is denoted by @xmath266 . note that the measure of @xmath267 is zero . if @xmath262 is a regular point of @xmath73 , then the image of the path @xmath268 , where @xmath269 , is a simple closed curve let @xmath270\to{{\mathbf \sigma}}_g$ ] be an injective path ( on @xmath271 ) , such that @xmath272 and its image is a simple closed curve defined above . define @xmath273\\ c_{3t-1}(x_i ) & \text { for } t\in \left [ \frac13,\frac23\right ] \\ s_{x_iz_i}(3t-2 ) & \text { for } t\in \left [ \frac23,1\right ] \end{cases}\ ] ] where the path @xmath274 was defined in section [ ss : gg+results ] . for almost every @xmath275 and for each @xmath124 the path @xmath276\ ] ] is in @xmath43 , where @xmath277\\ x_i & \text { for } t\in \left [ \frac13,\frac23\right ] \\ s_{x_iz_i}(3t-2 ) & \text { for } t\in \left [ \frac23,1\right ] \end{cases}.\ ] ] let @xmath278 , and let @xmath279 be a braid in @xmath16 represented by the path @xmath280 . since the connected components of @xmath263 and @xmath264 are disjoint for @xmath265 , the braids @xmath281 commute for each @xmath282 . for each @xmath283 the braid @xmath284 can be written as a product @xmath285 where @xmath286 is an integer which depends on @xmath287 , @xmath234 and @xmath262 , and the length of braids @xmath288 is bounded by some constant @xmath289 which depends only on @xmath3 . denote by @xmath290 the mapping class group of a surface @xmath0 with @xmath3 punctures . there is a following short exact sequence due to birman @xcite @xmath291 where @xmath292 is the mapping class group of a surface @xmath0 . hence the surface braid group @xmath16 may be viewed as an infinite normal subgroup of @xmath290 for @xmath14 . note that if @xmath293 then @xmath294 is no longer a subgroup of @xmath295 , see @xcite . hence the case of the torus is different and will not be treated in this paper . [ p : finite - conj - classes ] let @xmath14 . then there exists a finite set @xmath296 of elements in @xmath290 , such that for a diffeomorphism @xmath297 , which is generated by a morse function @xmath261 , and for each @xmath275 the braid @xmath279 is conjugated in @xmath290 to some element in @xmath296 . let @xmath275 . * suppose that @xmath279 is such that the image of the path @xmath269 is homotopically trivial in @xmath0 . hence there exists an embedded disc in @xmath0 whose boundary is the image of this curve . there exists some @xmath298 , such that there are exactly @xmath193 points from the set @xmath299 , that lie in this disc . it follows from the definition of @xmath279 that it is conjugate in @xmath16 ( and hence in @xmath290 ) to the braid @xmath300 shown in figure [ fig : braids - eta - j - n ] . braid @xmath301,height=153 ] * case 2 . * suppose that @xmath279 is such that the image of the path @xmath269 is homotopically non - trivial in @xmath0 . denote by @xmath302 a surface @xmath0 with @xmath303 punctures . we say that simple closed curves @xmath304 and @xmath305 are equivalent @xmath306 if there exists a homeomorphism @xmath307 such that @xmath308 . it follows from classification of surfaces that the set of equivalence classes @xmath309 is finite . denote by @xmath310 a simple closed curve which is an image of the path @xmath269 . since @xmath0 and @xmath310 are oriented , the curve @xmath310 splits in @xmath311 into two simple closed curves @xmath312 and @xmath313 which are homotopic in @xmath314 , see figure [ fig : path - split ] . part of the curve @xmath310 is shown in figure * a*. its splitting into curves @xmath312 and @xmath313 is shown in figure * b*. the left curve in figure * b * is @xmath312 and the right curve is @xmath313,height=144 ] the image of the braid @xmath279 in @xmath290 under the birman embedding of @xmath16 into @xmath290 is conjugated to @xmath315 , where @xmath316 and @xmath317 are dehn twists in @xmath311 about @xmath313 and @xmath312 respectively . note that if @xmath318 then there exists a homeomorphism @xmath319 such that @xmath320 , hence @xmath321 and @xmath322 . we have @xmath323 this yields @xmath324 in other words , the braid @xmath279 is conjugated in @xmath290 to some @xmath325 , where @xmath304 is some representative of an equivalence class in @xmath309 . let @xmath326 be a set of simple closed curves in @xmath0 , such that each equivalence class in @xmath309 is represented by some @xmath327 . let @xmath328 it follows that @xmath279 is conjugated to some element in @xmath296 . note that the set @xmath296 is independent of @xmath73 , @xmath287 and @xmath275 and the proof follows . let @xmath260 and @xmath75 be an autonomous flow generated by a function @xmath261 . let @xmath329 such that @xmath330 is a critical level set of @xmath73 . since @xmath75 is an autonomous flow , the set @xmath330 decomposes as a union of curves @xmath331 where the image of each curve @xmath332 is a point or a simple curve . we pick a point @xmath47 which satisfies the following conditions : * @xmath262 is a regular or critical point for each @xmath124 . * for each pair @xmath333 , where at least one of the points belongs to the regular level set of @xmath73 , we require the connected components of the sets @xmath263 and @xmath264 are disjoint for each @xmath265 . * for each pair @xmath333 , where both @xmath334 belong to the same critical level set @xmath330 , we require that they lie on disjoint curves in the decomposition of @xmath330 as a union of curves , see . such a set of points in @xmath43 is denoted by @xmath335 . note that the measure of @xmath336 is zero , and @xmath337 . let @xmath338 . if @xmath262 belongs to the regular level set of @xmath73 , then @xmath339 is defined in . suppose that @xmath262 belongs to the critical level set of @xmath73 . the image of the path @xmath268 , where @xmath269 , is a simple curve , because @xmath340 is an integral curve of a time - independent vector field . for each @xmath283 set @xmath341\\ h_{p(3t-1)}(x_i ) & \text { for } t\in \left [ \frac13,\frac23\right ] \\ s_{h_p(x_i)z_i}(3t-2 ) & \text { for } t\in \left [ \frac23,1\right ] , \end{cases}\ ] ] where the path @xmath274 was defined in section [ ss : gg+results ] . for almost every @xmath342 , for each @xmath124 and @xmath283 , the path @xmath343\ ] ] is in @xmath43 , where @xmath277\\ x_i & \text { for } t\in \left [ \frac13,\frac23\right ] \\ s_{x_iz_i}(3t-2 ) & \text { for } t\in \left [ \frac23,1\right ] \end{cases}.\ ] ] let @xmath344 be a braid in @xmath16 represented by the path @xmath345 . let @xmath342 . since the curves @xmath346 and @xmath347 are disjoint for @xmath265 , the braids @xmath348 commute for each @xmath282 . for each @xmath283 the braid @xmath284 can be written as a product @xmath349 where the length of braids @xmath288 is bounded by some constant @xmath289 which depends only on @xmath3 . moreover , if @xmath262 belongs to a critical level set @xmath330 , then : suppose that for some @xmath329 , the set @xmath330 is a figure eight curve . hence @xmath36 is a critical point , and suppose that @xmath262 is a regular point . the point @xmath36 can not be in the image of the curve @xmath350 and the curve @xmath340 does not reverse the orientation , because @xmath75 is an autonomous flow . it follows that the curve @xmath340 has a finite length.,height=115 ] * if @xmath262 is a critical point of @xmath73 , then the braid @xmath344 is the identity for every @xmath234 , since @xmath351 the flow @xmath352 . * if @xmath262 belongs to a critical set @xmath330 , but @xmath262 is not a critical point , then @xmath262 belongs to a simple curve in the decomposition of @xmath330 into union of curves . in this case , the length of the braid @xmath344 is bounded , uniformly for all @xmath234 , by some constant @xmath353 which depends only on @xmath3 . this follows from the fact that the curve @xmath340 has a finite length . for example , see figure [ fig : fig8-level - set ] . also note that if @xmath262 belongs to regular level set of @xmath73 , then the braid @xmath354 , where @xmath355 is the braid defined in . [ p : gen - finite - conj - classes ] let @xmath14 . then for @xmath297 , which is generated by a function @xmath261 and for @xmath356 there exists @xmath357 , such that for each @xmath262 the braid @xmath344 , defined in , is conjugated to @xmath358 , where @xmath274 is some element in the set @xmath296 which was defined in proposition [ p : finite - conj - classes ] . the proof is identical to the proof of proposition [ p : finite - conj - classes ] . recall that we view @xmath16 as a normal subgroup of @xmath290 and @xmath359 . for each @xmath260 denote by @xmath360 the space of homogeneous quasi - morphisms on @xmath16 so that : * for each @xmath361 there exists @xmath362 such that @xmath363 , and @xmath364 vanishes on the finite set @xmath296 . * there are infinitely many linearly independent homogeneous quasi - morphisms in @xmath360 that remain linearly independent when restricted to @xmath170 . in @xcite bestvina - bromberg - fujiwara classified all elements @xmath365 for which there is a quasi - morphism on @xmath290 which is unbounded on the powers of @xmath241 , in terms of the nielsen - thurston decomposition . as a corollary they proved the following [ p : bbf ] let @xmath366 be a subsurface that supports pseudo - anosov homeomorphisms . then the restriction map @xmath367 has infinite dimensional image . here @xmath368 is a mapping class group of a subsurface @xmath369 . let @xmath370 . we identify @xmath371 with a mapping class group of an @xmath3-punctured disc . since there are infinitely many pseudo - anosov braids in @xmath170 , it follows from proposition [ p : bbf ] that there exist infinitely many linearly independent homogeneous quasi - morphisms on @xmath290 that remain linearly independent when restricted to @xmath170 . combining this with the fact that the set @xmath296 is finite we have the following important [ cor : imp - inf - dim - cor ] let @xmath14 , @xmath370 . then @xmath372 . note that there are no non - trivial homomorphisms in @xmath360 . [ t : inf - diam ] for each @xmath14 the metric space @xmath373 has an infinite diameter . recall that @xmath374 is the set of all autonomous diffeomorphisms of @xmath0 . denote by @xmath375 the space of homogeneous quasi - morphisms on @xmath4 that vanish on the set @xmath376 . since the set @xmath376 generates @xmath4 , there are no non - trivial homomorphisms in @xmath375 . [ l : aut - zero ] let @xmath14 and @xmath370 . then for each @xmath377 and an autonomous diffeomorphism @xmath378 we have @xmath379 there are two possible proofs of this lemma . * first proof . * note that the set of morse functions on @xmath380 is @xmath381-dense in the set of smooth functions on @xmath380 , see @xcite . denote the set of morse functions from @xmath0 to @xmath382 by @xmath383 . arguing similarly as in the proof of ( * ? ? ? * theorem 3.4 ) one can prove the following [ t : morse - ham ] let @xmath261 be a hamiltonian function and @xmath384 be a family of functions so that each @xmath385 and @xmath386{}~h$ ] in @xmath381-topology . let @xmath287 and @xmath387 be the time - one maps of the hamiltonian flows generated by @xmath73 and @xmath388 respectively . then @xmath389 where @xmath390 is a quasi - morphism defined in . then , by theorem [ t : morse - ham ] , it is enough to prove this lemma for @xmath287 generated by some morse function @xmath73 . for almost every @xmath338 we have @xmath391 see . since all @xmath279 s commute we have @xmath392 by proposition [ p : finite - conj - classes ] each @xmath279 is conjugate to some element in @xmath296 , hence @xmath393 . since the length of braids @xmath394 and @xmath395 is bounded by @xmath289 , there exists a constant @xmath396 such that @xmath397 and @xmath398 . it follows that @xmath399 where the equality follows immediately from , and the proof follows . * second proof . * this proof does not require any continuity properties of quasi - morphisms . recall that for @xmath356 we have the following decomposition @xmath400 see . since all @xmath344 s commute we have @xmath401 we showed that if @xmath262 belongs to a critical level set of @xmath73 , then the length of @xmath344 is uniformly bounded by some @xmath353 . if @xmath262 belongs to a regular level set , then by proposition [ p : gen - finite - conj - classes ] , the braid @xmath344 is conjugate to some @xmath358 , where @xmath402 . since @xmath377 , then in both cases @xmath403 recall that @xmath397 and @xmath398 . it follows that @xmath404 this yields @xmath405 and the proof follows . it follows from corollary [ cor : inf - dim ] and from lemma [ l : aut - zero ] that @xmath406 by corollary [ cor : imp - inf - dim - cor ] we have @xmath407 hence @xmath408 [ r : unbounded ] the existence of a genuine ( that is not a homomorphisms ) homogeneous quasi - morphism @xmath409 that is trivial on the set @xmath374 implies that the autonomous norm is unbounded . indeed , for every @xmath410 we have that @xmath411 and hence for every natural number @xmath3 we get @xmath412 , provided @xmath413 . remark [ r : unbounded ] concludes the proof of the theorem . [ r : proof - without - braids ] in order to show that for each @xmath14 the metric space @xmath373 has an infinite diameter it is enough to consider the case when @xmath61 , i.e. the space @xmath414 , and the generalized gambaudo - ghys construction is not needed . for more details see @xcite . however , we will see that the proof of theorem [ t : main ] requires the generalized gambaudo - ghys construction . let us proceed with the proof of theorem [ t : main ] . choose @xmath415 . there exists a family @xmath416 of linearly independent homogeneous quasi - morphisms in @xmath417 , and a family of diffeomorphisms @xmath418 in @xmath4 , where the support of each @xmath419 is contained in some disc of area @xmath94 , such that @xmath420 where @xmath421 is the kronecker delta . this follows from theorem [ t : prop - gg - map ] combined with and with lemma 3.10 and lemma 3.11 in @xcite . since @xmath422 , there exists a family of diffeomorphisms @xmath423 in @xmath424 , such that @xmath425 and @xmath426 have disjoint supports for all different @xmath141 and @xmath193 between @xmath115 and @xmath427 . denote by @xmath428 . since all @xmath429 s have disjoint supports , they generate a free abelian group of rank @xmath427 . let @xmath430 where @xmath431 . it is obvious that @xmath432 is a monomorphism . in order to complete the proof of the theorem it is left to show that @xmath432 is a bi - lipschitz map , i.e. we are going to show that there exists a constant @xmath433 such that @xmath434 we have the following equalities @xmath435 the second equality follows from the fact that every homogeneous quasi - morphism is invariant under conjugation , and the third equality is . since all @xmath429 commute and @xmath436 , we obtain @xmath437 where @xmath438 is the defect of the quasi - morphism @xmath439 . the defect @xmath440 because each @xmath441 . let @xmath442 . we obtain the following inequality @xmath443 denote by @xmath444 . now we have the following inequality @xmath445 inequalities and conclude the proof of the theorem . recall that @xmath257 is a compact orientable surface of genus @xmath1 with @xmath258 boundary components . similarly to the case of closed surfaces the group @xmath446 admits a bi - invariant autonomous metric . the proof presented above , shows that for every @xmath14 theorem [ t : main ] holds for a group @xmath446 as well . at the beginning let us recall a construction , due to l. polterovich , of quasi - morphisms on @xmath88 . let @xmath447 . suppose that the group @xmath448 admits a _ non - trivial _ homogeneous quasi - morphism @xmath449 . for each @xmath450 let us choose an arbitrary geodesic path from @xmath35 to @xmath169 . in @xcite polterovich constructed the induced _ non - trivial _ homogeneous quasi - morphism @xmath451 on @xmath88 as follows : + for each @xmath450 and an isotopy @xmath452}$ ] between @xmath53 and @xmath74 , let @xmath453 be a closed loop in @xmath89 which is a concatenation of a geodesic path from @xmath169 to @xmath35 , the path @xmath454 and a described above geodesic path from @xmath455 to @xmath169 . denote by @xmath456 $ ] the corresponding element in @xmath448 and set @xmath457)\o^k\qquad\qquad \overline{\psi}(h):=\lim\limits_{p\to\infty}\frac{1}{p}\int\limits_{{{\mathbf m } } } \phi([(h^p)_x])\o^k,\ ] ] where @xmath458 . the maps @xmath108 and @xmath459 are well - defined quasi - morphisms because every diffeomorphism in @xmath88 is volume - preserving . they do not depend on a choice of the path @xmath211 because the evaluation map @xmath460 , where @xmath461 , induces a trivial map on @xmath462 , see @xcite . in addition , the quasi - morphism @xmath459 neither depends on the choice of a family of geodesic paths , nor on the choice of a base point @xmath169 . for more details see @xcite . note that if @xmath463 and @xmath14 , then this is exactly the generalized gambaudo - ghys construction for @xmath61 , i.e. @xmath62 and @xmath464 . an immediate corollary from this construction is [ cor : polt - qm ] let @xmath459 be a homogeneous quasi - morphism on @xmath88 given by polterovich construction , then @xmath465 for any @xmath466 which is supported in a ball . we start proving the case * 1 * , where @xmath463 and @xmath14 . there exist @xmath467 different points @xmath468 and @xmath467 disjoint embeddings @xmath469 of figure eight into @xmath0 such that : * an image of a singular point of figure eight under @xmath469 equals to some @xmath470 , * each @xmath469 induces an injective map @xmath471 , * the image of the induced map @xmath472 is infinite dimensional for each @xmath141 . these embeddings are shown in figure [ fig : fig8-emb ] . images of embeddings @xmath473 of figure eight are shown in this picture . the image of @xmath474 is a reflection of the image of @xmath469 over the horizontal plane for each @xmath475.,height=163 ] let @xmath476 denote generators of @xmath105 . since every @xmath188 is conjugate to every @xmath477 , then by ( * ? ? ? * lemma 3.10 ) there exists a family of quasi - morphisms @xmath478 and @xmath479\in\pi_1({{\mathbf \sigma}}_g , z_i)$ ] such that : * each @xmath479 $ ] is an image of some word in generators @xmath476 under the homomorphism @xmath471 , * we have @xmath480)=\delta_{ij}$ ] for @xmath481 , where @xmath421 is a kronecker delta , * and @xmath482 for @xmath483 . let @xmath131 and @xmath135 be simple closed oriented curves in @xmath0 which represent @xmath484 and @xmath485 respectively . denote by @xmath486,[\b_i])$ ] a reduced word in @xmath487 $ ] and @xmath488 $ ] which represents @xmath479 $ ] . let @xmath489 such that any @xmath490 matrix @xmath24 , where @xmath491 and @xmath492 ( @xmath493 ) is non - singular . in what follows we are going to construct , for each @xmath483 , autonomous hamiltonian flows @xmath494 such that * for each @xmath493 the support of the flow @xmath148 is disjoint from the supports of flows @xmath495 and @xmath496 , and the support of the flow @xmath497 is disjoint from the supports of flows @xmath495 and @xmath496 . * we have @xmath498 , where @xmath499 is a homogeneous quasi - morphism on @xmath4 induced by a quasi - morphism @xmath500 , and @xmath160 is a hamiltonian diffeomorphism @xmath501 , i.e. it is a word @xmath132 in diffeomorphisms @xmath502 and @xmath503 . * we have @xmath504 for @xmath493 . it follows from the tubular neighborhood theorem that for each @xmath131 and @xmath135 there exists @xmath505 , and area - preserving embeddings @xmath506 such that @xmath507 is the curve @xmath131 , @xmath508 is the curve @xmath135 . the area form on the annuli @xmath509 and @xmath510 is the standard euclidean area . in addition , we require that for each @xmath493 the image of @xmath511 is disjoint from the images of @xmath512 and @xmath513 , and the image of @xmath514 is disjoint from the images of @xmath512 and @xmath513 . it is straightforward to construct two hamiltonian isotopies of autonomous hamiltonian diffeomorphisms @xmath515 where @xmath516 such that : * for each @xmath147 the diffeomorphism @xmath75 equals to the identity in the neighborhood of @xmath517 , the diffeomorphism @xmath518 equals to the identity in the neighborhood of @xmath519 . * each diffeomorphism @xmath75 preserves the foliation of @xmath509 by the circles @xmath153 , and @xmath520 is constant on each circle . let @xmath521 . for every @xmath522 the restriction @xmath523 is the rotation by @xmath524 . it follows that the time - one map @xmath287 equals the identity on @xmath525 . * each diffeomorphism @xmath518 preserves the foliation of @xmath510 by the circles @xmath153 , and @xmath526 is constant on each circle . for every @xmath522 the restriction @xmath527 is the rotation by @xmath524 . it follows that the time - one map @xmath528 equals the identity on @xmath525 . * each diffeomorphism @xmath75 preserves the orientation on @xmath529 and @xmath530 in this region . each diffeomorphism @xmath518 preserves the orientation on @xmath531 and @xmath532 in this region . we identify @xmath509 with its image with respect to the embedding @xmath511 , and @xmath510 with its image with respect to the embedding @xmath514 . then we extend each isotopy @xmath75 and @xmath518 by the identity on @xmath533 and on @xmath534 obtaining smooth isotopies @xmath535 . the images of @xmath511 and @xmath514 are shown in figure [ fig : fig8-diff ] . recall that @xmath536 . let us compute @xmath537 . we have @xmath538)\o + \lim\limits_{p\to\infty}\frac{1}{p}\int\limits_{w_{g , j } } \phi_i([w^p(h_{1,j},g_{1,j})_x])\o\\ & + & \lim\limits_{p\to\infty}\frac{1}{p}\int\limits_{v_{h , j } } \phi_i([w^p(h_{1,j},g_{1,j})_x])\o + \lim\limits_{p\to\infty}\frac{1}{p}\int\limits_{v_{g , j } } \phi_i([w^p(h_{1,j},g_{1,j})_x])\o\\ & + & \lim\limits_{p\to\infty}\frac{1}{p}\int\limits_{u_{h , j}\setminus ( v_{g , j}\cup u_{g , j } ) } \phi_i([w^p(h_{1,j},g_{1,j})_x])\o\\ & + & \lim\limits_{p\to\infty}\frac{1}{p}\int\limits_{u_{g , j}\setminus ( v_{h , j}\cup u_{h , j } ) } \phi_i([w^p(h_{1,j},g_{1,j})_x])\o\\ & + & \lim\limits_{p\to\infty}\frac{1}{p}\int\limits_{u_{h , j}\cap u_{g , j } } \phi_i([w^p(h_{1,j},g_{1,j})_x])\o\thinspace .\end{aligned}\ ] ] note that if @xmath539 or @xmath540 then @xmath541 $ ] equals to @xmath542^p$ ] , hence @xmath543)=0.\ ] ] similarly , if @xmath544 or @xmath545 then @xmath541 $ ] is equals to @xmath546^p$ ] , hence @xmath543)=0.\ ] ] in addition , if @xmath547 then @xmath548 $ ] equals to @xmath549^p$ ] , hence @xmath550)=p\textrm { and } \phi_i([w^p(h_{1,j},g_{1,j})_x])=0\textrm { for } i\neq j.\ ] ] it follows that @xmath551)\o + \area ( u_{h , i}\cap u_{g , i})\\ \overline{\psi}_i(f_j)&=&\lim\limits_{p\to\infty}\frac{1}{p}\int\limits_{v_{h , j}\cup v_{h , j } } \phi_i([w^p(h_{1,j},g_{1,j})_x])\o , \rm { where } \thinspace\thinspace \textit{i}\neq \textit{j}\thinspace .\end{aligned}\ ] ] images of @xmath511 and @xmath514 are shown on the left and right respectively . the regions @xmath552 , @xmath553 , @xmath554 , and similarly @xmath555 , @xmath556 , @xmath557.,height=268 ] for @xmath558\in \im(\emb_j)_*$ ] denote by @xmath559)$ ] the length of @xmath558 $ ] in generators @xmath542,[b_j]$ ] . since @xmath500 is a homogeneous quasi - morphism and in addition @xmath560)=\phi_i([\b_j])=0 $ ] , we have @xmath561)|\leq d_{\phi_i}l_j([\g]).\ ] ] let @xmath562 , then from the construction of @xmath563 we have @xmath564)\leq p\cdot l_j([\g_j]).\ ] ] denote by @xmath565)}$ ] , @xmath566 , @xmath567 . it follows that @xmath568 and @xmath569 such that @xmath570 for each @xmath141 , and @xmath571 . now , quasi - morphisms @xmath572 satisfy the required properties . it follows from the properties of @xmath148 and @xmath497 that diffeomorphisms @xmath160 and @xmath573 commute for @xmath481 , and the @xmath490 matrix , whose @xmath574-th entry is @xmath537 , is non - singular . hence there exists a family @xmath575 of genuine homogeneous quasi - morphisms on @xmath4 such that @xmath576 . let @xmath577 where @xmath578 it is obvious that @xmath579 is a monomorphism . in order to complete the proof of the theorem in case of surfaces , it is left to show that @xmath579 is a bi - lipschitz map . since all @xmath160 commute and @xmath576 , we obtain @xmath580 where @xmath581 is the defect of the quasi - morphism @xmath582 . the defect @xmath583 because each @xmath582 is a genuine quasi - morphism . let @xmath584 . we obtain the following inequality @xmath585 denote by @xmath586 . now we have the following inequality @xmath587 this concludes the proof of the theorem in case of surfaces . * let @xmath588 . let @xmath589 different points in @xmath89 . since the dimension of @xmath89 is greater then two , and there exists a monomorphism @xmath590 such that the image of the induced map @xmath103 is infinite dimensional , there exist @xmath427 disjoint embeddings of figure eight into @xmath89 such that * an image of a singular point of the figure eight under the @xmath141-th embedding equals to some @xmath470 , * the @xmath141-th embedding induces an injective map @xmath591 , and the image of the induced map @xmath592 is infinite dimensional . as in case of surfaces we can construct , for each @xmath163 , autonomous hamiltonian flows @xmath593 , a diffeomorphism @xmath594 which can be written as a finite product of diffeomorphisms @xmath595 and their inverses , and homogeneous quasi - morphism @xmath596 such that * for each @xmath493 the support of the flow @xmath597 is disjoint from the supports of flows @xmath598 and @xmath599 , and the support of the flow @xmath600 is disjoint from the supports of flows @xmath598 and @xmath599 . * we have @xmath601 and @xmath602 for @xmath493 . the construction of these objects is very similar to one in the case of surfaces and is omitted . now we proceed exactly as in case * 1 * , and the proof follows . let @xmath14 and @xmath603 . recall that @xmath604 note that @xmath116 is a cyclic group generated by dehn twist around two punctures . since no positive power of the dehn twist about a simple closed curve , which bounds a disc with 2 punctures , is conjugate in @xmath605 to its inverse , then by ( * ? ? ? * theorem 4.2 ) there exists a homogeneous quasi - morphism in @xmath606 which is non - trivial on @xmath116 . the group @xmath607 contains a non - abelian surface group , see ( * ? ? ? * corollary 5.1 ) , and hence it is not virtually abelian . it is not a reducible subgroup of @xmath605 , because it is an infinite normal subgroup of @xmath605 , see ( * ? ? ? * corollary 7.13 ) . it follows from the proof of ( * ? ? ? * theorem 12 ) that the space of homogeneous quasi - morphisms on @xmath605 that remain linearly independent when restricted to @xmath607 is also infinite dimensional . hence there are infinitely many linearly independent homogeneous quasi - morphisms on @xmath607 which remain non - trivial when restricted to @xmath116 . denote this set by @xmath608 . it is infinite and it does not contain non - trivial homomorphisms , since every homomorphism on @xmath607 must be trivial on @xmath116 , see @xcite . let @xmath609 such that @xmath610 , where @xmath241 is a positive generator of @xmath116 . take a diffeomorphism @xmath378 which is supported in some disc @xmath611 . it follows from the generalized gambaudo - ghys construction that @xmath612 where @xmath613 is the calabi homomorphism defined in . the map @xmath614 is injective by corollary [ cor : inf - dim ] . this fact together with the fact that the set @xmath615 is infinite finishes the proof of the theorem . the most famous metric on the group @xmath88 of a symplectic manifold @xmath90 is the hofer metric , see @xcite . the associated norm is defined by @xmath616 where @xmath617 is a normalized hamiltonian function generating the hamiltonian flow @xmath75 from the identity to @xmath618 . let @xmath619 be a curve which represents a generator @xmath620 $ ] of @xmath125 defined in . let @xmath621 be a diffeomorphism generated by a function @xmath261 which satisfies the following conditions : it implies that all powers of @xmath74 are autonomous diffeomorphisms and hence @xmath625 for all @xmath28 . on the other hand , @xmath626 generates @xmath627 and @xmath628 , hence by a result of l. polterovich @xcite we have @xmath629 . on the other hand , for every @xmath630 , one can easily construct an autonomous diffeomorphism whose hofer norm is less then @xmath631 . this shows that the identity homomorphism between the autonomous metric and the hofer metric is not lipschitz in neither direction . in this paper we showed that there exists a hamiltonian diffeomorphism @xmath54 supported in some disc such that @xmath632{}\infty$ ] , but @xmath633 for all @xmath258 . on the other hand , one can show that there exists an autonomous diffeomorphism @xmath1 , such that @xmath634{}\infty$ ] . it follows that the identity homomorphism between the autonomous metric and the fragmentation metric is not lipschitz in neither direction as well . the author would like to thank mladen bestvina , danny calegari , louis funar , juan gonzalez - meneses , jarek kedra , chris leininger , dan margalit and pierre py for fruitful discussions . part of this work has been done during the author s stay in mathematisches forschungsinstitut oberwolfach and in max planck institute for mathematics in bonn . the author wishes to express his gratitude to both institutes . he was supported by the oberwolfach leibniz fellowship and max planck institute research grant . v. : _ mathematical methods of classical mechanics , _ translated from the 1974 russian edition by k. vogtmann and a. weinstein , graduate texts in math . , vol . 60 , springer - verlag , new york and berlin , 1978 . entov m. , polterovich l. , py p. , khanevsky m. : _ on continuity of quasi - morphisms for symplectic maps , _ perspectives in analysis , geometry , and topology , progress in mathematics volume 296 , 2012 , 169197 . mcduff d. : _ a survey of topological properties of groups of symplectomorphisms , _ in topology , geometry and quantum field theory , u.l . tillmann , ( ed . ) ( proceedings of 2002 symposium in honor of g.b . segal ) , cambridge university press , 2004 , 173193 .

a few galactic classical cepheids were observed in the programmes of space missions as _ coriolis _ , _ most _ and _ kepler_. an appealing opportunity was to detect additional nonradial modes , thus opening the possibility to perform asteroseismic studies and making the pulsational content of galactic cepheids more similar to that of magellanic clouds ones . however , only hints of cycle - to - cycle variations were found , without any strict periodicity . in this context the potential of the corot exoplanetary data base was not fully exploited despite the wide area covered on the galactic plane . therefore , we investigated all the candidate cepheids pointed out by the automatic classification of the corot curves . at the end we could identify seven bona - fide cepheids . the light curves were investigated to remove some instrumental effects . the frequency analysis was particularly delicate since these small effects can be enhanced by the large amplitude , resulting in the presence of significant , but spurious , peaks in the power spectrum . indeed , the careful evaluation of a very attracting peak in the spectra of corot 0102618121 allowed us to certify its spurious origin . once that the instrumental effects were properly removed , no additional mode was detected . on the other hand , cycle - to - cycle variations of the fourier parameters were observed , but very small and always within . among the seven cepheids , there are two pop . i first - overtone pulsators , four pop . i fundamental mode pulsators , and one pop . ii star . the corot colours allowed us to measure that times of maximum brightness occur a little earlier ( about 0.01 period ) at short wavelengths than at long ones . [ firstpage ] methods : data analysis stars : individual : v376 mon stars : individual : dp mon stars : interiors stars : oscillations stars : variables : cepheids

1508.07639

the use of cepheids as stellar candles requires a deep knowledge of their pulsational characteristics . the separation between fundamental ( @xmath1-mode ) and first - overtone ( @xmath2-mode ) radial pulsators @xcite is still the most relevant distinction to be taken into account to calibrate the period - luminosity relations . about the regularity of their variations , ground - based photometry of v473 lyr @xcite and polaris @xcite showed continuous changes in amplitude and phase . @xcite established that v473 lyr pulsates in the second radial overtone and shows strong amplitude and phase modulations on two long periods ( 1205@xmath33 and 5300@xmath3150 d ) . the slow changes in the light curve of polaris on a time baseline of 4.5 yr were precisely recorded by space photometry using the _ coriolis _ spacecraft . @xcite connected them to a quite fast evolutionary phase . however , @xcite proposed a different interpretation pointing out how the amplitude change is cyclic rather than monotonic and also suggesting an excess of power in the 2 - 6 d range due to the granulation . the situation was expected to change with the advent of high - precision space mission . a detailed analysis of the _ most _ observations of sz tau and rt aur revealed variations from cycle to cycle as a function of the pulsation phase @xcite . the stars were both monitored for less than seven consecutive cycles : the @xmath2-cepheid sz tau showed a pulsation less stable than the @xmath1-cepheid rt aur . the historical analysis of the times of maximum brightness proved that both stars experienced period variations on a time - scale of decades and that those of the overtone mode resulted to be more erratic than the @xmath4 ones . in the case of _ kepler _ , there was only one cepheid in the field - of - view , i.e. , v1154 cyg . @xcite provided a very detailed analysis of the data acquired in the first 600 d ( i.e. , about 120 cycles ) of the mission , detecting significant cycle - to - cycle fluctuations . a very slight correlation was also found between the fourier parameters and the o - c values , i.e. , the difference between the _ observed _ and _ calculated _ times of maximum brightness ( t@xmath5 ) . this suggested that the o - c variations ( up to 30 min ) might be due to instabilities in the light curve shape . on the basis of these observational results , @xcite suggested that convection and hot spots can explain the observed period jitter , thus supporting the @xcite interpretation of the polaris changes . the clues about the changes in the light curve of v1154 cyg were strengthened by the analysis of new _ kepler _ measurements ( up to q17 quarter ) : the analysis of the fourier parameters supplied a robust determination of a periodicity of 158.6 d @xcite . about ground - based observations , @xcite discovered significant modulations in the radial velocity curves of some cepheids . these modulations appear to be different from those reported on v1154 cyg , since they manifest as shape and amplitude variations on time - scales of years for short - period cepheids ( like v1154 cyg ) and cycle - to - cycle variations in the long - period cepheids . a common result among all these investigations is that no nonradial mode was identified in the light or radial velocity curves of galactic cepheids . hence the counterparts of the subclasses detected in the large magellanic cloud ( lmc ; for a review see * ? ? ? * ) are still missing . however , an important player has not yet contributed , i.e. , the corot data base of the light curves collected in the exoplanetary fields . such data base already proved to be a real treasure allowing the discovery of a unique triple - mode cepheid in the outskirts of the milky way @xcite . in this paper we describe the new elements brought by the corot photometry to the hot topic of the stability of the cepheid light curves . the first new cepheid discovered by corot was 2mass j06415168 - 0120059@xmath6corot 0102618121 , a 12@xmath7-mag star located at @xmath8 = @xmath9@xmath10@xmath11 , @xmath12 = 012005 ( j2000 ) . it was observed during the first long run in the anticentre direction ( lra01 ) and then in the lra06 four years later ( table [ runs ] ) . the classification of the type of variability was a little troublesome . the star was assigned to the rr lyr working group , but the preliminary analysis immediately pointed out that the star was not a rr lyr variable since the period was too long . the cepheid variability was much more probable , but some doubts still remained , based on the classification as m1v type reported in the header of the fits file of the lra01 data . indeed , the corot variability classifier " automated supervised method ( cvc ; * ? ? ? * ) suggested an active , rotational variable as first hypothesis and a classical cepheid as a second one . the spectral classification of corot 0102618121 was revised as a5v in the current version of the exo - dat @xcite catalogue and consequently reported in the header of the more recent lra06 data . while the analysis of corot 0102618121 was in progress , we proposed known cepheids as corot targets by answering the announcements of opportunity regularly issued before any run . in such a way v376 mon@xmath6corot 0300003751 and dp mon@xmath6corot 0221644967 were monitored in the lra02 and sra02 runs , respectively ( table [ runs ] ) . however , it should be expected that other cepheids would be discovered thanks to the deep survey performed by corot on the galactic plane . to search for these new cepheids , we first scrutinized all the new variables discovered with the berlin exoplanet search telescope ( best ) that have been classified as candidate cepheids @xcite and actually observed with corot . we noticed that the variability of v376 mon was recorded @xcite with the correct period , i.e. , 1.652 d @xcite , not the 0.623 d reported in the discovery paper @xcite . unfortunately none of other best candidates were confirmed as a new cepheid , since they turned out to be rotational variables . the misclassification is probably due to the too relaxed criterion used to classify a new variable as cepheid , i.e. , a periodic variability with an amplitude larger than 0.10 mag . indeed , such an amplitude threshold can be reached by active spotted stars . as a second step in the data mining of the corot data base , we selected all the variables for which the cvc method proposed a classification as cepheid , independently from rank and probability . the full light curves of about 250 candidates thus selected were inspected by eye . eclipsing variables , rr lyr and high - amplitude @xmath12 sct stars were found and not considered furthermore . we remained with a sample where the dichotomy was again between cepheid and rotational variables . the stars characterized by very large cycle - to - cycle variations were immediately considered as rotational variables . the stars showing more stable light curves were analysed in frequency and least - squares solutions of the mean light curves were obtained . we note that in this process the small amplitude was not considered as a rule of thumb since it could be reduced by an unresolved , blended star . on the other hand , a large amplitude ( say , greater than 0.20 mag ) was recognized as the signature of a pulsational mechanism . most of the stars were considered as rotational variables after that an ensemble of several parameters ( period , amplitude , stability of the light - curve shape , fourier parameters , ... ) was taken into account . however , we can not rule out that a few of them are actually cepheids , whose regular curve is perturbed by that of a blended irregular variable . we could identify some borderline cases ( e.g. corot 0102635101 , 0102640143 , 0605539616 , 0223971984 , 0102578739 , 0110827792 , and , overall , 0223971984 ) where the distinction between activity and pulsation was very difficult . new spectroscopic observations are necessary to establish the nature of these problematic stars in an unambiguous way . during the screening procedure we discovered the unique galactic triple - mode cepheid corot 0223989566 @xcite and at the end we could count seven bona - fide cepheids . in addition to corot 0102618121 , v376 mon , and dp mon we identified four new cepheids : corot 0221640090 , 0315221415 , 0211626074 , and 0659466327 ( table [ runs ] ) . the corot photometry of v376 mon and of other cepheids was analysed when the interpretation of the corot 0102618121 data was almost finished . therefore , we describe the procedure as applied to corot 0102618121 first and then how it has been extended to other cepheids later . rc lrrr lr r & & & & & & & & + 0102618121 & & lra01 & oct . 23 , 2007 & mar . 3 , 2008 & 131.46 d & chr & 32 s & 326634 + 0102618121 & & lra06 & jan . 12 , 2012 & mar . 29 , 2012 & 76.61 d & chr & 32 s & 196032 + 0300003751 & v376 mon & lra02 & nov . 16 , 2008 & mar . 8 , 2009 & 111.64 d & chr & 512 s & 16400 + 0221644967 & dp mon & sra02 & oct . 11 , 2008 & nov . 12 , 2008 & 31.75 d & chr & 32 s & 72144 + 0659466327 & & lrc09 & apr 12 , 2012 & july 5 , 2012 & 83.51 d & mon & 512 s & 11894 + 0315221415 & & sra03 & mar . 5 , 2010 & mar . 29 , 2010 & 24.73 d & mon & 512 s & 3670 + 0211626074 & & src01 & apr . 13 , 2007 & may 9 , 2007 & 25.54 d & mon & 512 s & 3996 + 0221640090 & nsvs 12485452&sra02&oct 11 , 2008 & nov . 12 , 2008 & 31.75 d & chr & 32 s & 73326 + [ runs ] to solve the ambiguous spectral classification of corot 0102618121 and to validate our classification criteria , we obtained three high - resolution spectra on 2012 january 10 and 11 with the harps spectrograph installed at the 3.6-m telescope of the eso - la silla observatory . the spectra were taken in the high - efficiency eggs configuration ( r=80,000 ) and they have signal - to - noise ratios ( s / n ) between 50 and 55 . we reduced and normalized the spectra with a semi - automated pipeline developed at the brera observatory @xcite . we then computed the mean line profiles for each spectrum using the lsd software @xcite and we used them to estimate the radial ( from 47.9 to 53.6 kms@xmath13 ) and projected rotational ( from 9.64 to 11.85 kms@xmath13 ) velocities by fitting the profiles with gaussians . the reduced and normalized spectra were shifted by their radial velocities and averaged in order to lessen the effect of the pulsation . the spectral synthesis software sme version 4.2.3 @xcite was used to estimate the physical parameters . the spectral regions 5160 - 5190 ( including the gravity sensitive mg - triplet ) and 4830 - 4900 ( the h@xmath14 region , sensitive to the temperature ) were selected . we left temperature ( @xmath15 ) , gravity ( @xmath16 , cgs units ) , metallicity ( [ fe / h ] ) , micro- and macroturbulence as free parameters , while @xmath17kms@xmath13 was imposed . following @xcite , we also left the abundances of na , si , ti , fe , and ni as free parameters . we performed an iterative process on the marcs stellar models @xcite : we started running the sme software on the mg region , then we switched to the h@xmath14 region keeping the @xmath16 previously found as a fixed parameter . after that we went back to the mg region keeping fixed the @xmath15 value found in the h@xmath14 region , and so on until the results converged . the resulting parameters are @xmath15= 6087@xmath344k , @[email protected] dex , [ fe / h]=@xmath18 dex . they confirm that corot 0102618121 is a pop . i supergiant and not a main - sequence star : its classification as a cepheid variable was definitely established . the original timeseries of corot 0102618121 consist of 348 702 and 206 528 data points , respectively . in both cases the exposure time was set to 32 s ( corresponding to the corot short cadence ) and the flux was recorded in three different spectral regions , i.e. , in the corot chromatic mode . the contamination factors ( in a scale ranging from 0 to 1 ) are 0.026 and 0.036 in the two corot runs , suggesting no relevant contribution from back- or fore - ground stars . the analysis of the original n2 data clearly detected a predominant peak at @xmath19=0.3574 d@xmath13 followed by several harmonics . figure [ lruns ] shows lra01 ( top panel of the upper part ) and lra06 ( top panel of the lower part ) corot n2 data on a 15-d time baseline . all data points are plotted , including those acquired during the crossing of the south atlantic anomaly ( saa ) : they appear as bright outliers on a vertical line . we avoided any cleaning procedure since the detection of instrumental trends and jumps and their subsequent correction is a crucial step to proceed in the analysis of the corot data . therefore , we computed a preliminary solution based on the set of frequencies @xmath19 , 2@xmath19 , 3@xmath19 , 4@xmath19 , and 5@xmath19 . the plots of the residuals ( middle panels of the two parts of fig . [ lruns ] ) show sudden jumps around jd 2454468 in lra01 and jd 2456003 in lra06 , the saa outliers , and continuous oscillations . outliers were iteratively removed at this stage . we still remained with the necessity to correct the sudden jumps and long - term drifts and to verify the origin of the residual oscillations . to do this , we applied the least - squares fits to each pulsation cycle , still using the set of frequencies from @xmath19 to 5@xmath19 . the 32-s cadence of the corot data warranted more than 7000 measurements for each cycle , making the solution very robust . this procedure returned the mean fluxes of the 47 cycles observed in the lra01 and the 27 ones in lra06 ( fig . [ lruns ] , bottom panels of the two parts ) . these mean values show the variations due to the jumps and the long - term trend . we notice that the long - term trends behave in different ways in the two runs . the effect of the jump around jd 2454468 was noticeable in the data until jd 2454485 , thus affecting the mean magnitudes of the subsequent six cycles . to remove the instrumental variations , we interpolated the values of the mean fluxes ( bottom panels of fig.[lruns ] ) at the times of each individual measurement and then we subtracted it . at this point two cleaned timeseries ( 326,634 and 196,032 points , respectively ) were ready to be analysed in frequency . by applying this procedure we probably modified any possible additional signal around 2.8 d and this fact has to be carefully considered when discussing the frequency content of the light curves of the two runs . we transformed the fluxes into magnitudes by calculating the average flux of each run . we also measured the level of the noise of the corot timeseries after subtracting the set of the @xmath19 frequency and harmonics . in the case of the lra01 data we obtained 0.065 mmag in the 0.01 - 1.0 d@xmath13 range , rapidly decreasing to 0.018 mmag in the 1.0 - 2.0 d@xmath13 region , and more slowly to 0.012 mmag in the 2.1 - 3.0 d@xmath13 region and to 0.007 mmag around 5.0 d@xmath13 . in the case of the lra06 data the four values are 0.104 , 0.030 , 0.020 , and 0.010 mmag , respectively . the number of measurements ( 326,634 versus 196,032 ) accounts for most of the difference between the noise levels of the lra01 and lra06 spectra . the frequency analysis was performed by means of the iterative sine - wave , least - squares fitting method @xcite and then refined with the mtrap algorithm @xcite , allowing us to include all the significant harmonics in the search for the best fit around @xmath19 . the best way to illustrate our final results is to follow the chronological order of their progressive consolidation . the analysis of the lra01 data in white light was concluded in 2011 and the results were apparently very promising . we found the harmonics up to 6@xmath19 ( amplitude 0.13 mmag ) , flanked by some modulation terms at @xmath30.007 d@xmath13 ( i.e. , 142 d ) . since the separation is very close to the length of the observations , the origin of these terms has to be ascribed to a residual , still imperfect correction of the instrumental drifts . moreover , a few peaks not very close to @xmath19 and 2@xmath19 were detected : 0.488 d@xmath13 ( amplitude 0.40 mmag ) , 0.392 d@xmath13 ( 0.14 mmag ) , and 0.288 d@xmath13 ( 0.15 mmag ) . only the peak at 0.488 d@xmath13 had a s / n above the threshold considered as the significativity one in the fourier analysis ( s / n=4.0 ; * ? ? ? * ) , i.e. , s / n=0.396/0.065=6.1 . the power spectra illustrating the bunch of low amplitude peaks still centred around @xmath19 , the 0.488-d@xmath13 peak , and the 6@xmath19 harmonic are shown in the top panel of fig . [ spectrares ] . however , the identification of the 0.488 d@xmath13 peak as an additional mode was not really convincing because no combination term was observed . why this non - detection is so important ? the combination or coupling terms are generally ascribed to the non - linear responses of the stellar atmospheres to the strong radial pulsations . the frequency analysis of the light curves of double - mode cepheids show several of them , also involving harmonics of the @xmath1 and @xmath2 radial modes @xcite . the fourier decomposition of the light curves also shows regularities in the distribution of the phase parameters that are not yet fully understood @xcite . the combination terms have been detected in the light curves of other multimode pulsators , e.g. hads stars @xcite , rr lyr stars @xcite , and white dwarfs @xcite . the corot photometry also allowed us to detect low - amplitude , nonradial modes in an hads star pulsating in a single radial mode and also in this case a lot of combination terms were detected @xcite . therefore , we should expect combination terms when additional modes , both radial and nonradial , are excited in a high - amplitude radial pulsator . the uncertainties on the real significativity of the peaks detected in the lra01 data slowed down our theoretical analysis and made us unsure of how to proceed further . the re - observation of the same field in the lra06 pointing supplied us with new data to verify the frequency content of corot 0102618121 . we applied the same procedure used for the lra01 data to remove the content of the main variation . the frequency analysis of the residuals of the lra06 data confirmed the presence of the 0.488 d@xmath13 peak ( fig . [ spectrares ] , bottom panel of the upper part ) . therefore , the confidence on the stellar origin of the 0.488 d@xmath13 grew . however , the ratio 0.357/0.488=0.732 would be quite unusual for a double - mode cepheid with a normal metallic content , both from observational data @xcite and theoretical models @xcite . therefore , such a ratio pointed out a possible nonradial mode . the identification of the peak as the rotational frequency is much more unlikely , since the corresponding rotational period ( 2.06 d ) appears too short for a supergiant star . the excitation of nonradial modes in a classical cepheid would be a very important ingredient in the modelling of the atmospheres of these supergiant stars , thus we submitted this result to further checks . at this stage , we detected the 0.488 d@xmath13 peak in the residuals of each long run . these sets of residuals were obtained by applying a global fit to the lra01 data ( grey points in the upper panel of fig . [ curveres ] ) and to the lra06 ones ( grey points in the lower panel of fig . [ curveres ] ) . indeed , variations in the shape of the residuals are clearly visible , especially after the jumps at jd 2454468 and jd 2456003 . they remind the oscillations noticed before removing the long - term trends and the jumps ( fig . [ lruns ] , middle panels ) . this suggested that the computation of the mean fluxes was very effective in cleaning the power spectra , but the interplay betweeen the large amplitude of corot 0102618121 and the jump correction could somehow leave spurious variability close to the significance level . for this reason we gave a closer look to the light curves of the residuals from which the 0.488 d@xmath13 peak was detected . these light curves show the same type of oscillations , suggesting that the global solution was not able to match the instrumental alterations of the stellar signal . thus we decided to calculate residuals following the same approach used to determine the mean fluxes . this implied to compute as many sets of parameters as the elapsed cycles were , i.e. , 47 for lra01 and 27 for lra06 . the analysis of these new sets of residuals were enlightening . the oscillations visible in fig . [ lruns ] ( middle panels ) had completely disappeared and the light curves became really flat , with just some clues of high - frequency noise ( fig . [ curveres ] , black points ) . indeed , the power spectra of the residuals from fitting one cycle at a time show the peak at the 6@xmath19 harmonic only . this was observed both in the lra01 data and in the lra06 data ( lower part of fig . [ spectrares ] ) . both spectra are characterized by an almost white noise . it could be argued that the signal at 0.488 d@xmath13 was removed when fitting the single cycles ( also taking into account the almost exact 3/4 ratio between the frequencies ) . however , the fact that the 6@xmath19 peak has remained clearly visible gave evidence that the fit with @xmath19 , 2@xmath19 , 3@xmath19 , 4@xmath19 , and 5@xmath19 of each cycle did not modified any signal component apart from these frequencies . we tried to come back to a new correction of the mean fluxes , but again the analysis of the global residuals thus obtained was unsatisfactory . we note that the technique of analysing single cycles were successfully applied to remove similar instrumental effects in the light curves of the high - amplitude @xmath12 sct star corot 101155310 @xcite and of the triple - mode cepheid corot 0223989566 @xcite . l llll c llll & & & + & & & & & & & & & + @xmath20[rad ] & 4.1101(7 ) & 4.220(2 ) & 4.070(2 ) & 3.934(2 ) & & 4.1081(9 ) & 4.191(2 ) & 4.012(6 ) & 3.925(6 ) + @xmath21[rad ] & 4.645(6 ) & 4.637(13 ) & 4.499(21 ) & 4.683(20 ) & & 4.689(7 ) & 4.674(15 ) & 4.814(84 ) & 4.920(64 ) + @xmath22 & 0.05395(4 ) & 0.05371(9 ) & 0.05260(12 ) & 0.05166(11 ) & & 0.05646(5 ) & 0.05717(14 ) & 0.05265(34 ) & 0.04961(30 ) + @xmath23[rad ] & 3.624(7 ) & 3.857(18 ) & 3.431(21 ) & 3.240(17 ) & & 3.573(7 ) & 3.712(26 ) & 3.290(34 ) & 3.193(32 ) + full ampl . [ mag ] & 0.222 & 0.191 & 0.250 & 0.306 & & 0.233 & 0.208 & 0.288 & 0.364 + res . rms [ mag ] & 0.0017 & 0.0036 & 0.0061 & 0.0063 & & 0.0018 & 0.0045 & 0.0150 & 0.0165 + [ colori ] once that we could rule out the excitation of other modes , we fitted the corot magnitudes by means of a cosine series ( see appendix [ appen ] ) . table [ sol ] lists the coefficients of the solutions of the light curves of both runs . the shape of the light curve could be defined by the fourier parameters ( see appendix [ appen ] ) and table [ colori ] lists those relevant to our discussion . the absence of any relevant peak once that the contribution of the main oscillation has been removed is not by itself an evidence that the light curve is stable . cycle - to - cycle variations as those observed in the _ kepler _ light curve of v1154 cyg or in the radial velocity curves could not result in a well defined peak in the power spectrum . moreover , at least in principle , the residual peaks left around @xmath19 when applying a global fit could be ascribed to cycle - to - cycle variations , maybe erratic ones . we analysed the @xmath22 , @xmath20 , and @xmath21 parameters of each cycle to quantify how much the light curve deviates from the average behaviour . the corot 32-s cadence allowed us to get about 7500 measurements per cycle , adequate to survey the harmonics 2@xmath19 and 3@xmath19 in a very accurate way , though their amplitudes are very small ( 6.0 and 0.7 mmag , respectively ) . the fourier parameters are shown in fig . [ fourier ] . the error bars are three times the standard deviations ( @xmath0 ) around the average of the parameters of the individual cycles . the immediate result is that the differences from the average values are small . we could notice that all the individual values are between 3@xmath24 , though some apparently regular variations are sometimes visible ( e.g. the @xmath22 behaviour in the lra01 ) . therefore , we performed a frequency analysis of the fourier parameters . we could not detect any periodicity since the power spectra of the @xmath22 , @xmath20 , and @xmath21 parameters do not show any common feature either in a single long run or when comparing lra01 and lra06 results . however , the best parameter to quantify the stability of the light curve is probably the o - c parameter . we determined the observed t@xmath5 s by means of cubic spline functions with a non - zero smoothing parameter that depends on the number of measurements to fit and on their scatter @xcite . we thus obtained the linear ephemeris from the least - squares method @xmath25 + ( e@xmath26 ) and hence we could calculate the o - c value for each t@xmath5 . the residual rms of the least - squares linear fit is 0.0052 d and the 3@xmath24 band is shown in fig . [ fourier ] . we also estimated the uncertainty on the individual t@xmath5 determination as the difference between the two times corresponding to the intersection of the spline function with the line @xmath27 ( fig . [ spline ] ) , where @xmath28 is the magnitude at maximum and @xmath24 is the standard deviation of the fit . such uncertainties are ranging from 0.010 to 0.016 d and hence the 3@xmath24 interval ( i.e. , 22.5 min ) matches the individual error bars , too . we analysed in frequency the o - c values too , again obtaining no clear indication of periodicities . as a final conclusion , looking at the different plots shown in fig . [ fourier ] and at the results of the various frequency analyses , we can not identify any additional periodicity in the variability of corot 0102618121 . the fourier parameters and the o - c values sometimes mimic an apparent regular behaviour . this could be a hint of cycle - to - cycle variations , much less regular than those observed in the _ kepler _ light curve of v1154 cyg @xcite . we merged the lra01 and lra06 measurements and we calculated a refined value of the period on the whole time baseline . the mtrap code supplied the final vale @[email protected] d , in excellent agreement with the value obtained from the t@xmath5 s values only . the most relevant difference in the datasets of the two long runs is a slight increase from 0.222 to 0.233 mag in the full - amplitude of the light curves . since we are dealing with absolute photometry , i.e. , without any comparison star , we have to be sure that the measurements were performed in the same instrumental system to infer that the amplitude difference has a stellar origin . we immediately noticed that the star was observed in the same ccd in the two runs , but with a different mask and at different pixel coordinates . as a results , the average white flux measured was of 1 030 069 electrons in lra01 and 944,911 electrons in lra06 . moreover , the flux subdvision in @xmath30 , @xmath31 , @xmath32 colours was also different : 65 , 14 , 21 per cent in lra01 , 78 , 10 , 12 per cent in lra06 . we conclude that the small difference in amplitude could very probably be ascribed to the different settings of the corot pointings in the two runs . the same effect occurs in _ kepler _ photometry : average fluxes and amplitudes change for the same star observed in the same ccd but with different masks . these effects are particularly disturbing when dealing with large amplitude variables @xcite . the procedure applied to white light measurements was repeated for each colour of both runs . other similar , subtle instrumental effects became evident during this process @xcite . then , the fourier decomposition was applied to the cleaned timeseries in each colour ( table [ colori ] ) . the parameters clearly show the expected dependence from the bandpass ( see * ? ? ? * for a comparison between @xmath33 and @xmath34 parameters ) . the different flux distributions between lra01 and lra06 also account for the small differences between parameters in the same colour . the procedure described in the previous sections for corot 0102618121 was applied to the other cepheids , whose folded light curves are shown in fig . [ curve ] ( six pop . i stars ) and fig . [ curvaii ] ( one pop . ii star ) . the fourier parameters are listed in table [ others ] . the frequency analysis of the v376 mon data supplied @xmath19=0.6060 d@xmath13 , refined to @xmath35 d with mtrap . table [ sol ] lists the least - squares solution of the corot white light curve . the frequency analysis did not detect any significant additional mode , but again the residual power spectra strongly depend on the procedure used to prewhiten the data . we obtained a bunch of peaks around @xmath19 when applying a global fit up to 7@xmath19 to the entire dataset . these peaks disappeared when the fit was calculated cycle - by - cycle ( fig . [ cef751 ] , upper part ) . the harmonic 8@xmath19 was detected in both cases , confirming that the single cycle fit did not alter the frequency content . the fourier parameters and the o - c values show a small scatter around the means ( fig . [ decocef751 ] ) , suggesting no periodic variation . the frequency analysis of corot 0659466327 data returned @xmath19=0.648 d@xmath13 and the period was refined to @xmath36 d. table [ solii ] lists the least - squares solution of the white light curve . the bump on the ascending branch ( fig . [ curvaii ] ) is typical of pop . ii cepheids . differently than in the previous cases , the peak in the power spectrum of the residuals from the global fit up to 9@xmath19 occurred around 2@xmath19 , but the bunch of peaks at @xmath19 were detected , too . the power spectrum of the residuals from the cycle - by - cycle fits is completely flat and it just shows the 10@xmath19 harmonic , not prewhitened from the data ( fig . [ cef751 ] , lower part ) . small - size scale variations are observed in the fourier parameters and o - c values , but again without any trace of periodicity ( fig . [ decocef327 ] ) . the power spectra of the residuals of corot 0102618121 , v376 mon and corot 0659466327 calculated by means of a global least - squares solution show the limits of this approach when applied to continuous and high - precision timeseries where systematic errors ( jumps and/or drifts ) can occur . when possible , e.g. in presence of a high - amplitude predominant mode , the frequency analysis of cycle - to - cycle residuals are preferable , taking care to keep a known frequency , e.g. an harmonic , as internal check of the signal preservation . the solutions of the light curves of the cepheids corot 0221640090 ( table [ soliii ] ) , corot 0211626074 ( table [ soliv ] , corot 0315221415 ( table [ solv ] ) , and dp mon ( table [ solvi ] were much less laborious since these stars were observed once in some short runs ( srs ) and hence followed for a restricted number of cycles . the o - c plots do not provide useful hints about periodic variations ( fig . [ ocallcep ] ) . only in the case of corot 0221640090 ( bottom panel ) the string of consecutive t@xmath5 shows a hint of regular behaviour , but the monitoring is too short to assess its repetitivity on a long time - scale . moreover , the standard deviation around the linear ephemeris is @xmath30.0033 d and all the maxima are well within 3@xmath24 . ll lll llll & & & & & & & & + & & & & & & & & + v376 mon & @xmath1 & 4785.4731 & 1.651969(1 ) & 0.372 & 4.1811(3 ) & 0.3745(1 ) & 2.2206 ( 1 ) & 0.172(1 ) + corot 0221640090 & @xmath2 & 4754.6755 & 2.19758(4 ) & 0.288 & 4.560(1 ) & 0.1397(1 ) & 3.641(5 ) & 2.670(9 ) + dp mon & @xmath1 & 4754.0407 & 3.08704(4 ) & 0.678 & 4.312(1 ) & 0.5201(1 ) & 2.4703(4 ) & 0.5439(6 ) + corot 0315221415 & @xmath1 & 5258.4566 & 4.7085(17 ) & 0.557 & 4.363(1 ) & 0.4650(5 ) & 2.478(3 ) & 0.434(5 ) + corot 0211626074 & @xmath1 & 4199.7103 & 5.4706(2 ) & 0.296 & 4.731(1 ) & 0.3147(3 ) & 3.242(3 ) & 1.391(35 ) + corot 0659466327 & pop . ii & 6031.4482 & 1.542353(2 ) & 0.492 & 4.394(1 ) & 0.2913(2 ) & 0.886(1 ) & 4.953(2 ) + the fourier decomposition of the light curves of cepheids allowed us to disentangle @xmath1 and @xmath2 radial pulsators ( fig . [ grafo ] ) . the @xmath20-sequence formed by the classical cepheids reproduces the well - known hertzsprung progression . it can be described by the linear relation @xmath37 @xcite . the progression is sharp , since a shift of @xmath30.30 rad pushes a star well outside it . another z - shaped sequence was put in evidence by applying the fourier decomposition to the light curves of galactic cepheids with @xmath388 d : the upper part of this sequence is composed of stars with @[email protected] rad , the lower part of stars with @xmath20@xmath40 4.0 rad ( fig . [ grafo ] , top panel ) . the @xmath21@xmath41 plane was the rosetta stone to link the upper and lower sequences , since it shows just one continuous sequence , well separated from the classical one ( fig . [ grafo ] , middle panel ) . the fourier decomposition of the light curves of double - mode ( @xmath42 ) cepheids provided the definitive evidence that the pulsation mode drives the separation between the different sequences @xcite . the @xmath22@xmath41 plot shows that the amplitude of the harmonic 2@xmath19 of the @xmath2-mode pulsators is much smaller than that of the main frequency @xmath19 ( fig . [ grafo ] , bottom panel ) . this fact makes the light curves of the @xmath2-mode pulsators much more ( but not exactly ) sine - shaped than those of @xmath1-mode pulsators . theoretical works demonstrated how the inclusion of convective energy transport has made possible to reproduce the observed sequence of @xmath2-pulsators and that the sharp features in the fourier coefficients are due to the resonance @xmath43 between the fourth and first overtones @xcite . the classification of corot 0211626074 and corot 0315221415 is straightforward , since the @xmath20 , @xmath22 , and @xmath21 parameters put both these long - period ( in our sample ) cepheids on the well - defined sequence of @xmath1-mode pulsators . the period of corot 0102618121 and dp mon are very particular in this context . they lie exactly on the discontinuity of the z - shaped sequence of the @xmath2 pulsators in the @xmath20@xmath41 plane , where the intersection with the @xmath1-sequence occurs . therefore , the @xmath20 parameter alone is not able to identify their pulsation mode . however , the much higher value of @xmath21 and the much smaller value of @xmath22 with respect to those of the classical cepheids clearly demonstrate that corot 0102618121 is a @xmath2 pulsator . on the other hand , the @xmath21 and @xmath22 parameters of dp mon match the @xmath1-sequence . the light curve of dp mon is very asymmetrical ( fig . [ curve ] ) and it is very similar to those of rrab stars . in particular , notice the kink before the minimum , that requires a large number of harmonics to be fitted . we classify corot 0221640090 as a @xmath2 pulsator , like corot 0102618121 . the case of v376 mon is more intriguing , since its period is very short also with respect to the full sample of galactic cepheids . the cepheids with @xmath382.5 d are quite rare and the space of fourier parameters is scarcely populated in this region . for this reason the values of @xmath20 and @xmath21 are not able to provide a clear mode identification . however , the @xmath22 value is quite high , and hence the star is very probably a @xmath44mode pulsator . the continuous photometry allowed us to verify the phases of maximum brightness ( @xmath45 ) in the corot colours with great accuracy . the analyses described in the previous sections were repeated for each colour of each star . the results are summarized in table [ coloriall ] . there is a common trend indicating that the maximum brightness occurs a little earlier at short wavelengths than at long ones . the only exception is dp mon , whose maxima are measured at the same phases ( within error bars ) in all colours . the observed phase shifts are very small , around 0.01 period . this result strongly corroborates the practice to merge all the maxima observed with different filters or techniques when studying secular period variations . indeed , this shift is usually smaller than the uncertainties on the t@xmath5 determination in case of ground - based , fragmented observations . this is the first measurement of the phase shifts among light curves of cepheids in different colours after the pioneering approach by @xcite . l rrrr & & & & + corot 0102618121 & 0.993 & 0.997 & 1.004 & 1.000 + in lra01 & @xmath30.003 & @xmath30.003 & @xmath30.003 & @xmath30.003 + corot 0102618121 & 0.994 & 0.996 & 1.003 & 1.000 + in lra06 & @xmath30.005 & @xmath30.005 & @xmath30.003 & @xmath30.003 + corot 0221640090 & 0.994 & 1.002 & 1.007 & 1.000 + & @xmath30.003 & @xmath30.003 & @xmath30.005 & @xmath30.003 + v376 mon & 0.996 & 0.998 & 1.002 & 1.000 + & @xmath30.003 & @xmath30.003 & @xmath30.003 & @xmath30.003 + dp mon & 0.998 & 1.000 & 1.001 & 1.000 + & @xmath30.003 & @xmath30.003 & @xmath30.003 & @xmath30.003 + the power spectra of the corot light curves of galactic cepheids did not show any convincing evidence of the excitation of nonradial modes , especially in the three monitored in the long runs . stars showing additional modes were found with an incidence of 10% among the @xmath2-cepheids of the lmc @xcite . the only two @xmath2-pulsators observed by corot can not yet tell us if galactic @xmath2-cepheids behave differently than lmc ones . on the other hand , the @xmath1-cepheids of the lmc do not show additional modes @xcite and the intensive corot photometry does not modify this issue . the detection of the same peak in two independent sets of residuals of corot 0102618121 seemed to be a decisive piece of evidence in favour of the excitation of a nonradial mode , but it not survived the change of the procedure for the computation of the residuals . in our opinion the instrumental effects seem to play an important role in the long time series of space photometry . our recommendation is to analyse residuals calculated from local solutions ( e.g. cycle - to - cycle when possible ) to search for additional periodicities . instrumental drifts and jumps ( fig . [ lruns ] ) or , in extreme cases , numerical errors or amplitude changes can enhance spurious peaks in the power spectrum of the residuals calculated from a global fit ( fig . [ spectrares ] ) . the residuals from the cycle - by - cycle fits are providing a much more reliable dataset to be analysed in frequency ( fig . [ curveres ] ) . omitting a high harmonic from the fit and verify that it is correctly recovered is an excellent litmus test of the whole procedure . moreover , the lack of combination terms between the main pulsation mode and the suspected additional component should be considered a warning against the identification of this component as a true excited mode . another compelling result expected from the analysis of the corot light curves was to verify the repetitivity and the amplitude of the cycle - to - cycle variations , as those observed in the _ kepler _ data of v1154 cyg @xcite and in the radial - velocity surveys @xcite . the fourier parameters of the light curves of corot cepheids show some variability , but in general all the changes are within @xmath0 . moreover , they seem to be erratic , not reproducing phenomena like activity cycle or long - term effects . no large scatter was observed around the maxima of the folded light curves and no large variations were observed in the o - c plots of the cepheids monitored with corot . in particular , the o - c values are all within @xmath30.01 d , i.e.,@xmath315 min , including the @xmath2-mode pulsators corot 0102618121 and corot 0221640090 ( figs . [ fourier ] , [ decocef751 ] , [ decocef327 ] , and [ ocallcep ] ) . this is particularly interesting since the space photometry obtained with the satellite _ most _ showed some instabilities in the pulsation cycle of the @xmath2-cepheid sz tau , while the @xmath1-mode rt aur repeated the light curve more precisely @xcite . therefore , we performed some additional tests on the folded light curves , evaluating the standard deviations as a function of phase bins . tiny increases of the standard deviations were measured in correspondence of the most rapid light variation ( i.e. , on the steep ascending branch ) ; they mostly reflect the variability due to the pulsation within the phase bins , as proven by the fact that they appear to be more relevant for the stars showing the largest amplitudes and the most asymmetrical light curves . this suggests that the standard deviations of the phase bins on the rising branch can not be used as a tool to search for cycle - to - cycle variations . we can conclude that in the case of the seven corot cepheids , the cycle - to - cycle variability is very limited and not greatly affecting the repetitivity of the light curves . the light variations observed around the maxima of sz tau are more promising , but on this context the low - frequency peaks in the power spectrum have to be carefully considered : the residuals observed in _ most _ photometry are very similar to the spurious ones detected in corot one ( fig . [ curveres ] ) . we also benefit from the unique opportunity provided by the corot multicolour photometry . very small shifts in the t@xmath5 s were observed for the first time in the optical range . because these shifts are around 0.01 period only , our analysis support the use of t@xmath5 s observed in different passbands to reconstruct secular period variations ( e.g. * ? ? ? * ) . the accurate values of the fourier parameters provided by the available space photometry up to very high harmonics can be used to improve the automatic classification of the cepheid subclasses , i.e. , pop . i , pop . ii , @xmath1-mode , and @xmath2-mode pulsators . the quantitative separation of these subclasses in the parameter space is of particular relevance when considering the large harvest of new variables that will be discovered by future , wide - field , and all - sky space missions like tess @xcite and plato 2.0 @xcite . the corot space mission has been developed and operated by cnes , with contributions from austria , belgium , brazil , esa ( rssd and science program ) , germany , and spain . the authors thank the anonymous referee for useful comments . ep acknowledges observatoire midi - pyrnes for the two - months grant allocated between 2014 may and july ; useful discussions with boris dintrans and pascal fouqu at the very beginning of the project are acknowledged . jmb thanks for the support of the nkfih grant k-115709 . mr acknowledges financial support from the fp7 project _ spaceinn : exploitation of space data for innovative helio- and asteroseismology _ , ep acknowledges prin - inaf 2014 _ galactic archaelogy _ for support to latest activities . this research has made use of the exodat database , operated at lam - oamp , marseille , france , on behalf of the corot / exoplanet program . the present study has used the simbad data base operated at the centre de donnes astronomiques ( strasbourg , france ) . mnras anderson r.i . , 2014 , , 566 , l10 antonello e. , poretti e. , reduzzi l. , 1990 , , 236 , 138 benk j.m . , plachy e. , szab r. , molnr l. , kollth z. , 2014 , , 213 , 31 berdnikov l.n . , kniazev a.yu . , sefako r. , kravtsov v.v . , zhujko s.v . , 2014 , astronomy letter , 40 , 125 bruntt h. , et al . , 2008 , , 683 , 433 buchler j.r . , 2008 , , 680 , 1412 buchler j.r , szab r. , 2007 , , 660 , 723 burki g. , mayor m. , benz w. , 1982 , , 109 , 258 carpino m. , milani a. , nobili a.m. , 1987 , a&a , 181 , 182 debosscher j. , et al . , 2009 , , 506 , 519 deleuil m. , et al . , 2009 , , 138 , 649 derekas a. , et al . , 2012 , , 425 , 1312 donati j .- f . , semel m. , carter b.d . , rees d.e . , collier cameron a. , 1997 , , 291 , 658 evans n.r . , sasselov d.d . , short i.s . , 2002 , , 567 , 1121 evans n.r . , et al . , 2015 , , 446 , 4008 feuchtinger m. , buchler j.r . , kollth z. , 2000 , , 544 , 1056 fruth t. , et al . , 2012 , , 143 , 140 gruberbauer m. , et al . , 2007 , , 379 , 1498 gustafsson b. , edvardsson b. , eriksson k. , jrgensen u.g . , nordlund . , plez b. , 2008 , , 486 , 951 kabath p. , eigmller p. , erikson a. , hedelt p. , rauer h. , titz r. , wiese t. , karoff c. , 2007 , , 134 , 1560 kabath p. , et al . , 2008 , , 136 , 654 kabath p. , et al . , 2009 , , 137 , 3911 kabath p. , et al . , 2009 , , 506 , 569 kanev e. , savanov i. , sachkov m. , 2015 , in 3@xmath46 corot symposium , 7@xmath7 kasc meeting , the space photometry revolution , toulouse , july 2014 , in press karoff c. , et al . , 2007 , , 134 , 766 kuschnig r. , weiss w.w . , gruber r. , bely p.y . , jenkner h. , 1997 , , 328 , 544 molnr l. , szabados l. , 2014 , , 442 , 3222 montgomery m.h . , odonoghue d. , 1999 , delta scuti star newsletter , 13 , 28 morgan s.m . , simet m. , bargenquast , s. , 1998 , , 48 , 509 moskalik p. , 2014 , in guzik j.a . , chaplin w.j . , handler g. , pigulski a. , eds . , iau proceedings s301 , vol . 9 , precision asteroseismology , p. 249 moskalik p. , koaczkowski z. , 2008 , comm . in asteroseismology , 157 , 343 moskalik p. , koaczkowski z. , 2009 , , 394 , 1649 neilson h.r . , ignace r. , 2014 , , 563 , l4 pardo i. , poretti e. , 1997 , , 324 , 121 poretti e. , 2000 , in ibanoglu c. , ed . , nato science series , vol . 544 , variable stars as essential astrophysical tools , kluwer academic publishers , p. 421 poretti e. , baglin a. , weiss w.w . , 2014 , , 795 , l36 poretti e. , le borgne j.f . , klotz a. , 2015 , in 3@xmath46 corot symposium , 7@xmath7 kasc meeting , the space photometry revolution , toulouse , july 2014 , preprint ( arxiv:1506.08122 ) poretti e. & pardo i. , 1997 , , 324 , 133 poretti e. , et al . , 2011 , , 528 , a147 poretti e. , et al , 2005 , , 440 , 1097 rainer m. , 2003 , laurea thesis , universit degli studi di milano rauer h. , et al . , 2014 , exp . astron . , 38 , 249 reinsch h. , 1967 , numerische mathematik 10 , 177 ricker g.r . , et al . , 2014 , spie proc . , 9143 , 20 spreckley s.a . , stevens i.r . , 2008 , , 388 , 1239 szildi k. , vink j. , poretti e. , szabados l. , kun m. , 2007 , , 473 , 579 valenti j.a . , fischer d.a . , 2005 , , 159 141 valenti j.a . , piskunov n. , 1996 , , 118 , 595 vaniek p. , 1971 , , 12 , 10 wachmann a.a . , 1964 , astron . abhandlungen hamburger sternwarte , 7 , 4 wisniewski w.z . , johnson h.l . , 1968 , communications of the lunar and planetary laboratory , volume 7 , part 2 , p. 57 - 78 wu y. , 2001 , , 323 , 248 we provide the solutions of the corot light curves of the cepheids observed in the long runs . we fitted the corot magnitudes by means of the formula @xmath47}\ ] ] and we calculated the fourier parameters @xmath48 ( e.g. @xmath49 ) and @xmath50 ( e.g. @xmath51 and @xmath52 ) from the amplitude ( @xmath53 ) and phase ( @xmath54 ) coefficients . the error bars were calculated at the @xmath55 level @xcite . r r ll c ll c ll & & & + & & & & & + & & & & & & & & & + & & & & & & & & & + @xmath19&&0.110500(4 ) & 3.06375(4 ) & & 0.116054(6 ) & 3.05962(5 ) & & 0.15852(2 ) & 2.5446(1 ) + 2@xmath19&&0.005962(4 ) & 3.95440(72 ) & & 0.006552(6 ) & 3.94414(90 ) & & 0.05937(2 ) & 2.9871(2 ) + 3@xmath19&&0.000699(4 ) & 1.26965(616 ) & & 0.000896(6 ) & 1.30168(655 ) & & 0.02715(2 ) & 3.5712(5 ) + 4@xmath19&&0.000621(4 ) & 3.31223(693 ) & & 0.000809(6 ) & 3.24492(726 ) & & 0.01096(2 ) & 4.0676(14 ) + 5@xmath19&&0.000287(4 ) & 4.55196(1499 ) & & 0.000389(6 ) & 4.51949(1509 ) & & 0.00435(2 ) & 4.4503(34 ) + 6@xmath19&&0.000135(4 ) & 5.74428(3180 ) & & 0.000120(6 ) & 5.64753(4876 ) & & 0.00156(2 ) & 4.6689(94 ) + 7@xmath19&&0.000049(4 ) & 0.55276(8822 ) & & 0.000078(6 ) & 0.09852(7486 ) & & 0.00077(2 ) & 4.5913(190 ) + 8@xmath19&&0.000024(4 ) & 2.10979(17710 ) & & 0.000025(6 ) & 1.97647(23114 ) & & 0.00049(2 ) & 4.6907(297 ) + 9@xmath19 & & & & & & & & 0.00036(2 ) & 4.9298(408 ) + 10@xmath19 & & & & & & & & 0.00028(2 ) & 5.3570(534 ) + 11@xmath19 & & & & & & & & 0.00016(2 ) & 5.8615(952 ) + 12@xmath19 & & & & & & & & 0.00010(2 ) & 6.1633(1511 ) + 13@xmath19 & & & & & & & & 0.00007(2 ) & 0.7590(2155 ) + 14@xmath19 & & & & & & & & 0.00005(2 ) & 0.7422(3075 ) + & & & & & + & & & & & + [ sol ] r r ll & + & + & & & + & & & + @xmath19 & & 0.19348(4 ) & 2.6323(2 ) + 2@xmath19 & & 0.05637(4 ) & 3.3752(7 ) + 3@xmath19 & & 0.03478(4 ) & 2.5001(12 ) + 4@xmath19 & & 0.02705(4 ) & 2.9155(16 ) + 5@xmath19 & & 0.01608(4 ) & 3.8907(26 ) + 6@xmath19 & & 0.01154(4 ) & 4.7951(36 ) + 7@xmath19 & & 0.00629(4 ) & 5.6173(67 ) + 8@xmath19 & & 0.00262(4 ) & 0.3086(161 ) + 9@xmath19 & & 0.00055(4 ) & 1.1627(767 ) + 10@xmath19 & & 0.00052(4 ) & 6.0158(832 ) + 11@xmath19 & & 0.00076(4 ) & 0.8167(555 ) + 12@xmath19 & & 0.00083(4 ) & 1.9510(508 ) + 13@xmath19 & & 0.00068(4 ) & 2.9567(614 ) + 14@xmath19 & & 0.00041(4 ) & 4.1419(1024 ) + 15@xmath19 & & 0.00019(4 ) & 5.2084(2196 ) + 16@xmath19 & & 0.00007(4 ) & 5.7712(6216 ) + 17@xmath19 & & 0.00009(4 ) & 5.4264(4543 ) + 18@xmath19 & & 0.00016(4 ) & 0.2652(2712 ) + & + & + [ solii ] r r ll & + & + & & & + & & & + @xmath19 & & 0.14079(1 ) & 2.9111(1 ) + 2@xmath19 & & 0.01967(1 ) & 4.0987(7 ) + 3@xmath19 & & 0.00290(1 ) & 6.0910(47 ) + 4@xmath19 & & 0.00156(1 ) & 1.7476(87 ) + 5@xmath19 & & 0.00102(1 ) & 3.1642(132 ) + 6@xmath19 & & 0.00055(1 ) & 4.3818(244 ) + 7@xmath19 & & 0.00035(1 ) & 5.5170(384 ) + 8@xmath19 & & 0.00017(1 ) & 0.4274(795 ) + 9@xmath19 & & 0.00009(1 ) & 1.3975(1468 ) + 10@xmath19 & & 0.00002(1 ) & 1.9638(6112 ) + 11@xmath19 & & 0.00006(1 ) & 3.2910(2157 ) + 12@xmath19 & & 0.00002(1 ) & 3.6209(6119 ) + & + & + [ soliii ] r r ll & + & + & & & + & & & + @xmath19 & & 0.12984(3 ) & 2.5207(3 ) + 2@xmath19 & & 0.04086(3 ) & 3.4897(8 ) + 3@xmath19 & & 0.01028(3 ) & 4.5213(32 ) + 4@xmath19 & & 0.00095(3 ) & 5.1938(341 ) + 5@xmath19 & & 0.00166(3 ) & 3.0263(195 ) + 6@xmath19 & & 0.00128(3 ) & 3.7305(253 ) + 7@xmath19 & & 0.00103(3 ) & 4.5825(315 ) + 8@xmath19 & & 0.00062(3 ) & 5.3574(527 ) + 9@xmath19 & & 0.00036(3 ) & 6.2663(912 ) + 10@xmath19 & & 0.00018(3 ) & 0.8783(1842 ) + 11@xmath19 & & 0.00010(3 ) & 2.1266(3298 ) + 12@xmath19 & & 0.00008(3 ) & 2.7998(4124 ) + & + & + [ soliv ] r r ll & + & + & & & + & & & + @xmath19 & & 0.21116(6 ) & 2.4327(3 ) + 2@xmath19 & & 0.09819(6 ) & 2.9456(6 ) + 3@xmath19 & & 0.04415(6 ) & 3.4927(14 ) + 4@xmath19 & & 0.02011(6 ) & 3.8819(31 ) + 5@xmath19 & & 0.00735(6 ) & 4.0916(84 ) + 6@xmath19 & & 0.00386(6 ) & 3.9842(159 ) + 7@xmath19 & & 0.00298(6 ) & 3.7086(207 ) + 8@xmath19 & & 0.00314(6 ) & 3.8961(196 ) + 9@xmath19 & & 0.00299(6 ) & 4.3491(205 ) + 10@xmath19 & & 0.00246(6 ) & 4.8484(248 ) + 11@xmath19 & & 0.00201(6 ) & 5.4295(305 ) + 12@xmath19 & & 0.00158(6 ) & 6.0745(387 ) + 13@xmath19 & & 0.00122(6 ) & 0.3769(501 ) + 14@xmath19 & & 0.00087(6 ) & 1.1612(706 ) + 15@xmath19 & & 0.00060(6 ) & 1.8127(1016 ) + 16@xmath19 & & 0.00045(6 ) & 2.6720(1363 ) + 17@xmath19 & & 0.00032(6 ) & 3.5748(1915 ) + 18@xmath19 & & 0.00016(6 ) & 4.2500(3960 ) + 19@xmath19 & & 0.00020(6 ) & 5.1683(3062 ) + 20@xmath19 & & 0.00016(6 ) & 0.3755(3767 ) + & + & + [ solv ] r r ll & + & + & & & + & & & + @xmath19 & & 0.23679(2 ) & 2.2927(1 ) + 2@xmath19 & & 0.12315(2 ) & 2.6145(1 ) + 3@xmath19 & & 0.07198(2 ) & 3.0652(2 ) + 4@xmath19 & & 0.04044(2 ) & 3.4315(4 ) + 5@xmath19 & & 0.02352(2 ) & 3.6996(7 ) + 6@xmath19 & & 0.01444(2 ) & 4.0036(12 ) + 7@xmath19 & & 0.00856(2 ) & 4.2711(20 ) + 8@xmath19 & & 0.00523(2 ) & 4.4842(33 ) + 9@xmath19 & & 0.00323(2 ) & 4.6872(54 ) + 10@xmath19 & & 0.00205(2 ) & 4.7897(85 ) + 11@xmath19 & & 0.00126(2 ) & 4.8674(137 ) + 12@xmath19 & & 0.00096(2 ) & 5.0046(180 ) + 13@xmath19 & & 0.00074(2 ) & 5.0513(234 ) + 14@xmath19 & & 0.00058(2 ) & 5.2112(296 ) + 15@xmath19 & & 0.00048(2 ) & 5.3839(359 ) + 16@xmath19 & & 0.00041(2 ) & 5.6321(425 ) + 17@xmath19 & & 0.00034(2 ) & 5.9751(515 ) + 18@xmath19 & & 0.00024(2 ) & 6.0823(712 ) + 19@xmath19 & & 0.00022(2 ) & 0.1627(799 ) + 20@xmath19 & & 0.00014(2 ) & 0.5256(1234 ) + 21@xmath19 & & 0.00016(2 ) & 0.7288(1113 ) + 22@xmath19 & & 0.00010(2 ) & 1.3601(1793 ) + 23@xmath19 & & 0.00009(2 ) & 1.1871(2008 ) + 24@xmath19 & & 0.00003(2 ) & 1.4777(5286 ) + 25@xmath19 & & 0.00005(2 ) & 1.8188(3629 ) + & + & + [ solvi ]

we present an algorithm that finds all toric noncommutative crepant resolutions of a given toric 3-dimensional gorenstein singularity . the algorithm embeds the quivers of these algebras inside a real 3-dimensional torus such that the relations are homotopy relations . one can project these embedded quivers down to a 2-dimensional torus to obtain the corresponding dimer models . we discuss some examples and use the algorithm to show that all toric noncommutative crepant resolutions of a finite quotient of the conifold singularity can be obtained by mutating one basic dimer model . we also discuss how this algorithm might be extended to higher dimensional singularities .

1104.1597

in @xcite van den bergh introduced the notion of a noncommutative crepant resolution of a singularity , which is an algebra that can act as a substitute of an ordinary commutative crepant resolution . a noncommutative crepant resolution of a singularity with coordinate ring @xmath0 is a homologically homogeneous algebra of the form @xmath1 where @xmath2 is a a reflexive @xmath0-module . in the case of three - dimensional terminal gorenstein singularities , the derived category of representations of @xmath3 is equivalent to the derived category of coherent sheaves of a commutative crepant resolution . the algebra @xmath3 can be seen as a path algebra of a quiver @xmath4 with relations . the vertices of this quiver correspond to the direct summands of @xmath2 and the arrows to a basic set of maps between them . one can also define a dimension vector @xmath5 which assigns to each vertex the rank of the corresponding summand . with these data the singular variety @xmath6 can be recovered as a moduli space parameterizing @xmath5-dimensional semisimple representations and in many cases a commutative resolution can be constructed by taking a moduli space parameterizing @xmath5-dimensional stable representations for some stability condition . if @xmath0 is a toric singularity , we would like our noncommutative resolution to carry an additional toric structure . this can be done by asking that all summands of @xmath2 are have rank 1 and are graded . this ensures that the moduli spaces can be constructed as toric quotients of toric varieties . if this is the case we call @xmath7 a toric noncommutative crepant resolution . an interesting problem is to classify all possible toric noncommutative crepant resolutions of a given singularity . in this paper we will describe an algorithm that does this for any toric three - dimensional gorenstein singularity . we then use a method by craw and quintero - velez @xciteto embed the quivers of these algebras inside a real 3-torus , such that the relations are precisely the homotopy relations . if the singularity is gorenstein the quiver can be projected to a 2-torus to obtain a dimer model . this is a combinatorial gadget that was originally introduced in string theory @xcite . we illustrate the power of the algorithm by looking at some special examples : singularities from reflexive polygons and abelian quotients of the conifold . in these cases we can prove that all the dimer models corresponding to such a singularity are connected by mutations . finally we study the possible generalization of this algorithm to the case where the toric singularity is not gorenstein or has higher dimension . i want to thank nathan broomhead , ben davison and travis schedler for the interesting discussions we had about this project during their visits to newcastle . i also want to thank lutz hille for the stimulating research stay i had in bielefeld . let @xmath0 be the coordinate ring of a singular variety @xmath8 . a proper surjective map @xmath9 is a _ resolution _ if it induces an isomorphism on the function fields and @xmath10 is a smooth variety . there are many different ways to resolve a singularity and therefore one wants to impose extra conditions . a resolution is called _ crepant _ if the pullback of the canonical divisor of @xmath8 under @xmath11 is the canonical divisor of @xmath10 . in particular if @xmath8 is a gorenstein singularity ( i.e. the canonical divisor is trivial ) then @xmath10 must be a calabi yau variety ( i.e. a smooth variety with a trivial canonical bundle ) . in 2 dimensions crepant resolutions are unique , in 3 dimensions a singularity can have more than one crepant resolution but they are closely related . if @xmath12 and @xmath13 are two crepant resolutions then bridgeland @xcite proved their bounded derived categories of coherent sheaves are the equivalent . in @xcite van den bergh introduced special algebras which can act as a noncommutative analogue of a crepant resolution . an algebra @xmath3 is a _ noncommutative crepant resolution _ of @xmath0 if it satisfies 2 conditions * @xmath14 where @xmath2 is a finitely generated reflexive @xmath0-module ( reflexive means @xmath15 ) , * @xmath3 is homologically homogeneous i.e. all simples have the same projective dimension . he used this definition to show that in 3 dimensions these algebras behave like crepant resolutions : [ vdb ] for a terminal 3-dimensional gorenstein singularity , the bounded derived category of finitely generated modules of a noncommutative crepant resolution is equivalent to the bounded derived category of coherent sheaves of a crepant resolution and vice versa . noncommutative crepant resolutions exist if and only if a commutative crepant resolution exists . from now on we will abbreviate the term commutative crepant resolution by ccr and noncommutative crepant resolution by nccr . a nice way to construct nccrs was introduced by iyama and wemyss in @xcite . for a positively graded commutative ring @xmath0 with @xmath16 and krull dimension @xmath17 we will call a module @xmath2 _ cohen - macaulay ( abbreviated as cm ) _ if @xmath18 for all @xmath19 . an algebra @xmath3 is called a _ modification algebra _ if it is of the form @xmath20 with @xmath2 a reflexive module and @xmath3 is cohen - macaulay as an @xmath0-module . @xmath20 is called a _ maximal modification algebra ( abbreviated as mma ) _ if for every reflexive module @xmath21 such that @xmath22 is cohen - macaulay then @xmath21 is a direct sum of summands of @xmath2 . in other words one can not add anything new to @xmath2 such that its endomorphism ring stays cohen - macaulay . if @xmath7 is a ( maximal ) modification algebra we will call @xmath2 a ( maximal ) modification module . an nccr is always an mma , but the converse is not true : it might be the case that a singularity has an mma that does not have finite global dimension . but as soon as a modification algebra has finite global dimension then it is an nccr ( see lemma 4.2 in @xcite ) . in three dimensions we have the following : [ mmaisnccr ] if @xmath0 is a 3-dimensional gorenstein singularity for which there exists an nccr , then the nccrs are precisely the mmas . let @xmath23 be a group isomorphic to @xmath24 and let @xmath25 , we use @xmath26 denote the pairing between @xmath23 and @xmath27 and between @xmath28 and @xmath29 . now let @xmath30 be vectors in @xmath27 such that the @xmath31 form the rays of an integral polyhedral cone @xmath32 . we also assume that @xmath33 , so the vectors are irreducible . we write @xmath34}:= { { \mathbb{r}}_+}v_1+\dots + { { \mathbb{r}}_+}v_k$ ] . the _ dual cone _ @xmath35 is defined as @xmath36 if we intersect this cone with @xmath23 we get a semigroup and we can use this semigroup to construct a semigroup algebra @xmath37 $ ] . this ring is positively graded and normal . a cone @xmath38}$ ] is called a _ face _ of the cone if and only if there is an @xmath39 such that @xmath40 . a face @xmath41 of @xmath42 will give an embedding @xmath43 and a inclusion @xmath44 and one can see @xmath45 as a localization of @xmath46 . this gives an embedding @xmath47 . a _ fan _ @xmath48 is a collection of integral polyhedral cones in @xmath29 such that * if @xmath41 is a face of @xmath49 then @xmath50 and * if @xmath51 then their intersection is a face of both @xmath52 . if @xmath48 is a fan we get a variety @xmath53 by taking the direct limit of the @xmath54 with respect to the embedding maps @xmath47 . to every cone @xmath42 we can associate its fan @xmath55 . the variety @xmath56 is @xmath57 . @xmath58 will be singular if and only if the @xmath30 do not form a basis for @xmath27 . to construct a resolution of @xmath59 we need to subdivide the cone @xmath42 in smaller cones that all correspond to smooth varieties . the fan @xmath48 containing all these cones will then give us a variety @xmath60 which is a resolution of @xmath59 . a toric singularity @xmath58 with @xmath61}$ ] will be called _ gorenstein _ if and only if there is an element @xmath62 in @xmath23 such that @xmath63 for all @xmath64 . this means that all @xmath64 lie in a common hyperplane and they form the vertices of a convex @xmath65-dimensional polytope . if the dimension of the singularity is @xmath66 this polytope is a convex polygon in a plane with integral coefficients . subdivide the polygon into triangles with size @xmath67 and take the cones over these triangles to obtain a subdivision of @xmath42 . the size of the triangles implies that the cones all give smooth varieties , so this subdivision gives rise to a resolution of @xmath58 . because the vectors that generate the subcones all lie in the same plane the resolution is calabi - yau and hence it is a ccr . as usual a _ quiver _ @xmath4 is an oriented graph . we denote the set of vertices by @xmath68 , the set of arrows by @xmath69 and the maps @xmath70 assign to each arrow its head and tail . a _ nontrivial path _ @xmath71 is a sequence of arrows @xmath72 such that @xmath73 . a _ trivial path _ is just a vertex . we will assume that our quivers are strongly connected i.e. there exist a path from each vertex to each other vertex . the _ path algebra _ @xmath74 is the complex vector space with as basis the paths in @xmath4 and the multiplication of two paths @xmath71 , @xmath75 is their concatenation @xmath76 if @xmath77 or else @xmath78 . the span of all paths of nonzero length form an ideal which we denote by @xmath79 . a _ path algebra with relations _ @xmath80 is the quotient of a path algebra by a finitely generated ideal @xmath81 . we will call a path algebra with relations @xmath82 _ positively graded _ if there exists a grading @xmath83 such that @xmath84 is generated by homogeneous relations . a _ dimer model _ @xmath85 is a strongly connected quiver @xmath4 enriched with a 2 disjoint sets of cycles of length at least @xmath66 : @xmath86 and @xmath87 , such that * * orientability condition . * every arrow is contained exactly once in one cycle in @xmath88 and once in one in @xmath87 . * * manifold condition . * the incidence graph of the cycles and arrows meeting a given vertex is connected . considering the cycles as polygons and gluing them together by the arrows we get a compact surface . we speak of a dimer model on a torus if this surface is a torus . a _ consistent @xmath89-charge _ is a grading @xmath90 such that * @xmath91 , * @xmath92 . given a dimer model we can construct its _ jacobi algebra _ @xmath93 the main results concerning dimers and nccrs the jacobi algebra of a dimer model is an nccr of its center if and only if it admits a consistent @xmath89-charge . in this case the center is a toric 3-dimensional gorenstein singularity . broomhead proved this theorem for geometrically consistent @xmath89-charges ( i.e. @xmath94 ) in @xcite . the method of this proof was later extended to include all consistent @xmath89-charges in @xcite . [ atleastonedimer ] for every toric 3-dimensional gorenstein singularity , there is a dimer model with a consistent @xmath89-charge that has this singularity as the center of its jacobi algebra . the proofs of gulotta @xcite and ishii and ueda @xcite are constructive : they provide an algorithm to construct at least one dimer model . varying the starting parameters of the algorithm one can obtain more than one dimer model for a given singularity , but it is far from clear whether you will obtain all possible dimer models . one of the aims in this paper is to construct a new algorithm that produces all dimer models for a given singularity . not every nccr of a toric 3-dimensional gorenstein singularity comes from a dimer model , but if we restrict to toric nccrs , a natural restriction we will introduce in the next section , one has the following : [ consistency ] every toric nccr of a toric 3-dimensional gorenstein singularity is isomorphic to the jacobi algebra of a dimer model that admits a consistent @xmath89-charge . now we can combine the concepts of the preliminary sections to obtain the notion of _ toric _ nccrs and mmas . let @xmath1 be an nccr of @xmath95 $ ] . the standard way to construct a commutative crepant resolution goes as follows . first one writes @xmath80 as the path algebra of a quiver with relations . the vertices of this quiver will be in one - to - one correspondence with the direct summands of @xmath2 . let @xmath96 be the direct summand corresponding to vertex @xmath97 to this quiver we can associate a dimension vector @xmath98 , which allows us to define the scheme @xmath99 , which parametrizes all @xmath5-dimensional presentations of the path algebra @xmath74 . this scheme contains a closed subscheme @xmath100 that parametrizes all @xmath74-representations that are also @xmath3-representations . on both schemes there is an action of @xmath101 and the quotient @xmath102 is isomorphic to @xmath103 ( this is because @xmath3 is an @xmath0-order , see @xcite ) . if @xmath8 is a toric variety we want the above story to happen in the toric world . this means that the main component of @xmath100 is a toric variety and the group @xmath104 is a torus . this is the same as asking that all @xmath96 are graded by @xmath23 ( where @xmath105 $ ] ) and have rank @xmath106 . an nccr or mma is a _ toric nccr _ ( _ toric mma _ ) if it is of the form @xmath20 where @xmath2 is a direct sum of @xmath23-graded rank 1 @xmath0-modules . the main questions we are interested in is the following : * * given a toric singularity can we find all toric nccrs / mmas ? * if the singularity is 3-dimensional and gorenstein can we find the dimer models corresponding to these nccrs ? if @xmath107 is a direct sum of reflexive graded rank one modules and @xmath21 is any reflexive graded rank one module then @xmath108 ( because @xmath21 is invertible ) , so by tensoring with @xmath109 we can make sure @xmath0 itself is one of the summands of @xmath2 . now @xmath110 and @xmath111 so if @xmath7 is an mma then it is cohen - macaulay so all direct summands @xmath96 are also cohen - macaulay . from @xcite we have the following theorem [ finite ] for a given toric singularity there are only a finite number of graded rank 1 cohen - macaulay modules up to isomorphism . for any toric singularity there are only a finite number of toric mmas and nccrs . this finiteness result implies that the problem of classifying all toric nccrs is feasible ( at least in principle ) and the first step will be to obtain a list of all graded rank 1 cohen - macaulay modules . in this section we discuss a method to determine the graded rank 1 cohen - macaulay modules . the main ideas go back to stanley , bruns and herzog in @xcite and @xcite . the method has also been described by perling in more detail in @xcite , @xcite . let @xmath112 a set of primitive vectors in @xmath27 that are the extremal rays of an @xmath17-dimensional cone @xmath42 . for every @xmath113 we fix an ordered basis @xmath114 for the space @xmath115 consisting of elements in @xmath116 . these bases can be used to define an incidence function @xmath117 let @xmath118 be any subset of @xmath119 and define @xmath120 to be the subset of @xmath118 containing all the faces of codimension @xmath121 ( @xmath122 ) and @xmath123 the free abelian group generated by this set . furthermore let @xmath124 it is easy to check that this turns @xmath125 into a complex and the homology of this complex is called the _ singular homology of @xmath126_. let @xmath42 be a cone . the singular homology of the fan @xmath119 is zero : @xmath127 . this is an easy consequence of the fact that a cone is contractible . let @xmath128 $ ] be the ring defined by the cone @xmath42 , for any object @xmath129 in the fan @xmath119 we put @xmath130 = r[u_1^{-1},\dots , u_k^{-1}]\text { where $ b_\varsigma=\{u_1,\dots , u_k\}$}\ ] ] we can also define a complex of @xmath0-modules @xmath131 with @xmath132 for any @xmath133 we set @xmath134 with @xmath135 where @xmath136 . this will give us again a complex and its homology is called the _ local homology of @xmath0_. if we tensor this complex with an @xmath0-module @xmath137 and take the homology we get the _ local homology of @xmath137_. @xcite[corollary 6.2.6 ] an @xmath23-graded @xmath0-module @xmath137 is cohen - macaulay if and only if the @xmath138 local homology of @xmath137 is trivial for @xmath139 , where @xmath140 is the dimension of @xmath0 . we are now going to apply this to graded rank @xmath106 reflexive submodules of @xmath0 . starting from a cone @xmath141}$ ] , we define for each @xmath142-vector of integers @xmath143 @xmath144 the first can be considered as a semigroup module over the semigroup @xmath145 , while the second is an algebra module over the algebra @xmath146 . for different @xmath147 s the @xmath148 s might be isomorphic as @xmath0-modules . indeed it is well known @xcite that * * an @xmath23-graded @xmath0-submodule of @xmath149 $ ] is reflexive if and only if it is of the form @xmath148 for some @xmath150 . * if @xmath151 then @xmath152 as @xmath0-modules if and only if there is an @xmath153 such that @xmath154 . because all cohen - macaulay modules are reflexive , the problem of determining graded rank @xmath106 cohen - macaulay modules reduces to checking which @xmath148 are cohen - macaulay or in other words when @xmath155 has trivial homology . if @xmath156 we can look at the degree @xmath62 part of the complex @xmath157 . for any module @xmath148 we have that @xmath158 where @xmath159 to check whether a given @xmath148 is cohen - macaulay we must check that @xmath160 or equivalently that @xmath161 for all the @xmath162 defined above . for a given @xmath156 and a @xmath163 we denote the signature of @xmath62 for @xmath64 @xmath164 if @xmath165 ( here positive includes zero ) and @xmath166 otherwise . given @xmath167 we define for every given signature @xmath168 the _ cell _ @xmath169 as @xmath170 the _ homology of the cell _ is defined as the homology of @xmath171 where @xmath172 contains only the faces in @xmath48 spanned by vectors with positive sign . @xmath173 : v_i\in \varsigma \implies s_i=+\}\ ] ] with this definition @xmath163 is cohen - macaulay if and only if the cells with nontrivial homology in degrees @xmath139 contain no lattice points of @xmath23 . in general checking the cohen - macaulay property reduces to checking whether integral solutions to a large set of inequalities exist ( see also @xcite[theorem 7.2 ] ) . this can become quite complex . in dimension 3 things can be done to simplify the problem a lot . let @xmath46 be a toric singularity with cone @xmath174}\subset \rl^3 $ ] . we can choose a plane that intersects the cone transversally in a polygon @xmath175 . without loss of generality we can choose the plane given by @xmath176 ( i.e. the third coordinate is 1 ) . if the singularity is gorenstein we can assume all vectors @xmath177 lie in this plane . in general the vertices of the polygon @xmath175 are of the form @xmath178 where @xmath179 and ( all @xmath180 in the gorenstein case ) . the standard orientation of @xmath176-plane will give us cyclic ( clockwise ) order to the vectors @xmath64 , which we will identify with the index @xmath181 . using this cyclic order , it is easy to see that @xmath182}\ } \cup \ { { [ v_i ] } | i \in { \mathbb{z}}/k{\mathbb{z}}\ } \cup \cup \ { { [ v_i , v_{i+1 } ] } | i \in { \mathbb{z}}/k{\mathbb{z}}\}\ ] ] this has an interesting consequence : [ segment ] if @xmath168 is a signature then @xmath172 has trivial singular homology if and only if the @xmath121 with positive signature @xmath183 consist of a sequence of numbers mod @xmath142 . @xmath184 it is clear that all @xmath185 of this form have trivial homology in degrees @xmath186 . if @xmath185 is not of this form @xmath172 consists of different connected components so the zeroth homology will be nontrivial . [ no - bounded - cells ] if the singularity is 3-dimensional a nonempty cell has nontrivial homology if and only if it is bounded . note that because no @xmath66 @xmath64 are sitting are linearly dependent , every @xmath187 planes @xmath188 will either go through a common point or bound a tetrahedron . suppose a cell @xmath169 has nontrivial homology then the sequence of @xmath64 with positive signature is not connected , so we can find @xmath189 with alternation signs @xmath190 . then @xmath169 is contained in @xmath191 we will now show that if this set is nonempty it is a solid tetrahedron or a point . so let @xmath192 be in this set , if @xmath193 is not a tetrahedron or a point the there would be a direction @xmath194 such that @xmath195 for all @xmath196 . i.e. @xmath197 for all @xmath198 . however because the cone spanned by @xmath199 and @xmath200 intersects the cone by @xmath201 and @xmath202 , there are positive @xmath203 such that @xmath204 , taking the inner product with @xmath194 gives a contradiction . on the other hand if a cell @xmath169 has trivial homology , the cone @xmath42 contains two cones : the one generated by the @xmath64 with positive signature , the one generated by the ones with negative signature . because of lemma [ segment ] these 2 cones only intersect in the zero and therefore we can find a plane through the origin in @xmath205 such that the two cones lie on different sides of the plane . the normal to this plane in the direction of the positive cone will give us an element @xmath206 such that @xmath207 for @xmath64 with positive signature and @xmath208 for @xmath64 with negative signature . therefore if @xmath209 then @xmath210 for all @xmath211 so @xmath169 is not bounded . the previous theorem can be used to show that the cohen - macaulay property remains true if we remove one of the @xmath64 from the cone of the singularity . [ remove - one ] if @xmath163 is a cohen - macaulay module for the 3-dimensional singularity generated by @xmath30 , then @xmath212 is a cohen - macaulay module for the singularity generated by @xmath213 . any cell @xmath214 for the singularity generated by @xmath213 is the union of two cells @xmath169 ( one of which might be empty ) for the original singularity . if @xmath214 was bounded and contained lattice points , then both @xmath169 would be bounded and at least one of them will contain a lattice point . in the other direction we can show that if we add an extra ray to the singularity every cohen - macaulay module lifts to a discrete interval of cohen - macaulay modules . [ interval ] let @xmath177 the vectors in cyclic order for a given 3-dimensional singularity and @xmath215 . if @xmath212 is a cohen - macaulay module for the singularity generated by @xmath213 , then there are numbers @xmath216 such that the module @xmath217 is cohen - macaulay for the original singularity if and only if @xmath218 . for each of the @xmath62 in @xmath23 we can look at the signs of @xmath219 and @xmath220 . in order to be cohen - macaulay @xmath221 must have the either the same sign @xmath222 as @xmath219 or @xmath220 . this gives us a condition of the form @xmath223 if both have positive sign and @xmath224 if both have negative sign . the intersection of all these conditions is an interval . this interval must be bounded because by theorem [ finite ] the number of cohen - macaulay graded rank 1 modules is a finite set . it is important to note that theorem [ no - bounded - cells ] and its 2 corollaries only hold in dimension 3 . we will discuss this further in section [ higherdims ] . an _ embedded quiver _ consists of a manifold @xmath225 and a quiver @xmath4 such that @xmath226 and every @xmath227 is a continuous map @xmath228 \to \mm$ ] with @xmath229 and @xmath230 . in this way every path in the quiver corresponds to a path in @xmath225 . the homotopy algebra of an embedded quiver is the quotient of the path algebra by the ideal generated by the expressions @xmath231 where @xmath71 and @xmath75 are homotopic paths . universal cover _ @xmath232 of an embedded quiver @xmath4 is by definition the pullback of the embedded quiver under the universal covering map @xmath233 . in general if @xmath234 is a direct sum of nonisomorphic graded reflexive rank 1 modules of a toric ring @xmath0 then @xmath235 is isomorphic to an embedded quiver . this can be seen as follows : choose for each @xmath236 a point in @xmath237 and make sure @xmath238 for @xmath239 . if we shift @xmath240 to @xmath241 with @xmath156 we get a new graded reflexive rank 1 module , which is isomorphic to @xmath236 . we will assign to this module the point @xmath242 . every homogeneous homomorphism from @xmath243 corresponds to shifting @xmath244 by a vector @xmath156 such that it lies in @xmath245 . therefore it makes sense to identify a homomorphism @xmath246 with the path @xmath247 in @xmath248 from @xmath242 to @xmath249 . we will denote the vector from @xmath242 to @xmath249 by @xmath250 . because @xmath251 and @xmath242 correspond to isomorphic modules , we can take the quotient of @xmath248 by @xmath23 to obtain an @xmath17-dimensional torus . now let @xmath68 be the set of the images of the @xmath251 in this quotient . fix a set of homogeneous morphism that generate @xmath235 and we take for @xmath69 the images of the paths @xmath250 in @xmath252 . it is easy to check that @xmath235 is isomorphic to the homotopy algebra of this embedded quiver and its universal cover is the infinite quiver we started from . although the embedding of the quiver depends on the choice of the @xmath251 , different choices will give rise to homotopic embeddings ( i.e. there is a one parameter family of embedded quivers connecting them ) . the aim is to find embeddings that look nice , i.e. such that the arrows do not wind more around the torus than needed . one way to do this is the following construction , which appears in work by craw and quintero - velez @xcite . given vectors @xmath30 that generate a cone in @xmath27 we have a map from @xmath253 by mapping @xmath254 to @xmath255 and a dual map @xmath256 . now @xmath257 will give us an embedding of @xmath23 in @xmath27 and @xmath258 will be an isomorphism . the map @xmath246 allows us to associate to each graded reflexive submodule @xmath167 a point in @xmath27 and the inverse @xmath259 embeds @xmath27 inside @xmath248 , so we can assign to @xmath167 the point @xmath260 this assignment has the property that if you shift @xmath261 by @xmath156 , the corresponding point will also shift by @xmath62 , which is precisely what we want . after factoring out @xmath23 we get a map @xmath262 . it is clear that if all @xmath245 are mapped by to different points in the torus under this map , @xmath263 will be isomorphic to homotopy algebra of the embedded quiver we constructed . in dimension 3 this is indeed true because of the following lemma . [ only - one - on - zero ] if @xmath264 are such that @xmath265 then @xmath163 can not be cohen - macaulay unless all the @xmath266 are zero . if @xmath163 were cm , then the @xmath267 must form a segment , so the cone spanned by the @xmath64 with nonnegative @xmath266 and the cone spanned by the @xmath64 with negative @xmath266 only intersect in the top . so @xmath268 and @xmath269 can only be the same if they are both zero . [ alldifferent ] if @xmath270 is a 3-dim nccr , then every @xmath245 will be mapped by @xmath271 to a different point in the torus . if @xmath245 and @xmath272 we mapped to the same point then @xmath273 would be mapped to the zero , but we know the only cohen - macaulay sitting on the zero is the trivial one so this would mean that @xmath274 . the first thing we need is a procedure to check whether two reflexives @xmath148 and @xmath275 are isomorphic . if we fix a basis for @xmath23 , we can look at the corresponding half open unit cube in @xmath276 . this cube is a fundamental domain for the quotient @xmath277 , so for every @xmath148 there is a unique @xmath156 such that @xmath278 lies inside @xmath279 . the module @xmath280 will be the unique graded reflexive module isomorphic to @xmath148 that is mapped to a point inside the unit cube . therefore it make sense to define the function @xmath281 two vectors in @xmath282 will define isomorphic modules if and only if there image under the map @xmath283 is the same . the algorithm now consists of 3 steps * generate the set @xmath284 . * find the maximal modifying sets : these the are subsets @xmath285 such that @xmath286 and @xmath287 . * for every maximal subset @xmath288 we construct the quiver of @xmath289 . we will now discuss these 3 steps in a bit more detail . because the 3-dimensional situation is special in the sense that the cohen - macaulay property is maintained after removing one of the @xmath64 , we can use an inductive procedure to generate @xmath290 . we order the @xmath64 cyclicly . the singularity generated by the first @xmath66 vectors is a quotient singularity @xmath291 where @xmath292 is abelian . this is because the cone is simplicial . this implies that every reflexive is cohen - macaulay , so @xmath293 . given @xmath294 we can construct @xmath295 as follows . for each @xmath296 we know from lemma [ interval ] that we there is an interval @xmath297 $ ] such that the module @xmath298 is cohen - macaulay if and only if @xmath299 $ ] . to find this interval we start with a given @xmath300 check whether @xmath301 is cohen - macaulay or not . this check is done by making looking at all @xmath187-tuples of planes @xmath302 where one of @xmath303 and check whether the tetrahedron they bound does not contain any lattice points . ( the @xmath304 is just a tiny number @xmath305 to compensate the fact that the cells are half open . ) * if it is cm we put @xmath306 in @xmath295 and we check @xmath307 until we get a vector that does not give a cohen - macaulay . then we do the same with @xmath308 and by lemma [ interval ] we can be sure all cohen - macaulays are found once we get to a vector that is not cohen - macaulay . * if @xmath301 is not cohen - macaulay , there were lattice points in some of the tetrahedra . chose such a lattice point @xmath62 and change @xmath194 to @xmath309 if @xmath310 or @xmath311 @xmath312 ( to remove @xmath62 from the tetrahedron ) . now check whether @xmath313 is cohen - macaulay . * * if it is cm proceed according to paragraph a. * * if it is not cm we take again a lattice point in a tetrahedron an use it to modify @xmath314 to @xmath315 as above , and continue like this until we hit a cohen - macaulay . + if the sequence @xmath316 is not monotone we stop . this means we can not make a cohen - macaulay of the form @xmath301 . note that the sequence can not continue monotonely indefinitely . indeed , we start with a finite number of lattice points in the tetrahedra and changing @xmath194 will remove some lattice points from the tetrahedra and maybe add some new ones , but in order to remove the new ones we have to move @xmath194 in the opposite direction . choose any indexing @xmath317 of the elements in @xmath290 such that @xmath318 . we are going to construct generations of pairs of subsets of @xmath290 , @xmath319 satisfying the following conditions : @xmath320 the first generation only contains the pair @xmath321 . given a generation we construct the next generation by constructing for every pair @xmath319 in this generation and for every @xmath322 with @xmath323 bigger than all indices in @xmath288 , a new pair @xmath324 as @xmath288 increases and @xmath2 decreases , there will be a last nonempty generation . the algorithm returns the set @xmath325 containing all first entries of the pairs of this last generation . following section [ embedded ] , we construct an embedded quiver for every @xmath326 . as vertices it will have the set @xmath327 ( which by corollary [ alldifferent ] are all different ) and the arrows @xmath69 come from a minimal set of graded algebra generators for @xmath328 . if the arrow @xmath329 corresponds to a graded homomorphism @xmath330 we define @xmath331 . using this notation we identify @xmath329 with the map @xmath332 \to m\otimes \rl / m : t \mapsto \kappa(b)+ t \vec a.\ ] ] many maximal modifying sets @xmath288 will give isomorphic homotopy algebras . in order to remedy this , we introduce the notion of affine equivalence . an affine transformation @xmath333 of @xmath248 ( considered as an affine space ) is compatible with the quotient @xmath334 if it maps fibers to fibers . this allows us to see it also as a map @xmath335 . we will call two embedded quivers @xmath4 and @xmath336 in @xmath334 affine equivalent if there is an affine transformation @xmath333 compatible with the quotient such that @xmath337 for each @xmath288 the algorithm will now construct the quiver and check whether it is affine equivalent to one of the quivers already constructed . if not this quiver will be added to the list @xmath338 . the complexity of checking affine equivalence of 2 embedded quivers is proportional to the number of arrows in the quiver because an affine transformation is fixed once we know the image of one arrows . [ effective ] given the toric data @xmath339 for a toric 3-dimensional gorenstein ring @xmath0 , the algorithm stops and gives 3 lists @xmath290 , @xmath325 and @xmath338 . 1 . the list @xmath290 contains vectors @xmath150 corresponding to all graded rank-1 cohen - macaulay modules @xmath163 up to isomorphism . the list @xmath325 will correspond to all maximal modifying modules that decompose as a direct sum of graded rank-1 cohen - macaulays , one of which is @xmath0 itself . the list @xmath338 contains all embedded quivers ( inside the three - dimensional torus ) for the toric nccrs up to affine equivalence . 1 . 2 . lemmas [ remove - one ] and [ interval ] and the monotony requirement for the sequence @xmath340 , ensure that step 1 stops and @xmath290 contains all graded rank-1 cohen - macaulay modules . steps 2 and 3 also clearly stop . 3 . if @xmath0 is gorenstein , then by theorem [ atleastonedimer ] we know that there is at least one toric nccr , which by theorem [ mmaisnccr ] is given by a maximal modifying module ( consisting of graded rank-1 cohen - macaulays because it is toric ) . this maximal modifying module will be contained in @xmath325 . by construction the @xmath325 contains sets of the same size , these must all be maximal modifying : if they were not , they were contained in a maximal modifying with more summands . the corresponding nccr would then have a higher rank grothendieck group than the one toric nccr we already had . this is impossible because by @xcite all nccrs are derived equivalent . this is by construction . we can also recover the dimer models from the embedded quivers in the following way : given any embedded quiver @xmath4 in the list @xmath338 , one can project it down to the twodimensional torus by forgetting the third coordinate . after this projection the arrows will cut the 2-torus in polygons which are bounded by cycles . these turn @xmath4 into a dimer model . as we already mentioned these dimer models are all consistent so this means there must exist consistent @xmath89-charges for these dimer models . [ charges ] choose a basis for @xmath23 such that the gorenstein vector is @xmath341 . let @xmath4 be any embedded quiver in the list @xmath338 . for any @xmath342 in the interior of @xmath35 we can construct a consistent @xmath89-charge for @xmath4 by putting @xmath343 where the inner product on @xmath248 comes from the basis and @xmath344 is as defined in [ consq ] . the proof of this theorem is postponed to the end of this section because it uses the techniques of perfect matchings and zigzag paths . a _ perfect matching _ is a set of arrows @xmath345 meeting every cycle in @xmath346 in precisely one arrow . it gives us a degree function on @xmath347 by giving the arrows in the matching degree @xmath106 and the others degree @xmath78 . we denote the degree of an element @xmath348 under @xmath349 by @xmath350 . @xmath349 gives a degree function on @xmath347 so by restriction also on @xmath351 . this degree function on @xmath352 $ ] comes from an element in @xmath27 which we denote by @xmath353 . note that different perfect matchings can have the same vector in @xmath27 . the convex hull of all these vectors @xmath353 forms a polygon in @xmath27 . we call a perfect matching extremal if its vector lies on a corner of this polygon . @xcite let @xmath85 be a consistent dimer model and @xmath347 its jacobi algebra . there is a one to one correspondence between the corners of the polygon that defines @xmath354 and the extremal perfect matchings . more precisely @xmath355}.\ ] ] the proof of the statement is done for geometrically consistent dimer models , but using section 8 of @xcite it works for all consistent dimer models . we order the extremal prefect matchings cyclicly according to the polygon @xmath356 and @xmath64 is the coordinate @xmath357 of the ith vertex of the polygon . let @xmath85 be a consistent dimer model and @xmath347 its jacobi algebra . fix a vertex @xmath358 . @xmath359 with @xmath360 where @xmath361 is any path from @xmath97 to @xmath362 . because @xmath347 is a toric nccr we know it is of the form @xmath363 for some @xmath364 . it is clear that @xmath365 . suppose we put @xmath366 . given any path @xmath361 from @xmath97 to @xmath362 , we can embed @xmath367 in @xmath368 by multiplication with @xmath361 . this shows that @xmath369 . to show that this is an equality we need to prove that for every extremal perfect matching @xmath370 there is a path in @xmath371 with @xmath349-degree zero . the proof of this follows from an adaptation of the proof of proposition 6.2 in @xcite and theorem 8.7 in @xcite which states that given a homology class of paths from @xmath372 ( on the @xmath373-torus ) we can find an extremal perfect matching @xmath370 and a path with that homotopy class @xmath374 such that @xmath375 . now from the construction in these proofs it is clear that every extremal perfect matching @xmath370 will occur if one varies the homology class . this shows that @xmath376 . [ arrows ] let @xmath85 be a consistent dimer model then every arrow @xmath329 is contained in at least @xmath106 extremal perfect matching and at most @xmath377 . moreover @xmath378 because @xmath329 is a path from @xmath379 to @xmath380 we know that if @xmath381 then @xmath382 . therefore @xmath383 . because @xmath384 , @xmath329 is contained in at least one extremal perfect matching . every arrow @xmath329 , is contained in a cycle @xmath385 in @xmath346 of length at least @xmath66 , because @xmath386 for all @xmath121 and the two other arrows in @xmath385 are each contained in at least @xmath106 extremal perfect matching , @xmath329 is contained in at most @xmath377 extremal perfect matchings . [ conspm ] let @xmath85 be a consistent dimer model . for every extremal perfect matching @xmath370 and every vertex we have the following property @xmath387 see the proof of theorem 8.7 in @xcite . we can rewrite the equation in the previous lemma as @xmath388 if we multiply this with @xmath64 , take the sum over @xmath121 and apply @xmath259 we get @xmath389 the special choice of basis gives @xmath390 if we now take the inner product of this with @xmath342 and rescale by @xmath391 , we get the consistency condition for the @xmath89-charge . note that @xmath342 needs to sit in @xmath35 for the charges to be positive . if a convex integral polygon in @xmath392 has exactly one internal lattice point it is called a reflexive polygon . up to an integral affine transformations , there are precisely @xmath393 reflexive polygons . we will choose the internal lattice point to be @xmath394 and let @xmath395 be the lattice points on the boundary of the polygon in cyclic order . the fan @xmath396}\,\dots , { [ v_k ] } , { [ v_1,v_2 ] } , \dots , { [ v_k , v_1 ] } \}\ ] ] in @xmath397 will define a projective smooth toric variety . this surface is a fano surface if all @xmath64 are corners of the polygon and a weak fano surface otherwise . to the fan we can associate a sequence of numbers @xmath398 and this sequence ( up to cyclic shifts and inversion of the order ) determines the isomorphism class of the fano variety . every divisor @xmath64 defines a line bundle @xmath399 and these generate the picard group . on the picard group we have the intersection form : @xmath400 in @xcite hille and perling studied full cyclic strongly exceptional sequences of line bundles associated to weak fano surfaces . full strongly exceptional sequences are infinite sequences of line bundles @xmath401 such that * @xmath402 if @xmath403 and @xmath404 , * @xmath405 if @xmath406 , * @xmath407 . here @xmath408 is the canonical bundle and @xmath142 is the rank of the grothendieck group ( which in the toric case equals the number of on dimensional cones in the fan ) . given a cyclic full strongly exceptional sequence @xmath409 on a toric surface , the sequence of numbers @xmath410 corresponds to the sequence a new reflexive polygon . we will call this sequence the type of the exceptional collection . note that this type can be different for different exceptional collections on the same fano . hille and perling also construct a table of types that can occur for each fano surface . this problem is closely related to our problem . starting from a fano variety we can also make a 3-dimensional gorenstein singularity by embedding the polygon in @xmath411 in the plane @xmath176 . we denote its cone by @xmath412}$ ] where @xmath413 . the vector @xmath414 gives rise to a grading on @xmath149 $ ] and hence also on @xmath415 and every reflexive module @xmath416 . for this grading @xmath417 gives us the fano variety and every graded module @xmath416 corresponds to a line bundle @xmath418 over @xmath417 . two isomorphic modules @xmath416 , @xmath419 can give the non - isomorphic line bundles if their grading is different . more specifically we have that @xmath420 let @xmath421 be a maximal modifying module for @xmath415 . without loss of generality we can assume that the @xmath266 are normalized ( @xmath422 ) and ordered by increasing @xmath423 . we now define a sequence of line bundles @xmath424 : @xmath425 this sequence is a cyclic full strongly exceptional sequence of line bundles . we can cover the fano variety by its basic affine open sets coming from the fan . this cover can be used to calculate the ech complex @xmath426 of a line bundle @xmath418 . the @xmath138 component of this complex can easily be seen to be the @xmath341-degree zero part of the @xmath427 component of @xmath428 . the fact that @xmath429 is cm for all @xmath430 gives us that @xmath431 if @xmath432 . now @xmath433_0 $ ] which is zero if @xmath434 because all homomorphisms have positive @xmath89-charge . finally because of poincar duality @xmath435 and this is zero if @xmath436 or @xmath437 . in the weak fano case a similar result holds but one needs to tweak the method to extract @xmath438 from @xmath439 : because @xmath440 is now singular and one needs to use its minimal resolution to pull back to the weak fano . this can be done using a moduli construction . for more info see @xcite and @xcite . using our algorithm we can determine all maximal modification modules of @xmath46 . for each maximal modification module we choose a compatible index and let @xmath441 be the vertex in corresponding the nccr according to @xmath236 . from @xcite we get @xmath442 the corresponding dimer models and their types can be found in the appendix and on www.algebra.ua.ac.be/dimers . an important thing that has been noticed is that it is possible to turn a consistent dimer model @xmath85 into another consistent dimer model @xmath443 such that @xmath444 and @xmath445 . this process is called mutation . it orginates from cluster algebras @xcite and is applied to algebras with a superpotential in @xcite . for dimer models on a torus the procedure restricts to the following construction . let @xmath97 be a vertex without loops or cycles of length @xmath373 and with exactly @xmath373 arrows @xmath446 leaving and two arrows @xmath447 arriving . the toricly mutated dimer @xmath443 has the same vertices as @xmath85 and to obtain a list of all arrows and faces for @xmath448 we apply the following procedures . * replace the arrows @xmath449 by @xmath450 in the opposite direction . * for each the path @xmath451 add an arrow @xmath452 with the same head and tail . * if @xmath453 is a face in @xmath454 change it to @xmath455 and add a cycle @xmath456 to @xmath457 . * if there are faces of length 2 , remove them and stick the 2 faced that bound them together . @xmath458 another way to describe this toric mutation procedure , which relates to the viewpoint of the dimer as an mma can be found in @xcite . if @xmath459 where the @xmath96 are the cohen - macaulays , a mutation of @xmath97 will correspond to construction a new algebra @xmath460 where @xmath461 is the kernel of the map @xmath462 if @xmath463 and @xmath464 then the kernel can be identified with the overlap of the two modules as submodules of @xmath96 , so @xmath465 . this second procedure gives the same result for the dimer model as the first , note however that the new vertex @xmath466 is not necessarily embedded on exactly the same spot as the original one . we have the following fact : the mutation of an mma is again an mma for the same singularity . from this and theorems [ mmaisnccr ] and [ consistency ] we can deduce that the procedure of toric mutation turns consistent dimer models into consistent dimer models . the analogy of this with the commutative case is remarkable . a toric crepant resolution of a 3-dimensional toric gorenstein singularity corresponds to a subdivision of its polygon in elementary triangles . given one such subdivision one can construct a new one by looking at a quadrangle consisting of 2 two elementary triangles and switching the diagonal so see this quadrangle as a union of 2 different triangles ( for a picture see section [ flop ] ) . this is called a toric flop @xcite and every two toric crepant resolutions can be transformed into each other by a sequence of flops . in the noncommutative situation we have the notion of a dimer model and again there is a procedure that can transform one dimer model into another : the toric mutation . the question now is very similar [ mutation ] can any 2 dimer models for the same singularity be transformed into each other by a sequence of toric mutations ? using our algorithm we can already answer this question affirmatively for the singularities from reflexive polygons . @xmath373 dimer models for a singularity coming from a reflexive polygon can transformed into each other by a sequence of toric mutations . the algorithm gives us a finite list of dimer models for which we can check the statement manually . in the two following subsections we are going to explore this question for another very special type of singularities : quotients of the conifold singularity . in this situation we will also be able to prove a positive result . suppose @xmath467 with @xmath174}\in n$ ] . we will identify @xmath27 with @xmath24 as column vectors and @xmath23 as row vectors . let @xmath21 be a matrix with integer coefficients which has a nonzero determinant . if all @xmath468 are primitive vectors , then we can define a new singularity with a cone @xmath469}$ ] . this new singularity is isomorphic to a quotient singularity of @xmath470 . let @xmath471 be the sub - lattice of @xmath23 of all @xmath472 for which @xmath153 . the group @xmath473 has an action on @xmath105 $ ] by setting @xmath474 . from the construction it is clear that the ring of invariant functions @xmath475 is isomorphic to the ring @xmath476 : @xmath477 choose representatives in @xmath478 for the elements in the quotient group @xmath479 . if @xmath416 is a graded reflexive @xmath0-submodule of @xmath149 $ ] , then it decomposes into graded reflexive @xmath475-submodules @xmath480 where @xmath481 is the @xmath476-module @xmath482 . [ cover ] let @xmath2 be a direct sum of graded rank one reflexive modules . * @xmath483 * if @xmath235 is an nccr for @xmath0 then @xmath484 is an nccr for @xmath475 . * if @xmath485 is the universal cover of the embedded quiver for @xmath235 , then @xmath486 will be the universal cover of the embedded quiver for @xmath487 . all @xmath416 are monomial submodules of @xmath149 $ ] have an action of @xmath488 and therefore @xmath235 also has a @xmath488-action and we can make the skew group ring @xmath489 . the orthogonal idempotents of @xmath489 are products of orthogonal idempotents in @xmath235 and orthogonal idempotents in @xmath490 . the former are parametrized by the direct summands of @xmath491 while the latter are parametrized by the @xmath492 : @xmath493 we have that @xmath494 if @xmath495 , so if we calculate the direct summands of @xmath489 we get @xmath496 which shows that @xmath497 . @xmath489 is homologically homogeneous if and only if @xmath235 is because the global dimension of @xmath489 and @xmath235 is the same ( see @xcite ) . the third statement follows from the easy to check fact that the map @xmath498 for the new singularity is the old @xmath498 multiplied with @xmath499 . here we will briefly examine the case where the polygon is a parallelogram . @xmath500}\ ] ] in this case the singularity can be seen as a quotient of the conifold singularity ( for which the polygon is the unit square @xmath501}$ ] ) . the conifold has a unique toric nccr which corresponds to a torus tiled by @xmath373 squares . @xmath502 by lemma [ cover ] this quotient singularity then has an nccr which corresponds to tiling the torus by @xmath503 squares , but in general this is not the only nccr . e.g. for @xmath504}$ ] we get 5 nccrs : + these dimers all look like the first one with some additional _ diagonal arrows _ that go around in curves , ignoring orientation issues . one can solve the orientation issue by reversing the directions of the diagonal arrows pointing in the `` north - south '' directions while keeping the orientation of the diagonal arrows in the `` east - west '' directions . if one looks at the homology of these curves with coefficients in @xmath505 , one sees that the total homology is zero . this is not a coincidence . consider @xmath46 with @xmath506}$ ] . * for each dimer model of this singularity , construct an unoriented graph by forgetting the orientations of the arrows . this graph consists of the tiling of the torus by @xmath503 squares together with some additional non - intersecting diagonals . * divide the diagonal arrows in two classes @xmath507 such that the directions of the diagonals in each class are parallel . then the homology class of @xmath508 on the torus is zero . * all dimer models of @xmath0 can be torically mutated to the dimer model which tiles the torus by @xmath503 squares ( i.e. the one without diagonals ) . one can compute that @xmath509 in the fundamental domain of the torus @xmath510 there are only @xmath503 points with the property @xmath511 , so for a dimer model , every such points must have a graded rank one cohen - macaulay sitting on it . there are @xmath187 classes of graded rank one reflexives according to the equivalence class of @xmath512 . note that the class to which a graded rank one reflexives belongs depends on the @xmath513 of its point in @xmath510 because @xmath514 . the directions of the arrows in the nccr will correspond to @xmath515 where by corollary [ arrows ] all @xmath516 except for one or two consecutive @xmath266 that are @xmath106 . so in total there are eight possible directions : the straight ones ( only one @xmath517 ) and the diagonal ones ( 2 @xmath517 ) . the diagonal ones we divide in 2 classes : @xmath518 ( @xmath519 or @xmath520 ) and @xmath521 ( @xmath522 or @xmath523 ) . note that after projection onto @xmath397 all arrows in each class are ( anti)parallel . the consistent @xmath89-charge @xmath524 assigns degree @xmath67 to the straight arrows and degree @xmath106 to the diagonal ones . the consistency condition @xmath525 implies that every vertex has precisely 2 straight arrows arriving and 2 straight arrows leaving , so ignoring orientation , the straight arrows form a square grid that tiles the torus . the number of diagonal arrows in a vertex is either @xmath373 or @xmath187 . now consider a diagonal arrow in @xmath518 with @xmath526 and look at the next diagonal arrow @xmath147 arriving in or leaving from @xmath97 , in clockwise direction from @xmath329 . if this arrow is in @xmath518 then @xmath526 because there are @xmath373 straight arrows between @xmath329 and @xmath147 . if this arrow is in @xmath521 then @xmath527 . this implies that @xmath528 is a sum of cycles . next we show that these cycles are boundaries . consider @xmath187 points on the torus that form a grid square made of straight arrows . if the dimer model has a diagonal in that square then the @xmath373 vertices of the square that are not on the diagonal have reflexives with different @xmath194-coordinates . indeed there is a path connecting these @xmath373 vertices consisting of @xmath373 straight arrows in the same direction and a diagonal arrow . the @xmath194-coordinate of such a path is in @xmath529 . if there is no diagonal in the square or there is a diagonal connecting the two vertices then there is a path between them consisting of 2 straight arrows . the @xmath194-coordinate of such a path is in @xmath505 . from this discussion we can conclude the following : * the diagonal arrows connecting cms with @xmath530 form the boundary between the region containing all cms with @xmath531 and the region containing all cms with @xmath532 . * the diagonals going between cms with @xmath533 form the boundary between the region containing all cms with @xmath534 and the region containing all cms with @xmath535 . because they are boundaries the total homology with coefficients in @xmath505 is zero . if one mutates a vertex with coordinates @xmath536 then the first two new coordinates @xmath537 must be the same because the other possibilities are occupied . the third coordinate must change by @xmath67 because if @xmath538 is the direction of a straight arrow and @xmath539 is a straight arrow in the opposite direction then @xmath540 . so mutation changes the equivalence class of the vertex . now look at the torus @xmath541 of the dimer model and cut it in pieces along the diagonal arrows . if there are diagonal arrows there must be at least 2 pieces because they form boundaries . if there is a piece that is simply connected then we can perform a mutation on an internal vertex contained in a triangle ( so it borders the boundary of the simply connected piece ) . after mutation this vertex lies outside the simply connected piece , so gradually we can shrink the piece away . if there is no simply connected piece , all pieces have the topology of a cylinder and are each bounded by two curves with either the same homology class or the opposite homology class . there must be at least one cylinder piece with opposite homology classes because the total homology of the boundaries is zero . using the same procedure as above we can shrink this piece away . eventually all pieces except one have shrunk away and we are left with the dimer model consisting only of squares . we have seen that the algorithm we proposed worked for the three dimensional gorenstein case . we are now going to discuss briefly what happens in other cases . the concepts of nccr and mma were originally thought of mainly in the context of gorenstein singularities . things work differently in the non - gorenstein world . already in the 2-dimensional situation , there is no straightforward connection between minimal commutative resolutions and nccrs . if we look at a 2-dimensional abelian quotient singularity usually the nccr , which is equal to the skew group ring , has a lot more vertices than the minimal resolution has components and therefore the derived categories can not be equivalent ( see @xcite ) . in the 3-dimensional case section [ somefacts ] still holds for non - gorenstein singularities . so the list @xmath290 produced by the algorithm will contain all graded rank @xmath106 cohen - macaulays . step 2 will now give a list of all toric modifying modules with the highest possible number of summands . however we do not know whether these modifying modules are maximal ( it might be that higher rank summands are needed to maximize them ) . we also do not know in general whether the endomorphism rings of these modules have global dimension @xmath66 , but often this is indeed what happens . let s have a closer look at a small example : @xmath542}\ ] ] which is the simplest example from toric geometry where flips occur ( see @xcite ) . the algorithm produces one algebra : @xmath543 the quiver of this algebra has 5 arrows : @xmath66 from the first to the second vertex and @xmath373 back . one can check that the global dimension of this algebra is @xmath66 . this algebra is also connected to the dimer model of the conifold . let @xmath449 be the @xmath187 arrows of the conifold dimer with @xmath544 and @xmath545 . by comparing generators and relations one can show that the endomorphism ring @xmath546 is isomorphic to the homotopy algebra of the embedded quiver that has the same vertices as the conifold dimer but as arrows @xmath547 it is well - known that the conifold algebra can be seen as the non - commutative algebra that governs the flop @xcite . if one looks at the moduli space of @xmath548-stable @xmath549-dimensional representations @xcite of the conifold dimer then the change from @xmath550 to @xmath551 will correspond to a flop in the moduli space .

we provide further evidence for the nontrivial interplay between geometry and temperature in the casimir effect . we investigate the temperature dependence of the casimir force between an inclined semi - infinite plate above an infinite plate in dimensions using the worldline formalism . whereas the high - temperature behavior is always found to be linear in in accordance with dimensional - reduction arguments , different power - law behaviors at small temperatures emerge . unlike the case of infinite parallel plates , which shows the well - known behavior of the force , we find a behavior for inclined plates , and a behavior for the edge effect in the limit where the plates become parallel . the strongest temperature dependence occurs for the casimir torque of inclined plates . numerical as well as analytical worldline results are presented .

0906.2313

the casimir effect @xcite is not only a field witnessing rapid experimental as well as theoretical progress , it also continues to offer surprising new features . the casimir effect derives its fascination from the fact that it originates from quantum fluctuations of the radiation field or of the charge distribution on the mesoscopic or macroscopic test bodies . moreover , it inspires many branches of physics , ranging from mathematical to applied physics , see @xcite for reviews and @xcite for experimental verifications . a distinctive feature of casimir forces between test bodies is the dependence on the geometry , i.e. , the shape and orientation of these bodies . for a comparison between theory and a real casimir experiment , a number of properties such as finite conductivity , surface roughness and finite temperature have to be taken into account in addition . generically , these latter corrections do not factorize but take influence on one another . for instance , the interplay between dielectric material properties and finite temperature @xcite is still a subject of intense theoretical investigations and has created a long - standing controversy @xcite . also the role of electrostatic patch potentials has been suggested as a potentially problematic issue @xcite , which has become a matter of severe debate @xcite . the present article is not meant to resolve these controversies . on the contrary , our work intends to draw attention to another highly nontrivial interplay which on the one hand needs to be accounted for when comparing theory and a real experiment and on the other hand is another characteristic feature of the casimir effect : the interplay between geometry and temperature . as first conjectured by jaffe and scardicchio @xcite , the temperature dependence of the casimir effect can be qualitatively different for different geometries , as both the pure casimir effect as well as its thermal corrections arise from the underlying spectral properties of the fluctuations . first analytical as well as numerical evidence of this `` geothermal '' interplay in a perpendicular - plates configuration has been found in @xcite using the worldline formalism . the physical reason for this interplay can be understood in simple terms : for the classical parallel - plate case , the nontrivial part of the fluctuation spectrum is given by the modes orthogonal to the plates . this relevant part of the spectrum has a gap of wave number @xmath6 , where @xmath7 is the plate separation . for small temperatures @xmath8 , the higher - lying relevant modes can hardly be excited , such that their thermal contribution to the casimir force remains suppressed : the resulting force law for the parallel - plates case scales like @xmath9 . this argument for a suppression of thermal contributions applies to all geometries with a gap in the relevant part of the spectrum ( e.g. concentric cylinders or spheres , casimir pistons , etc . ) . these geometries are called _ closed_. this reason for a suppression of thermal contributions is clearly absent for _ open _ geometries with a relevant gapless part of the spectrum . for these geometries , relevant modes of the spectrum can always be excited at any small temperature value . therefore , a stronger thermal contribution @xmath10 with @xmath11 can be expected . as experimentally important configurations such as the sphere - plate or the cylinder - plate geometry belong to this class of open geometries , a potentially significant geothermal interplay may exist in the relevant parameter range @xmath12 . in the present work , we provide further evidence for the geometry - temperature interplay in the casimir effect . for simplicity , we study the casimir effect induced by a fluctuating real scalar field obeying dirichlet boundary conditions ( dirichlet scalar ) . as an illustrative example , we concentrate on an inclined - plates configuration ; here , a semi - infinite plate is located above an infinite one , with an angle of inclination of @xmath13 , see fig . [ t0-ip-1 ] . this configuration generalizes geometries which have first been proposed and studied in the context of casimir edge effects @xcite . our results do not only generalize the findings of @xcite which hold for @xmath14 . most importantly , we identify regimes with fractional temperature dependences for certain geometries . moreover , we work in @xmath15 dimensional spacetime , yielding many analytical as well as numerical results for the casimir force and energy as well as for the torque . plane by an angle @xmath16 . as special cases , @xmath17 corresponds to the configuration of one semi - infinite plate parallel to an infinite plate ( 1si configuration ) , whereas @xmath14 yields the perpendicular - plates configuration . ] a reliable study of geothermal casimir phenomena requires a method that is capable of dealing with very general casimir geometries . for this , we use the worldline approach to the casimir effect @xcite , which is based on a mapping of field - theoretic fluctuation averages onto quantum - mechanical path integrals @xcite . for arbitrary backgrounds , this worldline integral representing the spacetime trajectories of the quantum fluctuations can straightforwardly be computed by monte carlo methods @xcite . as the computational algorithm is generally independent of the background , i.e. , the casimir geometry in our case , casimir problems can straightforwardly be tackled with this method . high - precision computations for dirichlet - scalar fluctuations have been performed , e.g. , for the sphere - plate and cylinder - plate case @xcite . in the present work , we demonstrate that the worldline approach can also be used to obtain novel analytical results ( see also @xcite for an analytical worldline approximation technique ) . in order to overcome standard approximative tools based , e.g. , on the proximity - force theorem @xcite , a variety of new field - theoretical methods for casimir phenomena have been developed in recent years , ranging from improved approximation methods @xcite to exact methods mainly based on scattering theory @xcite or a functional integral approach @xcite . it will certainly be worthwhile to generalize these methods to finite temperature for a study of the geometry - temperature interplay . in the remainder of this introduction , we summarize our most important results specializing to @xmath18 dimensional spacetime . in sect . [ sec : worldl - appr - casim ] , we briefly review the worldline approach to the casimir effect . [ sec : case - t=0 ] is devoted to a study of the zero - temperature casimir effect for the geometries under consideration . the finite - temperature case is described in sect . [ sec : finite - temperature ] . our conclusions are summarized in sect . [ sec : conclusions ] . let us already summarize our most important results , specializing to @xmath19 spacetime dimensions and concentrating on the casimir interaction energy ; the corresponding force can straightforwardly be derived by differentiation . at zero temperature , the classical casimir energy of two parallel dirichlet plates at a distance @xmath7 reads @xmath20 where @xmath21 is the area of the plates . from now on , we use natural units , setting @xmath22 . the casimir energy of inclined plates ( i.p . ) can be parameterized as @xmath23 where the coefficient @xmath24 is shown in fig . [ t0-ip-7 ] as a function of @xmath16 . the extent of the inclined plate in @xmath25 direction along the edge is @xmath26 . at @xmath17 , the energy per edge length ( [ int - sum-2 ] ) diverges and has to be replaced by @xmath27 where @xmath28 is the casimir energy ( [ int - sum-1 ] ) with @xmath29 being the semi - infinite plate s area and @xmath30 the so - called edge energy . the numerical value of @xmath31 is about @xmath32 in agreement with @xcite . the casimir torque is obtained from eq . ( [ int - sum-2 ] ) by @xmath33 . for @xmath19 , the torque @xmath34 as a function of @xmath16 is shown in fig . [ ctorque-1 ] . at @xmath17 , the casimir torque per unit length diverges as well but can be converted into finite torque per unit area . remarkably , for @xmath17 the standard torque obtained from the casimir energy of parallel plates ( [ int - sum-1 ] ) @xmath35 is reduced by a repulsive contribution @xmath36 arising from the edge effect . we encounter a similar subleading repulsive torque effect at finite temperature . thermal fluctuations modify the casimir energy , yielding the free energy @xmath37 where @xmath38 is the temperature correction . for @xmath39 , the correction @xmath40 to the well - known parallel - plates energy reads @xmath41 which is @xmath42 . note that only the @xmath43 term contributes to the force as the first term vanishes upon differentiation . for @xmath39 , our result for the thermal correction @xmath44 to the inclined - plates energy reads @xmath45 where @xmath46 is shown in fig . [ ym - integrals ] as a function of @xmath16 . the second term which is a purely analytical result is the generalization of a result for perpendicular plates , @xmath14 , found in @xcite ; numerically , this term evaluates to @xmath47 . again , eq . ( [ int - sum-6 ] ) denotes an energy per edge length and diverges as @xmath48 . it has to be replaced by the formula for the energy of a semi - infinite plate above a parallel one , @xmath49 . the thermal part of @xmath50 is as in ( [ int - sum-5 ] ) , where @xmath21 is the area of the semi - infinite plate . the leading thermal correction to the edge effect @xmath51 reads @xmath52 for @xmath53 , all thermal casimir energies increase linearly in @xmath1 due to dimensional reduction . for instance , the casimir energy @xmath54 for parallel plates becomes @xmath55 which is @xmath56 . note that @xmath57 is independent of @xmath58 as the dimensional analysis easily shows . the energy at large @xmath59 can therefore be interpreted as a classical effect . the same holds for the large @xmath59 behavior of the inclined - plates case as well as for semi - infinite plates . for inclined plates , we get @xmath60 where @xmath61 is shown in fig . [ ym - integrals ] as a function of @xmath16 . the edge effect reads at large @xmath59 @xmath62 in main part of this article , these results will be derived in detail in @xmath0 spacetime dimensions . let us briefly review the worldline approach to the casimir effect for a massless dirichlet scalar ; for details , see @xcite . consider a configuration @xmath63 consisting of two rigid objects with surfaces @xmath64 and @xmath65 . the worldline representation of the casimir interaction energy in @xmath15 dimensional spacetime reads @xmath66 \right\rangle.\end{aligned}\ ] ] here , the generalized step functional obeys @xmath67=1 $ ] if a worldline @xmath68 intersects both surfaces @xmath69 , and is zero otherwise . the expectation value in eq . is taken with respect to an ensemble of @xmath70-dimensional closed worldlines with a common center of mass @xmath71 and a gauian velocity distribution , @xmath72 here , we have already used the fact that the time component cancels out for static casimir configurations at zero temperature . ( [ int-1 ] ) has an intuitive interpretation : all worldlines intersecting both surfaces do not satisfy dirichlet boundary conditions on both surfaces . they are removed from the ensemble of allowed fluctuations by the @xmath73 functional and thus contribute to the negative casimir interaction energy . in the process of the auxiliary @xmath74 integration , the propertime parameter @xmath74 scales the extent of a worldline by a factor of @xmath75 . large @xmath74 correspond to long - wavelength or ir fluctuations , small @xmath74 to short - wavelength or uv fluctuations . introducing finite temperature @xmath76 by the matsubara formalism is equivalent to compactifying euclidean time on the interval @xmath77 $ ] . now , the closed worldlines live on a cylindrical surface and can carry a winding number . the worldlines @xmath78 winding @xmath79 times around the cylinder can be decomposed into a worldline @xmath80 with no winding number and a winding motion at constant speed , @xmath81 where the @xmath0th component corresponds to euclidean time . the casimir energy ( [ int-1 ] ) now becomes @xmath82 \right\rangle.\nonumber\end{aligned}\ ] ] the finite - temperature worldline formalism for static configurations thus boils down to a winding - number prefactor in front of the worldline expectation value together with a sum over winding numbers : @xmath83 the winding - number sum is directly related to the standard matsubara sum by a poisson resummation , @xmath84 this is already sufficient to understand the high - temperature limit of generic casimir configurations : at high temperatures @xmath85 , only the zeroth matsubara frequency survives as higher modes receive thermal masses of order @xmath86 and decouple . all remaining temperature dependence arises from the dimensional prefactor @xmath87 , and the dependence on the casimir geometry only enters the prefactor . the calculation of the latter is a dimensionally reduced problem in @xmath88 dimensions . this is a general mechanism of _ dimensional reduction _ in high - temperature field theories . the linear high - temperature asymptotics is also clear from the fact that the bose - einstein distribution governing the distribution of bosonic thermal fluctuations increases as @xmath89 in the high - temperature limit . finally , it is advantageous for numerical as well as analytical calculations to rescale the worldlines such that the velocity distribution becomes independent of @xmath74 , @xmath90 where the dot always denotes a derivative with respect to the argument , e.g. , @xmath91 . in terms of these normalized worldlines @xmath92 and the center - of - mass coordinate @xmath93 , the @xmath73 function reads more explicitly @xmath94\equiv\theta[\mathbf x_{\text{cm}}+\sqrt{{\mathcal{t}}}{\boldsymbol{\gamma}}(t)].\ ] ] the involved worldline integrals can be evaluated also numerically by monte carlo methods in a straightforward manner . for this , the path integral over an operator @xmath95 is approximated by a sum over a finite ensemble of @xmath96 loops , @xmath97 \rangle\ , \to\ , \frac{1}{{{n_{\text{l } } } } } \sum_{\ell=1}^{{{n_{\text{l } } } } } \mathcal o[{\boldsymbol{\gamma}}_\ell],\ ] ] where @xmath98 counts the worldlines in the ensemble . each worldline @xmath99 is furthermore discretized by a finite set of @xmath100 points per loop ( ppl ) , @xmath101 where @xmath102 are identified as the worldlines are closed . various efficient ab initio algorithms for generating discretized worldlines with gauian velocity distribution have been developed , see , e.g. , @xcite . with these comparatively simple prerequisites , we can now turn to an analysis of various non - trivial casimir configurations for the dirichlet scalar . let us first study parallel and inclined plates at zero temperature . the purpose of this section is on the one hand to review and generalize known results and on the other hand to exemplify how the casimir effect can be understood in terms of simple geometric properties of the worldlines . we start with casimir s classic configuration of two infinitely extended parallel plates . let the lower and upper plate lie in the @xmath103 and @xmath104 planes , respectively . in @xmath70 space dimensions the surface area @xmath21 of the plates is then @xmath105 dimensional . the @xmath73 functional for this configuration reads th moment of the maximum spatial extent @xmath106 of a worldline as a function of @xmath0 . this geometric object ( which is the same in any target dimension of the worldline ) is related to the casimir energy of the parallel - plates configuration in @xmath0 spacetime dimensions by eqs . and [ t0-pp-4 ] . the plot compares the exact analytical result with the worldline numerical computation based on 1000 worldlines with @xmath107 ppl ( points per loop ) each . ] @xmath108=&\,\theta\left(\sqrt{{\mathcal{t } } } \gamma_{{z_{\mathrm{max}}},\ell}+z_\mathrm{cm}\right ) \\ & \times\theta\left(-z_\mathrm{cm}-\sqrt{{\mathcal{t } } } \gamma_{{z_{\mathrm{min}}},\ell}-a\right),\notag\end{aligned}\ ] ] where @xmath109 is the @xmath110 coordinate of the @xmath98th worldline ( measured with respect to the center of mass ) . the quantities @xmath111 denote the worldline s maximal and minimal extent in the @xmath110 direction , respectively . the total maximal extent @xmath112 of the @xmath98th worldline then is @xmath113 . now , we can do the integral in eq . and obtain the casimir energy density ( suppressing the index @xmath98 from now on ) @xmath114 we observe that the @xmath0-dimensional parallel - plate casimir energy is related to the @xmath0th cumulant of the extent of the worldlines @xcite . this is a first example for a relation between casimir energies and geometric properties of the worldlines . instead of computing these cumulants directly , let us simply compare eq . with the well - known analytic result @xcite . @xmath115 yielding @xmath116 a comparison of the analytical result to a numerical evaluation of the cumulants is displayed in fig . [ t0-pp-5 ] . also , the casimir force density can straightforwardly be obtained as the derivative of eq . with respect to @xmath7 . incidentally , the connection between casimir energies and worldline properties also induces a relation between casimir energies and questions in polymer physics , as first observed in @xcite . the inclined - plates ( i.p . ) configuration consists of a perfectly thin semi - infinite plate above an infinite plate at an angle @xmath16 , see fig . [ t0-ip-1 ] . the semi - infinite plate has an edge with a ( @xmath117 dimensional ) length @xmath26 . the infinite plate has a ( @xmath105 ) dimensional area @xmath21 . and @xmath21 are considered in the limit @xmath118 . ] let @xmath7 be the minimal distance between the plates . in the following , we will omit the center - of - mass subscript cm . the @xmath73 functional for this configuration reads @xmath119 where the first @xmath120 function ensures the intersection of the worldline with the infinite plate . the remaining three ones account for the intersection with the semi - infinite plate . in eq . , we have used @xmath121 where @xmath122 parameterizes the worldline ; i.e. , in the discretized version , we have @xmath123 with @xmath100 being the number of points per worldline loop ( ppl ) . trivially , @xmath124 and @xmath125 holds . in other words , @xmath126 measures the minimal extent of the worldline in the @xmath127 direction of a coordinate system rotated by the angle @xmath16 . in eq . , we also encounter @xmath128 , denoting the @xmath127-dependent envelope of the worldline in positive @xmath110 direction . all these geometric properties of a worldline are displayed in fig . [ t0-ip - iloop ] . ) , is encoded in the function @xmath129 , which has to be integrated from @xmath130 to @xmath131 . ] the @xmath132 functional in eq . generalizes the case of perpendicular plates ( @xmath133 ) for @xmath14 and the case of one semi - infinite plate parallel to a infinite one ( 1si ) for @xmath134 ; both edge configurations were studied in detail in @xcite . let us define @xmath135 inserting @xmath132 for @xmath136 into eq . leads to the casimir energy density of the inclined plates of the inclined - plates ( i.p . ) configuration in @xmath19 versus the angle of inclination @xmath16 . for @xmath137 , this function can be evaluated analytically , yielding @xmath138 . we have used @xmath139 worldlines with @xmath140 ppl each . ] @xmath141 equation ( [ t0-ip-6 ] ) is shown as a function of @xmath16 in fig . [ t0-ip-7 ] for @xmath19 . for @xmath142 and @xmath19 , we rediscover the perpendicular plates result @xcite as a special case . incidentally , the integral in eq . can be done analytically for @xmath17 resulting in @xmath143 . together with the @xmath16-dependent prefactor , eq . diverges as @xmath48 as it should . this is because eq . corresponds to the energy per unit _ edge length _ , whereas for @xmath134 the casimir energy becomes proportional to the _ area _ of the semi - infinite plate . we devote the whole next section to analyzing how the limit @xmath144 yielding the 1si configuration can be obtained . it is instructive to study the limit of a semi - infinite plate parallel to an infinite plate ( 1si ) , @xmath144 , as it involves a subtle limiting process . recalling the general considerations of @xcite for the 1si case , the total casimir interaction energy decomposes into @xmath145 where @xmath146 is the usual casimir energy per unit area of two parallel plates eq . , with @xmath21 being now the area of the semi infinite plate . the so called edge energy @xmath30 measures the contribution that arises solely due to the presence of the edge . in the limit @xmath134 , this decomposition is naturally achieved by inserting @xmath147 of eq . into eq . and performing the @xmath110 integral first . this leads to @xmath148 this representation can straightforwardly be computed numerically @xcite . of course , for truly infinite plates , the edge effect being proportional to the length of the edge is completely negligible in comparison with @xmath28 , the latter being proportional to the area of the plates . however , dealing with finite plates , the edge effect contributes to the casimir force , effectively increasing the plate s area @xcite . of course , the same result has to arise from the general inclined - plates formula eq . in the limit @xmath134 . however , this representation naively exhibits a divergence in this limit . to find the origin of the divergence , we decompose eq . into the parts corresponding to the edge effect @xmath149 and the semi - infinite - plates energy @xmath150 , characterized by the integrals @xmath151 and @xmath152 , respectively . here @xmath153 denotes the ( infinite ) length of the semi - infinite plate in @xmath110 direction . the result for @xmath149 reads @xmath154 which becomes @xmath155 in eq . as @xmath48 ; eq . ( [ t0-ip - l-1 ] ) is therefore valid for @xmath17 . on the other hand , if we naively expand the result for @xmath150 for small @xmath16 , we obtain @xmath156 which is only valid for @xmath136 and does not reproduce eq . in the limit @xmath134 . instead of the energy per area , we have obtained the energy per length , which of course diverges in this limit . in order to rediscover the casimir energy for the 1si configuration , the limits @xmath144 and the implicit limit @xmath157 have to be taken in the right order . in eq . , the limit @xmath157 has implicitly been performed first , which precisely leads to the divergence of the energy per edge length . therefore , we need to first perform the limit @xmath134 at finite @xmath158 in order to obtain the desired energy per area . starting from @xmath150 at small @xmath16 , @xmath159 we do the @xmath160 integral _ first _ and obtain @xmath161 for small @xmath162 , i.e. , finite @xmath158 and @xmath144 , we can expand the last factor in @xmath16 , @xmath163 which for @xmath164 corresponds exactly to the parallel - plates contribution @xmath28 in eq . . from eq . , we also observe that the other order of limits , taking first @xmath165 while keeping @xmath16 finite , reproduces the divergent behavior of the energy per edge length in eq . ( as long as re@xmath166>2 $ ] ) . the proper order of limits is similarly important at finite temperature with the additional complication that another dimensionful parameter occurs . the casimir torque @xmath167 referring to rotations of one of the plates about the edge axis can easily be obtained by taking the derivative of the casimir energy ( [ t0-ip-6 ] ) , ( or eq . ( [ t0-ip - l-4 ] ) for small @xmath16 ) , with respect to the angle of inclination : @xmath168 for @xmath16 near @xmath169 , we can even set @xmath170 before taking the average with respect to the loop ensemble , simplifying the calculations . this is , because the derivative @xmath171 changes its sign for perpendicular plates @xmath14 if the worldline is rotated by an angle @xmath172 about the normal axis of the lower plate , see fig . [ t0-ip-1 ] . therefore , the sign correlates with the position of the minimum on the @xmath127 axis of the lower plate . but the position @xmath127 of the minimum @xmath173 does not correlate with the value of the integral in eq . ( [ t0-ip-6 ] ) leading to a mutual cancellation of terms involving @xmath171 . for @xmath174 , we have to rotate the worldline about the normal axis of the inclined lower plate . then , the correlation between the position of the minimum and the involved integrals does not vanish any more since the original and rotated worldline contribute differently to the integral . in general , expressions containing derivatives of @xmath173 can not be neglected even at @xmath14 . since the worldlines are not smooth , the convergence of averages of such expressions will be very slow . this is the case when calculating the coefficients of an expansion of eq . ( [ t0-ip - torque-1 ] ) near @xmath14 . since the second derivative already appears in the first expansion coefficient , more confident values are obtained by a numerical fit to eq . ( [ t0-ip - torque-1 ] ) . there , only the first derivative is present . for @xmath19 , we obtain ( see fig . [ ctorque-1 ] ) @xmath175 this should be compared to the worldline average based on the expansion of eq . ( [ t0-ip - torque-1 ] ) around @xmath169 : the linear coefficient in eq . ( [ t0-ip - torque-2 ] ) then yields 0.003 @xmath176 0.0002 . if we neglect all derivatives of @xmath173 the worldline result reads 0.00285 @xmath176 0.00003 . in all three cases @xmath177 worldlines with @xmath140 ppl were used . for @xmath48 , the casimir torque diverges . the expansion about @xmath17 can easily be obtained analytically from ( [ t0-ip - l-4 ] ) @xmath178 where we have used eq . ( [ t0-pp-4 ] ) . for @xmath19 , eq . ( [ t0-ip - torque-3 ] ) yields @xmath179 , being excellent approximation to eq . ( [ t0-ip - torque-1 ] ) for @xmath16 not too close to @xmath169 . the divergent casimir torque per length can be converted into finite torque per unit area by means of eq . ( [ t0-ip - l-4 ] ) . note that eq . ( [ t0-ip - l-4 ] ) leads to the classical result for the torque , @xmath180 where @xmath21 and @xmath158 denote the semi - infinite plate s area and extent in @xmath110 direction , respectively . for @xmath19 , eq . ( [ t0-ip - torque-4 ] ) becomes @xmath181 . a new characteristic contribution emerges from the edge effect eq . ( [ t0-ip - l-1 ] ) . unlike the total inclined - plate casimir energy @xmath182 , the edge energy ( [ t0-ip - l-1 ] ) decreases with the angle of inclination @xmath16 , see fig . [ t0-ip - torque-6 ] . this leads to a contribution which works against the standard torque ( [ t0-ip - torque-4 ] ) . for @xmath19 , the correction to eq . ( [ t0-ip - torque-4 ] ) emerging from the edge effect reads @xmath183 where we have used @xmath177 worldlines with @xmath140 ppl each . the coefficient in eq . ( [ t0-ip - torque-5 ] ) was calculated by expanding eq . ( [ t0-ip - l-1 ] ) around @xmath17 . equation ( [ t0-ip - torque-5 ] ) is shown in fig . [ t0-ip - torque-6 ] . we will see a similar subleading repulsive torque effect in the next section where we investigate finite - temperature contributions . decomposing the casimir energy at finite temperature @xmath76 into its zero - temperature part @xmath184 and finite - temperature correction @xmath38 , @xmath185 is straightforward in the worldline picture by using the relation ( [ int-4 ] ) . the finite - temperature correction is purely driven by the worldlines with nonzero winding number . as the winding - number sum does not take direct influence on the worldline averaging , the complicated geometry - dependent part of the calculation remains the same for zero or finite temperature . this disentangles the technical complications arising from geometry on the one hand and temperature on the other hand in a convenient fashion . the same statement holds for the casimir force @xmath186 . in order to demonstrate the simplicity of the worldline method , let us calculate the well - known thermal contribution to the casimir effect for parallel plates . in the following , we use the dimensionless parameter @xmath187 which distinguishes between the high - temperature @xmath188 and low - temperature @xmath189 parameter region . evaluating the general worldline formula for the casimir energy eq . using the parallel - plates @xmath73 functional of eq . results in ( @xmath190 ) @xmath191\right\rangle,\notag\end{aligned}\ ] ] where the exponential integral function @xmath192 is given by @xmath193 and @xmath106 again denotes the maximum extent of the worldline in the direction orthogonal to the plates . in the low - temperature limit , @xmath194 , the exponential integral functions vanish exponentially and can be neglected . we then obtain the small - temperature correction to @xmath195 in @xmath0 dimensions fully analytically : @xmath196 the term @xmath197 agrees with the standard textbook result @xcite . it dominates the thermal correction to the casimir force , yielding a comparatively suppressed power law dependence on the temperature , @xmath198 for small @xmath1 . this is an immediate consequence of the gap in the relevant part of the fluctuation spectrum in this closed geometry . this term can also be understood as an excluded - volume effect : the volume in between the plates can not be thermally populated by photons at low temperature due to the spectral gap . incidentally , the leading contribution to the energy @xmath199 is much less known . it does not contribute to the casimir force , since it is independent of @xmath7 when multiplied by the normalization prefactor @xmath200 . as we will see in section [ sec : ft - inclinedplates ] , @xmath7-independent terms in the energy should not be viewed as mere calculational artefacts but can also contribute to observables such as the casimir torque . to the best of our knowledge , ( [ ft - pp-3 ] ) represents the first exact analytic formula for this leading small temperature correction to the free energy . the high - temperature limit of ( [ ft - pp-1 ] ) can be obtained by a poisson resummation of the winding - number sum ( which is identical to returning to matsubara frequency space ) . our result agrees with @xcite and reads : @xmath201 for arbitrary @xmath202 and @xmath190 , eq . can be evaluated numerically . figure [ ft - pp-6 ] shows the worldline result together with the analytic asymptotics ( [ ft - pp-3])-([ft - pp-4 ] ) and the known exact analytic formula for @xmath19 , see e.g. @xcite : @xmath203 note that the result obtained in @xcite for the electromagnetic field is twice as large as the result for the scalar field ( [ ft - pp-5 ] ) . normalized to the zero - temperature result in @xmath19-dimensional spacetime versus the dimensionless temperature variable @xmath204 . the worldline result for 1000 worldlines with @xmath107 points each is plotted together with the analytic expressions ( [ ft - pp-3])-([ft - pp-5 ] ) . ] of course , taking the derivative of ( [ ft - pp-1 ] ) , ( [ ft - pp-3 ] ) or ( [ ft - pp-4 ] ) with respect to @xmath7 also gives immediate access to the thermal corrections @xmath205 to the casimir force . for instance , the low - temperature limit results in @xmath206 again revealing the power - law suppressed temperature dependence which is characteristic for a closed geometry . both magnitude and sign of the thermal force correction can be understood as an excluded - volume effect : as the temperature is small compared to the spectral gap , thermal modes in - between the plates can not be excited . hence , the thermal stefan - boltzmann energy density outside the plates is not balanced by a thermal contribution inside . thermal effects therefore enhance the attractive force between the plates . let us finally remark that the comparison between the small - temperature limit of ( [ ft - pp-5 ] ) ( calculated with the help of the poisson summation ) and our analytic formula ( [ ft - pp-3 ] ) closes a gap in the literature . with this comparison , we can find the exact value of the integral occurring in the prefactor of the leading low - temperature term in the energy , @xmath207=\frac{\zeta(3)}{2 \pi^2 } , \end{aligned}\ ] ] numerically corresponding to @xmath208 . this result has been observed numerically in the sum over odd reflection contributions to the parallel - plates casimir energy in the optical approach to the casimir effect @xcite . whereas the inclined - plate geometry is much more difficult to deal with than the parallel - plate case when using standard methods , there is comparatively little difference in the worldline language . inserting the inclined - plates @xmath73 functional ( [ t0-ip-2 ] ) with @xmath129 as in eq . ( [ t0-ip-5b ] ) into the general worldline formula ( [ int-3 ] ) yields @xmath209 where @xmath210\right\rangle\right ) \end{aligned}\ ] ] is the thermal contribution to the energy . here and in the following , we confine ourselves to spacetime dimensions @xmath211 where all expressions exhibit well - controlled convergence . in the low - temperature limit , the exponential integral functions can be neglected as long as @xmath212 for all @xmath127 . this is certainly the case for @xmath213 , but not necessarily for @xmath17 . the latter case is again identical to the semi - infinite plate parallel to a infinite one , and is being considered separately in the next section and also in the appendix . for @xmath213 , the low - temperature limit is then given by the first two terms ( first line ) of eq . ( [ ft - ip-2 ] ) . note that the first @xmath214 term does not contribute to the casimir force , since it is an @xmath7-independent contribution to @xmath215 if read together with the normalization prefactor . from the second term , we obtain the low - temperature thermal correction to the casimir force upon differentiation with respect to @xmath7 , @xmath216 the temperature dependence differs from the parallel - plates case by one power of @xmath1 , implying a significantly stronger temperature dependence at small temperatures . this is a direct consequence of the fact that we are dealing here with an open geometry . we emphasize that the result has been obtained fully analytically . in @xmath19 and @xmath14 , our result agrees with the perpendicular - plates study of @xcite where this nontrivial interplay between temperature and geometry has been demonstrated for the first time . as shown therein , the thermal correction for this open geometry at experimentally - relevant large separations can be an order of magnitude larger than for a closed geometry . whereas the @xmath7-independent first term of eq . ( [ ft - ip-2 ] ) does not contribute to the force , both terms in the first line of eq . ( [ ft - ip-2 ] ) contribute to the low - temperature limit of the casimir torque . the thermal contribution to the torque is @xmath217 which at low temperature reads @xmath218 \right . \notag \\ & \left.\ \ + a\zeta ( d-1 ) \left(2 t\right)^{d-1 } \gamma\left(\frac{d-1}{2}\right ) \sqrt{\pi}\right),\end{aligned}\ ] ] where @xmath219 has been used , and we have introduced @xmath220 . this expression depends on only one nontrivial worldline average . in limiting cases , this average can be given analytically , as it reduces to the case described by eq . ( [ t0-pp-4 ] ) : we find @xmath221 and @xmath222 . for arbitrary @xmath16 , this average can be well approximated by @xmath223 as we will explain in the following . with eq . ( [ t0-ip-3 ] ) , we can write @xmath224 where @xmath225 denotes the value of @xmath122 that satisfies the minimum condition in eq . ( [ t0-ip-3 ] ) . together with @xmath226 we can interpret @xmath227 as the coordinates of the point @xmath228 in the @xmath16-rotated system . since the @xmath229 and @xmath230 coordinates of each loop are generated independently of each other , @xmath231 and @xmath232 are not correlated . we therefore obtain @xmath233 by symmetry , the average @xmath234 vanishes , and we get from eq . ( [ ft - ip-6 - 2 ] ) @xmath235 on the other hand , @xmath236 . substituting eq . ( [ ft - ip-6 - 4 ] ) into the average of eq . ( [ ft - ip-6 - 1 ] ) leads to @xmath237 and @xmath238 , such that the desired eq . ( [ ft - ip6 ] ) can be motivated by eq . ( [ ft - ip-6 - 3 ] ) . we would like to stress that only the second term in eq . ( [ ft - ip6 ] ) has been estimated with the constraint imposed by the exactly known result for @xmath134 , see above . the first term is exact and dictates the behavior of the perpendicular - plates limit . this result is compared to the numerically obtained data in fig . [ ft - ip8 ] . for the v - loop algorithm @xcite used here to generate the loops , the expectation value of the maximal extent @xmath232 is systematically smaller . this error is about the average spacing @xmath239 , see eq . ( [ intro-10 ] ) . as a consequence , the systematic error of @xmath240 and @xmath241 is about @xmath242 at @xmath243 . we observe a good agreement of the data with eq.([ft - ip6 ] ) at @xmath17 and @xmath14 if the systematic error is taken into account . however , the agreement is actually perfect for all @xmath16 when using worldline estimates for the prefactors in eq . ( [ ft - ip6 ] ) instead of @xmath169 and @xmath244 , see the modified curve in fig . [ ft - ip8 ] . this shows that eq . ( [ ft - ip6 ] ) will well fit the data in the continuum limit @xmath245 . let us return to the calculation of the casimir torque . in the vicinity of the perpendicular - plates configuration , @xmath246 , we can now obtain an expression to first order in @xmath247 : @xmath248 \notag \\ & \ \ \ \ \ \ \ \quad + 2\,at\ , \zeta ( d-1 ) \gamma\left(\frac{d-1}{2}\right ) \sqrt{\pi}\big).\end{aligned}\ ] ] here we have used @xmath249 . apart from @xmath250 , eq . ( [ ft - ip9 ] ) is an analytical expression . using ( [ ft - ip6 ] ) , we obtain @xmath251 , which is about ten percent of the dominating analytical term in square brackets @xmath252 . we observe that the first term , which dominates in the limit @xmath253 , gives a contribution to the torque which drives the system away from the perpendicular - plates case @xmath14 . zero- and finite - temperature contributions thus have the same sign . the fact that @xmath14 is a repulsive fixed point is also in agreement with naive expectations . for @xmath19 , eq . ( [ ft - ip9 ] ) reads @xmath254 which should be compared with the first - order term arising from the @xmath255 contribution eq . ( [ t0-ip - torque-2 ] ) , which reads @xmath256 . thus , for @xmath19 we obtain to first order in @xmath257 @xmath258 in the validity regime of the low - temperature expansion , @xmath259 , the positive first term is always dominant , hence perpendicular - plates case remains a repulsive fixed point . most importantly , we would like to stress that the quadratic dependence of the torque on the temperature @xmath260 ( @xmath5 in the general case ) for the inclined - plates configuration represents the strongest temperature dependence of all observables discussed in this article . the high - temperature limit of eq . ( [ ft - ip-2 ] ) can again be obtained by poisson summation . the result is : @xmath261 the remaining worldline average in this expression yields some positive finite number . irrespective of its precise value for a given angle @xmath16 and @xmath0 ( the precise value of the integral for a specific @xmath16 can be read off , for instance , from fig . [ ym - integrals ] for either @xmath262 or @xmath19 and fig . [ t0-ip-7 ] for @xmath263 ) , we stress that we observe the same linear dependence on temperature @xmath264 as in the parallel plate case . this is nothing but the familiar phenomenon of the dominance of the zeroth matsubara mode at high temperatures , implying dimensional reduction , as discussed above . this mechanism is obviously geometry independent . also the casimir force remains attractive also for high temperatures . a particularly interesting example for the geometry - temperature interplay is given by the semi - infinite plate parallel to the infinite plate ( 1si configuration ) . in this case , the angle of inclination @xmath16 in fig . [ t0-ip-4 ] analogously to eq . ( [ t0-ip-4 ] ) , the finite - temperature casimir energy can be decomposed as @xmath265 where @xmath266 correspond to the standard parallel - plate formulas as given in eqs . ( [ t0-pp-3 ] ) and ( [ ft - pp-1 ] ) , with @xmath21 being now the surface of the semi - infinite plate . approaching the 1si limit of @xmath267 from the inclined - plates configuration in the limit @xmath144 is again a delicate issue , as the proper order of limits @xmath144 and @xmath165 has to be accounted for , see sect . [ sec : incl - plat - varph ] . as the analysis is technically involved ( but the outcome obvious ) , we defer it to the appendix . let us here concentrate on the temperature - dependent edge contribution @xmath268 . we set @xmath17 in eq . ( [ t0-ip-2 ] ) and evaluate eq . ( [ int-3 ] ) . the result is ( here and in the following , we confine ourselves to @xmath211 ) : @xmath269.\end{aligned}\ ] ] note that the first term , being the main contribution to the casimir energy at small @xmath1 , does not contribute to the casimir force since it is @xmath7 independent . contrary to the case with @xmath270 , the exponential integral functions can not be neglected in the low - temperature limit , since the argument of @xmath192 becomes zero at the lower bound of the integral for any @xmath271 . this results in a correction @xmath272 , with @xmath273 , to the low - temperature limit of the first term . here , however , we concentrate on the last term , as it gives rise to the thermal correction of the casimir force . in the low - temperature limit , we find @xmath274 for small @xmath202 , the main contribution to the integral comes from its lower bound as the exponential function rapidly decreases for large argument . at the lower bound , we can take the worldline average @xmath275 first , yielding a smooth function . we expand the latter in a power series , @xmath276 where the exponents @xmath277 do not necessarily have to be integers . inserting ( [ ft - si-4 ] ) into ( [ ft - si-3 ] ) leads to @xmath278 where we have neglected exponentially suppressed contributions . for the lowest - order term , we obtain @xmath279 , since for a given worldline @xmath280 there exists a corresponding worldline in the ensemble with @xmath281 . we conclude that the coefficient of the @xmath282 term vanishes . we determine the higher coefficients @xmath283 from computing @xmath275 in the vicinity of @xmath284 by worldline numerics . figure [ ft - si-7 ] depicts the form of @xmath285 near the lower bound @xmath284 . a global fit to this function including two coefficients @xmath286 is given by @xmath287 where we have kept @xmath288 fixed . the resulting thermal correction to the force is shown in fig . ( [ fig - si-9 ] ) for @xmath19 , where we compare the full numerical solution with different orders of the expansion and the high - temperature asymptotics , see below . as the low - temperature asymptotics is directly related to the lowest nonvanishing coefficient @xmath289 , we have also performed local fits to the function @xmath285 in the vicinity of @xmath284 . depending on the fit window , the leading exponent can grow up to @xmath290 . ( of course , the fit window must be large enough to avoid that the worldline discretization becomes visible ; otherwise , the exponent trivially but artificially approaches @xmath291 as the discretized worldline is a polygon on a microscopic scale ) . in any case , we conclude that the low - temperature regime of the 1si edge effect is well described by a non - integer power law , @xmath292 , where the fractional exponent arises from the geometry - temperature interplay in this open geometry . of course , our numerical analysis can not guarantee to yield the true asymptotic behavior in the limit @xmath293 , but our data in the low - temperature domain @xmath294 is well described by the non - integer scaling at next - to - leading order . let us finally turn to the high - temperature limit of eq . which can again be obtained by poisson summation . the result for the edge energy reads @xmath295 where the worldline average is subject to numerical evaluation . the resulting high - temperature limit of the casimir force is shown in fig . ( [ ft - si-9 ] ) for @xmath19 . the high - temperature limit is again linear in @xmath1 in accordance with general dimensional - reduction arguments . in this work , we have provided further numerical as well as analytical evidence for the nontrivial interplay between geometry and temperature in the casimir effect . whereas closed geometries such as the parallel - plates case exhibit a comparatively strong suppression of thermal corrections at low temperatures , open geometries such as the general inclined - plates geometry reveal a more pronounced temperature dependence in this regime . the terminology _ open _ and _ closed _ corresponds to the absence or presence of a gap in the relevant part of the spectrum of fluctuations which gives rise to the casimir effect . in closed geometries , the spectral gap inhibits sizable fluctuations at temperatures below the scale set by the gap . by contrast , open geometries allow for sizable thermal fluctuations at any value of the temperature . concentrating on the inclined - plates geometry in @xmath0 dimensions , the temperature dependence of the casimir force can become stronger by one power in the temperature parameter ( implying thermal corrections which can be an order of magnitude larger than for a closed geometry ) . the inclined - plates geometry is particularly interesting as the limit of a semi - infinite plate parallel to an infinite plate ( 1si configuration ) is somewhat in - between open and closed geometries : the open part of the spectrum only arises due to the edge of the semi - infinite plate . interestingly , the resulting thermal correction numerically shows a power - law temperature dependence with a non - integer exponent @xmath296 . the strongest temperature dependence @xmath5 in the low - temperature limit occurs for the casimir torque of the inclined - plates configuration . this is , because it arises from the leading thermal correction of the interaction energy which contributes to the torque but not to the casimir force . our results have been derived for the case of a fluctuating scalar field obeying dirichlet boundary conditions on the surfaces . whereas this model system should not be considered as a quantitatively appropriate model for the real electromagnetic casimir effect , our general conclusions about the geometry - temperature interplay are not restricted to the dirichlet scalar case . on the contrary , all our arguments based on the presence or absence of a spectral gap will also be valid for the electromagnetic case . whether or not the case of neumann or electromagnetic boundary conditions leads to different power - law exponents for the temperature dependence of the `` geothermal '' phenomena remains an interesting question for future research . in view of the fact that most ( strictly speaking all ) experiments are performed in open geometries , e.g. , the sphere - plate geometry , at room temperature , an analysis of the geometry - temperature interplay of these experimentally relevant configurations is most pressing . we have analyzed the @xmath48 behavior of inclined plates at zero temperature in sect . [ sec : incl - plat - varph ] . here , we consider the same limit for the thermal correction to the energy . the decomposition of the 1si casimir energy into bulk and edge contributions can also be performed for the thermal corrections , @xmath297 where whereas @xmath301 remains finite , @xmath302 shows a divergent behavior as @xmath48 . let us therefore concentrate on @xmath303 , in order to isolate the source of the apparent divergence which is related to the order of limits of @xmath304 and @xmath305 . in the case of inclined plates , @xmath306 is infinite for all @xmath136 , resulting in a @xmath307 divergent energy density ( energy per length ) for @xmath48 . parallel plates , on the other hand have a finite energy density ( energy per area ) and @xmath308 . in the following , we show how to obtain an analytic transition from @xmath309 to @xmath310 for small @xmath16 by working with large but finite @xmath158 , and taking @xmath311 at the end of the calculation . the first limit results in a divergent energy density per unit edge length of the inclined plates as @xmath48 , in eqs . ( [ wpir - iplaft - f1]-[wpir - iplaft - f3 ] ) , the @xmath110 integration was performed first . let us now do the proper time integration first . the @xmath120 function ( [ t0-ip - l-3 ] ) , valid for small @xmath16 , reflects itself in the lower end of the proper time integral : @xmath313\right\rangle . \end{aligned}\ ] ] the proper time integration yields for re@xmath166>1 $ ] : @xmath314 inserting eq . ( [ wpir - iplaft - f4b ] ) into eq . ( [ wpir - iplaft - f4 ] ) leads to where @xmath316 and @xmath317 . one can show that the first three terms of eq . ( [ wpir - iplaft - f5 ] ) are of order @xmath318 . the forth term is clearly @xmath319 . the last line can be converted into the parallel - plates energy density ( [ ft - pp-1 ] ) by neglecting @xmath320 with respect to @xmath7 and using the identity @xmath321 ; the error is of order @xmath318 . the first - order correction to the parallel - plates case is therefore encoded in the fourth term . the second - order correction is in the first three terms since the @xmath322 terms cancel each other in the remainder . in this limit ( @xmath323 ) , all sums converge for @xmath324>2 $ ] . let us rearrange ( [ wpir - iplaft - f5 ] ) so as to investigate the @xmath325 case with @xmath16 being small but finite : the large-@xmath320 behavior of the first two terms can be obtained through poisson summation limit of @xmath327 yields @xmath328 where @xmath329 is the fourier transform of @xmath330 . ] and reads @xmath331 where @xmath332 are constants , the values of which are of no importance . we see that eq . ( [ wpir - iplaft - f7 ] ) vanishes for @xmath190 . for @xmath211 the terms vanish even if multiplied by the infinite length @xmath158 . remember that the inclined - plates formulae at finite temperature are valid for @xmath211 as well . in order to keep the remaining terms of eq . ( [ wpir - iplaft - f6 ] ) finite , we multiply both sides with the infinite length @xmath158 converting the vanishing casimir energy per area into the finite energy per length . the poisson summation of the second line of eq . ( [ wpir - iplaft - f6 ] ) results in where @xmath334 are constants . these terms containing @xmath335 s vanish for @xmath211 and @xmath325 . applying the identity @xmath336 to the last three terms in eq . ( [ wpir - iplaft - f6 ] ) , we rediscover the inclined - plates formula ( [ wpir - iplaft - f3 ] ) from eqs . ( [ wpir - iplaft - f6 ] ) , ( [ wpir - iplaft - f8 ] ) valid for small angles @xmath16 . from eqs . ( [ wpir - iplaft - f7 ] ) and ( [ wpir - iplaft - f8 ] ) , one can infer that the first correction to eq . ( [ wpir - iplaft - f3 ] ) is of order @xmath337 . the authors are grateful to klaus klingmller for interesting discussions . aw acknowledges support by the landesgraduiertenfrderung baden - wrttemberg , by the heidelberg graduate school of fundamental physics , and by the dfg under contract gi 328/3 - 2 . hg was supported by the dfg under contract no . gi 328/1 - 4 ( emmy - noether program ) , gi 328/5 - 1 ( heisenberg program ) . casimir , kon . wetensch . * 51 * , 793 ( 1948 ) . m. bordag , u. mohideen and v. m. mostepanenko , phys . rept . * 353 * , 1 ( 2001 ) ; r. onofrio , new j. phys . * 8 * , 237 ( 2006 ) [ arxiv : hep - ph/0612234 ] ; s. y. buhmann and d. g. welsch , prog . quant . electron . * 31 * , 51 ( 2007 ) [ arxiv : quant - ph/0608118 ] . k. a. milton , `` the casimir effect : physical manifestations of zero - point energy , '' _ river edge , usa : world scientific ( 2001)_. s. k. lamoreaux , phys . * 78 * , 5 ( 1997 ) ; u. mohideen and a. roy , phys . * 81 * , 4549 ( 1998 ) ; h.b . chan _ et al . _ , science 291 , 1941 ( 2001 ) ; r.s . decca _ et al . _ , phys . rev . d * 68 * , 116003 ( 2003 ) ; phys . * 94 * , 240401 ( 2005 ) . m. bostrm and bo e. sernelius , phys . * 84 * , 4757 ( 2000 ) . v. m. mostepanenko _ et al . _ , j. phys . a * 39 * , 6589 ( 2006 ) [ arxiv : quant - ph/0512134 ] . i. brevik , s. a. ellingsen and k. a. milton , arxiv : quant - ph/0605005 . g. bimonte , phys . rev . a * 79 * , 042107 ( 2009 ) [ arxiv:0903.0951 [ quant - ph ] ] . ingold , a. lambrecht , s. reynaud , arxiv:0905.3608 [ quant - ph ] ( 2009 ) . a. scardicchio and r. l. jaffe , nucl . b * 743 * ( 2006 ) 249 [ arxiv : quant - ph/0507042 ] . h. gies and k. klingmuller , j. phys . a * 41 * , 164042 ( 2008 ) . h. gies and k. klingmuller , phys . lett . * 97 * , 220405 ( 2006 ) [ arxiv : quant - ph/0606235 ] . h. gies , k. langfeld and l. moyaerts , jhep * 0306 * , 018 ( 2003 ) ; arxiv : hep - th/0311168 . m. b. halpern and w. siegel , phys . rev . d * 16 * , 2486 ( 1977 ) ; a. m. polyakov , `` gauge fields and strings , '' harwood , chur ( 1987 ) z. bern and d.a . kosower , nucl . * b362 * , 389 ( 1991 ) ; * b379 * , 451 ( 1992 ) m.j . strassler , nucl . phys . * b385 * , 145 ( 1992 ) . m. g. schmidt and c. schubert , phys . b * 318 * , 438 ( 1993 ) [ arxiv : hep - th/9309055 ] ; for a review , see c. schubert , phys . rept . * 355 * , 73 ( 2001 ) . h. gies and k. langfeld , nucl . b * 613 * , 353 ( 2001 ) ; int . j. mod . a * 17 * , 966 ( 2002 ) . h. gies and k. klingmuller , j. phys . a * 39 * 6415 ( 2006 ) [ arxiv : hep - th/0511092 ] . h. gies and k. klingmuller , phys . lett . * 96 * , 220401 ( 2006 ) [ arxiv : quant - ph/0601094 ] . h. gies and k. klingmuller , phys . d * 74 * , 045002 ( 2006 ) [ arxiv : quant - ph/0605141 ] . m. schaden , phys . lett . * 102 * , 060402 ( 2009 ) . derjaguin , i.i . abrikosova , e.m . lifshitz , q.rev . * 10 * , 295 ( 1956 ) ; j. blocki , j. randrup , w.j . swiatecki , c.f . tsang , ann . ( n.y . ) * 105 * , 427 ( 1977 ) . a. scardicchio and r. l. jaffe , nucl . b * 704 * , 552 ( 2005 ) ; phys . lett . * 92 * , 070402 ( 2004 ) . r. balian and b. duplantier , annals phys . * 112 * , 165 ( 1978 ) . a. bulgac , p. magierski and a. wirzba , phys . d * 73 * , 025007 ( 2006 ) [ arxiv : hep - th/0511056 ] ; a. wirzba , a. bulgac and p. magierski , j. phys . a * 39 * ( 2006 ) 6815 [ arxiv : quant - ph/0511057 ] . t. emig , r. l. jaffe , m. kardar and a. scardicchio , phys . rev . lett . * 96 * ( 2006 ) 080403 . m. bordag , phys . d * 73 * , 125018 ( 2006 ) ; phys . d * 75 * , 065003 ( 2007 ) . o. kenneth and i. klich , phys . lett . * 97 * , 160401 ( 2006 ) ; arxiv:0707.4017 . t. emig , n. graham , r. l. jaffe and m. kardar , arxiv:0707.1862 ; arxiv:0710.3084 . r. b. rodrigues , p. a. maia neto , a. lambrecht and s. reynaud , phys . lett . * 96 * , 100402 ( 2006 ) [ arxiv : quant - ph/0603120 ] ; phys . a * 75 * , 062108 ( 2007 ) . f. d. mazzitelli , d. a. r. dalvit and f. c. lombardo , new j. phys . * 8 * , 240 ( 2006 ) ; d. a. r. dalvit , f. c. lombardo , f. d. mazzitelli and r. onofrio , phys . a * 74 * , 020101 ( 2006 ) . k. a. milton and j. wagner , phys . d * 77 * , 045005 ( 2008 ) [ arxiv:0711.0774 [ hep - th ] ] ; j. phys . a * 41 * , 155402 ( 2008 ) [ arxiv:0712.3811 [ hep - th ] ] . k. a. milton , p. parashar and j. wagner , arxiv:0806.2880 [ hep - th ] . m. bordag , d. robaschik and e. wieczorek , annals phys . * 165 * , 192 ( 1985 ) . t. emig , a. hanke and m. kardar , phys . * 87 * ( 2001 ) 260402 .

we have used 2d back - projection doppler tomography as a tool to examine the influence of gravitational and magnetic phenomena in interacting binaries which undergo mass transfer from a magnetically - active star onto a non - magnetic main sequence star . this multi - tiered study of over 1300 time - resolved spectra of 13 algol binaries involved calculations of the predicted dynamical behavior of the gravitational flow and the dynamics at the impact site , analysis of the velocity images constructed from tomography , and the influence on the tomograms of orbital inclination , systemic velocity , orbital coverage , and shadowing . the h tomograms revealed eight sources : chromospheric emission , a gas stream along the gravitational trajectory , a star - stream impact region , a bulge of absorption or emission around the mass - gaining star , a keplerian accretion disk , an absorption zone associated with hotter gas , a disk - stream impact region , and a hot spot where the stream strikes the edge of a disk . we described several methods used to extract the physical properties of the emission sources directly from the velocity images , including s - wave analysis , the creation of simulated velocity tomograms from hydrodynamic simulations , and the use of synthetic spectra with tomography to sequentially extract the separate sources of emission from the velocity image . in summary , the tomography images have revealed results that can not be explained solely by gravitational effects : chromospheric emission moving with the mass - losing star , a gas stream deflected from the gravitational trajectory , and alternating behavior between stream state and disk state . our results demonstrate that magnetic effects can not be ignored in these interacting binaries .

1408.0224

the effects of gravitational forces and magnetic fields have been recognized as playing an important role in the mass accretion process in a variety of astrophysical systems , including protostars , binaries containing compact objects , and massive galaxies with supermassive black holes . however , the consequences for accretion onto normal stars in close binary star systems are not as well understood . in this paper , we focus on binaries containing normal stars , specifically the algol - type binaries with prototype @xmath1 per ( algol ) , which are progenitors of the cataclysmic variables @xcite . in these systems , the cool star displays enhanced magnetic activity associated with rapid rotation induced by tidal synchronization of the binary system , and the resulting dynamic phenomena are many times more powerful than those found on the sun . since the cool star is the source of the mass transfer stream , then this enhanced activity will influence the mass transfer process in a direct way . in the algols , no magnetic activity is linked to the hot b - a - type mass - gaining star @xcite . however , both gyrosynchrotron radio emission and thermal x - ray emission are expected from the cool f - k - type mass - losing companion , mostly due to coronal activity @xcite . additional evidence of magnetic activity can be derived from the spectra ( e.g. , caii h & k or h@xmath0 emission ) , however such evidence is typically sparse because these cool stars contribute only @xmath2 5 - 15% to the total luminosity of the binary , depending on wavelength , with the highest contribution in the infrared . moreover , any magnetic components in the h@xmath0 line are masked by contributions from circumstellar gas flowing between and around the stars @xcite . these magnetic phenomena observed in algol binaries are similar to those detected in the double - cool star detached rs cvn binaries @xcite . the presence of an active magnetic field on the mass - losing star can lead to magnetic threading of the gas flow from the l1 point towards the mass - gaining star when the ram pressure of the accretion flow exceeds the magnetic stress at the l1 point @xcite . as a consequence , the magnetic field will be dragged along with the fluid flow and eventually surround the mass gainer . this process is associated with the superhump phenomenon , which typically arises from the prograde apsidal precession of an eccentric accretion disk and is usually identified in cataclysmic variables @xcite . superhumps were first detected in algol binaries by @xcite using 2.3 ghz and 8.3 ghz observations of @xmath1 per derived from the 5.6-year continuous radio flare survey of @xcite . this prediction of the superhump phenomenon in @xmath1 per was confirmed by @xcite using 15 ghz vlbi observations . astrometric methods have also been used to confirm the association between the source of the radio emission and the physical location of the k - type mass - losing star , with a precision of @xmath30.5 mas @xcite . indirect images of interacting binaries created with the aid of doppler tomography have led to a better understanding of the role played by gravitational forces and magnetic fields . this image reconstruction technique was introduced by @xcite and then applied to cataclysmic variables and x - ray binaries @xcite and also to direct - impact and long - period algol binaries @xcite . the images illustrate the movement of gas from the inner lagrangian point towards and around normal or compact mass - gaining stars , and show that a well - defined accretion disk is formed around the mass - gaining star in the compact binaries . however , in the _ direct - impact _ systems the mass gainer is a normal main sequence star , and the stars are very close to each other relative to the orbital separation . the first 2d tomography images of these latter systems have shown that the path of the flow is usually governed by gravitational and coriolis forces , and there is additional evidence that the active cool mass - losing star can have substantial influence on the mass transfer process @xcite . in this paper , we assess the progress that has been made in 2d tomography of the direct - impact and long - period algols to identify patterns and similarities between systems with orbital periods from 1.2 days to 11.1 days ; some preliminary results have appeared in conference proceedings ( e.g. , @xcite ) . in 2 , we describe the predicted dynamical behavior of the gravitational flow along the gas stream and beyond ; in 3 , we describe the technique of doppler tomography , and 4 provides the application to the direct - impact systems . we also examine the influence of various system properties on the 2d images in 5 , specifically the effects of orbital inclination , systemic velocity , orbital phase coverage , inadequate velocity resolution , and the effect of the gas opacity known as shadowing . in 6 , we describe the techniques that can be used to extract physical properties of the gas flows directly from the images via four methods : an s - wave analysis ; using hydrodynamic simulations to create simulated tomograms ; using synthetic spectra to model the contributions of the accretion disk , gas stream , and other sources of emission ; and using synthetic spectra with tomography to confirm visually that the separate accretion structures have been modeled accurately . the results and conclusions are given in 7 and 8 , respectively . the semidetached interacting algol binaries contain a cool f - k iii - iv secondary star which has expanded to fill its roche lobe and is transferring material through a gas stream onto a hot b - a v primary star ( the mass gainer ) ; in this context , the primary is the brighter and more massive star . a radius vs. mass ratio or @xmath4 diagram , as illustrated in figure [ fig1 ] , can be used to predict the systems that should undergo direct impact of the gas stream onto the stellar photosphere . this figure also shows the roche geometries for the direct - impact systems and some wider binaries based on the properties listed in table [ tab1 ] , derived from @xcite . here , @xmath5 is the orbital period of the binary ; @xmath6 is the orbital inclination ; @xmath7 is the mass ratio of the binary ; @xmath8 is the radius of the mass gainer , @xmath9 , relative to the binary separation @xmath10 ; @xmath11 and @xmath12 are the velocity semi - amplitudes of the primary and secondary , respectively ; and @xmath13 is the systemic velocity of the binary . until these systems are resolved by an instrument with absolute phase angle calibration or direct imaging , the values of inclination will have an @xmath6 vs. @xmath14 ambiguity , corresponding to prograde vs. retrograde orbits . for the close binary in @xmath1 per , @xcite found that the orbit of the binary is retrograde , so the inclination should be @xmath15 . however , we retain the usual convention here with angles up to @xmath16 corresponding to prograde orbits . in figure [ fig1 ] , the direct - impact binaries will be found above the solid curve labeled @xmath17 , corresponding to the radius of a stable accretion disk ( e.g. , @xmath18 lib , rz cas , u sge ) since the mass gainer is larger in size than the predicted disk radius . however , the wider binaries that always form a stable accretion disk ( e.g. , s cnc ) will be found below the dashed curve labeled @xmath19 , corresponding to the distance of closest approach of the gas stream , hence no impact is predicted since the size of the star is smaller than @xmath19 ; these curves are based on the semi - analytical ballistic calculations of @xcite . it is noticeable that systems between the two curves may have a grazing impact between the gas stream and the mass gainer ( e.g. , s equ , sw cyg , cx dra ) while no impact is expected in other cases ( e.g. , tt hya ) . in this scenario , the direct - impact systems typically have @xmath20 days @xcite while the grazing - impact systems have @xmath21 days . lcccrcr ' '' '' system & @xmath5 & @xmath6 & @xmath22 & @xmath23 & @xmath11 & @xmath13 + & ( days ) & ( @xmath24 ) & & & ( km ) & ( km ) + ' '' '' rz cas & 1.195 & 83.0 & 0.33 & 0.232 & 70.9 & @xmath2545.5 + @xmath18 lib & 2.327 & 78.6 & 0.35 & 0.300 & 76.6 & @xmath26 4.0 + rw tau & 2.769 & 90.0 & 0.23 & 0.179 & 53.0 & @xmath2520.2 + @xmath1 per & 2.867 & 81.4 & 0.22 & 0.206 & 44.0 & @xmath26 3.8 + tx uma & 3.063 & 81.0 & 0.30 & 0.154 & 54.8 & @xmath2513.0 + u sge & 3.381 & 89.0 & 0.33 & 0.225 & 70.0 & @xmath2510.0 + s equ & 3.436 & 87.4 & 0.12 & 0.192 & 23.0 & @xmath2548.0 + u crb & 3.452 & 79.1 & 0.29 & 0.170 & 59.7 & @xmath25 6.7 + rs vul & 4.478 & 78.7 & 0.31 & 0.267 & 54.0 & @xmath2520.1 + sw cyg & 4.573 & 87.1 & 0.21 & 0.155 & 32.3 & @xmath25 7.5 + cx dra & 6.696 & 53.0 & 0.23 & 0.128 & 34.1 & @xmath26 3.54 + tt hya & 6.953 & 84.4 & 0.27 & 0.089 & 23.9 & @xmath26 0.0 + au mon & 11.113 & 82.0 & 0.20 & 0.114 & 32.2 & @xmath2611.8 + v711 tau & 2.838 & 33.0 & 1.25 & 0.113 & 61.7 & @xmath2514.01 + [ tab1 ] the dynamics of the direct impact can be understood in terms of the parameters given in table [ tab2 ] , which lists the thermal or kelvin - helmholtz timescale , @xmath27 , where @xmath28 ; the dynamical or free fall timescale , @xmath29 ; the circularization timescale or the time for the gas to circle the mass gainer at a distance @xmath30 from the center of the star , @xmath31 , where @xmath32 and the values in the table correspond to @xmath33 ; the synchronous velocity of each binary , @xmath34 ; the escape velocity for the mass gainer , @xmath35 ; and the velocity of the gas stream at the impact site or shock region , @xmath36 , derived from the calculation of the gas stream trajectory . for these calculations , the mass , radius , luminosity , and temperature all refer to the mass - gaining star , and @xmath5 is the orbital period . the values of @xmath37 in table [ tab2 ] suggest that the mass gainer in a typical direct - impact system would radiate its reserves of thermal energy within @xmath38 yr . however , the star would take only @xmath2 0.8 - 3.9 hr to respond to the inflow of gas from the companion star ( @xmath39 in table [ tab2 ] ) . moreover , the gas beyond the impact site would circle the star in only @xmath40 hr at the stellar surface , @xmath33 ( see @xmath41 in table [ tab2 ] ) or @xmath42 hr at a distance @xmath43 from the center of the star . moreover , the impact of the gas should spin up the star since the stream impact velocities of @xmath44 km are much higher than the @xmath45 km synchronous velocities of the mass gainer . the stream impact velocity is lower than the predicted escape velocity ( @xmath46 km ) except in the cases of rw tau , rs vul , and au mon . lcccccc ' '' '' system & @xmath37 & @xmath39 & @xmath41 & @xmath47 & @xmath48 & @xmath36 + & ( @xmath49yr ) & ( hr ) & ( hr ) & ( km ) & ( km ) & ( km ) + [ 0.5ex ] ' '' '' rz cas & 1.620 & 0.84 & 3.75 & 62.3 & 673.8 & 514 @xmath3 9 + @xmath18 lib & 0.530 & 2.42 & 10.73 & 89.6 & 659.6 & 434 @xmath3 3 + rw tau & 1.070 & 1.27 & 5.63 & 40.2 & 671.3 & 739 @xmath3 11 + @xmath1 per & 0.943 & 1.61 & 7.14 & 51.2 & 697.5 & 557 @xmath3 7 + tx uma & 1.790 & 1.15 & 5.11 & 38.0 & 772.6 & 644 @xmath3 11 + u sge & 0.736 & 2.26 & 10.03 & 62.9 & 719.4 & 558 @xmath3 13 + s equ & 0.735 & 1.67 & 7.40 & 41.2 & 649.8 & 566 @xmath3 11 + u crb & 1.430 & 1.49 & 6.60 & 44.0 & 781.1 & 719 @xmath3 22 + rs vul & 0.195 & 3.85 & 17.11 & 62.2 & 552.3 & 581 @xmath3 8 + sw cyg & 0.597 & 1.66 & 7.38 & 28.8 & 605.5 & 534 @xmath3 8 + cx dra & 1.400 & 1.85 & 8.24 & 30.2 & 834.2 & 576 @xmath3 17 + tt hya & 1.340 & 1.13 & 5.03 & 14.8 & 697.2 & - + au mon & 0.621 & 2.52 & 11.20 & 21.0 & 705.3 & 691 @xmath3 65 + [ tab2 ] to illustrate the expected behavior of the gas flow in the velocity domain , we compared plots of the cartesian representation of the binary with the corresponding velocity frame ( see figure [ fig2 ] ) . since the direct - impact systems have eccentricities that are very close to zero , then the binaries can be assumed to be tidally locked . the assumption of synchronous rotation ( @xmath50 ) provides a relationship between the length and velocity scales that permits the locations of the stars to be plotted in the velocity frame . figure [ fig2 ] also displays two large circles around the binary corresponding to the locus of a keplerian accretion disk in the velocity frame if that disk fills the space between the photosphere and the roche surface of the mass gainer ; the largest solid circle and the smaller dashed circle mark the inner and outer edge of the accretion disk , respectively ; and the plus sign marks the center of mass of the binary . in both the cartesian and velocity frames , the solid trajectory is the gravitational free - fall path of the gas stream from the l1 point under the influence of the coriolis force , in the frame of reference of the binary . the trajectory has been divided into ten parts marked by small solid circles to permit a visual comparison of the simultaneous position and velocity of a specific location along the gas stream trajectories . in addition , the large solid red circle and the extended blue region along the trajectory represent the median and range , respectively , of the predicted location of the stream - photosphere impact site . the values of the stream velocity at the impact site are listed under @xmath36 in the last column of table [ tab2 ] . the impact site was examined in greater detail by @xcite by calculating the ram pressure in the gas stream and the depth of the shock region relative to the stellar photosphere . the depth at which the shock forms is sensitive to the assumed valid of the mass transfer rate . over a range of mass transfer rates , the stream reached a depth of @xmath51 below the photosphere to @xmath52 above the photoshere in the case of @xmath1 per , suggesting that most of the flow is reflected away from the shock region in this binary . the tangential component of the impact velocities listed in table [ tab2 ] can produce substantial differential rotation and redistribution of the gas flow around the mass gainer ; and observational evidence of such high rotational velocity regions was found in ultraviolet spectra by @xcite and at h@xmath0 by @xcite . the impact also produces heating of the gas to temperatures of @xmath53 k , and evidence of the shock region was detected as a high temperature accretion region in ultraviolet spectra @xcite . similar behavior is expected for the other direct - impact systems . so , purely in terms of gravitational and coriolis forces , the gas flow from the l1 point will make direct impact with the surface of the mass gainer in most of the systems under study . since the flow velocities are much higher than the synchronous velocity of the star , the impact should create an equatorial bulge on the surface of the star , and in a few cases , the gas may be expelled from the system . the conditions at the impact site will be turbulent , and the tangential component of the flow will propel the gas around the star well beyond the impact site on timescales of hours to days , until it reaches the location of the incoming gas stream . the image reconstruction technique of tomography can be implemented in two steps : ( 1 ) the acquisition of a set of views of the object at directions around the object over the 360@xmath54 range of angles ; these views are called slices or projections that can be represented mathematically by the radon transform @xcite . ( 2 ) the recovery of an image of the original object may be obtained through a summation process called back projection , accomplished by taking each image projection and returning it along the path from which it was acquired . the quality of the reconstruction depends on the number of views acquired and the angular distribution of these views . in general , back - projection tomography uses ( @xmath55)-dimensional projections to calculate an @xmath56-dimensional image of an object ( e.g. , construct a 3d image using 2d projections , like cat scans in medicine ) . moreover , the technique is robust in that it can be applied to different types of measurements , for example : x - rays ( medicine or archaeology ) , radar ( satellite aperture radar - sar ) , sonar ( oceanography ) , seismic waves ( geophysics ) , and spectra ( astronomy ) . tomography can be applied only if there exist different angular views of a system . for example , since the surface of venus is obscured by thick clouds and is opaque at optical wavelengths , the reconstruction of its surface was achieved at radio frequencies using radar data collected by the nasa magellan satellite which orbited the planet for four years from 1990 - 1994 . eclipsing binaries and rotating stars also provide us with changing views of the system and the reconstruction requires the use of spectra . in these latter cases , the gas motions detected through doppler shifts of spectral lines are used to provide an image of the gas flows in velocity coordinates , so the technique is termed _ doppler tomography_. in two - dimensional back - projection doppler tomography , the radon transform of the function , @xmath57 , is a set of projections , @xmath58 representing the _ line profile _ , with doppler shift , @xmath59 , at each orbital phase , @xmath60 . @xmath61 the reconstruction creates the 2d image @xmath57 directly from the 1d doppler shifts of the line profile by inverting the equation for the radon transform @xcite . @xmath62 where @xmath63 . this filtered back - projection formulation can be extended to the next dimension to calculate @xmath64 by changing @xmath58 to @xmath65 , where @xmath66 is the orbital inclination , setting @xmath67 , and integrating over @xmath68 . the 3d case reverts to the 2d case for @xmath69 , corresponding to total eclipses , compared to partial eclipses for lower inclinations . moreover , while orbital inclinations close to @xmath70 are preferred for 2d reconstructions , inclinations close to @xmath71 are optimal for 3d reconstructions @xcite . the application of doppler tomography requires one main assumption , namely that the line profiles are broadened only by doppler motions ; which is a good first order approximation , although turbulent motions may contribute ( see @xcite and the extensive discussion in @xcite ) . in addition , we need numerous spectra distributed around the orbit of the binary , with high wavelength resolution and adequate coverage in orbital phase . the technique has been applied to spectra that are dominated by emission , but the generality of the equations suggests that the method may also be applied with prudence to absorption spectra . in practice , tomography is a summation process and the velocity images are calculated by summing the normalized intensities of the spectra for all velocity pixels after they have been converted to the separate @xmath72 and @xmath73 components , and then sweeping through all phases around the orbit . in the calculation , the doppler shifts typically range from -800 to + 800 km relative to the rest wavelength of the line ( e.g. h@xmath0 ) . the images are constructed in the velocity domain and can only be converted to the cartesian domain if the separate velocity fields of all accretion structures ( e.g. , gas stream , accretion disk , etc . ) are known . figure [ fig3 ] illustrates the full orbit data collection procedure in the case of u crb , with 294 phases over several epochs . a representative set of these spectra for epoch 1994 were stacked in order of orbital phase from 0.0 to 1.0 in preparation for the summation calculation . since the spectra of the direct - impact systems are usually dominated by absorption near primary eclipse ( @xmath74 ) , these eclipse spectra were typically excluded from the analysis for these systems . the tomography calculations were implemented using both fortran and c++ codes with matlab as the platform for creating the plots of the images . the speed of the code depends sensitively on the number of orbital phases multiplied by the velocity resolution of the spectrum in pixels . doppler tomography has been applied to eclipsing systems which contain white dwarfs and neutron stars ( e.g. , cataclysmic variables or cvs , nova - like systems , and soft x - ray binaries ) , as well as non - compact main sequence stars ( e.g. , algol binaries ) . the main differences between these systems are : ( 1 ) the accretion structures in the compact systems are bright relative to the stars , while the structures in the algols are faint relative to the luminous main sequence mass - gaining star ; and ( 2 ) the large size of the mass gainer in the algols leads to the direct impact of the gas stream onto the stellar surface in the short - period algols , while this type of impact does not occur in the long - period algols or in compact systems . hence , composite accretion structures should form in the short - period algols compared to the classical accretion disks that have been detected in other systems . in all cases , the cool magnetically - active mass - losing star is expected to influence the resulting accretion structures . numerous time - resolved spectra of the target systems are required for the tomography calculation in order to recover a well - resolved image of the gas flows ( see figure [ fig3 ] ) . images are usually created from spectral lines in the optical regime since the expected gas temperatures are @xmath75k . however , since the h@xmath0 line is saturated in the spectra of cvs and x - ray binaries , relatively weaker lines have been used , e.g. , he i @xmath765015 , he ii @xmath764686 , h@xmath1 . many doppler tomograms have been created for cvs ( e.g. , @xcite ) , x - ray binaries ( e.g. , @xcite for smc x-1 and her x-1 ) , and the nova - like binary v3885 sgr @xcite . the first tomogram of a black - hole candidate , cyg x-1 , was made by @xcite . most of the direct - impact algol binaries have weak emission - line spectra , so the emission from non - photospheric gas can be enhanced by subtracting the composite spectrum of the stars from the observed spectrum to create a difference spectrum ( e.g. , @xcite ) . this subtraction process assumes that the circumstellar gas is optically thin , which is a good first order approximation . however , this assumption can be examined by comparing the tomography images derived separately from the observed and difference spectra ( see section 6 ) . for the direct - impact systems , tomograms have been made predominantly from the h@xmath0 line , and other images have been constructed from the h@xmath1 , he i @xmath766678 , and si ii @xmath766371 lines ( e.g. , cx dra ; @xcite ) , and also from the si iv @xmath761063,1073 doublet and si iv @xmath761394 line in the ultraviolet @xcite . table [ tab3 ] provides a summary of spectra of algol binaries ( epochs 1976 to 2003 ) that were processed into images , and it lists the number of orbits , n(orbits ) , the number of phases , n(phi ) , and the velocity resolution of the spectra , n(vel ) . these data were derived primarily from the extensive database of direct - impact systems described by @xcite for which the spectra have high signal - to - noise ( typically s / n @xmath77 ) , high wavelength resolution ( see table [ tab3 ] ) , and coverage of the entire orbit of each binary ( see figure [ fig3 ] ) . other spectra were included for three systems : @xmath1 per ( epochs 1976 - 1977 ; @xcite ) , tt hya ( epochs 1985 - 2001 ; @xcite ) , and au mon ( epochs 1984 - 2003 ; @xcite ) . the best images are obtained from data with the highest number of phases distributed around the orbit ( i.e. , high n(phi ) ) , and from spectra with the highest resolutions ( i.e. , high n(vel ) ) . moreover , datasets collected over the smallest number of orbits are preferred since they show the state of the gas over fewer dynamical timescales ( see table [ tab2 ] ) . llcccc ' '' '' system & epoch & n(nights ) & n(orbits ) & n(phi ) & n(vel ) + [ 0.5ex ] ' '' '' rz cas & 1994 & 7 & 5.9 & 28 & 262 + @xmath18 lib & 1993 & 7 & 3.0 & 51 & 275 + rw tau & 1994 & 7 & 2.5 & 30 & 275 + @xmath1 per & 1976 - 1977 & 23 & 156.9 & 59 & 728 + @xmath1 per & 1992 & 7 & 5.6 & 135 & 503 + @xmath1 per & 1994 & 7 & 2.4 & 36 & 275 + tx uma & 1992 & 15 & 4.9 & 81 & 468 + tx uma & 1993 & 29 & 27.4 & 114 & 468 + tx uma & 1994 & 14 & 62.0 & 27 & 275 + u sge & 1993 & 12 & 9.5 & 106 & 256 + u sge & 1994 & 7 & 2.1 & 48 & 275 + s equ & 1993 & 12 & 9.3 & 57 & 275 + u crb & 1993 & 41 & 24.3 & 161 & 459 + u crb & 1994 & 14 & 55.0 & 47 & 275 + rs vul & 1993 & 12 & 7.1 & 81 & 275 + sw cyg & 1994 & 7 & 1.5 & 36 & 275 + cx dra & 1994 & 7 & 1.0 & 54 & 275 + tt hya & 1994 - 1997 & 22 & 158.3 & 45 & 446 + au mon & 1988 - 2003 & 64 & 458.2 & 83 & 217 + v711 tau & 1994 & 7 & 2.5 & 27 & 209 + [ tab3 ] the 2d velocity images derived from back - projection doppler tomography of the h@xmath0 line are shown in four parts in figure 4 : [ fig4a - f ] , [ fig4g - l ] , [ fig4m - r ] , and [ fig4s - t ] , for thirteen systems listed in order of orbital period from @xmath78 days ( rz cas ) to @xmath79 days ( au mon ) . in addition , the images of four systems are shown at multiple ( 2 - 3 ) epochs ( @xmath1 per , tx uma , u sge , u crb ) . these figures show the cartesian representation of each binary , the trailed spectrogram , color and contour tomograms and the corresponding color bar for the images . a reference template is overlaid on all images to show the expected locations of the stars , the gas stream gravitational trajectory , and the locus of a keplerian accretion disk ; these were illustrated in figure [ fig2 ] and described in section 2 . in addition , since the mass - losing star in many interacting binaries is a cool magnetically - active star , we used the tomogram of the detached rs cvn binary v711 tau ( hr 1099 ) to illustrate the location of h@xmath0 emission from a magnetically active star ( see table [ tab3 ] ) . v711 tau consists of a g star with a k star that nearly fills its roche lobe ; and the tomogram shown in figure 4 ( t ) illustrates that most of the emission arises from the k - star component , which is similar to the mass loser in the direct - impact systems . the tomograms of algol binaries in figure 4 reveal several distinct sources of h@xmath0 emission which result from a combination of gravitational forces and magnetic activity . the main sources in the image are in accordance with the expected behavior of the gas flow from the l1 point under the influence of gravity and coriolis forces , as described in section 2 . in addition , nearly all systems display evidence of magnetic activity associated with the mass - losing star . this latter result is surprising given that the mass - losing star contributes a small fraction of the total luminosity of the binary ( only 5 - 15% , depending on wavelength ) , and it was considered to be relatively unimportant in the mass transfer process , besides being the source of the gas flow through the l1 point . the sources of h@xmath0 emission seen in the 2d images are described below : ( 1 ) _ a source of magnetic activity associated with the cool mass - losing star _ , including emission from the chromosphere , prominences , flares , and coronal mass ejections . the tomogram of v711 tau illustrated that the emission from the cool star has the same velocity as that star . this type of emission was detected in almost all systems , including rz cas , @xmath18 lib , @xmath1 per , tx uma , u sge , s equ , u crb , rs vul , cx dra , and au mon ; the exceptions were rw tau , sw cyg , and tt hya . the byproducts of magnetic activity revealed in the velocity images of so many systems should not be surprising since the mass - losing stars have spectral types from f0 to k2 and there is independent evidence of magnetic activity in these systems @xcite . \(2 ) _ a gas stream along the predicted gravitational trajectory _ ( e.g. , rz cas , @xmath18 lib , rw tau , @xmath1 per , tx uma , u sge , s equ , u crb , rs vul , cx dra , au mon ) . in most of these systems , the gas stream flows along or close to the predicted gravitational trajectory . the exception is tx uma in which the stream path appears to be deflected during epochs 1992 and 1994 but not during epoch 1993 . it is interesting also that the emission associated with the cool star is weaker during epoch 1993 relative to the other epochs . tx uma is known to be magnetically active @xcite , and therefore , it should not be surprising that enhanced magnetic activity in the form of prominences or coronal mass ejections could deflect the gas stream flow away from the expected gravitational trajectory . with increased room between the stars , the gas stream can make its way around the mass - gaining star to form a transient or stable accretion disk ( e.g. , tt hya and au mon ) . \(3 ) _ a star - stream impact region _ where the gas stream strikes the surface of the mass - gaining star . the expected location of this shock region was discussed in section 2 and shown in figure [ fig2 ] for different values of the size of the mass gainer . in most cases , the shock should occur close to the intersection of the gas stream path and the circle representing the keplerian velocity of the mass gainer . extended emission at this location is visible in the velocity images for u sge ( epoch 1993 ) , u crb ( epoch 1993 ) , and s equ . \(4 ) _ a bulge of extended absorption or emission around the mass - gaining star _ produced by the impact of the high velocity stream onto the slowly rotating photosphere ( e.g. , rz cas , @xmath18 lib , tx uma ( epochs 1992 and 1994 ) , u sge ( both epochs ) , s equ , u crb ( both epochs ) , and rs vul ) . this is the same region as the circumprimary bulge detected via photometry as a hot band around the equator of the mass gainer ( e.g.,rz cas @xcite ; u sge @xcite ) . the impact increases the rotational velocity of the star beyond the synchronous value @xcite . spectroscopic studies of tx uma by @xcite and @xcite also found that the mass - gainer in this binary rotates @xmath80 times the synchronous velocity as a result of the impact . so , the images have confirmed the photometric and spectroscopic evidence of the circumprimary bulge . \(5 ) _ a transient or permanent accretion disk _ around the mass - gaining star . accretion disks were found in all systems with @xmath81 ( e.g. , sw cyg , cx dra , tt hya , au mon ) . disks were also found in u sge ( epoch 1994 ) and u crb ( epoch 1994 ) , however the gas distribution in these two systems can be transformed from a disk - like state to a stream - like state within a few orbital cycles . the evidence of this variability can be seen in the trailed spectrograms and in the individual spectra through the change from a double - peaked profile to a single - peaked profile . \(6 ) _ an absorption zone _ where the gas flows are seen in absorption ( e.g. , u sge , s equ , u crb , rs vul , sw cyg , cx dra ) . a tomogram created from the siiv @xmath761394 line of u sge is seen in figure [ fig7 ] and shows that the main source of emission in the ultraviolet overlaps with the absorption zone , which in turn represents the locus of hotter gas ( @xmath82k ) . \(7 ) _ a disk - stream impact region _ where the circling gas strikes the incoming stream and is slowed by the impact ( e.g. , rw tau , @xmath1 per , tx uma , u sge , s equ , u crb , rs vul , cx dra ) . the result is a region of emission nearly along the line of centers with velocities lower than that of a keplerian disk , i.e. , within the locus of the disk in the velocity image . therefore , in the systems with @xmath83 , except for sw cyg , after the gas stream strikes the star it has enough angular momentum to flow around the star until it comes in contact with the incoming stream . in these systems , the density of the gas stream must be higher than the density of the gas that has circled the mass gainer . this region has been called the localized region ( e.g. , @xmath1 per , @xcite ; rw tau , @xcite ; other systems , @xcite ) . in wider binaries , the gas stream makes a more grazing impact and forms a more extensive flow in an accretion disk , so it is more difficult to determine where the disk intersects with the incoming stream in the velocity image . a pictorial representation of the variety of emission sources is shown in figure [ fig5 ] . this figure illustrates that the gas stream emission was sufficiently bright to be detected in the images of most systems . figure [ fig6 ] summarizes the transition in the line profile at quadrature , the emission sources in the velocity image , and the trailed spectrogram , with increasing binary separation ( semi - major axis , @xmath10 ) . the progression from the upper to lower parts of this figure illustrate the transformation from the _ stream state _ to the _ disk state_. in essence , the closest binaries ( e.g. , @xmath18 lib ) typically display only emission associated with the mass - losing star , and hence evidence of magnetic activity , because there is insufficient room for the gas stream to make its way around the mass gainer . if there is some extra room between the stars , the gas stream can progress farther along , and we find evidence of extended gas streams along the projected gravitational path in several systems ( e.g. , s equ , rs vul , u sge , u crb ) . in the widest systems ( e.g. , tt hya , au mon ) , there is sufficient room to form a keplerian accretion disk . over the range of systems , the line profile at quadrature shows a progression from nearly single - peaked profiles indicative of gas streams in the closest systems ( smaller @xmath10 ) , to distinct double - peaked profiles indicative of accretion disks in the wider systems ( larger @xmath10 ) . potential variability in the images was examined in four systems for which data were collected at multiple epochs ( i.e. , @xmath1 per , tx uma , u crb , u sge ) , although data were also collected over an extensive range of epochs for cx dra , tt hya , and au mon . the images of @xmath1 per suggest that this system was relatively stable over many orbital cycles ( epochs 1976 to 1994 ) , however there was a slight change in the direction of the gas stream in tx uma between epochs 1992 and 1994 , suggesting the influence of magnetic activity on the cool star in that system ( see earlier discussion in section 4.1 ) . in addition , u crb and u sge are unusual because they display both stream - like and disk - like states at different epochs ( see figure [ fig6 ] and section 4.1 ) , and a closer examination of their spectra suggest that these binaries may alternate regularly between the two states . consequently , u crb and u sge have been designated as _ alternating systems _ , and the source of the variability may be linked to magnetic activity on the mass - losing star ( see @xcite and the discussion in section 6 ) . the images of accretion disks in the wide grazing - impact systems are similar to those of the cataclysmic variables ( e.g. , @xcite ) . hence , stable accretion disks are found in binaries containing both compact and main sequence mass gainers . multiwavelength spectra of algol binaries have been collected at wavelengths 4800 @xmath25 6700 , including the @xmath84 , @xmath85 , he i @xmath766678 , and si ii @xmath766371 lines . in addition , ultraviolet spectra of 17 algol binaries can be found in the iue archives from epochs 1978 to 1996 . the first ultraviolet image of u sge was derived from the si iv @xmath761394 doublet ( epoch 1983 - 1994 ) after the contribution of the stars was extracted from the spectra . figure [ fig7](a ) shows that the si iv emission overlaps with the absorption zone identified in the h@xmath0 image for epoch 1993 , and can be interpreted as the locus of gas that was heated beyond the @xmath86k regime associated with h@xmath0 to temperatures of @xmath87k in the uv region @xcite . figure [ fig7](b ) displays the multiwavelength images of cx dra , and demonstrates that the he i @xmath766678 and si ii @xmath766371 emission arise from an extended accretion disk , while the h@xmath0 and h@xmath1 emission arise primarily from a compact region where the gas stream strikes the outer edge of the disk @xcite . these results suggest that a more comprehensive understanding of the gas flows can be obtained when images are constructed for different temperature regimes . the attractiveness of tomography is that it is a robust technique ; however it is useful to discover how various criteria influence the resulting 2d image constructed from the back - projection technique . it should be noted that the tomograms shown in figure 4 were computed without any filtering , smoothing , denoising or other regularization of the images . the only additional processing was to interpolate the images in orbital phase to reduce the radial spoke - like streaks produced by the back - projection procedure when there are gaps in orbital phase . in spite of the minimal processing , the amount of detail manifested in the images is truly remarkable . in this section , we examine the effects of ( 1 ) orbital inclination , ( 2 ) systemic velocity , ( 3 ) orbital phase coverage , ( 4 ) velocity resolution , and ( 5 ) shadowing on the tomography results . the effects of adopting different values of orbital inclination are shown in figure [ fig8 ] for a direct - impact binary ( e.g. , u crb ) and a wider binary ( e.g. , tt hya ) . the main result is that the image size decreases as the inclination increases from @xmath88 to @xmath70 . moreover , the measured inclinations of u crb ( @xmath89 ) and tt hya ( @xmath90 ) lead to images that are consistent with the template of predicted locations for the gas stream and disk in the velocity frame . figure [ fig8 ] illustrates the effect of adopting a systemic velocity for the binary , @xmath13 , that is inconsistent with the measured value ; the result is a distortion of the image . the example cases of u crb and tt hya have systemic velocities of @xmath91 km and @xmath92 km , respectively , and the corresponding 2d images are therefore similar to the image for @xmath93 km . figure [ fig8 ] illustrates the effect of phase coverage on the resulting 2d image using two subsets the u crb and tt hya data . the images were made from three sets of data : a subset of the full dataset ( left frame ) , after this subset was interpolated to create a grid of 100 evenly - spaced phases ( middle frame ) , and the full dataset ( right frame ) . insufficient orbital phase coverage creates radial spoke - like streaks across the diagram . however , an interpolation in orbital phase adequately represents the outcome when the full dataset is used to construct the image . the resolution of the h@xmath0 line was typically 0.17 / pixel ( n(vel ) = 275 ) and was as fine as 0.09 / pixel ( n(vel ) = 503 ) . since the wavelength or velocity resolution of the spectra is already very high , an increase in the resolution does not result in any significant improvement in the resulting 2d image . however , poor wavelength resolution in the spectrum would lead to a coarser image . the effect of shadowing results when the passage of light through a medium is obstructed because of the opacity of the medium . hence , the view from one direction may not be identical to that from the opposite direction @xmath94 away . a similar effect called an acoustic shadow in radar tomography occurs when radio waves fail to propagate due to some topological obstruction , or in medical tomography because of the attenuation of light by denser parts of the medium @xcite . figure [ fig8 ] shows images produced from data at phases @xmath95 ( left ) , @xmath96 ( middle ) , and the complete data set ( right ) to check for the effect of shadowing in the example cases of u crb and tt hya . the rough resemblance between the images derived from half the orbit and that from the full orbit suggest that shadowing , if it exists , does not play a significant factor in the resulting tomograms . instead , the uneven sampling of the data for each half orbit and the number of spectra in each dataset have larger effects on the resulting image . an alternative way to examine the opacity of the medium is to compare the tomograms derived from the observed and difference spectra , since the subtraction of the stellar spectra from the observed spectrum assumes that the circumstellar gas is optically thin . in the case of cx dra , the observed and difference h@xmath0 images shown in figure [ fig7](b ) are not similar in appearance , which suggests that the circumstellar gas in cx dra was not optically thin . however , there is evidence from similar comparisons that the circumstellar gas is optically thin in tt hya and au mon ( see section 6.3 ) , in agreement with the results of the shadowing test for tt hya ( figure [ fig8 ] ) . the 2d velocity images have provided visual confirmation of the existence of both gravitational and magnetic structures in the direct - impact and grazing - impact algol binaries . in addition , the physical properties of the gas flows can be derived from the data and images using ( 1 ) an analysis of the s - wave patterns in the trailed spectrograms , ( 2 ) the creation of simulated tomograms derived from hydrodynamic simulations , ( 3 ) the computation of synthetic spectra of the accretion structures , and ( 4 ) a systematic extraction of the separate accretion structures using synthetic spectra with tomography . these procedures require substantial effort and we present here some examples of the results that have been achieved so far . these results have enhanced our interpretation and understanding of the gas flows in both the velocity and cartesian frames . the trailed spectra of the direct - impact binaries display one or two sinusoidal patterns called _ s - waves _ which represent the orbital radial velocity variations of the emission peaks ( see figures 4 and [ fig6 ] ) . these patterns can be used to identify the locations of the most intense sources of emission in the velocity image . figure [ fig9 ] shows that the h@xmath0 emission peaks roughly follow the velocity curves of the stars , sometimes with a smaller amplitude , in direct - impact systems when they are in the stream state ( e.g. , @xmath18 lib , rw tau , @xmath1 per , u sge , s equ , u crb , rs vul ) . however , the emission peaks appear to be independent of the velocity curves when the gas is in a disk state ( e.g. , tt hya ) . in the direct - impact binaries , the larger variation is associated with the mass - losing star so the ratio of velocity semi - amplitudes of the mass loser and the s - wave source should be the same as the ratio of their distances from the center of mass . as a result , this s - wave source is located at a distance @xmath97 from the c.m . of the binary , where @xmath98 is the velocity semi - amplitude of the mass loser , @xmath99 is the distance of the mass loser from the c.m . of the binary , and @xmath100 is the velocity semi - amplitude of the s - wave . a similar argument follows for the other s - wave associated with the mass - gaining star . s - wave analyses for @xmath1 per ( epochs 1976 - 1977 ) and cx dra ( epochs 1975 - 1998 ) have identified the location of the most intense sources of h@xmath0 emission in these binaries . figure [ fig10 ] illustrates the dominant s - wave solution in the velocity frame ( large red dot ; left frame ) , the s - wave superimposed on the trailed spectrum ( middle frame ) , and the predicted source of the dominant s - wave in the cartesian frame ( large red dot ; right frame ) for these binaries . for the case of @xmath1 per ( epoch 1976 - 1977 ) , the source of the dominant s - wave was located nearly along the line of centers between the stars , with coordinates @xmath101 km relative to the center of mass of the binary @xcite . the corresponding source in cartesian coordinates was found at a distance @xmath102 compared to the location of the @xmath103 point at @xmath104 from the c.m . of the binary , for @xmath105 km and @xmath106 . this source is consistent with the location of the localized region in the cartesian map predicted by @xcite . the analysis of cx dra ( epochs 1975 - 1998 ) identified the dominant s - wave source at @xmath107 km relative to the c.m . of the binary @xcite . the derived value of @xmath100 = @xmath108 km led to @xmath109 from the c.m . of the binary compared to @xmath110 = @xmath111 , for @xmath112 km and @xmath113 . figure [ fig10 ] shows the position of the s - wave source relative to the predicted locus of the accretion disk derived from the hei @xmath766678 line ( shown in figure [ fig7 ] ) . this s - wave solution is consistent with the h@xmath0 velocity image and suggests that the h@xmath0 emission along the gas stream path was interrupted by the presence of a slightly hotter accretion disk . hence , this source represents the location of the _ stream - disk impact site or `` hot spot '' _ and should be added to the list of seven emission sources described in section 4.1 . similar analyses can be performed for other systems . hydrodynamic simulations are useful in the study of accretion flows because they illustrate the evolution of the flow under a variety of initial conditions , and can be used to constrain the properties of the gas flows derived from spectroscopic analysis . these simulations can be used to model the h@xmath0 emissivity and to study the evolution of the flow simultaneously in spatial and velocity dimensions . this was achieved by @xcite for two algol binaries : @xmath1 per and tt hya , which span the range from direct - impact to disk - state systems . they used the vh1 ( virginia hydrodynamics-1 ) 2d hydrodynamics code , which incorporates the effects of gravity and coriolis forces , and assumes optically - thin radiative cooling of the gas ; no magnetic effects were included in these simulations . for each binary , a _ simulated velocity tomogram _ can be constructed by summing the velocity images derived from the simulations over a range of epochs , @xmath114 , where @xmath5 is the orbital period , for comparison with the doppler tomograms computed from observed spectra . figure [ fig11 ] shows the simulated velocity tomograms of @xmath1 per and tt hya for two values of the initial density of the gas stream at the l1 point : @xmath115 @xmath116 and @xmath117 @xmath116 . in the case of @xmath1 per , only the gas stream would be observed at the lower density because the remaining gas is nearly a factor of @xmath118 fainter than the gas stream , while both the gas stream and the disk would be observed at the higher density because they have comparable intensities . similarly , the gas stream and disk are substantially brighter in tt hya at the higher density . the similarity between the h@xmath0 doppler tomograms of the alternating systems at different epochs ( specifically u crb and u sge ) and the simulated tomograms of @xmath1 per at the two densities suggest that the variable and alternating behavior of the gas flows may be associated with changes in the density of the gas stream , and hence with changes in the mass transfer rate @xcite . moreover , since the change between stream state and disk state in u crb and u sge can occur within days , which is much shorter than the kelvin - helmholtz timescales listed in table [ tab2 ] , @xcite proposed that the alternating behavior may be triggered by flaring activity on the mass - losing star and could be associated with magnetic activity on the cool star . the significant flaring cycles for @xmath1 per ( a direct - impact system ) and for v711 tau and ux ari ( two rs cvn binaries ) at radio frequencies of 2.3 ghz and 8.3 ghz are given in table [ tab4 ] based on a 5.6-year nearly - continuous survey of radio flares @xcite . these results suggest that flares could influence the mass transfer process regularly , as often as every 8 - 42 orbital cycles . this survey demonstrated that radio flares from @xmath1 per occurred with a predictable cycle of 49 days , which could result in sporadic mass ejections through the l1 point on that timescale . an alternative explanation is that the change in the image could result from the interaction between the incoming gas stream and the gas that has circled the star since the variability timescale of days is comparable to the circularization timescale @xmath119 ( see table [ tab2 ] ) . in the case of tt hya , the hydrodynamic simulations produced an accretion disk that was asymmetric in both cartesian space and velocity coordinates ( figure [ fig11 ] ) . this important result went unnoticed for nearly 10 years until a synthetic spectrum of a symmetric accretion disk was modeled , extracted from the observed spectrum , and used to construct a difference doppler tomogram of tt hya @xcite . this procedure revealed a small arc of emission within the locus of the accretion disk in the doppler tomogram exactly at the location of the disk asymmetry seen in the simulated tomogram ( see figure [ fig12 ] and section 6.3 ) . in summary , the construction of simulated tomograms from hydrodynamic calculations under controlled conditions have improved our understanding of the geometry and dynamics of the accretion flows in both velocity and cartesian dimensions . lcll ' '' '' system & @xmath120 ( d ) & flaring cycles , @xmath5 & @xmath121 + [ 0.5ex ] ' '' '' @xmath1 per & 2.87 & 48.9 @xmath3 1.7 & 17.1 + v711 tau & 2.84 & 120.7 @xmath3 3.4 , 80.8 @xmath3 2.5 & 42.5 , 28.5 + ux ari & 2.33 & 141.4 @xmath3 4.5 , 52.6 @xmath3 0.7 & 22.0 , 8.2 + [ tab4 ] the computation of synthetic spectra provides us with a tool that can be used to extract the separate sources of emission from the composite tomograms of the direct - impact and disk - state algol systems . the shellspec code can create composite spectra and light curves of interacting binaries which contain moving transparent or non - transparent 3d structures such as a disk , stream , spot , jet , or shell , in addition to the stars . this code solves the simple radiative transfer along the line of sight in lte , in an optional optically thin 3d moving medium ( see @xcite for details ) . in addition , @xcite used the code to examine the effects of various free parameters and how double - peaked emission from an accretion disk is formed . they found that the overall strength of the emission is influenced mainly by the density and temperature of the disk , while the position and separation of the emission peaks are influenced by the outer disk radius , the inclination , and the radial density profile . furthermore , the depth of the central absorption in the profile is very sensitive to the temperature , inclination , geometry , and dynamics of the disk . the shellspec code was used to sequentially model the composite synthetic spectra of the ( i ) stars , ( ii ) stars and disk , ( iii ) stars , disk , and gas stream , or ( iv ) stars , disk , gas stream , and spot ( e.g. , localized region ) . these calculations require the availability of optimal values for the orbital elements and stellar properties , and well as initial estimates of the temperature , mass density and electron number density of the disk . the estimated density of the stream at the l1 point was chosen to improve the fit between the observed and synthetic emission at the quadratures of the orbit and during the eclipses . the density was allowed to vary along the stream to satisfy the continuity equation , and the electron number densities in the disk and stream were calculated from the temperature , density , and chemical composition using the newton - raphson linearization method . the resulting composite synthetic spectrum provides a good comparison with the observed h@xmath0 spectra for tt hya @xcite and au mon @xcite . the comparison between the observed and synthetic spectra can provide us with the physical properties of the accretion structures . for tt hya , @xcite adopted an initial disk temperature of @xmath122 k and found that the disk extends up to about 10 solar radii and has a vertical thickness comparable to the diameter of the primary star , with mass density of @xmath2@xmath123 g@xmath116 and electron number density of @xmath2@xmath124 @xmath116 . the initial gas stream temperature was set to @xmath125 k , corresponding to the peak in h@xmath0 emissivity , and the density of the gas stream at the @xmath126 point was estimated to be @xmath127g@xmath116 . finally , a mass transfer rate of @xmath128 was derived from the velocity , density , and cross - section of the gas stream above and below the disk . a similar analysis for au mon revealed a mass transfer rate of @xmath129 @xcite . the synthetic spectra can also be used in combination with tomography to demonstrate visually that the separate sources of emission have been adequately modeled . this procedure has been accomplished for two disk - state systems , tt hya and au mon , and the results are displayed in figure [ fig12 ] . the refined properties derived from the comparison between the observed and synthetic spectra were used to generate composite synthetic spectra of the disk , stream , and other features for input to the tomography code . the sequential modeling and extraction of the spectral contributions of the stars and various accretion structures resulted in the identification of an asymmetric accretion disk in both tt hya and au mon @xcite . the residual emission in the tt hya image ( figure [ fig12]d ; upper frames ) is located within the locus of the accretion disk and it represents the asymmetric portion of the disk that was not included in the synthetic spectrum based on a symmetric disk ; hence the accretion disk in tt hya is asymmetric . an asymmetric disk and gas moving at sub - keplerian velocities were also discovered in au mon from the tomograms ( see figure [ fig12]d ; lower frames ) . a localized region was also detected in the au mon image ; this emission was modeled as a spot " and removed from the image ( see figure [ fig12]e ; lower frames ) ; the density of this region was @xmath130 g@xmath116 , which is comparable to that of the gas stream . in summary , the shellspec code has proven to be an effective tool in the modeling of various accretion structures and for the calculation of critical parameters , such as the mass transfer rate , needed to compute accurate stellar evolution models . moreover , the use of the synthetic spectra with tomography has led to the isolation and discovery of asymmetric accretion disks in two disk - state systems which have stable accretion disks . moreover , this analysis has led to the successful modeling and extraction of the localized region from a composite spectrum , in the case of au mon . in this paper , we have derived the behavior of the accretion flows in a range of direct - impact and disk - state algol binaries with orbital periods from 1.2 to 11.1 days . these systems undergo mass transfer from a magnetically - active star onto a normal , non - magnetic main sequence star . we have applied the technique of tomography and a variety of other procedures , both observational and theoretical , to derive a better understanding of the gravitational and magnetic phenomena associated with the mass transfer process in interacting binary stars . \(1 ) the @xmath4 diagram was used as a predictive tool to determine which systems will undergo direct - impact accretion and which will form an accretion disk ( figure [ fig1 ] ) . in order to understand the predicted dynamical behavior of the gravitational flow along the gas stream , several key dynamical parameters were calculated , namely the thermal , dynamical , and circularization timescales which were then compared with the synchronous velocity , escape velocity for the mass gainer , and the velocity of the gas stream at the impact site . if only gravitational and coriolis forces are considered , the gas flow from the l1 point will make direct impact with the surface of the mass gainer in the short - period systems , and the impact can create an equatorial bulge on the surface of this star because the velocity of the flow is typically much higher than the synchronous velocity of the star . the gas at the impact site will be turbulent , and the tangential component of the flow will propel the gas around the star well beyond the impact site on timescales of hours to days . the predicted behavior of the gas at the shock site was calculated for both the cartesian and velocity frames for comparison with the doppler tomograms ( figure [ fig2 ] ) . \(2 ) the image reconstruction technique of back - projection doppler tomography was described mathematically , along with its advantages and constraints . in essence , this technique creates velocity images by summing the normalized intensities of the spectra at all velocities across the spectrum and for all viewing positions around the orbit . this procedure requires the availability of numerous spectra with high spectral resolution and high resolution in orbital phase ( figure [ fig3 ] ) . the 2d images shown in figure 4 were computed without any filtering , smoothing , denoising or other regularization procedures , except for interpolation in orbital phase . \(3 ) for the application of tomography , we analyzed over 1300 time - resolved spectra of 13 algol binaries collected at multiple wavelengths and multiple epochs ( figure 4 ) . most of the tomograms were constructed from h@xmath0 spectra , and we included images derived from the h@xmath1 , he i @xmath766678 , and si ii @xmath766371 lines in the optical region , as well as the si iv @xmath761394 line in the ultraviolet region . the velocity images produced via tomography display seven distinct sources of emission that provide evidence of both gravitational and magnetic effects ( figure [ fig5 ] ) : ( i ) a source of magnetic activity associated with the cool mass - losing star , including emission from the chromosphere , prominences , flares , and coronal mass ejections ; ( ii ) a gas stream along the predicted gravitational trajectory , which was identified in the images of nearly all of the systems in our survey ; ( iii ) a star - stream impact region where the gas stream strikes the surface of the mass - gaining star ; ( iv ) a bulge of extended absorption or emission around the mass - gaining star produced by the impact of the high velocity stream onto the slowly rotating photosphere ; ( v ) a transient or permanent accretion disk around the mass - gaining star ; ( vi ) an absorption zone which may be the locus of hotter gas beyond the @xmath86 k regime associated with h@xmath0 emission ; and ( vii ) a disk - stream impact region where the circling gas strikes the incoming stream and is slowed by the impact to sub - keplerian velocities . in addition , ( viii ) a hot spot where the gas stream strikes an already well - formed disk was found in one system by means of s - wave analysis . the images , trailed spectrograms , and representative spectra at quadrature provide us with a consistent progression from the stream state to the disk state when the binaries are displayed in order of increasing orbital separation ( figure [ fig6 ] ) . this progression reveals the existence of the alternating systems which have been found in both the stream state and the disk state at different epochs . in addition , the multiwavelength spectra provide us with views of the gas flows at different temperatures ( e.g. , h@xmath0 vs. he i spectra ; figure [ fig7 ] ) . \(4 ) we examined the influence of various system properties on the resulting 2d velocity images ( see figure [ fig8 ] ) . orbital inclination causes the image size to decrease as the inclination increases from @xmath88 to @xmath70 . the adoption of an inaccurate systemic velocity results in the distortion of the image . insufficient orbital phase coverage creates radial spoke - like streaks across the diagram which can be rectified if the data are interpolated in orbital phase . inadequate velocity resolution across the spectrum can also be addressed by interpolation , but the observed spectra used in the analysis already have high resolution and do not require any interpolation . finally , shadowing , or the presence of partially opaque gas , will influence the passage of light through a medium and this effect can be noticed if we view the binary from opposite directions . this latter effect was not detected in u crb or tt hya ( see figure [ fig8 ] ) . however , non - transparent gas was detected through a comparison between the observed and difference spectra of cx dra . \(5 ) finally , we summarized four methods that have been used to extract the physical properties and other characteristics of the gas flows directly from the velocity images : ( a ) s - wave analysis ; ( b ) creation of simulated velocity tomograms from hydrodynamic simulations ; c ) sequential extraction of the separate emission sources using synthetic spectra of the disk , gas stream , and other sources of emission such as the localized region ; and ( d ) using synthetic spectra with tomography to confirm that the separate accretion structures have been modeled accurately . \(a ) a comparison of the trailed spectrograms with the velocity curves of the stars showed that there are compact structures within the gas flow that nearly follow the velocity curve of one of the stars ; the orbital variation of these s - waves can be used to derive the location of the source in both the velocity and cartesian frames . hence , the s - wave analysis provides a means by which the information in the velocity map can be translated onto the cartesian frame without knowledge of the composite velocity fields in the tomogram ( figures [ fig9 ] , [ fig10 ] ) . \(b ) the construction of simulated tomograms from hydrodynamic calculations under controlled conditions provide a tool for comparison with the doppler tomograms derived from the observations , and provide a direct connection between the geometry and dynamics of the accretion flows in both velocity and cartesian dimensions under these controlled conditions ( figure [ fig11 ] ) . the simulated tomograms permitted an interpretation of the alternating behavior of u crb and u sge as due to periodic changes in the density of the gas stream , and hence in the mass transfer rate . results from a radio flare survey suggest that the timescale of variability in radio flares from magnetically - active algol binaries is comparable to that of the alternating systems , and hence the alternating behavior may be triggered by magnetic activity on the cool mass - losing star . \(c ) the code has been used effectively to compute synthetic spectra in moving transparent or non - transparent 3d structures such as a disk , stream , spot , jet , or shell , in addition to the stars . it can be used to sequentially model the composite synthetic spectra of combinations of these structures ( e.g. , stars , disk , and gas stream ) for comparison with the observed spectra . as a result , critical physical parameters , including the mass transfer rate , of these structures have been obtained for tt hya and au mon with this code . these parameters are needed to compute accurate stellar evolution models . \(d ) in combination with tomography , the synthetic spectra have been used to visually confirm the quality of the computed model . this procedure has led to the discovery of asymmetric accretion disks in tt hya and au mon , as well as the isolation of a sub - keplerian velocity region and a localized region in au mon . the multi - tiered analysis performed in this work involved both theoretical tools and spectroscopic observations of algol binaries from the stream state to the disk state . this analysis has provided observational evidence of the effects of gravitational forces on the accretion flows in these interacting binaries . moreover , we have used images derived from 2d doppler tomography to identify associations with magnetic activity of the cool star , which is the source of the accretion flow . these associations are distinct from the effects of gravitational effects on the gas flows and demonstrate that magnetic effects can not be ignored in any of these interacting binary star systems . the evidence is in the form of ( i ) chromospheric emission with the same velocity as that of the mass - losing star arising from prominences , flares , and coronal mass ejections , and this emission was detected in 11 of the 13 systems studied ; ( ii ) detection of a gas stream deflected from the predicted gravitational trajectory ( e.g. , tt hya ) ; and ( iii ) alternating behavior between stream state and disk state possibly caused by flaring activity ( e.g. , u sge , u crb ) . the extension to 3d doppler tomography is beginning to provide additional evidence of the influence of magnetic activity on the gas flows beyond the central orbital plane of the binary . the spectra employed in this study were used to calculate 3d tomograms of three systems : u crb , rs vul , and @xmath1 per @xcite . these 3d images suggest that magnetic threading of the gas , the superhump phenomenon , may alter the direction and intensity of the gas flow in the cases of rs vul and @xmath1 per , and may be responsible for deflecting the gas stream away from the central plane to produce a tilted accretion disk in u crb @xcite . these magnetic effects were mimicked using 3d hydrodynamic simulations by boosting the magnitude of the gas stream velocity to mach 30 and by deflecting the initial gas stream at the l1 point by 45@xmath24 from the orbital plane in u crb and rs vul @xcite . these simulations concluded that a deflected stream is a viable mechanism for producing the strong out - of - plane flows seen in the 3d tomographic images . we thank the referee for helpful comments on the manuscript . this research involved undergraduate students ( jf , mc ) and a graduate student ( ac ) , and it was partially supported by national science foundation grants ast-0908440 and dms-1309808 , and by the institut fr angewandte mathematik , ruprecht - karls - universitt heidelberg , germany .

we study on the tripartite entanglement dynamics when each party is initially entangled with other parties , but they locally interact with their own markovian or non - markovian environment . first , we consider three ghz - type initial states , all of which have ghz symmetry provided that the parameters are chosen appropriately . however , this symmetry is broken due to the effect of environment . the corresponding-tangles , one of the tripartite entanglement measure , are analytically computed at arbitrary time . for markovian case while the tripartite entanglement for type i exhibits an entanglement sudden death , the dynamics for the remaining cases decays normally in time with the half - life rule . for non - markovian case the revival phenomenon of entanglement occurs after complete disappearance of entanglement . we also consider two w - type initial states . for both cases the-tangles are analytically derived . the revival phenomenon also occurs in this case . on the analytical ground the robustness or fragility issue against the effect of environment is examined for both ghz - type and w - type initial states .

1601.00273

entanglement@xcite is one of the important concepts from fundamental aspect of quantum mechanics and practical aspect of quantum information processing . as shown for last two decades it plays a crucial role in quantum teleportation@xcite , superdense coding@xcite , quantum cloning@xcite , and quantum cryptography@xcite . it is also quantum entanglement , which makes the quantum computer outperform the classical one@xcite . quantum mechanics is a physics , which is valid for ideally closed system . however , real physical systems inevitably interact with their surroundings . thus , it is important to study how the environment modifies the dynamics of given physical system . there are two different tools for describing the evolution of open quantum system : quantum operation formalism@xcite and master equation approach@xcite . both tools have their own merits . since it is known that quantum system loses quantum properties by contacting the environment@xcite , we expect that the degradation of entanglement occurs@xcite . sometimes entanglement exhibits an exponential decay in time by successive halves . sometimes , however , the entanglement sudden death ( esd ) occurs when the entangled multipartite quantum system is embedded in markovian environments@xcite . this means that the entanglement is completely disentangled at finite times . this esd phenomenon has been revealed experimentally@xcite . when the esd occurs , it is natural to ask where the lost entanglement goes . it was found that when the entanglement of given quantum system suddenly disappears , the reservoir entanglement suddenly appears , which is called entanglement sudden birth ( esb ) @xcite . since we do not consider the degrees of freedom for the environment , we do not examine the esb phenomenon in this paper . the dynamics of entanglement was also examined when the physical system is embedded in non - markovian environment@xcite . it has been shown that there is a revival of entanglement after a finite period of time of its complete disappearance . this is mainly due to the memory effect of the non - markovian environment . this phenomenon was shown in ref.@xcite by making use of the two qubit system and concurrence@xcite as a bipartite entanglement measure . subsequently , many works have been done to quantify the non - markovianity@xcite . in this paper we consider the entanglement dynamics when the qubit system interacts with the markovian or non - markovian environment . so far this issue was investigated by making use of the bipartite system . recently , the tripartite entanglement dynamics was also explored in ref.@xcite numerically . since entanglement is an important physical resource in the quantum information processing , it is important to control the entanglement dynamics when the environment is present . in order to control the entanglement it is crucial to derive the entanglement analytically in the entire range of time . for example , analytic derivation for the bipartite entanglement dynamics enables us to explore the entanglement invariants@xcite . it is also possible to discuss the robustness or fragility issue against the environment by exploiting the analytical results . thus , we will explore the tripartite entanglement dynamics in this paper on the analytical ground . for simplicity , we choose the physical setting , i.e. there is no interaction between qubit and each qubit interacts with its own reservoir . we will compute the entanglement at arbitrary time for three - types of initial greenberger - horne - zeilinger(ghz ) states@xcite and for two types of initial w - states@xcite in the presence of the markovian or non - markovian environment . typical tripartite entanglement measures are residual entanglement@xcite and @xmath0-tangle@xcite . for three - qubit pure state @xmath1 the residual entanglement @xmath2 becomes @xmath3 where @xmath4 thus , the residual entanglement of any three - qubit pure state can be computed by making use of eq . ( [ residual-1 ] ) . although the residual entanglement can detect the ghz - type entanglement , it can not detect the w - type entanglement : @xmath5 where @xmath6 for mixed states the residual entanglement is defined by a convex - roof method@xcite as follows : @xmath7 where the minimum is taken over all possible ensembles of pure states . the pure state ensemble corresponding to the minimum @xmath2 is called the optimal decomposition . it is in general difficult to derive the optimal decomposition for arbitrary mixed states . hence , analytic computation of the residual entanglement can be done for rare cases@xcite . furthermore , recently , three - tangle three - tangle and @xmath8 residual entanglement . ] @xmath9 of the whole ghz - symmetric states@xcite was explicitly computed@xcite . the @xmath0-tangle defined in ref.@xcite is easier for analytic computation than the residual entanglement ( or three tangle ) because it does not rely on the convex - roof method . the @xmath0-tangle is defined in terms of the global negativities @xcite . for a three - qubit state @xmath10 they are given by @xmath11 where @xmath12 , and the superscripts @xmath13 , @xmath14 , and @xmath15 represent the partial transposes of @xmath10 with respect to the qubits @xmath16 , @xmath17 , and @xmath18 respectively . then , the @xmath0-tangle is defined as @xmath19 where @xmath20 the remarkable property of the @xmath0-tangle is that it can detect not only the ghz - type entanglement but also the w - type entanglement : @xmath21 as commented earlier we will examine the tripartite entanglement dynamics of the three - qubit states in the presence of the markovian or non - markovian environment . we will adopt the @xmath0-tangle as an entanglement measure for analytic computation as much as possible . in section ii we consider how the three - qubit initial state is evolved when each qubit interacts with its own markovian or non - markovian environment@xcite . in section iii we explore the entanglement dynamics of three ghz - type initial states . the initial states are local unitary ( lu ) with each other . thus , their entanglement are the same initially . furthermore , if the parameters are appropriately chosen , they all have ghz - symmetry , i.e. they are invariant under ( i ) qubit permutation ( ii ) simultaneous three - qubit flips ( iii ) qubit rotations about the @xmath22-axis . however , this symmetry is broken due to the environment effect . as a result , their entanglement dynamics are different with each other . in section iv we examine the entanglement dynamics of two w - type initial states . they are also lu to each other . however , the dynamics is also different because of the environment effect . in section v a brief conclusion is given . we consider three - qubit system , each of which interacts only and independently with its local environment . we assume that the dynamics of single qubit is governed by hamiltonian @xmath23 where @xmath24 in eq . ( [ hamiltonian-2 ] ) @xmath25 is a transition frequency of the two - level system ( qubit ) , and @xmath26 are the raising and lowering operators . the index @xmath27 labels the different field modes of the reservoir with frequencies @xmath28 , creation and annihilation operators @xmath29 , @xmath30 , and coupling constants @xmath31 . in the interaction picture the dynamics is governed by the schrdinger equation @xmath32 where @xmath33 the hamiltonian ( [ hamitonian-1 ] ) represents one of few exactly solvable model@xcite . this means that the schrdinger equation ( [ schrodinger-1 ] ) can be formally solved if @xmath34 is given . then , the reduced state of the single qubit @xmath35 is given by@xcite @xmath36 where @xmath37 and @xmath38 denotes the partial trace over the environment . the function @xmath39 satisfies the differential equation @xmath40 and the correlation function @xmath41 is related to the spectral density @xmath42 of the reservoir by @xmath43.\ ] ] we choose @xmath42 as an effective spectral density of the damped jaynes - cummings model@xcite @xmath44 where the parameter @xmath45 defines the spectral width of the coupling , which is connected to the reservoir correlation time @xmath46 by the relation @xmath47 and the relaxation time scale @xmath48 on which the state of the system changes is related to @xmath49 by @xmath50 . by making use of the residue theorem in complex plane the correlation function can be easily computed in a form @xmath51 inserting eq . ( [ correlation-1 ] ) into eq . ( [ density-2 ] ) and making use of laplace transform , one can compute @xmath39 explicitly . while in a weak coupling ( or markovian ) regime @xmath52 @xmath39 becomes @xmath53\ ] ] with @xmath54 , in a strong coupling ( or non - markovian ) regime @xmath55 @xmath39 reduces to @xmath56\ ] ] with @xmath57 . since , in the markovian regime @xmath58 , @xmath39 in eq . ( [ pt - m ] ) exhibits an exponential decay in time , it seems to make exponential decay of entanglement or esd phenomenon . however , in the non - markovian regime @xmath59 , @xmath39 in eq . ( [ pt - nm ] ) exhibits an oscillatory behavior in time with decreasing amplitude . it seems to be responsible for the revival phenomenon of entanglement@xcite , after a finite period of time of its complete disappearance . the state @xmath60 at time @xmath61 of whole three - qubit system , each of which interacts only and independently with its own environment , can be derived by the kraus operators@xcite . introducing , for simplicity , @xmath62 , the diagonal parts of @xmath60 are @xmath63 \nonumber \\ & & \hat{\rho}^t_{22 } ( t ) = p_t^2 \left[\hat{\rho}^t_{22 } ( 0 ) + \left\ { \hat{\rho}^t_{33 } ( 0 ) + \hat{\rho}^t_{66 } ( 0 ) \right\ } ( 1 - p_t^2 ) + \hat{\rho}^t_{77 } ( 0 ) ( 1 - p_t^2)^2 \right ] \nonumber \\ & & \hat{\rho}^t_{33 } ( t ) = p_t^4 \left[\hat{\rho}^t_{33 } ( 0 ) + \hat{\rho}^t_{77 } ( 0 ) ( 1 - p_t^2 ) \right ] \\ \nonumber & & \hat{\rho}^t_{44 } ( t ) = p_t^2 \left[\hat{\rho}^t_{44 } ( 0 ) + \left\ { \hat{\rho}^t_{55 } ( 0 ) + \hat{\rho}^t_{66 } ( 0 ) \right\ } ( 1 - p_t^2 ) + \hat{\rho}^t_{77 } ( 0 ) ( 1 - p_t^2)^2 \right ] \\ \nonumber & & \hat{\rho}^t_{55 } ( t ) = p_t^4 \left[\hat{\rho}^t_{55 } ( 0 ) + \hat{\rho}^t_{77 } ( 0 ) ( 1 - p_t^2 ) \right ] \\ \nonumber & & \hat{\rho}^t_{66 } ( t ) = p_t^4 \left[\hat{\rho}^t_{66 } ( 0 ) + \hat{\rho}^t_{77 } ( 0 ) ( 1 - p_t^2 ) \right ] \\ \nonumber & & \hat{\rho}^t_{00 } ( t ) = 1 - \sum_{i=1}^7 \hat{\rho}^t_{ii } ( t ) \end{aligned}\ ] ] and the non - diagonal parts are @xmath64 \nonumber \\ & & \hat{\rho}^t_{02 } ( t ) = p_t \left[\hat{\rho}^t_{02 } ( 0 ) + \left\ { \hat{\rho}^t_{13 } ( 0 ) + \hat{\rho}^t_{46 } ( 0 ) \right\ } ( 1 - p_t^2 ) + \hat{\rho}^t_{57 } ( 0 ) ( 1 - p_t^2)^2 \right ] \nonumber \\ & & \hat{\rho}^t_{04 } ( t ) = p_t \left[\hat{\rho}^t_{04 } ( 0 ) + \left\ { \hat{\rho}^t_{15 } ( 0 ) + \hat{\rho}^t_{26 } ( 0 ) \right\ } ( 1 - p_t^2 ) + \hat{\rho}^t_{37 } ( 0 ) ( 1 - p_t^2)^2 \right ] \nonumber \\ & & \hat{\rho}^t_{03 } ( t ) = p_t^2 \left[\hat{\rho}^t_{03 } ( 0 ) + \hat{\rho}^t_{47 } ( 0 ) ( 1 - p_t^2 ) \right ] \hspace{.5 cm } \hat{\rho}^t_{05 } ( t ) = p_t^2 \left[\hat{\rho}^t_{05 } ( 0 ) + \hat{\rho}^t_{27 } ( 0 ) ( 1 - p_t^2 ) \right ] \nonumber \\ & & \hat{\rho}^t_{06 } ( t ) = p_t^2 \left[\hat{\rho}^t_{06 } ( 0 ) + \hat{\rho}^t_{17 } ( 0 ) ( 1 - p_t^2 ) \right ] \hspace{.5 cm } \hat{\rho}^t_{12 } ( t ) = p_t^2 \left[\hat{\rho}^t_{12 } ( 0 ) + \hat{\rho}^t_{56 } ( 0 ) ( 1 - p_t^2 ) \right ] \nonumber \\ & & \hat{\rho}^t_{13 } ( t ) = p_t^3 \left[\hat{\rho}^t_{13 } ( 0 ) + \hat{\rho}^t_{57 } ( 0 ) ( 1 - p_t^2 ) \right ] \hspace{.5 cm } \hat{\rho}^t_{14 } ( t ) = p_t^2 \left[\hat{\rho}^t_{14 } ( 0 ) + \hat{\rho}^t_{36 } ( 0 ) ( 1 - p_t^2 ) \right ] \nonumber \\ & & \hat{\rho}^t_{15 } ( t ) = p_t^3 \left[\hat{\rho}^t_{15 } ( 0 ) + \hat{\rho}^t_{37 } ( 0 ) ( 1 - p_t^2 ) \right ] \hspace{.5 cm } \hat{\rho}^t_{23 } ( t ) = p_t^3 \left[\hat{\rho}^t_{23 } ( 0 ) + \hat{\rho}^t_{67 } ( 0 ) ( 1 - p_t^2 ) \right ] \\ \nonumber & & \hat{\rho}^t_{24 } ( t ) = p_t^2 \left[\hat{\rho}^t_{24 } ( 0 ) + \hat{\rho}^t_{35 } ( 0 ) ( 1 - p_t^2 ) \right ] \hspace{.5 cm } \hat{\rho}^t_{26 } ( t ) = p_t^3 \left[\hat{\rho}^t_{26 } ( 0 ) + \hat{\rho}^t_{37 } ( 0 ) ( 1 - p_t^2 ) \right ] \\ \nonumber & & \hat{\rho}^t_{45 } ( t ) = p_t^3 \left[\hat{\rho}^t_{45 } ( 0 ) + \hat{\rho}^t_{67 } ( 0 ) ( 1 - p_t^2 ) \right ] \hspace{.5 cm } \hat{\rho}^t_{46 } ( t ) = p_t^3 \left[\hat{\rho}^t_{46 } ( 0 ) + \hat{\rho}^t_{57 } ( 0 ) ( 1 - p_t^2 ) \right ] \\ \nonumber & & \hat{\rho}^t_{07 } ( t ) = \hat{\rho}^t_{07 } ( 0 ) p_t^3 \hspace{.5 cm } \hat{\rho}^t_{16 } ( t ) = \hat{\rho}^t_{16 } ( 0 ) p_t^3 \hspace{.5 cm } \hat{\rho}^t_{17 } ( t ) = \hat{\rho}^t_{17 } ( 0 ) p_t^4 \hspace{.5 cm } \hat{\rho}^t_{25 } ( t ) = \hat{\rho}^t_{25 } ( 0 ) p_t^3 \\ \nonumber & & \hat{\rho}^t_{27 } ( t ) = \hat{\rho}^t_{27 } ( 0 ) p_t^4 \hspace{.5 cm } \hat{\rho}^t_{34 } ( t ) = \hat{\rho}^t_{34 } ( 0 ) p_t^3 \hspace{.5 cm } \hat{\rho}^t_{35 } ( t ) = \hat{\rho}^t_{35 } ( 0 ) p_t^4 \hspace{.5 cm } \hat{\rho}^t_{36 } ( t ) = \hat{\rho}^t_{36 } ( 0 ) p_t^4 \\ \nonumber & & \hspace{1.3cm}\hat{\rho}^t_{37 } ( t ) = \hat{\rho}^t_{37 } ( 0 ) p_t^5 \hspace{.5 cm } \hat{\rho}^t_{47 } ( t ) = \hat{\rho}^t_{47 } ( 0 ) p_t^4 \hspace{.5 cm } \hat{\rho}^t_{56 } ( t ) = \hat{\rho}^t_{56 } ( 0 ) p_t^4 \\ \nonumber & & \hspace{2.7cm}\hat{\rho}^t_{57 } ( t ) = \hat{\rho}^t_{57 } ( 0 ) p_t^5 \hspace{.5 cm } \hat{\rho}^t_{67 } ( t ) = \hat{\rho}^t_{67 } ( 0 ) p_t^5\end{aligned}\ ] ] with @xmath65 . now , we are ready to explore the tripartite entanglement dynamics in the presence of the markovian or non - markovian environment . in this section we examine the tripartite entanglement dynamics when the initial states are ghz - type states . all initial states have ghz - symmetry@xcite if the parameters are appropriately chosen . however , this symmetry is broken due to the effects of environment . let us choose the initial state in a form @xmath66 where @xmath67 with @xmath68 . as commented before @xmath69 has a ghz - symmetry when @xmath70 and @xmath71 . then the spectral decomposition of @xmath72 can be read directly from eqs . ( [ diagonal-1 ] ) and ( [ non - diagonal-1 ] ) as a form : @xmath73 where @xmath74 ^ 2 - 4 b^4 p_t^6 ( 1 - p_t^2)^2 } \right]\ ] ] and @xmath75 with @xmath76 ^ 2 - 4 b^4 p_t^6 ( 1 - p_t^2)^2 } \nonumber \\ & & y = 2 a b p_t^2\end{aligned}\ ] ] and @xmath77 is a normalization constant . since @xmath72 is a full rank , it seems to be highly difficult to compute the residual entanglement ( or three - tangle ) analytically . however , from eq . ( [ type1-ghz-2 ] ) one can realize the upper bound of @xmath2 as @xmath78 \frac{4 x^2 y^2}{(x^2 + y^2)^2}.\ ] ] it is worthwhile noting that @xmath72 does not have the ghz - symmetry even at @xmath70 and @xmath71 . thus , the symmetry which @xmath79 has is broken due to the effect of environment . in order to explore the tripartite entanglement dynamics on the analytical ground , we compute the @xmath0-tangle of @xmath72 . using eq . ( [ negativity-1 ] ) it is straightforward to compute the induced bipartite entanglement quantities @xmath80 , @xmath81 , and @xmath82 . one can show that they are all the same with @xmath83,\end{aligned}\ ] ] where @xmath84 one can also show the two - tangles completely vanish , i.e. @xmath85 , easily . thus the @xmath0-tangle of @xmath72 is @xmath86 -tangle of @xmath72 as a function of the parameters @xmath87 and @xmath88 when the state interacts with the markovian and non - markovian environments . we choose @xmath45 as ( a ) @xmath89 and ( b ) @xmath90 . , title="fig:",height=185 ] -tangle of @xmath72 as a function of the parameters @xmath87 and @xmath88 when the state interacts with the markovian and non - markovian environments . we choose @xmath45 as ( a ) @xmath89 and ( b ) @xmath90 . , title="fig:",height=185 ] eq . ( [ ghz-1-pi-1 ] ) guarantees that regardless of markovian or non - markovian environment @xmath91 becomes zero if an inequality @xmath92 holds because @xmath93 becomes negative in this condition . now , let us examine the dynamics of the tripartite entanglement for @xmath72 when the quantum system interacts with markovian environment . since @xmath39 in eq . ( [ pt - m ] ) decays exponentially in time , one can expect that the tripartite entanglement exhibits an asymptotic decay , i.e. decay with the half - life rule , similarly . in fact , this is true when the inequality ( [ add1 ] ) is violated . if the inequality holds at @xmath94 , the tripartite entanglement becomes zero at @xmath95 abruptly . this is an esd phenomenon . if the inequality does not hold for all time , the tripartite entanglement decays with the half - life rule as expected . this is shown clearly in fig . 1(a ) , where @xmath91 is plotted as a function of @xmath87 and @xmath88 . in this figure we choose @xmath89 . as expected , the tripartite entanglement decreases with increasing @xmath87 . when @xmath96 ( blue line ) it decays exponentially in @xmath87 with the half - life rule . for @xmath97 ( red line ) , however , it becomes zero in the region @xmath98 . for non - markovian regime the decay behavior of the tripartite entanglement in time is completely different . this difference arises due to combination of the inequality ( [ add1 ] ) and difference form of @xmath39 . since @xmath39 in eq . ( [ pt - nm ] ) exhibits an underdamping behavior in time with zeros at @xmath99 \hspace{.2 cm } ( n = 1 , 2 , \cdots)$ ] , one may expect that the tripartite entanglement also decays with oscillatory behavior . this is true when the inequality ( [ add1 ] ) is violated for all time . this behavior is shown as a blue line ( @xmath96 ) of fig . in this figure we choose @xmath90 . if the inequality holds for some time interval @xmath100 , the tripartite entanglement becomes zero in this interval . after this time interval , however , nonzero tripartite entanglement reappears , which makes a revival of entanglement after a finite period of time of its complete disappearance . this is shown as a red line ( @xmath101 ) of fig . 1(b ) . let us choose the initial state in a form @xmath102 where @xmath103 with @xmath68 . since @xmath104 , @xmath105 has a ghz - symmetry provided that @xmath70 and @xmath71 . using eqs . ( [ diagonal-1 ] ) and ( [ non - diagonal-1 ] ) one can show that the spectral decomposition of @xmath106 becomes @xmath107 { \lvert 0 \rangle } { \langle 0 \lvert } + b^2 p_t^2 ( 1 - p_t^2 ) \left ( { \lvert 2 \rangle } { \langle 2 \lvert } + { \lvert 4 \rangle } { \langle 4 \lvert } \right)\ ] ] where @xmath108 unlike the case of type i @xmath106 is rank four tensor . from eq . ( [ type2-ghz-2 ] ) one can derive the upper bound of @xmath2 for @xmath106 , which is @xmath109 the negativities @xmath80 , @xmath81 , and @xmath82 of @xmath106 can be computed by making use of eq . ( [ negativity-1 ] ) . the final expressions are @xmath110 ^ 2 + 4 a^2 b^2 p_t^6 } - ( 1 - p_t^2 ) \left [ a^2 + b^2 ( 1 - p_t^2 ) \right].\end{aligned}\ ] ] it is also easy to show @xmath85 . thus the @xmath0-tangle of @xmath106 is @xmath111.\ ] ] when @xmath112 , @xmath113 becomes @xmath114 and it reduces to zero as @xmath115 . of course , the entanglement of @xmath106 is completely disentangled at @xmath116 in the non - markovian regime . let us choose the initial state in a form @xmath117 where @xmath118 with @xmath68 . since @xmath119 , @xmath120 has a ghz - symmetry provided that @xmath70 and @xmath71 . using eqs . ( [ diagonal-1 ] ) and ( [ non - diagonal-1 ] ) one can show that the spectral decomposition of @xmath121 becomes @xmath122 { \lvert 0 \rangle } { \langle 0 \lvert } + a^2 p_t^2 ( 1 - p_t^2 ) \left ( { \lvert 1 \rangle } { \langle 1 \lvert } + { \lvert 2 \rangle } { \langle 2 \lvert } \right)\ ] ] where @xmath123 unlike the case of type i @xmath121 is rank four tensor . from eq . ( [ type3-ghz-2 ] ) one can derive the upper bound of @xmath2 for @xmath121 , which is @xmath124 the negativities @xmath80 , @xmath81 , and @xmath82 of @xmath121 can be computed by making use of eq . ( [ negativity-1 ] ) , whose explicit expressions are @xmath125 ^ 2 + 4 a^2 b^2 p_t^6 } - ( 1 - p_t^2 ) \left[a^2 ( 1 - p_t^2 ) + b^2 \right ] \nonumber \\ & & { \cal n}_{b(ac ) } = { \cal n}_{(ab)c } = \sqrt{a^4 p_t^4 ( 1 - p_t^2)^2 + 4 a^2 b^2 p_t^6 } - a^2 p_t^2 ( 1 - p_t^2).\end{aligned}\ ] ] it is of interest to note that @xmath80 and @xmath81 of type iii is the same with @xmath82 and @xmath80 of type ii with @xmath126 respectively . it is easy to show @xmath85 . thus the @xmath0-tangle of @xmath121 is @xmath127.\ ] ] one can also consider different types of initial ghz - type states . for example , one can consider @xmath128 , where @xmath129 . although , in this case , @xmath130 is different from @xmath106 , one can show that its @xmath0-tangle is exactly the same with that of type ii . thus , this case is not discussed in detail . -tangle for the initial states ( a ) @xmath131 and ( b ) @xmath132 as a function of the parameters @xmath87 and @xmath88 . we choose @xmath45 as a @xmath90 . , title="fig:",height=188 ] -tangle for the initial states ( a ) @xmath131 and ( b ) @xmath132 as a function of the parameters @xmath87 and @xmath88 . we choose @xmath45 as a @xmath90 . , title="fig:",height=188 ] as shown in eqs . ( [ ghz-2-pi-2 ] ) and ( [ ghz-3-pi-2 ] ) the dynamics of the tripartite entanglements for type ii and type iii are not expressed in terms of an inequality like eq . ( [ add1 ] ) in type i. thus , if @xmath133 and @xmath134 interact with the markovian surroundings , these entanglements decay exponentially with the half - life rule . this means that there is no esd phenomenon in these cases . if @xmath133 and @xmath134 interact with the non - markovian environment , @xmath135 and @xmath136 should exhibit an oscillatory behavior with rapid decrease of amplitude due to @xmath39 in eq . ( [ pt - nm ] ) . this can be seen in fig . 2 , where @xmath135 and @xmath136 are plotted as a function of dimensionless parameter @xmath87 and @xmath88 . we choose @xmath45 as a @xmath90 . as expected the tripartite entanglement reduces to zero with increasing time with oscillatory behavior . dependence of @xmath137 ( red solid ) , @xmath135 ( black dashed ) , and @xmath136 ( blue dotted ) when ( a ) @xmath138 and ( b ) @xmath139 . we choose @xmath45 as a @xmath140 . , title="fig:",height=188 ] dependence of @xmath137 ( red solid ) , @xmath135 ( black dashed ) , and @xmath136 ( blue dotted ) when ( a ) @xmath138 and ( b ) @xmath139 . we choose @xmath45 as a @xmath140 . , title="fig:",height=188 ] the @xmath0-tangles @xmath137 , @xmath135 , and @xmath136 are compared in fig . 3 when @xmath141 . they are represented by red solid , black dashed , and blue dotted lines respectively . fig . 3(a ) and fig . 3(b ) correspond to @xmath138 and @xmath139 . both figures clearly show the revival of the tripartite entanglement , after a finite period of time of complete disappearance . the revival phenomenon seems to be mainly due to the memory effect of the non - markovian environment . it is of interest to note that while @xmath142 when @xmath138 , the order is changed as @xmath143 when @xmath139 . in this section we examine the tripartite entanglement dynamics when the initial states are two w - type states . both initial states are lu to each other . however , their entanglement dynamics are different due to eqs . ( [ diagonal-1 ] ) and ( [ non - diagonal-1 ] ) . in this subsection we choose the initial state as @xmath144 where @xmath145 with @xmath146 . then , it is straightforward to show that the spectral decomposition of @xmath147 is @xmath148 eq . ( [ type1-w-2 ] ) guarantees that the residual entanglement and three - tangle of @xmath147 are zero because the spectral decomposition exactly coincides with the optimal decomposition . by making use of eq . ( [ negativity-1 ] ) one can compute the induced bipartite entanglement quantities @xmath80 , @xmath81 , and @xmath82 of @xmath149 directly , whose expressions are @xmath150 also , the two tangles @xmath151 , @xmath152 , and @xmath153 become @xmath154 ^ 2 + 4 b^2 c^2 p_t^4 } - \left [ ( 1 - p_t^2 ) + a^2 p_t^2 \right ] \nonumber \\ & & { \cal n}_{ac } = \sqrt{\left [ ( 1 - p_t^2 ) + b^2 p_t^2 \right]^2 + 4 a^2 c^2 p_t^4 } - \left [ ( 1 - p_t^2 ) + b^2 p_t^2 \right ] \\ \nonumber & & { \cal n}_{bc } = \sqrt{\left [ ( 1 - p_t^2 ) + c^2 p_t^2 \right]^2 + 4 a^2 b^2 p_t^4 } - \left [ ( 1 - p_t^2 ) + c^2 p_t^2 \right].\end{aligned}\ ] ] thus , using eqs . ( [ pi-1 ] ) and ( [ pi-2 ] ) one can compute the @xmath0-tangle of @xmath147 , whose explicit expression is @xmath155 \sqrt{\left[(1 - p_t^2 ) + a^2 p_t^2 \right]^2 + 4 b^2 c^2 p_t^4 } \nonumber \\ & & \hspace{2.0 cm } + 2 \left[(1 - p_t^2 ) + b^2 p_t^2 \right ] \sqrt{\left[(1 - p_t^2 ) + b^2 p_t^2 \right]^2 + 4 a^2 c^2 p_t^4 } \nonumber \\ & & \hspace{2.0 cm } + 2 \left[(1 - p_t^2 ) + c^2 p_t^2 \right ] \sqrt{\left[(1 - p_t^2 ) + c^2 p_t^2 \right]^2 + 4 a^2 b^2 p_t^4 } \\ \nonumber & & \hspace{2,0 cm } - ( 1 - p_t^2 ) \bigg\ { \sqrt{(1 - p_t^2)^2 + 4 a^2 ( b^2 + c^2 ) p_t^4 } \\ \nonumber & & \hspace{2.0 cm } + \sqrt{(1 - p_t^2)^2 + 4 b^2 ( a^2 + c^2 ) p_t^4 } + \sqrt{(1 - p_t^2)^2 + 4 c^2 ( a^2 + b^2 ) p_t^4 } \bigg\ } \\ \nonumber & & \hspace{4.0 cm } - 2 ( a^4 + b^4 + c^4 ) p_t^4 - ( 1 - p_t^2 ) ( 3 + p_t^2 ) \bigg].\end{aligned}\ ] ] when @xmath112 , eq . ( [ w-1-pi-3 ] ) reduces to @xmath156,\ ] ] which exactly coincides with a result of ref.@xcite . of course , when @xmath157 and @xmath158 , the entanglement of @xmath147 is completely disentangled in the non - markovian regime . in this subsection we choose the initial state as @xmath159 where @xmath160 with @xmath146 . this initial state is lu to @xmath161 because of @xmath162 . then , by making use of eqs . ( [ diagonal-1 ] ) and ( [ non - diagonal-1 ] ) it is straightforward to show that @xmath163 is @xmath164 where @xmath165.\end{aligned}\ ] ] the spectral decomposition of @xmath166 can not be derived analytically . also , analytic computation of @xmath0-tangle for @xmath163 is impossible . thus , we have to reply on the numerical approach for computation of @xmath0-tangle . dependence of @xmath167 ( red line ) and @xmath168 ( blue line ) when @xmath161 and @xmath169 interact with the markovian environment . we choose @xmath89 and @xmath170.,height=188 ] however , some special cases allow the analytic computation . in this paper we consider a special case @xmath170 . in this case the spectral decomposition of @xmath166 can be derived as @xmath171 where @xmath172 thus , eqs . ( [ type2-w-2 ] ) and ( [ type2-w-4 ] ) imply that @xmath163 with @xmath170 is rank-@xmath173 tensor , three of them are w - states and the remaining ones are fully - separable and bi - separable states . thus , its residual entanglement and three - tangles are zero . using eq . ( [ negativity-1 ] ) one can show that @xmath80 , @xmath81 , and @xmath82 are all identical as @xmath174.\ ] ] also @xmath151 , @xmath152 , and @xmath153 are all identical as @xmath175 thus , the @xmath0-tangle for @xmath163 with @xmath170 is given by @xmath176 . in fig . 4 we plot @xmath177 ( red line ) and @xmath178 ( blue line ) as a function of @xmath87 when @xmath161 and @xmath169 interact with the markovian environment . we choose @xmath89 and @xmath170 . as expected both reduce to zero with the half - life rule . it is of interest to note @xmath179 in full range of time . this means that @xmath161 is more robust than @xmath169 against the markovian environment . and @xmath87 dependence of @xmath180 when @xmath181 . we choose @xmath90 . ( b ) the @xmath87 dependence of @xmath182 ( solid line ) and @xmath183 ( dashed line ) when @xmath170 . we choose @xmath140 . this figure implies that @xmath149 is more robust against the environment than @xmath163 . , title="fig:",height=188 ] and @xmath87 dependence of @xmath180 when @xmath181 . we choose @xmath90 . ( b ) the @xmath87 dependence of @xmath182 ( solid line ) and @xmath183 ( dashed line ) when @xmath170 . we choose @xmath140 . this figure implies that @xmath149 is more robust against the environment than @xmath163 . , title="fig:",height=188 ] in fig . 5(a ) we plot @xmath180 as a function of @xmath88 and @xmath87 when @xmath161 is embedded in the non - markovian environment . we choose @xmath181 and @xmath184 . as expected the @xmath0-tangle reduces to zero as @xmath115 with an oscillatory behavior . to compare @xmath180 with @xmath185 we plot both @xmath0-tangles as a function of @xmath87 in fig . in this figure we choose @xmath170 and @xmath141 . the @xmath0-tangles @xmath180 and @xmath185 are plotted as solid and dashed lines respectively . in this case , as in the other cases , the revival of entanglement occurs after complete disappearance . it is interesting to note that like a markovian case @xmath149 is more robust than @xmath163 against non - markovian environment . in this paper we have examined the tripartite entanglement dynamics when each party is entangled with other parties initially , but they locally interact with their own markovian or non - markovian environment . first , we have considered three ghz - type initial states @xmath186 , @xmath187 , and @xmath188 . all states are lu to each other . it turns out that the ghz symmetry of the initial states is broken due to the effect of environment . we have computed the corresponding @xmath0-tangles analytically at arbitrary time @xmath61 in eqs . ( [ ghz-1-pi-2 ] ) , ( [ ghz-2-pi-2 ] ) , and ( [ ghz-3-pi-2 ] ) . it was shown that while the esd phenomenon occurs for type i , the entanglement dynamics for the remaining types exhibits an exponential decay in the markovian regime . in the non - markovian regime the @xmath0-tangles completely vanish when @xmath189 \hspace{.3 cm } ( n = 1 , 2 , \cdots)$ ] and @xmath115 . as shown in fig . 3 the revival phenomenon of entanglement occurs after complete disappearance of entanglement . based on the analytical results it was shown that while the robustness order against the effect of reservoir is @xmath69 , @xmath133 , @xmath134 for large @xmath88 region , this order is reversed for small @xmath88 region . dependence of concurrences eq.([bipartite1 ] ) and eq . ( [ bipartite2 ] ) when @xmath170 . ( a ) in this figure we choose @xmath89 . this shows that while bipartite entanglement dynamics for type i ( red line ) decays exponentially with the half - life rule , that for type ii ( blue line ) exhibits an esd . ( b ) in this figure we choose @xmath90 . although both entanglements decay in time , the decay rate for type ii ( blue line ) is much faster than that for type i ( red line).,title="fig:",height=188 ] dependence of concurrences eq.([bipartite1 ] ) and eq . ( [ bipartite2 ] ) when @xmath170 . ( a ) in this figure we choose @xmath89 . this shows that while bipartite entanglement dynamics for type i ( red line ) decays exponentially with the half - life rule , that for type ii ( blue line ) exhibits an esd . ( b ) in this figure we choose @xmath90 . although both entanglements decay in time , the decay rate for type ii ( blue line ) is much faster than that for type i ( red line).,title="fig:",height=188 ] we also have examined the tripartite entanglement dynamics for two w - type initial states @xmath190 and @xmath191 with @xmath146 . like ghz - type initial states they are lu to each other . for initial @xmath161 state the @xmath0-tangle is analytically computed in eq . ( [ w-1-pi-3 ] ) . since , however , @xmath169 propagates to higher - rank state with the lapse of time , the analytic computation is impossible except few special cases . thus , we have computed the @xmath0-tangle analytically for special case @xmath170 . in fig . 4 and fig . 5 it was shown that @xmath161 is more robust than @xmath169 against the markovian and non - markovian environments . the bipartite entanglements measured by the concurrence@xcite for @xmath147 and @xmath163 are @xmath192 and @xmath193 \nonumber \\ & & { \cal c}^{ii}_{ac } ( t ) = 2 p_t^2 \max \left[0 , |a c| -|b| \sqrt{(1 - p_t^2 ) ( 1 - b^2 p_t^2 ) } \right ] \\ \nonumber & & { \cal c}^{ii}_{bc } ( t ) = 2 p_t^2 \max \left[0 , |a b| - |c| \sqrt{(1 - p_t^2 ) ( 1 - c^2 p_t^2 ) } \right ] . \end{aligned}\ ] ] one can show @xmath194 in the entire range of time like a tripartite entanglement regardless of markovian or non - markovian environment . the @xmath87-dependence of the concurrences is plotted in fig . 6 as red line for type i and blue line for type ii when ( a ) markovian ( @xmath89 ) and ( b ) non - markovian ( @xmath90 ) environments are introduced . 6(a ) shows that while the entanglement for type i exhibits an exponential decay with the half - life rule , that for type ii exhibits an esd . for non - markovian case the decay rate for type ii is much faster than that for type i although both exhibit a revival phenomenon of entanglement . it is of interest to study the effect of non - markovian environment when the initial state is a rank-@xmath195 mixture @xmath196 where @xmath197 and @xmath198 . the residual entanglement of @xmath199 is known as @xmath200 where @xmath201{2}}{3 + 4 \sqrt[3]{2 } } = 0.626851\cdots \hspace{1.0 cm } p_1 = \frac{1}{2 } + \frac{3 \sqrt{465}}{310 } = 0.70868\cdots \\ \nonumber & & g_i ( p ) = p^2 - \frac{8 \sqrt{6}}{9 } \sqrt{p ( 1 - p)^3 } \hspace{1.0 cm } g_{ii}(p ) = 1 - ( 1 - p ) \left(\frac{3}{2 } + \frac{1}{18 } \sqrt{465 } \right).\end{aligned}\ ] ] it is interesting , at least for us , how the non - markovian environment modifies coffman - kundu - wootters inequality @xmath202 \geq { \cal c}(\rho_{ab})^2 + { \cal c}(\rho_{ac})^2 $ ] in this model . similar issue was discussed in ref . @xcite . since we have derived the @xmath0-tangles analytically , we tried to find the entanglement invariants@xcite , which was originally found in four - qubit system . in our three - qubit systems we can not find any invariants it is of interest to examine the entanglement invariants in the higher - qubit and qudit systems . 99 m. a. nielsen and i. l. chuang , _ quantum computation and quantum information _ ( cambridge university press , cambridge , england , 2000 ) . r. horodecki , p. horodecki , m. horodecki , and k. horodecki , _ quantum entanglement _ , * 81 * ( 2009 ) 865 [ quant - ph/0702225 ] and references therein . c. h. bennett , g. brassard , c. crepeau , r. jozsa , a. peres , and w. k. wootters , _ teleporting an unknown quantum state via dual classical and einstein - podolsky - rosen channles _ , phys.rev * 70 * ( 1993 ) 1895 . c. h. bennett and s. j. wiesner , _ communication via one- and two - particle operators on einstein - podolsky - rosen states _ , * 69 * ( 1992 ) 2881 . v. scarani , s. lblisdir , n. gisin , and a. acin , _ quantum cloning _ , * 77 * ( 2005 ) 1225 [ quant - ph/0511088 ] and references therein . a. k. ekert , _ quantum cryptography based on bell s theorem _ , * 67 * ( 1991 ) 661 . c. kollmitzer and m. pivk , applied quantum cryptography ( springer , heidelberg , germany , 2010 ) . t. d. ladd , f. jelezko , r. laflamme , y. nakamura , c. monroe , and j. l. obrien , _ quantum computers _ , nature , * 464 * ( 2010 ) 45 [ arxiv:1009.2267 ( quant - ph ) ] . g. vidal , _ efficient classical simulation of slightly entangled quantum computations _ * 91 * ( 2003 ) 147902 [ quant - ph/0301063 ] . breuer and f. petruccione , _ the theory of open quantum systems _ ( oxford university press , oxford , new york , 2002 ) . w. h. zurek , _ decoherence , einselection , and the quantum origins of the classical _ , rev . mod . * 75 * ( 2003 ) 715 [ quant - ph/0105127 ] . t. yu and j. h. eberly , _ phonon decoherence of quantum entanglement : robust and fragile states _ , * b 66 * ( 2002 ) 193306 [ quant - ph/0209037 ] . c. simon and j. kempe , _ robustness of multiparty entanglement _ , phys . rev . * a 65 * ( 2002 ) 052327 [ quant - ph/0109102 ] . w. dr and h. j. briegel , _ stability of macroscopic entanglement under decoherence _ , lett . * 92 * ( 2004 ) 180403 [ quant - ph/0307180 ] . t. yu and j. h. eberly , _ finite - time disentanglement via spontaneous emission _ , * 93 * ( 2004 ) 140404 [ quant - ph/0404161 ] . t. yu and j. h. eberly , _ sudden death of entanglement : classical noise effects _ , opt . commun . * 264 * ( 2006 ) 393 [ quant - ph/0602196 ] . t. yu and j. h. eberly , _ quantum open system theory : bipartite aspects_. phys . * 97 * ( 2006 ) 140403 [ quant - ph/0603256 ] t. yu and j. h. eberly , _ sudden death of entanglement _ , science , * 323 * ( 2009 ) 598 [ arxiv:0910.1396 ( quant - ph ) ] . et al _ , _ environment - induced sudden death of entanglement _ , science * 316 * ( 2007 ) 579 [ quant - ph/0701184 ] . j. laurat , k. s. choi , h. deng , c. w. chou , and h. j. kimble , _ heralded entanglement between atomic ensembles : preparation , decoherence , and scaling _ , physics . * 99 * ( 2007 ) 180504 [ arxiv:0706.0528 ( quant - ph ) ] . c. e. lpez , g. romero , f. lastra , e. solano , and j. c. retamal , _ sudden birth versus sudden death of entanglement in multipartite systems _ , * 101 * ( 2008 ) 080503 [ arxiv:0802.1825 ( quant - ph ) ] . b. bellomo , r. lo franco , and g. compagno , _ non - markovian effects on the dynamics of entanglement _ , phys . * 99 * ( 2007 ) 160502 [ arxiv:0804.2377 ( quant - ph ) ] . s. hill and w. k. wootters , _ entanglement of a pair of quantum bits _ , * 78 * ( 1997 ) 5022 [ quant - ph/9703041 ; w. k. wootters , _ entanglement of formation of an arbitrary state of two qubits _ , * 80 * ( 1998 ) 2245 [ quant - ph/9709029 ] . breuer , e. -m . laine , and j. piilo , _ measure for the degree of non - markovian behavior of quantum processes in open systems _ * 103 * ( 2009 ) 210401 [ arxiv:0908.0238 ( quant - ph ) ] . b. vacchini , a. smirne , e. -m . laine , j. piilo , and h. -p . breuer , _ markovian and non - markovian dynamics in quantum and classical systems _ , new j. phys . * 13 * ( 2011 ) 093004 [ arxiv:1106.0138 ( quant - ph ) ] . d. chruciski , a. kossakowski , and a. rivas , _ measures of non - markovianity : divisibility versus backflow of information _ , phys . rev . * a 83 * ( 2011 ) 052128 [ arxiv:1102.4318 ( quant - ph ) ] . a. rivas , s. f. huelga , and m. b. plenio , _ quantum non - markovianity : characterization , quantification and detection _ * 77 * ( 2014 ) 094001 [ arxiv:1405.0303 ( quant - ph ) ] . m. j. w. hall , j. d. cresser , l. li , and e. andersson _ canonical form of master equations and characterization of non - markovianity _ , phys . rev . * a 89 * ( 2014 ) 042120 [ arxiv:1009.0845 ( quant - ph ) ] . kim , h .-m . li , and b. -k . zhao , _ genuinetripartite entanglement dynamics and transfer in a triple jaynes - cummings model _ , int * 55 * ( 2016 ) 241 . m. ynac , t. yu , and j. h. eberly , _ pairwise concurrence dynamics : a four - qubit model _ , j. phys . b : at . mol . * 40 * ( 2007 ) 545 [ quant - ph/0701111 ] . d. m. greenberger , m. horne , and a. zeilinger , _ bell s theorem , quantum theory , and conceptions of the universe _ , edited by m. kafatos ( kluwer , dordrecht , 1989 ) . w. dr , g. vidal , and j. i. cirac , _ three qubits can be entangled in two inequivalent ways _ , phys * a62 * ( 2000 ) 062314 [ quant - ph/0005115 ] . v. coffman , j. kundu , and w. k. wootters , _ distributed entanglement _ , phys . rev . * a 61 * ( 2000 ) 052306 [ quant - ph/9907047 ] . y. u. ou and h. fan , _ monogamy inequality in terms of negativity for three - qubit states _ , phys . rev . * a75 * ( 2007 ) 062308 [ quant - ph/0702127 ] . c. h. bennett , d. p. divincenzo , j. a. smokin , and w. k. wootters , _ mixed - state entanglement and quantum error correction _ , phys . rev . * a 54 * ( 1996 ) 3824 [ quant - ph/9604024 ] . a. uhlmann , _ fidelity and concurrence of conjugate states _ , phys . rev . * a 62 * ( 2000 ) 032307 [ quant - ph/9909060 ] . r. lohmayer , a. osterloh , j. siewert , and a. uhlmann , _ entangled three - qubit states without concurrence and three - tangle _ , phys . lett . * 97 * ( 2006 ) 260502 [ quant - ph/0606071 ] ; c. eltschka , a. osterloh , j. siewert , and a. uhlmann , _ three - tangle for mixtures of generalized ghz and generalized w states _ , new j. phys . * 10 * ( 2008 ) 043014 [ arxiv:0711.4477 ( quant - ph ) ] ; e. jung , m. r. hwang , d. k. park , and j. w. son , _ three - tangle for rank-@xmath203 mixed states : mixture of greenberger - horne - zeilinger , w and flipped w states _ , phys . rev . * a 79 * ( 2009 ) 024306 [ arxiv:0810.5403 ( quant - ph ) ] ; e. jung , d. k. park , and j. w. son , _ three - tangle does not properly quantify tripartite entanglement for greenberger - horne - zeilinger - type state _ rev . * a 80 * ( 2009 ) 010301(r ) [ arxiv:0901.2620 ( quant - ph ) ] ; e. jung , m. r. hwang , d. k. park , and s. tamaryan , _ three - party entanglement in tripartite teleportation scheme through noisy channels _ , quant . inf . comp . * 10 * ( 2010 ) 0377 [ arxiv:0904.2807 ( quant - ph ) ] . c. eltschka and j. siewert , _ entanglement of three - qubit greenberger - horne - zeilinger - symmetric states _ , phys . * 108 * ( 2012 ) 020502 [ arxiv:1304.6095 ( quant - ph ) ] . j. siewert and c. eltschka , _ quantifying tripartite entanglement of three - qubit generalized werner states _ * 108 * ( 2012 ) 230502 . g. vidal and r. f. werner , _ computable measure of entanglement _ , phys . rev . * a65 * ( 2002 ) 032314 [ quant - ph/0102117 ] . b. m. garraway , _ nonperturbative decay of an atomic system in a cavity _ , phys . rev . * a55 * ( 1997 ) 2290 . s. maniscalco and f. petruccione , _ non - markovian dynamics of a qubit _ , phys . rev . * a73 * ( 2006 ) 012111 [ quant - ph/0509208 ] . k. kraus , _ states , effect , and operations : fundamental notions in quantum theory _ ( springer - verlag , berlin , 1983 ) . a. c. s. costa , r. m. angelo , and m. w. beims , _ monogamy and backflow of mutual information in non - markovian thermal baths _ , phys . rev . * a 90 * ( 2014 ) 012322 [ arxiv:1404.6433 ( quant - ph ) ] .

we study quantum interference effects on the transition strength for strongly localized electrons hopping on two - dimensional ( 2d ) square and three - dimensional ( 3d ) cubic lattices in the presence of a magnetic field . these effects arise from the interference between phase factors associated with different electron paths connecting two distinct sites . for electrons confined on a square lattice , with and without disorder , we obtain _ closed - form _ expressions for the tunneling probability , which determines the conductivity , between two _ arbitrary _ sites by exactly summing the corresponding phase factors of _ all _ forward - scattering paths connecting them . by analytically summing paths which allow backward excursions in the forward - scattering direction , we find that the interference patterns between the dominant winding paths are not drastically different from those between the directed paths . an _ analytic field - dependent _ expression , valid in _ any dimension _ , for the magnetoconductance ( mc ) is derived . a _ positive _ mc is clearly observed when turning on the magnetic field . in 2d , when the strength of reaches a certain value , which is inversely proportional to twice the hopping length , the mc is increased by a factor of two compared to that at zero field . the periodicity in the flux of the mc is found to be equal to the superconducting flux quantum . we also investigate transport on the much less - studied and experimentally important 3d cubic lattice case , where it is shown how the interference patterns and the small - field behavior of the mc vary according to the orientation of the applied field . at very small fields , for two sites diagonally separated a distance , we find that the mc behaves as : in quasi-1d systems , in 2d with , and [ ] in 3d with parallel [ perpendicular ] to the direction . furthermore , for a 3d sample , the effect on the low - flux mc due to the randomness of the angles between the hopping direction and the orientation of is examined analytically .

cond-mat9611226

electrons moving on a lattice immersed in a magnetic field have attracted much attention due to their relevance to many physical problems . in particular , quantum interference ( qi ) effects between different electron paths in disordered electron systems have been a subject of intense study because they play an important role in quantum transport . for instance , the qi of closed loops and their time - reversed paths is central to weak - localization phenomena.@xcite indeed , during the past decade and half , many fascinating phenomena including universal conductance fluctuations as well as magnetic - field and spin - orbit scattering effects on the conductivity observed in the _ weakly _ localized , metallic regime have been understood in terms of the qi between different feynman diffusive paths in backscattering loops ( i.e. , paths bringing an electron back to the starting point ) . recently , interest has grown in the effects of a magnetic field on _ strongly _ localized electrons@xcite with variable - range hopping ( vrh ) where striking qi phenomena has been observed in mesoscopic and macroscopic insulators or strongly disordered compounds : anomalous magnetoresistance , pronounced conductance fluctuations , aharonov - bohm oscillations with periods of @xmath8 and @xmath1 , and the hall effect . this strongly localized regime@xcite is less well - understood than the weak - localization case . in the strongly localized regime , the major mechanism for transport is thermally activated hopping between the localized sites . in the vrh introduced by mott,@xcite localized electrons , whose wavefunctions decay exponentially with a localization length @xmath9 , hop a distance which is many times larger than @xmath9 . as a result of the balance between the probabilities for hopping and thermal activation , mott derived that in @xmath10 dimensions the hopping length changes with temperature as @xmath11 , where @xmath12 is a characteristic temperature . therefore , the lower the temperature is , the further away the electron tunnels in order to find a localized site of closer energy . according to the critical path analysis"@xcite arguments , the conductance of the sample is governed by one critical hopping event . during this critical phonon - assisted tunneling process , the electron traverses many other impurities since the hopping length is very large at low temperatures . while encountering these intermediate scatterers , the electron preserves its phase memory . this elastic multiple - scattering is the origin of the qi effects associated with a single hopping event between the initial ( @xmath13 ) and final ( @xmath14 ) sites . the overall tunneling amplitude @xmath15 between the sites @xmath13 and @xmath14 is therefore the sum of the contributions from all possible paths connecting them.@xcite in other words , the tunneling probability of one distant hop is determined by the interference of many electron paths between @xmath13 and @xmath14 . this leads to mott s law for the temperature dependence of the conductivity in @xmath10 dimensions:@xcite @xmath16.\ ] ] it is worthwhile to note that , in the limit of strong localization , the dominant contribution to @xmath15 comes from the shortest paths between @xmath13 and @xmath14 ( i.e. , the directed path model " ) . in other words , only interference between _ forward - scattering _ paths need to be taken into account . this is in contrast with weak localization which results from _ backscattering _ processes on closed paths . the focus of this paper is on the qi effects on @xmath15 and relevant physical quantities due to the presence of an external magnetic field . we will use the model proposed in ref . 3 , which is used in most of the recent theoretical work in this area . in this model , the impurities are arranged on a regular square ( cubic ) lattice in 2d ( 3d ) and a nearest - neighbor tight - binding anderson hamiltonian is employed . in this work we investigate the qi of strongly localized electrons by doing _ summations over _ all _ forward - scattering paths between two _ arbitrary _ sites . we derive compact _ closed - form _ expressions for various physical quantities ( e.g. , the transition strength which determines the conductivity ) for electrons propagating on a square lattice subject to an external magnetic field , with and without random impurities . we also obtain an explicit formula for an experimentally important case that has been much less studied theoretically so far : the interference between paths on a 3d cubic lattice . in the disordered case , by analytically computing the moments for the tunneling probability and employing the replica method , we derive _ analytic _ results for the magnetoconductance ( mc ) in terms of _ sums - over - paths _ , which are applicable in _ any dimension_. our explicit field - dependent expressions for the mc provide a precise description of the mc in terms of the magnetic flux . a _ positive _ mc , with a saturation value slightly larger than twice the mc at zero field , is observed when turning on the field @xmath0 . in 2d , the saturated value of the magnetic field @xmath17 ( i.e. , the first field that makes the mc become twice the value at zero field ) is inversely proportional to twice the hopping length : the larger the system is , the smaller @xmath17 will be . in other words , as soon as the system , with hopping distance @xmath2 , is penetrated by a total flux of @xmath18 , the mc reaches the saturation value . the period of oscillation of the mc is found to be equal to @xmath1 , which is the superconducting flux quantum . furthermore , at very small fields , for two sites diagonally separated a distance @xmath2 , the mc scales as follows : ( i ) @xmath19 for quasi-1d ladder - type geometries with @xmath5 , ( ii ) @xmath20 in 2d with @xmath5 , ( iii ) @xmath19 in 3d with @xmath0 parallel to the @xmath7 direction , and ( iv ) @xmath20 in 3d with @xmath0 perpendicular to the @xmath7 direction . the general expressions presented here : ( i ) _ contain , as particular cases _ , several qi results@xcite derived during the past decade ( often by using either numerical or approximate methods ) , ( ii ) include qi to arbitrary points @xmath21 on a square lattice , instead of only diagonal sites @xmath22 , ( iii ) focus on 2d _ and _ 3d lattices , and ( iv ) can be extended to also include _ backward _ excursions ( e.g. , side windings ) in the directed paths . exact results in this class of directed - paths problems are valuable and can be useful when studying other systems , for instance : ( 1 ) directed polymers in a disordered substrate ( see , e.g. , refs . 12 and 13 ) , ( 2 ) interfaces in 2d ( see , for instance , ref . 14 ) , ( 3 ) light propagation in random media,@xcite and ( 4 ) charged bosons in 1d.@xcite to study the magnetic field effects on the tunneling probability of strongly localized electrons , we start from the tight - binding hamiltonian @xmath23 where @xmath24 refers to the nearest - neighbor sites and the phase @xmath25 is @xmath26 times the line integral of the vector potential along the bond from @xmath13 to @xmath27 in units of the normal flux quantum @xmath8 . in the strongly localized regime , @xmath28 , the electron energy can be set to zero.@xcite consider two states , localized at sites @xmath13 and @xmath14 which are @xmath2 bonds apart . by using a locator expansion , the transition amplitude ( green s function ) @xmath15 between these two states can be expressed as@xcite @xmath29 where @xmath30 and @xmath31 is the sum over phases of the bonds on the path @xmath32 of @xmath33 steps connecting sites @xmath13 and @xmath14 . in general , @xmath32 contains closed loops . in the strongly localized regime ( i.e. , @xmath28 ) , the dominant contribution to @xmath15 is @xmath34 , where @xmath35 in other words , only the shortest - length paths ( with _ no _ backward excursions ) connecting them are taken into account , namely , the _ directed - path model _ of refs . 2 - 8 this directed - path model provides an excellent approximation to @xmath15 since @xmath36 is quite small in the extremely localized regime.@xcite it is important to stress that the _ conductivity_@xcite between @xmath13 and @xmath14 is proportional to @xmath37 . quantum interference , contained in @xmath38 , arises because the phase factors of different paths connecting the initial and final sites interfere with each other . we will first focus on the computation of @xmath39 , which is the essential qi quantity for electrons deep in the localized regime . in 2d , we also analytically compute @xmath40 which becomes important when electrons are not so strongly localized . this paper is organized as follows . in sec . ii , we study qi on a square lattice under a uniform potential , which is related to the decay of gap states into the bulk.@xcite here we derive an elegant , general , and very compact closed - form expression for @xmath39 . intriguing properties associated with the behavior of @xmath39 on diagonal sites are discussed in detail . it will be shown later ( in section iii ) that the effect of a magnetic field on the mc is governed by the behavior of @xmath39 . as a step towards the understanding of interference between non - directed paths , we also go beyond the directed - path model by exactly computing analytic results for @xmath40 . in sec . iii , we investigate the tunneling in a random impurity potential , which is relevant to the conductivity of , for example , lightly doped semiconductors and strongly disordered compounds.@xcite closed - form results for the tunneling probability , which determines the conductivity , are obtained . we then analytically compute the moments for the tunneling probability . from them , we derive analytic field - dependent expressions , valid in any dimension , for the mc . the full behavior of the mc as a function of the magnetic flux including the scaling in the low - field limit and the occurrence of saturation is discussed in detail . the close relationship between the qi quantity @xmath39 and the corresponding mc is illustrated . comparison of our results with experimental observation and other theoretical work is also made . in sec . iv , we examine the qi on a 3d cubic lattice and provide a general formula for @xmath39 . we show how the interference patterns and the small - field behavior of the mc vary according to the orientation of the applied field . furthermore , we investigate the effect on the low - flux mc due to the randomness the angles between the directions of the critical hops and the orientation of the applied field . in sec . v , we conclude by addressing several relevant issues and summarize our results . let @xmath21 denote the site coordinates . without loss of generality , we choose @xmath41 to be the initial site and focus on @xmath42 . for forward - scattering paths of @xmath2 steps , which exclude backward excursions ( i.e. , only moving upward and to the right is allowed ) , ending sites @xmath21 satisfy @xmath43 . let @xmath44 ( = @xmath39 ) be the sum over all directed paths of @xmath2 steps on which an electron can hop from the origin to the site @xmath21 , each one weighted by its corresponding phase factor . employing the symmetric gauge @xmath45 and denoting the flux through an elementary plaquette ( i.e. , with an area corresponding to the square of the average distance , which is typically equal to or larger than the localization length @xmath9 , between two impurities ) by @xmath46 , it is straightforward to construct the recursion relation : @xmath47 this equation states that the site @xmath21 can be reached by taking the @xmath2th step from neighboring sites to the left or below . the factors in front of the @xmath48 s , namely , @xmath49 and @xmath50 , account for the presence of the magnetic field . enumerating the recursion relations for @xmath51 ( @xmath52 ) successively and using @xmath53 for any @xmath54 , we obtain the following relation @xmath55 here @xmath44 is expressed as a sum of the @xmath48 s one row below ( i.e. , on the line @xmath56 ) . the physical meaning of eq . ( 6 ) is clear : the site @xmath21 can be reached by moving one step upward from sites @xmath57 with @xmath58 , acquiring the phase @xmath59 ; then traversing @xmath60 steps from @xmath61 to @xmath21 , each step with a phase @xmath62 . by applying eq . ( 6 ) recursively ( and utilizing @xmath63 for any @xmath64 ) , @xmath44 for @xmath65 can be written as @xmath66 where @xmath67 with @xmath68 . if we use the landau gauge instead , the expression for the sum - over - paths @xmath39 will read @xmath69 ; namely , @xmath44 employs the _ s_ymmetric gauge , while @xmath69 uses the _ l_andau gauge . notice that each term in the summand corresponds to the overall phase factor associated with a directed path . in the absence of the magnetic flux ( @xmath70 ) , @xmath71 which is just the total number of @xmath2-step paths between @xmath41 and @xmath21 . after some calculations we obtain one of our main results , a very compact and elegant closed - form expression for @xmath72 : @xmath73 where @xmath74 notice that the symmetry @xmath75 [ apparent in eq . ( 10 ) ] is due to the square lattice geometry . moreover , we also obtain @xmath76 \left[\,\prod_{k=1}^{n}\left(1-e^{ik\phi}\right)\right]}\,.\ ] ] previous work on qi in the vrh regime obtained particular cases , mostly numerical , of sums to diagonal points @xmath77 , while the general result eq . ( 10 ) is valid for arbitrary ( e.g. , non - diagonal ) sites . to illustrate the quantum interference originating from sums over phase factors associated with directed paths , we show the five possible ending sites @xmath21 for @xmath78 and their corresponding @xmath44 in fig . in addition , the six different paths connecting @xmath41 and @xmath79 and their separate phase factor contributions to @xmath80 are shown in fig . 1(b ) . in the very - low - flux limit @xmath81 , the logarithm of @xmath44 , calculated exactly to order @xmath82 , is @xmath83 thus we obtain the familiar@xcite harmonic shrinkage of the wave function with _ expressions for all the prefactors . this result can be interpreted as follows . the effective cigar - shape " area exposed to the field has an effective length @xmath84 : @xmath85 ( i.e. , the square root of the area enclosed by the paths ) and an effective width @xmath86 : @xmath87 ( i.e. , the square root of the length of the paths ) . for the special case @xmath88 , @xmath89^{2}}-\frac{1}{96}r^{2}(r+1)\,\phi^{2}\,,\ ] ] which is consistent with , and generalizes , the results in ref . 4 since it gives the exact prefactor . thus , the effective length is @xmath90 , while the effective width is @xmath91 . furthermore , for a ladder - type quasi-1d system , ( e.g. , @xmath92 and @xmath93 ) , we have @xmath94 in this case , the effective length and width are both @xmath95 . this result remains valid for small values of @xmath54 ( narrow stripes or multiladders ) . the fourth - order contribution to @xmath96 can also be computed exactly as @xmath97\,\phi^4.&\end{aligned}\ ] ] among the @xmath48 s for an even number of steps , those located along the diagonal corners contain the richest interference effects since the number of paths ending at @xmath22 and the areas they enclose are both the largest . we therefore examine more closely the behavior of the quantities @xmath98 for irrational flux @xmath99 , it can be proved that @xmath100 for any @xmath64 . a _ particular case _ ( asymptotic behavior ) of our very compact general expression eq . ( 16 ) for @xmath101 is investigated in detail by fishman , shapir , and wang in ref . for @xmath102 ( @xmath103 and @xmath104 are positive integers with @xmath105 and @xmath103 being prime to @xmath104 ) , we obtain ( @xmath106 ) for @xmath107 @xmath108 and @xmath109 furthermore , for those @xmath64 satisfying @xmath110 @xmath111 in other words , the zeros of @xmath112 are given by @xmath102 for @xmath113 with @xmath114 and the @xmath103 s are prime to each allowed @xmath104 . from a physical viewpoint , these flux values produce the complete cancelation of all phase factors ( i.e. , _ fully destructive interference _ ) and result in the _ vanishing _ of the tunneling probability ( and conductivity ) . indeed , we will see in section iii that if we also consider the effects of the on - site impurity scattering , these flux values lead to the largest ( i.e. , saturated ) value for the positive mc . in fig . 2 , we show the zeros for @xmath115 obtained by using eqs . ( 21 ) and ( 22 ) . note that the smallest value of @xmath46 satisfying @xmath116 is always @xmath117 and the number of zeros rapidly increases when @xmath64 becomes larger . the @xmath101 can be expressed as sums of trigonometric cosines . for instance , the first few ones are ( with @xmath118 ) : @xmath119 notice that @xmath101 depends only on the even ( odd ) harmonics of @xmath120 when @xmath64 is even ( odd ) . @xmath112 obeys the following properties : ( i ) @xmath26 ( @xmath121 ) periodicity in @xmath99 for even ( odd ) @xmath64 , namely , @xmath122 in other words , the period of @xmath101 corresponds to @xmath8 when @xmath64 is even , and @xmath123 when @xmath64 is odd . ( ii ) with @xmath64 even @xmath124 for @xmath125 . also @xmath126 ^ 2}.\ ] ] ( iii ) with @xmath64 odd @xmath127 for @xmath128 . from the properties described above , we can draw a general picture of the behavior of @xmath101 . let @xmath129 . @xmath130 , which is a enormous number for large @xmath64 . as the magnetic field is turned on , @xmath101 rapidly drops to its first zero at @xmath131 . @xmath101 then shows distinct behaviors depending on @xmath64 . for even @xmath64 , @xmath101 exhibits many small - magnitude fluctuations around zero for @xmath132 . @xmath101 then monotonically climbs from @xmath133 to a large positive value , @xmath134 ^ 2 $ ] , for @xmath135 . it is evident that @xmath136 is still very small compared to @xmath137 . within the period @xmath138 ( i.e. , @xmath128 ) , @xmath101 has mirror symmetry with respect to @xmath139 ( i.e. , @xmath140 ) . for odd @xmath64 , @xmath101 exhibits many small - magnitude fluctuations around zero for @xmath141 . in addition , @xmath101 always equals @xmath133 at @xmath139 ( i.e. , @xmath140 ) . @xmath101 then monotonically drops from @xmath133 to @xmath142 for @xmath143 . for @xmath138 , @xmath101 has inversion symmetry with respect to @xmath139 . within the period @xmath144 ( i.e. , @xmath145 ) , @xmath101 has mirror symmetry around @xmath146 ( i.e. , @xmath147 ) . recall that for any @xmath64 , @xmath100 , for irrational values of @xmath148 . in fig . 3 , we plot @xmath149 through @xmath150 , @xmath151 , @xmath152 , @xmath153 and @xmath154 . these figures show very interesting interference patterns of @xmath101 and clearly reflect the general description given above . it is worthwhile to keep in mind that the properties embedded in @xmath44 described above play a central role in determining the behavior of the mc obtained in section iii . up to now , we have focused on the computation of @xmath39 and presented a detailed investigation of their properties . when the electrons are less strongly localized , the next higher - order contribution to @xmath15 ( i.e. , @xmath155 , which is the dominant term _ including backward excursions _ ) becomes important . therefore , quantum interference effects between phase factors of paths with backward recursions ( i.e. , moving downward and to the left is also included ) need to be taken into consideration . notice that paths in @xmath40 , though include backscattering processes , do not involve closed loops enclosing flux ( e.g. , elementary square plaquettes ) . in this section we present the computation of the second - order contribution , namely @xmath40 , to the transition amplitude @xmath15 . let @xmath156 ( @xmath157 ) denote the sums over paths of @xmath158 steps starting from @xmath41 and ending at @xmath21 . we assume that the electrons are confined on a square lattice with non - negative @xmath159 and @xmath160 coordinates . we can divide the contribution to @xmath156 into five parts . first , hopping directly to site @xmath161 , with @xmath162 , electrons take one step back to @xmath163 , then hop @xmath164 steps to @xmath21 . second , hopping directly to site @xmath165 , with @xmath166 , electrons take one step back to @xmath167 , then hop @xmath168 steps to @xmath21 . third , directly hopping to site @xmath169 , with @xmath170 , electrons move one - step downward to @xmath171 gaining a phase factor @xmath172 , then hop @xmath173 steps to @xmath21 . fourth , directly hopping to site @xmath174 , with @xmath175 , electrons move one - step downward to @xmath176 gaining a phase factor @xmath177 , then hop @xmath178 steps to @xmath21 . fifth , directly hopping to @xmath179 with @xmath180 and @xmath181 , electrons take one step back to either @xmath182 or @xmath183 , accompanied by the phase factor @xmath177 or @xmath172 , then hop @xmath184 steps to the ending site @xmath21 . therefore @xmath156 can be written as @xmath185 where @xmath186 is the sum over phase factors of all directed paths ( i.e. , containing @xmath187 steps ) starting from @xmath179 and ending at @xmath21 . after some calculation we obtain @xmath188\phi}{2}\right\ } \,s_{m - p , n - q}.\ ] ] by substituting eq . ( 24 ) into eq . ( 23 ) , we derive @xmath189\phi/2}\,s_{m - p+1,n - q } + e^{-i[m(q-1)-(n-1)p]\phi/2}\,s_{m - p , n - q+1}\right\}. \end{aligned}\ ] ] in the special case @xmath190 , @xmath191\,s_{j , m } + 2\,\sum_{j=0}^{m}\cos\left[\frac{(m^2-mj+j)\phi}{2}\right]\,s_{j , m+1 } + 2\,\sum_{j=0}^{m-1}\cos\left(\frac{j\phi}{2}\right)s_{m - j , m - j}\,s_{j , j+1 } \nonumber \\ & & \mbox{}+2\,\sum_{j=1}^{m-1}\sum_{k=0}^{j-1}s_{m - j , m - k}\left\ { \cos\left[\frac{(mj - mk - j)\phi}{2}\right]\,s_{j , k+1 } + \cos\left[\frac{(mj - mk+k)\phi}{2}\right]\,s_{k , j+1}\right\}. \end{aligned}\ ] ] the explicit expressions for the first few @xmath192 are ( with @xmath193 ) : @xmath194 these @xmath192 s are plotted in fig . note that @xmath192 depends only on the even ( odd ) harmonics of @xmath120 and has a period @xmath26 ( @xmath121 ) for even ( odd ) @xmath64 . the expressions for @xmath192 are obviously more complicated than the corresponding @xmath101 . however , by comparing figs . 3 and 4 , we find that the general features in the interference behaviors are surprisingly similar . we thus infer that the relevant physical quantities are not significantly changed by the addition of interference between the dominant winding paths . to incorporate the effects of random impurities , we now replace the on - site energy part in eq . ( 1 ) ( first term in @xmath195 ) by @xmath196 , where the @xmath197 s are now independent random variables . the hamiltonian now takes the form @xmath198 we have studied two commonly used models : ( i ) @xmath197 can take two values : @xmath199 and @xmath200 with equal probability ; and ( ii ) @xmath197 is randomly chosen from a uniform distribution of width @xmath201 and zero mean . we found that both models yield the same results for the mc . we now start with the general case of the first model , namely , @xmath197 can have two values : @xmath199 with probability @xmath202 and @xmath200 with probability @xmath203 , where @xmath204 . due to disorder , the transition amplitude becomes @xmath205 with @xmath206e^{i\phi_{\gamma}},\ ] ] where @xmath32 runs over all directed paths of @xmath2 steps connecting sites @xmath41 and @xmath21 , and @xmath27 over sites on each path . for all directed paths ending at @xmath21 , electrons traverse @xmath207 sites ( the initial site @xmath41 is excluded ) . each site visited now contributes an additional multiplicative factor of either @xmath208 or @xmath209 to the phase factor . therefore , for a given path @xmath32 , the probability for obtaining @xmath210 is @xmath211 it is then clear that @xmath212 where @xmath213 denotes averaging over all impurities . by exploiting eqs . ( 7 ) and ( 8) , we derive the following general expressions valid for _ any _ @xmath202 and @xmath203 , @xmath214 where @xmath215 also , the disorder average of the tunneling probability ( i.e. , the transmission rate ) @xmath216 is @xmath217 the physical origin of eqs . ( 29 ) and ( 31 ) becomes clearer by rewriting them as @xmath218\ ] ] and @xmath219,\ ] ] where @xmath220 and we have used @xmath221 the first terms in eqs . ( 32 ) and ( 33 ) account for contributions from pairs of identical paths : @xmath222 in @xmath223 , and @xmath224 in @xmath225 . the second terms in eqs . ( 32 ) and ( 33 ) account for contributions from pairs of distinct paths . note that @xmath226 and @xmath227 , when @xmath228 . we then have in the absence of magnetic flux @xmath229 furthermore , in the special case @xmath230 , since @xmath231 we then obtain @xmath232 for @xmath230 ( the most studied case so far ) , we can obtain analytical expressions for the moments @xmath233 and @xmath234 for any value of @xmath235 . only a few of these will be presented here . from now on , @xmath236 stands for @xmath237 and @xmath238 stands for @xmath72 . the derivation of these moments is given in appendix a. @xmath239 and @xmath240 these moments satisfy the consistency check @xmath241 and odd moments vanish by symmetry . the moments provide an analytical view of the structure of the qi in the tunneling process . since @xmath216 , each @xmath242 represents @xmath243 _ forward _ paths to @xmath21 , each one with its corresponding _ reversed _ path back to the origin . also , @xmath244 averages over the contributions of @xmath245 such pairs of paths . in general , @xmath246 consist of terms involving @xmath247 the above explicit expressions for the moments will allow us to deduce general formulae for the first few leading ( i.e. , dominant ) terms in the moments . we first focus on the leading terms ( @xmath248 ) , since they provide the most significant contribution to the moments when @xmath243 is large . recall that @xmath249 , therefore we need to consider all terms involving @xmath250 in @xmath246 . we derive ( see appendix b for more details ) @xmath251^{2k}\right\}. \end{aligned}\ ] ] furthermore , by considering all the second leading terms ( @xmath252 ) , we obtain @xmath253 \,n^{p-1 } , \\ \langle\ , |j(\phi)|^{2p } \,\rangle & = & -\frac{p!\,n^{p-1}}{6}\ , \left\{\sum_{k=0}^{\infty } \frac{(2k+1)!}{(2^k\,k!)^{2}}\left [ ( 3p+2k-3)\,c^{p}_{2k+1 } + 2k\,c^{p}_{2k+2}\frac{s(4\phi)}{n}\right]\left [ \frac{s(2\phi)}{n}\right]^{2k}\right\}. \end{aligned}\ ] ] also , when @xmath254 with @xmath255 ( e.g. , at @xmath256 , @xmath257 ) , the third leading terms ( @xmath258 ) in the moments are @xmath259\,n^{p-2}\ , , \\ \langle\ , |j(\phi)|^{2p } \,\rangle & = & \frac{1}{72}\,p\,(p-1)\,(p-2 ) \,(9p+5)\,(p!)\,n^{p-2}\,.\end{aligned}\ ] ] the above general expressions for the moments are of value since they enable us to _ analytically _ obtain the dominant contributions to the quantity we are interested in : the magnetoconductance . we now use the replica method : @xmath260 to compute the log - averaged mc with respect to the zero - field log - averaged mc ( denoted by @xmath261 ) , defined as @xmath262 taking into account only the first leading terms in the moments , shown in eqs . ( 37 ) and ( 38 ) , we derive the @xmath261 as @xmath263^{2k},\ ] ] where we have @xmath264 exploiting the following identity@xcite for @xmath265 @xmath266 which reduces to eq . ( 46 ) for @xmath267 , we thus obtain a very concise exact expression for the @xmath261 as @xmath268.\end{aligned}\ ] ] the typical mc of a sample @xmath269 is then given by , normalized by the zero - field mc @xmath270 , @xmath271 ^ 2}\,.\ ] ] eq . ( 49 ) is one of our main results . it provides a concise closed - form expression for the mc , as an _ explicit _ function of the magnetic flux . from eq . ( 49 ) it becomes evident that a magnetic field leads to an increase in the _ positive _ mc : @xmath272 increases from @xmath273 to a saturated value @xmath274 ( since @xmath275 decreases from @xmath243 to @xmath133 ) when the flux is turned on and increased . @xmath276 at the field @xmath99 that satisfies @xmath277 . furthermore , it is clear that the mc varies _ periodically _ with the magnetic field and the periodicity in the flux is equal to the superconducting flux quantum @xmath1 . it is important to point out that eqs . ( 48 ) and ( 49 ) are valid in any dimension as long as we use the corresponding d - dimensional sum @xmath39 . it is illuminating to draw attention to the close relationship between the behaviors of @xmath278 and the corresponding @xmath279 . when @xmath228 , @xmath280 , which is the _ largest _ value of @xmath281 as a function of @xmath99 , and the mc is equal to the _ smallest _ value @xmath270 . when the magnetic field is increased from zero , @xmath281 quickly approaches ( more rapidly as @xmath64 becomes larger ) its _ smallest _ value , which is zero , at @xmath282 . at the same time , the mc rapidly increases to the _ largest _ value @xmath283 . the physical implication of this is clear : fully constructive interference in the case without disorder leads to the smallest hopping conduction in the presence of disorder . while fully destructive interference in the case without disorder yields the largest hopping conduction in the presence of disorder . moreover , when @xmath64 ( the system size is @xmath284 ) is large , @xmath272 remains in the close vicinity of @xmath274 for @xmath285 in spite of the strong very - small - magnitude fluctuations of @xmath286 around zero . the saturated value of the magnetic field @xmath17 ( i.e. , the first field that makes @xmath276 ) is inversely proportional to twice the hopping length : the larger the system is , the smaller @xmath17 will be . in other words , as soon as the system , with hopping distance @xmath287 , is penetrated by a total flux of @xmath288 ( in units of the flux quantum @xmath8 ) , the mc reaches the saturation value @xmath283 . defining the relative mc , @xmath289 , as @xmath290 and utilizing eq . ( 49 ) , we show @xmath289 versus @xmath99 for several different hopping lengths in fig . the behavior of @xmath289 described above can be clearly observed in these figures . now let us examine the behavior of the mc in the low - flux limit . from eqs . ( 14 ) and ( 15 ) , it follows then that , for very small fields , in 2d @xmath291 and in ladder - type quasi-1d structures @xmath292 in fig . 6 , we plot @xmath289 computed directly from eq . ( 49 ) , for various small values of @xmath99 , versus @xmath293 , with @xmath294 , in ( a ) and versus @xmath295 , with @xmath296 , in ( b ) , respectively for 2d and quasi-1d systems . it is seen that , both in ( a ) and ( b ) , all the data nicely collapses into a straight line , which verifies the scaling of the low - flux mc in eqs . ( 50 ) and ( 51 ) . if we consider the second leading terms in the moments , namely eqs . ( 39 ) and ( 40 ) , the second - order contribution to the @xmath261 is @xmath297\right\ } & \mbox{when $ s(2\phi)/n\neq \pm 1 $ } \end{array } \right . \nonumber \\ & = & \left\ { \begin{array}{ll } 0 & \mbox{when $ \lambda=0 $ } \\ \frac{1}{6n}\,\left\{1-\frac{1-\lambda}{\lambda^3(1+\lambda ) } \left[\left(1-\frac{s(4\phi)}{n}\right ) ( 1 + 2\lambda)-\lambda^2(2 + 3\lambda)\right]\right\ } & \mbox{when $ \lambda\neq 0 $ } \end{array } \right.,\end{aligned}\ ] ] where @xmath298 and we have used @xmath299 the principal features in the behavior of the mc are not significantly modified by the addition of the contribution from @xmath261 in eq . ( 52 ) : while the magnitude of @xmath289 is slightly increased for @xmath300 , the period of the mc remains unchanged . for a 2d system and in the low - flux limit , we derive from eq . ( 52 ) for diagonal sites @xmath301 : @xmath302 comparing this result with eq . ( 50 ) , we see that the dependence of the small - field mc on the hopping length and the field is the same except for different prefactors . summing up both contributions , we have for small @xmath99 @xmath303 in addition , when @xmath254 , we have from eqs . ( 37 - 42 ) @xmath304 this indicates that the magnitude of the positive mc is gradually increased ( e.g. , the saturation value of @xmath305 is raised above @xmath273 ) when contributions from higher - order terms ( i.e. , terms @xmath306 with @xmath307 ) are included , though they are negligibly small . our results for the mc are in good agreement with experimental measurements.@xcite for instance , a positive mc is observed in the vrh regime of both macroscopically large in@xmath308o@xmath309 samples@xcite and compensated n - type cdse.@xcite moreover , saturation in the mc as the field is increased is also reported in ref . the results for @xmath289 presented in this work are consistent with theoretical studies based on an independent - directed - path formalism@xcite and a random matrix theory of the transition strengths.@xcite the advantages of our results include : ( i ) they provide explicit expressions for the first two dominant contribution to the mc , as a function of the magnetic field ; ( ii ) they provide straightforward determination of the period of the oscillation of the mc ; ( iii ) they provide explicit scaling behaviors ( i.e. , the dependence on the hopping length and the orientation and strength of the field ) of the low - flux mc in quasi 1-d , 2d and 3d systems ; and ( iv ) they allow us to make quantitative comparison with experimental data . finally , it is important to emphasize that our analytic results [ eqs . ( 48 ) , ( 49 ) and ( 52 ) ] for the mc are equally applicable to _ any dimension _ , since the essential ingredient in our expressions is the qi quantity @xmath39 , which takes into account the dimensionality . in appendix b , we outline the computational scheme using the second model of disorder , i.e. , @xmath310 is randomly chosen from the interval @xmath311 $ ] . the moments obtained in this case are the same as those presented in eqs . ( 37 ) and ( 38 ) . therefore , the result for the mc remains unchanged . let @xmath312 ( @xmath313 in 3d ) be the sum over all phase factors associated with directed paths of @xmath314 steps along which an electron may hop from ( 0,0,0 ) to the site @xmath315 . again we assume @xmath64 , @xmath54 , and @xmath316 . in other words , electrons can now also hop in the positive @xmath317 direction . the vector potential of a general magnetic field @xmath318 can be written as @xmath319 also , @xmath320 , @xmath321 and @xmath322 represent the three fluxes through the respective elementary plaquettes on the @xmath323- , @xmath324- and @xmath325-planes . to compute @xmath312 , we start from the following recursion relation @xmath326 where @xmath327 is the sum over all directed paths starting from @xmath328 and ending at @xmath315 . the physical meaning of eq . ( 55 ) is as follows . the site @xmath315 is reached by taking one step from @xmath329 to @xmath328 , acquiring the phase @xmath330 , then traversing from @xmath328 to @xmath315 on the @xmath331 plane . after some calculation , we find that @xmath332\right\}\,s_{m - p , n - q}(c),\ ] ] where @xmath333 is defined as shown in eq . ( 10 ) . by applying eq . ( 56 ) @xmath334 times , we obtain a general formula of @xmath312 for @xmath335 in terms of the fluxes @xmath336 , @xmath337 and @xmath338 as @xmath339\,{\cal l}_{m , n , l}(a , b , c),\ ] ] where @xmath340\right\}\,l_{p_{j+1}-p_j , q_{j+1}-q_j}(c)\right]\right\ } \,l_{p_1,q_1}(c),\ ] ] with @xmath341 , @xmath342 , and the @xmath343 s are defined as in eq . ( 12 ) . it is clear that @xmath344 , @xmath345 , and @xmath346 . also , the following symmetries hold : @xmath347 when there is no magnetic flux , @xmath348 gives the total number of @xmath349-step paths connecting @xmath350 and @xmath315 . we have obtained explicit expressions for many @xmath312 , and here we explicitly present only the first few @xmath351 , since @xmath351 have long expressions for larger @xmath64 , @xmath54 , and @xmath334 . @xmath352 , \\ { \cal s}_{2,1,1}&=&2\left[\cos\frac{a}{2}+\cos\left(\frac{a}{2}-b\right ) + \cos\left(\frac{a}{2}-c\right)+\cos\left(\frac{a}{2}\pm b\mp c\right ) + \cos\left(\frac{a}{2}-b - c\right)\right ] , \\ { \cal s}_{2,2,1}&=&2 + 2\sum\ ! { ^{(1)}}\cos\alpha+4\cos(a - b)+2\cos(a - c ) + 2\cos(b - c)+2\cos(a-2c)+2\cos(b-2c ) \\ & & \mbox{}+2\cos(a - b\pm c)+2\cos(a+b-2c)+2\cos(a - b\pm 2c ) , \\ { \cal s}_{2,2,2}&=&6+\sum\ ! { ^{(1)}}\left[4\cos\alpha+2\cos2\alpha\right ] + \sum\ ! { ^{(2)}}\left[4\cos(\alpha-\beta)+4\cos2(\alpha-\beta ) + 2\cos(\alpha-2\beta)+2\cos(2\alpha-\beta)\right ] \\ & & \mbox{}+2\sum\ ! { ^{(3)}}\left[\cos(\alpha+\beta-\gamma ) + \cos2(\alpha+\beta-\gamma)+\cos(\alpha+\beta-2\gamma ) + \cos(\alpha\pm 2\beta\mp 2\gamma)\right].\end{aligned}\ ] ] here @xmath353 denote sums over @xmath354 ; @xmath355 ; and @xmath356 ; for @xmath357 , @xmath274 , and @xmath358 respectively ; and , for instance , the term @xmath359 means @xmath360 . in the very - low - flux limit , and calculated exactly to second - order in the flux , we obtain the logarithm of @xmath312 , the 3d analog of the harmonic shrinkage of the wave function , as @xmath361 . \nonumber\ ] ] this generalizes the 2d harmonic - shrinkage of the wave function obtained in eq . ( 13 ) . when @xmath362 , we have @xmath363 ^ 3}-\frac{1}{216 } \left\{r^{2}(a^2+b^2+c^2 ) + \frac{r^{3}}{3 } \left[(b - c)^2+(c - a)^2+(a - b)^2\right]\right\}. \nonumber\ ] ] these results generalize to 3d the 2d results obtained in sec . ii b. in order to see how the interference patterns vary according to the orientation of the applied field , we focus on @xmath364 ( i.e. , @xmath351 on the body diagonals ) . we now examine two special cases : @xmath365 and @xmath366 , namely fields parallel and perpendicular to the @xmath7 direction , respectively . their @xmath364 s are denoted respectively by @xmath367 and @xmath368 and have been computed to high orders . here we only present the first few : @xmath369 and @xmath370 where @xmath193 . it can be seen that @xmath367 and @xmath368 exhibit quite different behaviors as shown in fig . 7 where we plot @xmath371 through @xmath372 and @xmath373 through @xmath374 . notice that the period in @xmath99 for @xmath367 is @xmath26 ( @xmath121 ) for even ( odd ) @xmath64 , while the period for @xmath368 is @xmath121 for any @xmath64 . therefore , the periodicity for the mc in 3d is identical to that in 2d . we have also computed @xmath367 and @xmath368 ( @xmath375 ) for @xmath376 and @xmath377 and find that their behaviors are insensitive to the commensurability of @xmath99 , unlike the case on a square lattice . physically , this can be understood because two randomly chosen paths have a higher probability of crossing ( and thus interfering ) in 2d than in 3d ; thus making qi effects less pronounced in 3d than in 2d . a similar situation occurs classically ( e.g. , multiply - scattered light in a random medium@xcite ) . for a 3d system , the relative mc , @xmath378 , now reads @xmath379 ^ 2}\,.\ ] ] the above general expression is valid for any ending site as well as arbitrary orientation and strength of the magnetic field . from eq . ( 61 ) , in the small - field limit and at ending site @xmath380 , we have @xmath381}\,,\ ] ] which is applicable for any orientation of the field . below we focus on two special orientations of the field : @xmath382 and @xmath383 . for very small @xmath99 , we have from eq . ( 61 ) @xmath384 ^ 3 } -\frac{1}{144}\,r^{2}(r+1)\,\phi^{2}\,,\ ] ] and @xmath385 ^ 3 } -\frac{1}{72}\,r^{2}\,\phi^{2}\,.\ ] ] the 3d behavior of the low - flux mc thus becomes clear : for @xmath382 @xmath386 and for @xmath383 @xmath387 these results can be interpreted as follows : the effective area @xmath388 exposed to @xmath382 is larger , @xmath389 similar to the 2d case where @xmath390 ; while the effective area @xmath391 exposed to @xmath383 is smaller , @xmath392 thus closer to our quasi-1d ladder case with @xmath393 . as a numerical test of eqs . ( 66 ) and ( 67 ) , in fig . 8 we show @xmath305 calculated directly from eq . ( 62 ) , versus @xmath293 in ( a ) and versus @xmath295 in ( b ) , respectively for several values of @xmath382 and @xmath383 with hopping length @xmath394 . the collapse of all the data into a straight line verifies the scaling of the low - flux @xmath305 presented above . in a macroscopic sample , the conductance may be determined by _ a few _ ( instead of one , as considered before ) critical hopping events . as a result of this , the observed mc of the whole sample should be the average of the mc associated with these critical hops . thus , in 3d systems it is also important to take into account the randomness of the angles between the hopping direction and the orientation of the applied magnetic field.@xcite to theoretically investigate the effect of the average over angles on the mc , we consider all possible relative hopping directions with respect to that of the magnetic field , or equivalently , the continuously - varing orientation of the field with respect to a fixed hopping direction . we adopt the latter below : the ending site of all hopping events ( with the same hopping length @xmath2 ) is located at the diagonal point @xmath380 and the magnetic field can be adjusted between the parallel and perpendicular directions with respect to the vector @xmath395 . our interest here is in the mc averaged over angles , denoted by @xmath396 , in the low - field limit . recall that the magnetic field is @xmath397 , and from eq . ( 63 ) , we have @xmath398 where @xmath399 is the magnitude of the field and @xmath400 is the angle between @xmath0 and @xmath401 . by averaging over the angle @xmath400 , we obtain @xmath402 where @xmath403 is the complete elliptic integral of the second kind . when @xmath2 is large , @xmath404 and we therefore have @xmath405 equation ( 70 ) means that the dominant contribution to the mc stems from the critical hop which is _ perpendicular _ to the field . this is understandable through our earlier observation that the effective area enclosed by the electron is largest when @xmath0 is perpendicular to @xmath401 . from the above analysis , we conclude that in 3d macroscopic samples the low - field mc should in principle behave as @xmath4 . in closing we briefly address four issues . first , although relevant measurable quantities such as @xmath406 and @xmath407 are gauge - invariant , the transition amplitudes are gauge dependent . as an illustration , the transition amplitude will be @xmath69 [ eq . ( 12 ) ] if we use the landau gauge @xmath408 on a square lattice . the notation @xmath44 ( @xmath69 ) refers the use of the @xmath48ymmetric ( @xmath409andau ) gauge . similarly , the transition amplitude will read @xmath410 [ eq . ( 58 ) ] if we use the gauge @xmath411 on a cubic lattice . second , returns to the origin ( see , e.g. , refs . 1 and 22 - 24 ) become important for less strongly localized electrons , and their qi effects@xcite can be incorporated in our approach . third , the main limitations of our study in the case with impurities are the following : no inclusion of spin - orbit scattering effects ( for this see , e.g. , refs . 7 - 9 and references therein ) , and no explicit inclusion of the correlations between crossing paths , as discussed in refs . 4 and 7 . however , these correlations are negligible when spin - orbit scattering is present.@xcite fourth , besides analytical closed - form results in 2d , this work presents exact results for 3d systems , e.g. , @xmath412 [ eq . these results can provide further tests of the quantum interference effects . this can be done by measuring the mc of bulk samples ( which are small enough that only a single critical hop is allowed ) in various orientations of the field . by doing this , one can then _ determine _ the values of @xmath2 and @xmath400 ( and , hence , also the direction of this critical hop ) . therefore , the small - field behaviors of the mc with fields parallel and perpendicular to the direction of this critical hop can be measured and compared to our predictions . this could potentially be very useful . in summary , we present an investigation of quantum interference phenomena and the magnetic - field effects on the mc resulting from sums over directed paths on resistor networks in 2d and 3d . the principal results include : ( 1 ) an exact and explicit closed - form expression for the sum over forward - scattering paths @xmath39 to any site on a square lattice , which is the essential qi quantity in both uniform and disordered cases , ( 2 ) an explicit formula for @xmath39 for electrons hopping on a cubic lattice , ( 3 ) the low - flux behaviors of @xmath39 in both 2d and 3d , ( 4 ) the exact summation of the dominant winding paths in 2d , ( 5 ) compact , analytic results for the positive mc as explicit functions of the magnetic flux which are valid in any dimension , ( 6 ) the small - field behaviors of the mc in quasi-1d , 2d and 3d , and ( 7 ) an analytic result for the small - field mc in 3d incorporating the randomness in the relative angles between the hops and the applied field . they provide analytical and explicit closed - form results concerning the hopping transport of strongly localized electrons subject to an external magnetic field in the macroscopic regime . we hope that our results stimulate further work ( e.g. , inclusion of spin - orbit effects on lattice path integrals ) on exact results in 2d and 3d systems . it is a great pleasure to thank b.l . altshuler , m. kardar , e. medina , y. meir , y. shapir and j. stembridge for their very useful suggestions . f.n . acknowledges partial support from the ucsb institute for theoretical physics through the nsf grant no . phy89 - 04035 , the nato office of scientific affairs , and the office of the vice provost for academic and multicultural affairs , through a faculty career development award during the fall of 1994 . in this appendix , we outline the derivation of the moments shown in eqs . ( 32 ) , ( 33 ) , ( 35 ) , and ( 36 ) . first , let us derive eqs . ( 32 ) and ( 33 ) . let @xmath415 , where @xmath416 is the sum over all phases along path @xmath13 and @xmath417 . note that for _ every _ path , the probability for obtaining the overall phase factor @xmath418 is @xmath419 . now @xmath420 and @xmath421 where @xmath422 with probability @xmath419 . if the number of @xmath423 is @xmath424 , the overall probability is @xmath425 and there are @xmath426 combinations among @xmath427 ( @xmath428 ) . for @xmath429 and @xmath430 , there is only one combination producing @xmath431 for @xmath432 and also only one combination producing @xmath433 for @xmath242 . when @xmath434 , for @xmath432 there are @xmath435 combinations producing @xmath436 and @xmath437 combinations producing @xmath431 . also , when @xmath434 , for @xmath242 there are @xmath438 combinations producing @xmath439 and @xmath440 combinations producing @xmath441 . in @xmath438 , the factor @xmath274 comes from the two possible ways : @xmath442 and @xmath443 . @xmath444 comes from arranging @xmath445 s minus signs among the @xmath446 @xmath447 s left ( @xmath448 ) . therefore , the overall average is @xmath449 we thus have @xmath450 where we have used the relation @xmath451 similarly , @xmath452 by exploiting @xmath453 and @xmath454 we thus obtain eqs . ( 32 ) and ( 33 ) . for @xmath230 , ( namely , each site contributes either @xmath208 or @xmath209 with equal probability ) , @xmath455 for _ every _ path @xmath13 . the total number of sites that can be visited is @xmath456 ( the initial site is not counted ) . therefore , the total number of all possible impurity configurations is @xmath457 . let us now focus on the @xmath458 with @xmath459 the total number of configuration sets is @xmath460 since each @xmath461 can be either @xmath208 or @xmath209 with equal probability . among the @xmath457 impurity configurations , there are _ always _ @xmath462 producing a set of @xmath463 for _ every _ possible set of @xmath463 . for instance , @xmath464 and @xmath465 in the simplest case @xmath466 . among @xmath467 impurity configurations , two of them give @xmath468 , @xmath469 , @xmath470 , and @xmath471 . we thus have @xmath472,\ ] ] and similarly , @xmath473.\ ] ] the summation in @xmath474 is over all possible configuration sets of @xmath461 . in other words , the average of @xmath475 over disorder ( i.e. , @xmath476 ) means to sum over all possible @xmath477 for the desired quantity @xmath475 , and then divide the sum by @xmath460 . we thus obtain @xmath478 and @xmath479 noticing that @xmath480 , it can be derived that : @xmath481 , \\ \sum_{i\neq j}s_i^4\,s_j^2 & = & s(2\phi)\,s(4\phi)-s(6\phi ) , \\ \sum_{i\neq j}s_i^4\,s_j^4 & = & \frac{1}{2}[s^2(4\phi)-s(8\phi ) ] , \\ \sum_{i\neq j}s_i^6\,s_j^2 & = & s(2\phi)\,s(6\phi)-s(8\phi ) , \\ \sum_{i\neq j}s_i^8\,s_j^2 & = & s(2\phi)\,s(8\phi)-s(10\phi ) , \\ \sum_{i\neq j}s_i^6\,s_j^4 & = & s(4\phi)\,s(6\phi)-s(10\phi ) , \\ \sum_{i\neq j\neq k}s_i^2\,s_j^2\,s_k^2 & = & \frac{1}{6}\,s^3(2\phi)-\frac{1}{2}\,s(2\phi)\,s(4\phi ) + \frac{1}{3}\,s(6\phi ) , \\ \sum_{i\neq j\neq k}s_i^4\,s_j^2\,s_k^2 & = & \frac{1}{2}\,s^2(2\phi)\,s(4\phi)-s(2\phi)\,s(6\phi ) -\frac{1}{2}\,s^2(4\phi)+s(8\phi ) , \\ \sum_{i\neq j\neq k}s_i^6\,s_j^2\,s_k^2 & = & \frac{1}{2}\,s^{2}(2\phi ) \,s(6\phi)-s(2\phi)\,s(8\phi ) -\frac{1}{2}\,s(4\phi)\,s(6\phi)+s(10\phi ) , \\ \sum_{i\neq j\neq k}s_i^4\,s_j^4\,s_k^2 & = & \frac{1}{2}\,s(2\phi)\,s^{2}(4\phi)-\frac{1}{2}\,s(2\phi)\,s(8\phi ) -s(4\phi)\,s(6\phi)+s(10\phi ) , \\ \sum_{i\neq j\neq k\neq l}s_i^2\,s_j^2\,s_k^2\,s_l^2 & = & \frac{1}{24}\,s^4(2\phi)-\frac{1}{4}\,s^2(2\phi)\,s(4\phi ) + \frac{1}{3}\,s(2\phi)\,s(6\phi)+\frac{1}{8}\,s^2(4\phi ) -\frac{1}{4}\,s(8\phi),\\ \sum_{i\neq j\neq k\neq l}s_i^4\,s_j^2\,s_k^2\,s_l^2 & = & \frac{1}{6}\,s^{3}(2\phi)\,s(4\phi)-\frac{1}{2}\,s^2(2\phi)\,s(6\phi ) -\frac{1}{2}\,s(2\phi)\,s^2(4\phi ) , \\ & & \mbox+s(2\phi)\,s(8\phi)+\frac{5}{6}\,s(4\phi)\,s(6\phi)-s(10\phi ) , \\ \sum_{i\neq j\neq k\neq l\neq m}s_i^2\,s_j^2\,s_k^2\,s_l^2\,s_m^2 & = & \frac{1}{120}\,s^5(2\phi)-\frac{1}{12}\,s^3(2\phi)\,s(4\phi ) + \frac{1}{6}\,s^2(2\phi)\,s(6\phi)+\frac{1}{8}\,s(2\phi)\,s^2(4\phi ) \\ & & \mbox{}-\frac{1}{4}\,s(2\phi)\,s(8\phi)-\frac{1}{6}\,s(4\phi)\,s(6\phi ) + \frac{1}{5}\,s(10\phi ) , \\ \sum_{i\neq j}\frac{s_i^2}{s_j^2 } & = & s^2(2\phi)-n , \\ \sum_{i\neq j}\frac{s_i^4}{s_j^4}&= & s^2(4\phi)-n , \\ \sum_{i\neq j\neq k}\frac{s_i^4}{s_j^2\,s_k^2}&= & \sum_{i\neq j\neq k}\frac{s_j^2\,s_k^2}{s_i^4 } = \frac{1}{2}s^2(2\phi)\,s(4\phi)-s^2(2\phi)-\frac{1}{2}s^2(4\phi)+n , \\ \sum_{i\neq j\neq k\neq l}\frac{s_i^2\,s_j^2}{s_k^2\,s_l^2 } & = & \frac{1}{4}\,s^4(2\phi)-(n-2)\,s^2(2\phi)-\frac{1}{2}\,s^2(2\phi)\,s(4\phi ) + \frac{1}{4}\,s^2(4\phi ) + \frac{1}{2}n\,(n-3).\end{aligned}\ ] ] by utilizing the above relations , we have obtained the results for the moments presented in eqs . ( 35.1)-(35.4 ) and ( 36.1)-(36.4 ) . in this appendix , we study another model of diagonal disorder : @xmath197 uniformly distributed between @xmath482 and @xmath483 . our focus is on the analytical computation of the leading terms ( terms @xmath484 ) in the moments @xmath485 and @xmath486 . indeed , the scheme presented below applies equally to our first model of disorder , where @xmath310 can take two values , @xmath199 and @xmath200 , with equal probability . let us write @xmath487 where for each path @xmath13 @xmath488 it is reasonable to choose @xmath489 from a gaussian distribution of zero mean and unit standard deviation . we therefore have @xmath490 and @xmath491 . for @xmath228 , we have @xmath492 it is found that the leading term in @xmath493 comes from all terms having the form @xmath494 and there are @xmath495 distinct terms of this type , each term has a coefficient @xmath496 . therefore , we obtain for the leading term @xmath497 for @xmath300 , we have now @xmath498\,.\end{aligned}\ ] ] the contributions to the leading terms involve different factors , as shown below . there are @xmath495 terms like @xmath499 and they contribute @xmath500 there are @xmath501 terms like @xmath502 and they contribute @xmath503 similarly , there are @xmath504 terms like @xmath505 and they contribute @xmath506 in general , there are @xmath507 terms of the following type @xmath508 each one with a coefficient @xmath509 . the contribution to the moment is therefore @xmath510 notice that we have utilized the fact that @xmath511 always contains @xmath512 . totally , we thus obtain for the leading terms @xmath513^{2k}\right\}.\ ] ] for reviews and further references , see : _ hopping transport in solids _ , m. pollak and b.i . shklovskii , eds . ( elsevier science , 1991 ) ; _ hopping conduction in semiconductors _ , m. pollak and b.i . shklovskii , eds . ( north - holland , amsterdam , 1990 ) ; _ hopping and related phenomena _ , h. fritzche and m. pollak , eds . ( world scientific , 1990 ) ; y. shapir and x.r . wang , mod . phys . lett . b * 4 * , 1301 ( 1990 ) ; b.i . shklovskii and a.l . efros _ electronic properties of doped semiconductors _ ( springer - verlag , berlin , 1984 ) .

the kepler mission is dramatically increasing the number of planets known in multi - planetary systems . many adjacent planets have orbital period ratios near resonant values , with a tendency to be larger than required for exact first - order mean - motion resonances . this intriguing feature has been shown to be a natural outcome of orbital circularization of resonant planetary pairs due to star - planet tidal interactions . however , this feature holds in multi - planetary systems with periods longer than ten days , for which tidal circularization is unlikely to provide efficient divergent evolution of the planets orbits . gravitational interactions between planets and their parent protoplanetary disk may instead provide efficient divergent evolution . for a planet pair embedded in a disk , we show that interactions between a planet and the wake of its companion can reverse convergent migration , and significantly increase the period ratio from a near - resonant value . divergent evolution due to wake - planet interactions is particularly efficient when at least one of the planets opens a partial gap around its orbit . this mechanism could help account for the diversity of period ratios in kepler s multiple systems comprising super - earth to sub - jovian planets with periods greater than about ten days . diversity is also expected for pairs of planets massive enough to merge their gap . the efficiency of wake - planet interactions is then much reduced , but convergent migration may stall with a variety of period ratios depending on the density structure in the common gap . this is illustrated for the kepler-46 system , for which we reproduce the period ratio of kepler-46b and c.

1301.0779

planetary astrophysics is undergoing an epoch of explosive growth driven by the discovery of about 850 exoplanets in two decades . continuous improvement in detection techniques will uncover many more planets in the near future which are too small or too far from their host star to be detected at present . observations suggest that planet formation is ubiquitous , with more than 40% of nearby sun - like stars harboring at least one planet less massive than saturn @xcite . these prolific discoveries have revealed the amazing diversity of exoplanetary systems and indicated that planets may form under a wide range of conditions and have considerable mobility . the diversity of exoplanets has been particularly highlighted by the kepler s space mission . to date , kepler has detected over 2300 planet candidates @xcite , from which more than a hundred are confirmed planets ( http://kepler.nasa.gov ) . about one third of kepler s candidates are associated with multiple transiting systems and the vast majority of them are expected to be real multi - planetary systems @xcite . some of these systems show remarkable properties , like the five coplanar , alternating rocky and icy planets orbiting kepler-20 in less than 80 days @xcite . interactions between planets and their parent protoplanetary disk are likely to have played a major role in shaping such compact system . a striking feature of kepler s multi - planetary systems is the large number of planet pairs far from mean - motion resonances ( hereafter mmr , * ? ? ? planet pairs near resonances show some tendency to have period ratios slightly greater than resonant values ( see figure [ fig : kepler ] ; period ratios are defined as the outer planet s orbital period divided by the inner planet s ) . for instance , there are almost twice as many planet pairs with period ratios between 2.0 and 2.1 than between 1.9 to 2.0 . this intriguing feature could naturally arise from tidal orbital circularization of close - in resonant planet pairs @xcite . however , as stressed by the histograms in figure [ fig : kepler ] , the same trend holds for planet pairs where the orbital period of the inner planet exceeds 10 days , most particularly near first - order resonances . for such systems , tidal circularization is unlikely to have caused significant divergent evolution of the planets orbits , as shown in section [ sec : tidal ] of the appendix . an example of a planetary system with a period ratio slightly greater than resonant is the kepler-46 system ( previously known as koi-872 ) . the 0.9 solar - mass star is surrounded by a @xmath0 transiting planet on a 33.6 day period ( kepler-46b ) . large transit time variations ( ttv ) of kepler-46b are caused by the presence of an outer planet , kepler-46c , whose mass ( @xmath1 ) and orbital period ( 57.0 days ) have been determined by @xcite . the same authors have also reported the detection of a @xmath2 planet candidate on a 6.8 day period ( kepler-46d , not yet confirmed ) . the main properties of the kepler-46 planetary system are summarized in table [ tab : tablekoi ] . the low - eccentricity and quasi - coplanar orbits in this system suggest that the planets reached their present location most probably through disk - planet interactions . still , the ratio or orbital periods between kepler-46c and kepler-46b is @xmath3 , which is slightly greater than @xmath4 . capture in the 5:3 mmr is not a natural outcome of planet migration driven by disk - planet interactions . it requires a rate of convergent migration that is large enough to cross the 2:1 mmr , but not too large so that the planets do not reach their 3:2 mmr . capture into a second - order resonance also requires small but not zero eccentricities , which disk - planet interactions tend to damp quite efficiently . should disk - planet interactions have led kepler-46b and kepler-46c to reach their 5:3 mmr , it is unclear how their period ratio would have then increased to their present value . the initial motivation of this work is to investigate under what circumstances disk - planets interactions may account for the orbital period ratio between kepler-46c and kepler-46b . in section [ sec : hydro ] , we present results of hydrodynamical simulations modeling the early evolution of both planets as they were embedded in their parent protoplanetary disk . these saturn - mass planets are expected to open a partial gap around their orbit . a rapid convergent migration causes the planets to merge their gap and to evolve into a common gap . the planets convergent migration is found to stall with period ratios between 1.5 and 1.7 , depending on the disk s density profile inside the common gap . the observed period ratio can be reproduced without the planets being in resonance . furthermore , we show that disk - driven migration of partial gap - opening planets may lead to significant divergent evolution of planet pairs over typical disk lifetimes . such divergent evolution is shown to arise from the interaction between planets and the wakes of their companions . this mechanism is illustrated for super - earth - mass planets in section [ sec : results_nbody ] with both hydrodynamical simulations and customized three - body integrations . divergent evolution of gap - opening planet pairs mediated by wake - planet interactions could partly account for the diversity of period ratios in kepler s multi - planetary systems . concluding remarks are provided in section [ sec : conclusion ] . bottom : ratio of orbital periods of all planet pairs among kepler s candidate multi - planetary systems detected from quarters 1 to 6 ( x - axis ) versus the orbital period of the inner planet in each pair ( in days , y - axis ) . bins are @xmath5 wide . from left to right , vertical dashed lines show the location of the 5:4 , 4:3 , 3:2 , 5:3 and 2:1 mean - motion resonances . top : histograms of the orbital period ratio for all planet pairs ( black ) , and for planet pairs where the inner planet s period is greater than 10 days . data were extracted from http://planetquest.jpl.nasa.gov / kepler/. ] .[tab : tablekoi]kepler-46 planetary system @xcite [ cols="<,<,<,<",options="header " , ] we model in this section the early evolution of the kepler-46 planetary system when the planets were embedded in their parent protoplanetary disk . our study does not address the formation of the planets . we focus instead on the planets orbital evolution due to disk - planet and planet - planet interactions . two - dimensional hydrodynamical simulations of disk - planets interactions were carried out using the code fargo @xcite . the physical model and numerical setup of the hydrodynamical simulations are described in section [ sec : setup ] . results of simulations follow in section [ sec : results_hydro ] . _ disk model _ we adopt a two - dimensional disk model for the protoplanetary disk in which the planets orbiting kepler-46 formed . disk - planets interactions determine the rate of convergent migration of the planets and the damping rate of their eccentricity . both quantities are sensitive to the disk s surface density , temperature , turbulent viscosity and radiative properties . the parameter space is therefore particularly large , and the need to assess the planets orbital evolution over typically a thousand planet orbits has led us to make simplifying assumptions for the disk model and to fix some of the disk parameters . disk self - gravity is discarded given the low disk masses adopted in this work ( the toomre - q parameter associated with the unperturbed density profile does not go below 7 ) . a locally isothermal equation of state is used where the vertically - integrated pressure @xmath6 and density @xmath7 satisfy @xmath8 , with the sound speed @xmath9 being specified as a fixed function of the cylindrical radius @xmath10 . the sound speed is related to the disk s pressure scale height @xmath11 through @xmath12 with @xmath13 the keplerian angular velocity . the ( fixed ) temperature profile is taken proportional to @xmath14 so that the disk s aspect ratio @xmath15 is uniform . we take the standard value @xmath16 . a locally isothermal equation of state is generally not appropriate to model the migration of an embedded low - mass planet , as it underestimates the magnitude of the corotation torque that the planet experiences from the disk ; see @xcite for a recent review on planet migration . this is not an issue in this study , however , since we consider partial gap - opening planets for which the corotation torque has a rather weak effect on migration . this may be quantified through the dimensionless parameter @xmath17 ( @xmath18 denotes the planet - to - star mass ratio ) which is @xmath19 in our simulations . this lower limit implies that the direction and speed of planet migration are weakly sensitive to the choice for the equation of state @xcite . the effects of turbulence are modeled by a constant shear kinematic viscosity @xmath20 . we take @xmath21 in standard code units ( defined below ) which translates into a viscous alpha parameter @xmath22 at the initial location of the inner planet . this value of @xmath23 is typical of the magnetically active regions of protoplanetary disks with magneto - hydrodynamic turbulence driven by the magneto - rotational instability . the present orbital periods of the planets in the kepler-46 system imply that the planets have probably interacted with such active regions in their parent disk . note , however , that the disk s magnetic field is not included in our simulations . we also point out that the planets that we consider are of large enough mass so that small - scale turbulent fluctuations can be safely ignored when modeling the interaction with their parent disk . in other words , the viscous diffusion approximation is expected to be a reasonable approximation for the large planet masses we consider @xcite . the disk is set up in radial equilibrium , with the centrifugal acceleration and the radial acceleration related to the pressure gradient balancing the gravitational acceleration due to the central star . the initial gas surface density is taken proportional to @xmath24 . its value at the initial location of the inner planet is taken as a free parameter , which we denote by @xmath25 . + _ planet parameters _ we model the orbital evolution of the outer two planets in the kepler-46 system : kepler-46b ( hereafter refereed to as the inner planet ) and kepler-46c ( as the outer planet ) . the @xmath26 unconfirmed planet candidate with a period of 6.8 days is assumed to have had a negligible impact on the orbital evolution of kepler-46b and kepler-46c . the planets are assumed to have already formed at the beginning of our simulations , they therefore take their present mass . kepler-46b has a physical radius @xmath27 however , its mass could not be constrained from ttv ; see @xcite . the mass - radius diagram of exoplanets detected by transiting methods indicates that the mass of kepler-46b most probably lies between @xmath28 and @xmath29 . its planet - to - primary mass ratio ( @xmath30 ) is therefore varied in the range @xmath31 $ ] . the ttv signal of kepler-46b constrains the mass of kepler-46c to be @xmath32 @xcite . the planet - to - primary mass ratio of kepler-46c is thus fixed to @xmath33 . the initial orbital radius of kepler-46b , @xmath34 , is taken to be the code s unit of length . that of kepler-46c is arbitrarily set to @xmath35 , so that the ratio of orbital periods @xmath36 initially . for the disk aspect ratio and viscosity taken in our study , the two planets are expected to open a partial gap around their orbit @xcite . + _ numerical setup _ the hydrodynamical equations are solved in a cylindrical coordinate system @xmath37 centered on to the star with @xmath38r_{\rm inner}$ ] and @xmath39 $ ] . the computational grid has @xmath40 zones evenly spaced along the radial direction and @xmath41 azimuthal sectors . the frame rotates with the keplerian frequency at the inner planet s location , and the indirect terms that account for the acceleration of the central star by the disk and the planets are included in the equations of motion . a standard outflow boundary condition is taken at the grid s inner edge , while damping is used in a so - called wave killing - zone extending from @xmath42 to @xmath43 ( disk quantities are damped towards their instantaneous axisymmetric profile ) . to avoid a violent relaxation of the disk due to the sudden introduction of the planets , the mass of the planets is gradually increased over 10 orbital periods . the gravitational potential of the planets is smoothed over a softening length , @xmath44 , equal to @xmath45 with @xmath11 evaluated at the planets location . the calculation of the force exerted by the disk on the planets excludes the content of the planets circumplanetary disk , the size of which being about @xmath46 of the planets hill radius @xcite . the hill radius of each planet is resolved by about 10 grid cells along each direction initially . + _ code units _ results of simulations are expressed in the standard units of disk - planet calculations : the mass unit is the mass of the central star ( denoted by @xmath47 and equal to @xmath48 for the kepler-46 system ) . the length unit is the initial orbital radius of the inner planet ( @xmath34 ) , and the time unit is the initial orbital period of the inner planet ( @xmath49 ) divided by @xmath50 . whenever time is expressed in orbits , it refers to the orbital period at the initial location of the inner planet . the free parameters in our simulations are the mass of the inner planet and the initial surface density of the disk . both quantities determine the rate of convergent migration and the evolution of the orbital eccentricities . we consider three different masses for the inner planet : @xmath51 , @xmath52 and @xmath29 ( recall that the mass of the outer planet is fixed at @xmath53 ) . for each planet mass , a set of 9 simulations was carried out varying @xmath25 , the unperturbed surface density at the initial orbital radius of the inner planet . we took @xmath25 in the range @xmath54 $ ] , which corresponds to an initial disk mass between @xmath55 and @xmath56 . typical outcomes of the simulations are described in section [ sec : illust ] . we find that disk - planets interactions may lead to significant divergent evolution of the planet orbits , which we interpret in section [ sec : divergent ] . to obtain evolution of the planets orbital elements to a steady state , a simple model for disk dispersal is included and its results presented in section [ sec : disp ] . the robustness of our results to varying some physical and numerical parameters is discussed in section [ sec : robust ] of the appendix . figure [ fig : ex ] displays typical outcomes of our hydrodynamical simulations of the kepler-46 system . from left to right in the figure we show the time evolution of the planets semi - major axes , orbital period ratio , eccentricities and resonant angles . here @xmath57 and @xmath58 denote the mean longitude and longitude of periastron , with the subscripts i and o referring to the inner and outer planets , respectively . the top row shows results with @xmath59 and @xmath60 . as the planets progressively clear a partial gap around their orbit , convergent migration does not occur at a constant rate . in particular , the inward migration of the outer planet slightly accelerates at early times due to a stage of type iii migration @xcite before slowing down upon approaching the 2:1 mmr with the inner planet . the planets are locked in the 2:1 mmr from about 300 orbits onwards , as can be seen from the libration of the resonant angles associated with this resonance . the resonance capture increases the inner planet s eccentricity to about 0.035 . the increase in the outer planet s eccentricity is much more modest , reaching about @xmath61 , even though the mass ratio of the planets is near unity . this modest increase in the outer planet s eccentricity arises from the contribution of the indirect term nearly cancelling the contribution from the direct term to the term that is first order in the eccentricities in the development of the disturbing function , which governs the response of the outer planet at a 2:1 mmr ( see , e.g. , * ? ? ? * ; * ? ? ? damping of the eccentricities follows from the gravitational interaction with the background gas disk . interestingly , the ratio of orbital periods rises from 2.02 to 2.15 in only 1500 orbits , while the eccentricities are damped . the physical origin of this divergent evolution is discussed in section [ sec : divergent ] . note that resonant coupling is maintained throughout this divergent evolution , the width of the 2:1 mmr increasing as the eccentricities decrease . disk s surface density at 400 orbits for @xmath62 and @xmath60 ( top ) , and for @xmath63 and @xmath64 ( bottom ) . the outer planet s mass is @xmath65.,title="fig : " ] disk s surface density at 400 orbits for @xmath62 and @xmath60 ( top ) , and for @xmath63 and @xmath64 ( bottom ) . the outer planet s mass is @xmath65.,title="fig : " ] the second row of panels in figure [ fig : ex ] displays results for @xmath66 and @xmath67 . this case is particularly interesting as it shows that convergent migration may stall with an orbital period ratio very close to the observed value ( @xmath68 ) . the accelerating inward migration of the outer planet is particularly visible in the first 100 orbits . this runaway migration is fueled by the dense gas region located between the two planet gaps . this dense material is continuously funneled beyond the outer planet s orbit upon embarking on horseshoe u - turns with respect to the outer planet . the density between the two planet gaps therefore decreases , and so does the migration rate of the outer planet . convergent migration then proceeds at a slower pace and the planets end up merging their gap , thereby forming a common gap in the disk . this is illustrated in figure [ fig : dens ] , which compares the disk s surface density at 400 orbits for the two aforementioned simulations which have @xmath59 and @xmath69 ( top panel ) , and @xmath66 and @xmath67 ( bottom panel ) . while in the former case a narrow ring of gas is left between the two planet gaps , the latter case shows that both planets end up in a common gap . back to the second row of panels in figure [ fig : ex ] , we see that the orbital period ratio remains approximately stationary from about 400 orbits , and so do the eccentricities . the time - averaged eccentricity of the inner planet is close to 0.01 ; that of the outer planet is @xmath70 . these values are consistent , albeit slight larger than those inferred from observations ( see table [ tab : tablekoi ] ) . note from the critical angles that the planets are not locked in the 5:3 mmr . lastly , the third row of panels in figure [ fig : ex ] is for @xmath71 and @xmath72 . here , convergent migration is rapid enough that the planets approximately reach the nominal location of their 3:2 mmr ( the minimum value of the period ratio is @xmath73 at 170 orbits ) . the planet eccentricities increase to about 0.03 ( inner planet ) and 0.015 ( outer planet ) . subsequent damping of the eccentricities occurs along with a slight divergent evolution of the planets . the rate of divergent evolution is much reduced compared to the 2:1 mmr case shown in the top row of panels in figure [ fig : ex ] . we will analyse this result further in section [ sec : divergent ] . we also point out that no steady - state has been reached at the end of the simulation . for this simulation , as for all simulations of the kepler-46 system , results are shown until the inner planet reaches @xmath74 , below which the proximity of the grid s inner edge ( at @xmath75 ) starts affecting the planets orbital evolution . the final period ratio obtained in all our simulations is displayed in figure [ fig : finalopr ] as a function of @xmath25 for the three masses assumed for kepler-46b . by final period ratio , we mean either its value obtained at the end of the simulation ( after between 1500 and 2000 planet orbits ) , or that obtained when the inner planet s orbital radius falls below @xmath76 . figure [ fig : finalopr ] highlights the three major outcomes of our simulations : ( i ) capture in 2:1 mmr generally followed by rapid divergent evolution , ( ii ) capture in 3:2 mmr followed by slow divergent evolution , and ( iii ) convergent migration stalling with a period ratio between 1.6 and 1.7 , depending on the disk s density distribution inside the planets common gap . several simulations can reproduce the observed period ratio of kepler-46b and kepler-46c . the best agreement with observations is obtained for @xmath66 , our upper mass value . this mass would imply a mean density for kepler-46b about 20% larger than jupiter s ( that is , about 1.6 g @xmath77 ) . the results of hydrodynamical simulations in section [ sec : illust ] show that the orbital period ratio of two partial gap - opening planets may increase from a near resonant value . this divergent evolution is reminiscent of the late evolution of short - period planet pairs through star - planet tidal interactions . in that case , tidal circularization causes a slow divergent evolution of the orbits @xcite , a mechanism known as tidally - driven resonant repulsion @xcite . it is based on energy dissipation of the planet pair while its total angular momentum remains constant @xcite . do we have an analogous resonant repulsion mechanism due to disk - planets interactions ? to address this question , it is instructive to first look at the evolution of two embedded planets that do not interact gravitationally . in figure [ fig : fono ] we compare the results of two hydrodynamical simulations : one in which the gravitational interaction between the planets is switched off ( both the direct and the indirect interaction terms are discarded ) and another where the planet - planet interaction is fully accounted for . the inner planet s mass is @xmath78 and the initial disk surface density at @xmath79 is @xmath80 . the other disk and planet parameters are those described in section [ sec : setup ] . when planet - planet interactions are on , capture into the 2:1 mmr is followed by rapid divergent evolution and damping of the eccentricities . this case is similar to that presented in the top row of panels in figure [ fig : ex ] . quite surprisingly , we see that a very similar divergent evolution is also obtained when planet - planet interactions are switched off , while the planets are not resonantly coupled ( the resonant angles do not librate ) and their eccentricities take vanishingly small stationary values . this comparison highlights that , when disk - planet interactions are considered , eccentricity damping is not necessarily responsible for the planets divergent evolution . the divergent evolution of a planet pair may instead be driven by changes to the disk - planet interaction brought about through the proximity of the planets to each other , such as the development of an interaction between each planet and the wake of its companion . we argue below that the damping of the eccentricities can actually be a consequence of divergent evolution driven by wake - planet interaction . before describing the process of wake - planet interaction in more detail , we briefly report that , in some other simulations , not illustrated here , with planet - planet interactions switched off , a tiny increase in the orbital eccentricities is observed when the orbital period ratio takes near resonant values . eccentricities may reach a few @xmath81 before being damped by the disk . although this is a small effect , it suggests that planets could be weakly resonantly coupled through interaction with the wakes and/or the circumplanetary disk of their companions . in our simulations , the mass of the circumplanetary disks does not exceed one percent of the planets mass . + _ wake - planet interactions _ the results of simulations shown in figure [ fig : fono ] make clear that disk - driven divergent evolution of a planet pair is not necessarily driven by eccentricity damping and it may continue to operate even when the eccentricities are very small . we here indicate how the interaction between a planet and the wake of its companion could act as a driving mechanism for divergent evolution . consider for instance the inner planet . its outer wake carries positive fluxes of angular momentum and energy . a fraction of this angular momentum flux can be deposited in the coorbital region of the outer planet through the dissipation of shocks , and then transferred to the outer planet via an effective positive corotation torque ( see * ? ? ? associated with the angular momentum flux @xmath82 deposited in the coorbital region of the outer planet is an energy flux @xmath83 , where @xmath84 denotes the inner planet s mean motion ( or angular velocity ) . from this energy flux , @xmath85 is used to keep the outer planet and its coorbital region nearly circular ( see figure [ fig : fono ] ; here , @xmath86 is the outer planet s mean motion ) . the remaining energy flux , @xmath87 , is dissipated in the outer planet s coorbital region . reciprocally , a fraction of the negative inward flux of angular momentum carried by the inner wake of the outer planet can be deposited in the inner planet s coorbital region and subsequently transmitted to the inner planet . again , this exchange is associated with energy dissipation in the disk at a rate equal to the ( absolute value of ) the product of the flux of angular momentum transmitted to the inner planet by the outer one and the difference in the planets angular velocities . this dissipation occurs , as is usual in disk - planet interactions , through the tidally excited density waves steepening into shocks ( e.g. , * ? ? ? numerically , shocks are dealt with through a combination of numerical diffusion and the application of an artificial viscosity . as a locally isothermal equation of state is adopted in our simulations , this dissipated energy is effectively lost to the system through cooling . while the total orbital energy of the two planets is dissipated due to the wake - planet interactions described above , their total angular momentum remains unaltered . as studied in the context of star - planet tidal interactions , energy dissipation at constant angular momentum leads to the divergent evolution of a planet pair @xcite . we now derive an equation satisfied by the period ratio of the two planets as a direct consequence of considering the evolution of their total orbital energy and angular momentum . we denote by @xmath88 , @xmath89 and @xmath90 the semi - major axis , eccentricity and mass of the inner planet ( with the subscript changed to o denoting the same quantities for the outer planet ) . the total angular momentum of the two planets is @xmath91 which gives , correct to second order in the eccentricities , @xmath92 we are looking for changes in the planets semi - major axes and eccentricities that occur on similar timescales , we may therefore discard the eccentricity terms in the expressions for @xmath93 @xmath94 and @xmath95 . denoting by @xmath96 and @xmath97 the torques exerted by the disk on the inner and outer planets , respectively , conservation of angular momentum reads @xmath98 the total orbital energy is given by @xmath99 and conservation of energy reads @xmath100 where @xmath101 the planets mean motion . here the right hand side of eq . ( [ eq_e ] ) contains three contributions . the first arises from the energy dissipation in the disk resulting from the action of the torques @xmath102 as these are negative this energy is removed from the orbital motion . the second originates from orbital circularization due to disk - planet interactions . the quantities @xmath103 and @xmath104 are the circularization times for the planets that arise from the disk torque and energy dissipation acting on each planet through the effects of its own wake @xcite . the third and final contribution @xmath105 stems from the energy dissipation due to wake - planet interactions . we remark that wake dissipation in the coorbital region of the outer planet contributes @xmath106 to @xmath107 , in addition to a corresponding contribution arising from the coorbital region of the inner planet , as indicated above . note that terms of order @xmath108 are retained above even though the eccentricities are small as it is assumed that @xmath109 are small enough so that dissipation through circularization could be as important as , or even dominate the other effects under some circumstances . combining eqs . ( [ eq_j ] ) and ( [ eq_e ] ) , we find @xmath110 when wake - planet interactions are negligible ( @xmath111 ) , eq . ( [ eq_pr ] ) shows that , far from resonance , where the eccentricities are very small , the planet s period ratio @xmath112 decreases as a consequence of convergent migration , provided that @xmath113 as expected , the latter condition is that the inward migration rate of the outer planet should exceed the inward migration rate of the inner planet . as the eccentricities increase as resonance is approached , a steady state may be reached in which the planets period ratio and eccentricities take stationary values . if the eccentricity damping timescales are sufficiently short , disk - driven circularization of the orbits can reverse convergent migration , that is the planets period ratio can increase . ( [ eq_pr ] ) highlights that wake - planet interactions may also lead to divergent evolution of a planet pair . divergent evolution driven by wake - planet interactions does not require the planets to be resonantly coupled , as already inferred from the simulations shown in figure [ fig : fono ] . similar divergent evolution of a non - resonant planet pair embedded in a disk , comprising a jupiter and a super - earth , was reported by @xcite . when divergent evolution is primarily due to wake - planet interactions , as the period ratio of the planet pair increases away from a resonant value , the distance to resonance increases , and the forced eccentricities therefore decrease ( e.g. , * ? ? ? it implies that orbital circularization is a natural consequence of divergent evolution driven by wake - planet interactions . this is actually a major difference between the divergent evolutions driven by wake - planet interactions and by orbital circularization ( through the disk or stellar tides ) : eccentricity damping is the consequence of the former mechanism , while it is the cause of the latter . note that , with both mechanisms , the divergent evolution of a resonant planet pair maintains the libration of the resonant angles by damping the eccentricities . we call _ disk - driven resonant repulsion _ the divergent evolution of a resonant planet pair embedded in a disk , following the terminology introduced by @xcite for the divergent evolution of a resonant planet pair due to star - planet tidal interactions . we stress that disk - driven resonant repulsion can have two origins : ( i ) disk circularization of the orbits and ( ii ) wake - planet interactions . in other words , disk - driven repulsion comprises ( i ) _ circularization - driven repulsion _ and ( ii ) _ wake - driven repulsion_. the results of hydrodynamical simulations shown above indicate that , for the disk and planet parameters that we have considered , wake - planet repulsion dominates circularization - driven repulsion . this point will be further illustrated in sect . [ sec : nbody_method ] . + _ when is wake - driven repulsion important? _ we have seen above that the period ratio of a planet pair embedded in a disk may increase due to angular momentum transfer between the planets through their wakes . we further discuss in this paragraph under what conditions wake - planet interactions may lead to efficient divergent evolution of a planet pair . we separate the planet pair into a donor and a recipient . the bigger the donor s mass , the stronger its wake , and the easier for the wake to deposit angular momentum in the coorbital region of the recipient planet through the dissipation of shocks . this requires the dimensionless parameter @xmath17 of the donor planet to typically exceed unity ( @xmath18 is the planet - to - star mass ratio and @xmath114 the disk s aspect ratio ) . the donor planet can thus be a partial or a deep gap - opening planet . the angular momentum that is deposited by the donor s wake is ( at least partly ) transferred to the recipient planet through an effective corotation torque @xcite . if the recipient s mass is too big , the density in its coorbital region will be too low for this effective corotation torque to matter . in other words , the recipient could be a partial gap - opening planet or even a type - i migrating planet . these considerations show why divergent evolution is neither expected for two type - i migrating planets ( the donor s wake being too weak ) , nor for two type - ii migrating planets ( the recipient s gap being too deep ) . we have checked this with hydrodynamical simulations , not illustrated here . the above considerations also help understand the different rates of divergent evolution obtained in figure [ fig : ex ] . the fact that both planets open a partial gap is a favorable situation for wake - planet interactions to play a significant role . in the top row of panels , where the planets become locked in the 2:1 mmr , the high - density region between the two gaps ( see upper panel in figure [ fig : dens ] ) causes significant deposition of angular momentum in the planets coorbital region and therefore very efficient divergent evolution . in contrast , in the bottom row of panels , where the planets merge their gap and become locked in the 3:2 mmr , the moderate disk density left between the planets can only trigger slow divergent evolution . in the middle row of panels , where the period ratio stalls near 1.7 , a steady state is reached with background convergent migration balanced by wake - planet divergent evolution . + _ evolution towards a steady state _ when the planets are widely separated , they are expected to interact with the disk independently of each other and they can then enter a convergent migration phase . as they approach each other both planet - planet interactions , and wake - planet interactions where the planets interact with each others wakes can occur . both these processes involve exchanges of orbital energy and angular momentum between the planets . planet - planet interactions can result in resonant trapping associated with the growth of eccentricities , while wake - planet interactions act to produce divergent migration without the growth of eccentricities . from eq . ( [ eq_pr ] ) we see that both circularization- and wake - driven resonant repulsions act together to increase the period ratio . if wake - driven repulsion dominates after the eccentricities have damped to low values , with @xmath115 remaining constant , the period ratio should increase linearly with time , which is roughly what is obtained in the simulation presented in the top row of panels in figure [ fig : ex ] . however , in the longer term @xmath116 should decrease with time as the planets separate , which indicates that a steady state should be ultimately reached with convergent migration driven by background disk torques balancing wake - driven repulsion . we stress that wake - planet interactions cause a non - linear evolution of the disk s density profile near the planets , which may affect the time evolution of the background torques responsible for convergent migration . at least in the simulation without planet - planet interaction of figure [ fig : fono ] , a slow decrease in the background torques could explain why the period ratio decreases then increases , instead of reaching a stationary non - resonant value . another possibility is that convergent migration and disk - driven repulsion behave differently as the planet pair moves closer to the star , if the torques responsible for both mechanisms have different radial dependences . investigation of these aspects is deferred to a future study . the above simulations of the kepler-46 system show that the planets period ratio can take a range of values , depending on the convergent migration rate during the early evolution of the system . the limited duration of the simulations ( up to a few thousand orbits ) and the fact that planets still migrate at the end of the simulations however raise the question of the long - term evolution of the system . even when the inward migration of the planets gets stalled , further evolution of the period ratio can be maintained by disk - driven resonant repulsion . a steady state may _ a priori _ be achieved after depletion of the protoplanetary disk . disk dispersal typically occurs between @xmath117 and @xmath118 years after formation of the central star . photoevaporation driven by extreme ultraviolet radiation from the star opens a gap in the disk at separations of typically a few au for sun - like stars ( e.g. , * ? ? ? * ; * ? ? ? * ) . for planets below that separation the background disk will be depleted on up to a viscous timescale at the orbital separation where photoevaporation sets in . assuming a local viscous alpha parameter of a few @xmath119 , viscous draining of the inner disk could operate in a few @xmath120 years , that is typically a few @xmath121 orbital periods at the present location of kepler-46b and kepler-46c . to assess the impact of disk evaporation on our results , we have restarted two simulations adopting a simple exponential decay of the surface density profile . this is done by solving @xmath122 in addition to the hydrodynamical equations , where @xmath123 is the azimuthally - averaged density profile at restart time ( from which evaporation switches on ) , @xmath124 is an arbitrarily small density profile in a steady state ( taken to be not zero for numerical convenience ) , and @xmath125 is the evaporation timescale . for illustrative purposes , we have considered three short evaporation timescales : 100 , 500 and 2000 orbits , simulations being restarted at 1000 orbits . results are shown in figure [ fig : disp ] . the upper panels are obtained with @xmath126 and @xmath60 , for which a rapid divergent evolution occurs after capture in the 2:1 mmr . we see that the evolution is frozen out at a few evaporation timescales after restart , the planets reaching stationary eccentricities and semi - major axes . similarly , the lower panels in figure [ fig : disp ] show the results for @xmath127 and @xmath67 , for which convergent migration stalls at about the observed period ratio in the kepler-46 system . in that case , including disk evaporation stalls the migration of both planets , as expected , but it does not significantly affect the planets eccentricities and period ratio . the final period ratio remains very close to the observed value . in section [ sec : hydro ] we have presented results of hydrodynamical simulations modeling the early evolution of kepler-46b and kepler-46c as they were embedded in their parent protoplanetary disk . the variety of period ratios obtained in our simulations raises the question of whether disk - planets interactions could partly explain the diversity of period ratios in kepler s multi - planetary systems . we review below some of the processes that might account for such diversity . as highlighted in figure [ fig : kepler ] , kepler s multi - planetary systems show a clear tendency for planet pairs near resonances to feature period ratios slightly greater than strict commensurability . divergent evolution of a short - period resonant planet pair due to star - planet tidal interactions could partly explain this feature @xcite . although the efficiency of tidal dissipation remains largely uncertain , it is thought that efficient tidal circularization requires the inner planet to orbit its star in less than a few days . as can be seen in figure [ fig : kepler ] many of kepler s candidate systems have an inner planet beyond 10 days , for which tidal circularization is unlikely to have caused significant resonant repulsion ( see section [ sec : tidal ] of the appendix ) . for these systems we propose that resonant repulsion driven by disk - planets interactions could have played a prominent role in determining the observed period ratios . the vast majority of the planets in kepler s multi - planetary systems have physical radii between 0.1 and 0.3 jupiter radii , which places them in the super - earth to neptune - mass range . from the mass - radius diagram of kepler candidates followed up by radial velocity , the median mass of a 0.2 jupiter - radius planet is about 10 to 15 earth - masses ( see , e.g. , http://exoplanets.org ) . based on generally considered temperatures and viscosities in protoplanetary disks ( corresponding to @xmath128 and @xmath129 a few @xmath81 ) such planets are expected to experience type i migration . recent n - body experiments by @xcite have shown that convergent migration of type - i migrating planets yields very little departure from strict commensurability in contrast with kepler s data . he showed that the inclusion of stochastic forces in addition to type i migration could reproduce the observed distribution of period ratios quite nicely . such agreement has been obtained assuming that the initial period ratio distribution is the observed one , and taking the same convergent migration timescale and the same amplitude of stochastic forces for all kepler s multiple systems , which remains quite uncertain given the expected diversity of physical properties in protoplanetary disks . still , his results indicate that type i migration alone may not be able to account for kepler s diversity of period ratios . based on our results of hydrodynamical simulations for the kepler-46 system , we propose that some of kepler s planet candidates in multi - planetary systems could have opened partial gaps in their parent disk , thereby escaping the type i migration regime . this could be the case if , for instance , planets formed and/or migrated in regions of low turbulent activity ( dead zones ; see , e.g. , * ? ? ? * ) or if the disk s aspect ratio takes smaller values than commonly adopted . aspect ratios @xmath130 are likely typical of the short orbital separations at which kepler s candidate systems are detected . to address this possibility , we carried out hydrodynamical simulations with two planets in the super - earth mass range . the inner planet s mass @xmath131 , the outer planet s mass @xmath132 and the central star is one solar mass . the planet - to - primary mass ratios are therefore @xmath133 and @xmath134 , that is an order - of - magnitude smaller than those of kepler-46b ( median value ) and kepler-46c . in order to get similar gap depths in the super - earths and kepler-46 simulations , we adopted values of @xmath114 and @xmath23 such that the dimensionless parameters in the gap - opening criterion of @xcite take the same values . for the super - earths simulations , this yields @xmath135 and @xmath136 . all other simulation parameters are otherwise identical , except the grid resolution that we increased to @xmath137 . figure [ fig : m10 ] displays the results of two simulations . in the top panels , the unperturbed disk s surface density is @xmath138 . we see that the planets lock themselves in the 3:2 mmr from about 1300 orbits . divergent evolution proceeds with a decrease in the planets eccentricities . after reaching exactly 1.5 at about 1500 orbits , the period ratio increases to 1.54 at 5000 orbits . as illustrated by the screenshot of the disk surface density , the proximity of the planets allows their wakes to penetrate into each others coorbital region , where they can deposit part of their angular momentum flux . as we have shown in section [ sec : divergent ] , transfer of angular momentum between planets through wake - planet interactions will result in a loss rate of the total orbital energy of the two planets . this has the consequence of increasing the period ratio and decreasing the eccentricities . in the bottom panels of figure [ fig : m10 ] , the unperturbed disk surface density parameter is decreased to @xmath139 . here , convergent migration due to the background disk torques is slow enough to lock the planets in the 2:1 mmr , but this time the resonant capture does not lead to divergent evolution . instead , a steady state is reached in which both the eccentricities and the orbital period ratio take stationary values . this steady state is a consequence of negligible wake - planet interactions in this case . as illustrated in the bottom - right panel of figure [ fig : m10 ] , the planets are sufficiently far from each other that their wakes primarily transfer energy and angular momentum to regions of the disk that are not coorbital with the planets . the results of section [ sec : m10 ] support the idea that disk - driven resonant repulsion of partial gap - opening super - earths may lead to period ratios that substantially differ from nominal resonant values . as shown in section [ sec : divergent ] , this mechanism has two origins : circularization of the orbits and wake - planet interactions . its efficiency depends on the planets masses and the evolution of the disk s surface density . the survey of such a large parameter space is out of reach of hydrodynamical simulations , but three - body simulations with customized prescriptions for migration , eccentricity damping and disk dispersal can help illustrate the diversity of period ratios that can be achieved . this is the strategy that we have adopted . for illustration purposes we have considered fixed planet masses : @xmath140 , @xmath132 and @xmath141 ( as in section [ sec : m10 ] ) . our three - body simulations solve the equations of motion for the two planets and the central star using a standard technique . planets are assumed to be coplanar . disk - planet interactions are incorporated by applying appropriate dissipative forces ( for details , see * ? ? ? inspection at eq . ( [ eq_pr ] ) shows that , in order to model the planets evolution , we need to specify ( i ) the rate of convergent migration before disk - driven repulsion sets in , ( ii ) the damping rate of the eccentricities , ( iii ) the efficiency of wake - driven divergent evolution , and ( iv ) how disk dispersal is modeled . point ( iii ) remains uncertain at this stage , as the term @xmath142 in the right - hand side of eq . ( [ eq_pr ] ) , which models energy dissipation due to wake - planet interactions , should depend sensitively on the planet masses , their mutual separation and distance from the central star , and several disk quantities ( including its temperature and viscosity ) . a systematic study of this dependence goes beyond the scope of this paper , and is therefore left for future work . in order to do simple modeling , and to minimize the number of free parameters , we do not attempt to parameterize the @xmath142 term . instead , we model disk - driven repulsion as circularization - driven repulsion only . in other words , in eq . ( [ eq_pr ] ) we compensate the absence of the @xmath142 term in our three - body integrations by considering circularization timescales @xmath103 and @xmath143 that are shorter than in the hydrodynamical simulations . damping of the planets eccentricity is modeled as an exponential decay with a constant damping timescale to orbital period ratio , which we take as a free parameter ( specified below ) . to model the convergent migration prior to resonant repulsion , we adopt a simple prescription that reproduces the results of hydrodynamical simulations . denoting by @xmath144 the ratio of semi - major axes ( @xmath145 ) , we found good agreement using @xmath146 with @xmath147 \times \frac { ( t / \tau)^3 } { \sqrt{1+(t/\tau)^6 } } , \label{eq : f}\ ] ] where pr and @xmath148 are fitting parameters . the expression in eq . ( [ eq : prescmig ] ) tends to a constant value for @xmath149 , which mimics the fact that convergent migration is found to nearly stall before resonant repulsion sets in . in eq . ( [ eq : f ] ) , pr denotes the period ratio at which convergent migration approximately stalls , and @xmath148 the time at which this occurs . the good agreement between this migration prescription and our hydrodynamical simulations is illustrated in figure [ fig : fit ] . note that in the three - body simulations , disk - driven migration is applied to the outer planet only . we display in figure [ fig : john1 ] the results of a three - body simulation with @xmath150 and @xmath151 orbits , which mimics the convergent migration obtained in the hydrodynamical simulation of section [ sec : m10 ] with @xmath152 ( see right panel of figure [ fig : fit ] ) . in the top - left panel of figure [ fig : john1 ] , only the semi - major axis of the outer planet varies in the first 1500 orbits as our prescription for disk - driven migration is applied to the outer planet . once resonant repulsion sets in near 1500 orbits , the semi - major axis of the inner planet decreases slightly , that of the outer planet increases slightly as well . the eccentricity damping timescale here is 20 planet orbits , and this rather short timescale is found to give very similar divergent evolutions in the three - body and hydrodynamical simulations ( see top - right panel in figure [ fig : john1 ] ) . at 5000 orbits , the period ratio is 1.53 in the three - body run , and 1.54 in the hydrodynamical run . this small difference can be qualitatively explained by the fact that the repulsion occurs at a slightly larger pace in the hydrodynamical simulation than has been modelled in the three - body run . furthermore , we point out that the eccentricity peaks near the 2:1 and 3:2 mmrs are about a factor of 2 to 3 smaller in the three - body simulation ( compare with the eccentricities of the hydrodynamical run shown in the second upper panel of figure [ fig : m10 ] ) . the reason for this difference is most likely a reflection of the fact that the actual circularization time in the hydrodynamical run is longer than that adopted in the three - body integration , consistent with the operation of wake - planet repulsion in addition to circularization - driven repulsion as discussed above . we also incorporated a simple model for disk dispersal , which is taken to occur between @xmath120 and @xmath117 orbits . since planets are assumed to be already formed at the beginning of the simulations , this range of timescales is meant to account for different planet formation timescales . the longest time ( @xmath117 orbits ) corresponds to a case where both planet formation and capture into resonance following convergent migration occur rapidly . the shortest one ( @xmath120 orbits ) stands for the opposite . once disk dispersal switches on , the eccentricity damping timescale is progressively increased as an exponential growth with characteristic timescale @xmath153 orbits . this timescale is shorter than the typical viscous draining timescale of a disk inside the photoevaporation radius . it implies that the efficiency of disk - driven resonant repulsion is nearly entirely determined by the time from the beginning of the simulations at which dispersal switches on . we display in figure [ fig : john2 ] the results of a series of three - body simulations using the model described in section [ sec : nbody_method ] . recall that @xmath154 , @xmath132 and @xmath155 , which can be thought of being representative of kepler s sample of multi - planetary systems ( the planets mass ratio is arbitrary , it takes the same value as in the hydrodynamical simulations in section [ sec : m10 ] ) . planets are initiated with a period ratio of 2.4 . the prescription for convergent migration given at eq . ( [ eq : prescmig ] ) is used with @xmath156 and @xmath157 orbits . this value of pr is meant to explore the diversity of period ratios that can be achieved by disk - driven resonant repulsion away from the 3:2 mmr . the eccentricity damping timescale is varied from 20 to 200 orbits , and the time at which disk dispersal switches on from @xmath120 to @xmath117 orbits ( see section [ sec : nbody_method ] ) . given the duration of disk - driven resonant repulsion , our results are found to have very little dependence on the `` convergent migration timescale '' @xmath148 . simulations were carried out over @xmath158 orbits , and the period ratio time - averaged over the last 1000 orbits of the simulations is indicated by the color bar . for a given eccentricity damping timescale , the later disk dispersal occurs , the longer planets eccentricities are damped , and therefore the larger the final period ratio . at a given time prior to disk dispersal , the shorter the eccentricity damping timescale , the larger the final period ratio , again as expected . for the shortest damping timescales that we have considered , period ratios can reach @xmath159 1.75 . with such short timescales , planets can evolve continuously from the 3:2 mmr to the 2:1 mmr ( that is , the 2:1 resonant angles start to librate when the 3:2 resonant angles no longer librate ) . had we taken even shorter damping timescales or delayed disk dispersal even more , resonant repulsion would have led to period ratios exceeding 2 . the main conclusion that can be drawn from figure [ fig : john2 ] is that partial gap - opening super - earths may experience significant disk - driven resonant repulsion away from the nominal 3:2 mmr . under very favorable circumstances ( particularly efficient repulsion maintained over a particularly long timescale ) period ratios can easily exceed nominal resonant values . although we have considered fixed planet masses for illustration purposes , we believe that disk - driven resonant repulsion is a generic mechanism that may have likely occurred among kepler s multi - planetary systems . results of three - body simulations with prescribed convergent migration and eccentricity damping . the central star has a solar mass . the mass of the inner and outer planets is @xmath160 and @xmath161 , respectively . convergent migration is modeled using the expression in eq . ( [ eq : prescmig ] ) with pr = 1.5 and @xmath162 orbits . the eccentricity damping timescale is varied from 20 to 200 orbits ( y - axis ) , and it remains fixed until disk dispersal is switched on . the timescale at which disk dispersal switches on ranges from @xmath120 to @xmath117 orbits after the beginning of the simulations ( x - axis ) , and the evaporation timescale is set to @xmath163 orbits . the orbital period ratio obtained after @xmath158 orbits , and time - averaged over the last 1000 orbits , is colour - coded.,title="fig : " ] the multi - planetary systems detected by the kepler mission have very diverse architectures , which reflect the wide range of conditions under which planets may form and evolve . the quasi - coplanar orbits in kepler s multiple systems suggest that interactions between planets and their parent protoplanetary disk should play a prominent role in shaping such diverse architectures . in particular , close planet pairs feature a variety of orbital period ratios and many of them are not in mean - motion resonance . convergent migration of multiple planets due to disk - planet interactions may lead to resonant systems in disks where turbulence is low enough not to disrupt resonances @xcite . resonant planet pairs formed by disk - planet interactions may experience divergent evolution of their orbits due to tidal orbital circularization @xcite . this so - called _ tidal resonant repulsion _ may increase the orbital period ratio of close - in planet pairs quite substantially depending on the efficiency of star - planet tidal interactions @xcite . tidal resonant repulsion could help account for the diversity of period ratios in kepler s close - in planet pairs . we have shown in this paper that disk - planets interactions could also form planet pairs with a variety of period ratios . we have first focused on the kepler-46 planetary system , in which the two saturn - mass planets kepler-46b and kepler-46c have a period ratio @xmath164 @xcite . we have presented results of hydrodynamical simulations modeling the early evolution of these planets as they were embedded in their parent disk . for typical disk temperatures and turbulent viscosities , both planets are expected to open a partial gap around their orbit . a rapid convergent migration causes the planets to merge their gaps and to evolve into a common gap . depending on the density s structure inside the common gap , the planets convergent migration is found to stall with a variety of period ratios between 1.5 and 1.7 ( see figure [ fig : finalopr ] ) . the observed period ratio of kepler-46b / c can be reproduced without the planets being in mean - motion resonance . our results also highlight that disk - planets interactions may significantly increase the orbital period ratio of a planet pair from a near - resonant value , a mechanism that we term _ disk - driven resonant repulsion_. it has two origins : ( i ) orbital circularization by the disk , and ( ii ) the interaction between each planet and the wake of its companion . as the wake of a planet turns into a shock at some distance from the planet , it may deposit angular momentum in the coorbital region of another planet through an effective corotation torque . we have shown that this mechanism can dissipate the total orbital energy of a planet pair without changing its total angular momentum , a configuration that increases the planets period ratio and decreases their eccentricity . resonant repulsion caused by wake - planet interactions can be particularly efficient when at least one of the planets opens a partial gap around its orbit ( see section [ sec : divergent ] ) . for the disk and planet parameters adopted in this work , wake - planet interactions are much more efficient at driving the repulsion of a planet pair than disk circularization . however , there may be circumstances for which this balance is different . the strength of wake - planet interactions increases as the radial separation between planets decreases , independently of their eccentricity . if wake - planet interactions are the dominant process for the divergent evolution of a planet pair , then one could expect the period ratio to reach a constant value , which could be different from a resonant ratio . our results of simulations report , however , many cases where the period ratio nearly stalls but then increases . there could be two reasons for this behavior . first , wake - planet interactions cause a ( non - linear ) evolution of the disk s density profile near the planets , which causes the torques driving convergent migration and wake - planet interaction to change over time . furthermore , independently of the previous consideration , preliminary results indicate that the torques responsible for convergent migration and for wake - driven repulsion have different dependences on the stellar distance as the planets migrate . both effects are found to strengthen wake - driven repulsion as compared to convergent migration as planets migrate inwards . this would explain why , in our simulations , the period ratio of inwardly migrating planet pairs increases after reaching a nearly constant value . the above considerations indicate that wake - planet interactions may naturally increase the period ratio of planet pairs away from resonant values . one of our low - mass disk models for kepler-46 , in which the saturn - mass planets are captured in the 2:1 mmr , a period ratio of 2.3 is attained in as short as a few thousand orbits . we note that a significant divergent evolution of sub - jovian planets was obtained in some of the hydrodynamical simulations of @xcite and @xcite . the similarity between their results of simulations and ours ( convergent migration nearly stalled close to resonance , followed by divergent evolution of the planets orbits alongside the damping of their eccentricity ) suggests that the divergent evolution obtained in these simulations is likely to arise from wake - planet interactions . many planets in kepler s multiple systems are in the super - earth to neptune - mass range , and have orbital periods below 100 days . at these short periods the aspect ratio of a disk should be rather small , thus many kepler s planet candidates could have opened partial gaps in their parent disk . in contrast to the saturn - mass planets in the kepler-46 system , super - earths are not expected to merge their gaps and to evolve in a common gap , unless a very fast convergent migration brings them to very short mutual separations . we have argued instead that close - in super - earths are good candidates for efficient resonant repulsion caused by wake - planet interactions . for instance , one of our simulations shows that the orbital period ratio of two @xmath159 15 earth - mass planets increases from 1.5 to 1.55 in few thousand orbits . disk - driven repulsion via wake - planet interactions could explain why many planet pairs in kepler s multiple systems have period ratios slightly greater than resonant . while resonant repulsion driven by stellar tides should only be relevant to planet pairs with orbital periods below a few days , disk - driven repulsion may also apply to planets on longer periods . the ttv signals of kepler s short - period planet pairs indicate that some of these planets have small but not zero free eccentricities @xcite . free eccentricities get quickly damped by tidal orbital circularization . however , the presence of a disk of planetesimals or a non - smooth gas disk dispersal could give planets some free eccentricity . close - in planet pairs that seem to have undergone resonant repulsion , but which have not zero free eccentricities , like kepler-23b / c and kepler-28b / c , could indicate that repulsion was primarily mediated by disk - planets interactions rather than by star - planet tidal interactions . we foresee at least three directions to pursue work on disk - driven resonant repulsion . firstly , the effects of disk turbulence , which we have modeled with a constant viscosity , need to be addressed carefully . the levels of stochastic fluctuations required to disrupt the process as planet masses decrease , and to produce some residual free eccentricity should be explored . secondly , planet mass growth should be taken into account as it may also lead to some divergent evolution @xcite . thirdly , a monte - carlo approach with three - body integrations using simple prescriptions for the effects of disk - planet interactions should be used to compare the distribution of orbital elements among kepler s planet pairs with the synthetic distributions predicted by disk - driven resonant repulsion . cb is supported by a herchel smith postdoctoral fellowship of the university of cambridge . we thank gwenal bou , cathie clarke , jrme guilet , cristobal petrovich and yanqin wu for useful discussions . we also thank the referee for a very helpful report . hydrodynamical simulations were performed on the darwin supercomputer of the university of cambridge high performance computing service using strategic research infrastructure funding from the higher education funding council for england . we show here that star - planet tidal interactions are unlikely to cause significant divergent evolution of two planets in the super - earth mass range when the orbital period of the inner planet exceeds about 10 days . in the constant time lag model , taking tidal dissipation within the planet to be much larger than within the star , the orbital circularization timescale , @xmath165 , is given by equation ( 25 ) of @xcite and can be expressed as follows : @xmath166 where @xmath167 is the tidal quality factor of the inner planet , @xmath168 its mean density , @xmath169 its physical radius , @xmath170 its orbital period and @xmath47 is the star s mass . divergent evolution of two resonant planets may occur as a result of tidal dissipation of orbital energy at approximately constant orbital angular momentum @xcite . the timescale for divergent evolution , which we denote by @xmath171 , satisfies @xmath172 with @xmath88 and @xmath89 the semi - major axis and forced eccentricity of the inner planet , and @xmath173 its change in semi - major axis during divergent evolution ( it can be taken to be the planet s distance from resonance ) . we have @xmath174 , where @xmath175 is the outer planet s mass , @xmath90 the inner planet s mass ( @xmath176 denotes the outer - to - inner planet mass ratio ) , and @xmath177 is a constant of order unity that varies with the resonance ( see , e.g. , equation 38 of * ? ? ? * ) . using eq . ( [ eq_app1 ] ) and taking @xmath178 , we obtain @xmath179 this order of magnitude estimate shows that planet pairs orbiting sun - like stars should experience negligible tidally - driven resonant repulsion over gigayear timescales when the inner planet is a super - earth , with @xmath180 @xmath181 and orbital period greater or equal to 10 days , even if @xmath182 and @xmath183 . moreover , if the inner planet is a jupiter - like planet , then assuming @xmath184 and @xmath185 g@xmath77 , eq . ( [ eq_app2 ] ) shows that divergent evolution of more than a few percent away from resonance is expected over a few gigayears as long as the inner planet s orbital period does not exceed 10 days . in the course of our numerical experiments , we have tested the robustness of our results by varying the assumptions made in the physical model and by varying the grid resolution . _ physical model _ we performed several simulations reducing the disk s aspect ratio to 4% and 3% while keeping the same viscous alpha parameter . although not illustrated here , these simulations show again a variety of outcomes , but none reproducing the observed period ratio between kepler-46c and kepler-46b . decreasing the disk s aspect ratio makes the gaps carved by the planets deeper . when convergent migration is rapid enough for the planets to cross their 2:1 mmr and to evolve in a common gap , we find that the planets generally lock themselves into the 3:2 , 4:3 or even 7:5 mmr with very little departure from exact commensurability . in these simulations , the density in the common gap is too low for wake - planet interactions to cause significant divergent evolution ( see section [ sec : divergent ] ) . these simulations thus suggest that partial gap opening is required to obtain a period ratio consistent with the observed value . furthermore , as the structure of the gaps plays a prominent role in the final evolution of the planets period ratio , we carried out additional simulations where the planets were first held on fixed circular orbits for 500 orbits before migrating . this preliminary stage allowed enough time for each planet to build up a gap with a stationary density structure . the results of the simulations are shown in the left panel of figure [ fig : gapres ] for @xmath78 and three values of @xmath25 . starting with pre - evolved gaps changes the initial rate of convergent migration . notwithstanding quite different initial evolutions , two out of these three models end up following the same evolution with or without pre - evolved gaps , and reach the same final period ratio at the end of the simulations . the model with @xmath186 reaches a slightly smaller period ratio ( decreased from 1.65 to 1.6 when starting with pre - evolved gaps ) . + _ grid resolution _ the effect of resolution was checked by doubling the number of grid cells along each direction for a few simulations . the results of three high - resolution runs ( @xmath187 ) are compared to the nominal - resolution runs ( @xmath188 ) in the right panel of figure [ fig : gapres ] . doubling the grid resolution slightly changes the initial rate of convergent migration , leading to slightly different initial evolution . as for the above series of simulations with pre - evolved gaps , the model with capture into 2:1 mmr followed by resonant repulsion is not affected by doubling the resolution . the two models with common gap evolution seem again more susceptible to a variation in the initial convergent migration , but lead to small differences in the final outcome . interestingly , both high - resolution models with @xmath189 and @xmath190 yield a period ratio @xmath191 , in very good agreement with the observed value .

we have studied the properties of a sample of 67 very blue and likely young massive clusters in m31 extracted from the bologna revised catalog of globular clusters , selected according to their color [ and/or to the strength of their spectral index ( ) . their existence in m31 has been noted by several authors in the past ; we show here that these blue luminous compact clusters ( blccs ) are a significant fraction (% ) of the whole globular cluster system of m31 . compared to the global properties of the m31 globular cluster system , they appear to be intrinsically fainter , morphologically less concentrated , and with a shallower balmer jump and enhanced absorption in their spectra . empirical comparison with integrated properties of clusters with known age as well as with theoretical ssp models consistently indicate that their typical age is less than gyr , while they probably are not so metal - poor as deduced if considered to be old . either selecting blccs by their colors or by the strength of their index the cluster sample turns out to be distributed onto the outskirts of m31 disc , sharing the kinematical properties of the thin , rapidly rotating disc component . if confirmed to be young and not metal - poor , these clusters indicate the occurrence of a significant recent star formation in the thin disc of m31 , although they do not set constraints on the epoch of its early formation .

astro-ph0503188

globular clusters ( gcs ) are ubiquitous stellar systems living in any kind of galaxies , from dwarf to giants , from the earliest to the latest types . their integrated properties carry crucial information on the physical characteristics of their host galaxy at the time of their formation . hence , the study of globular clusters systems is a fundamental tool to understand the evolutionary history of the barionic component of distant galaxies ( see @xcite and references therein ) . in this framework , the globular cluster system of the andromeda galaxy ( m31 ) plays a twofold role , as a natural reference to compare the milky way ( mw ) gc population , and as a fundamental testbed for the techniques to be applied to systems in more distant galaxies ( see * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * ; * ? ? ? * and references therein ) . indeed , the comparison of the gc system of m31 and the mw has revealed both fundamental similarities and interesting differences whose complete understanding may have a deep impact on our knowledge of galaxy formation and evolution @xcite . among the latter ones , in the present contribution we will especially focus on the claimed presence in m31 of stellar systems similar to mw globulars in luminosity and shape but with integrated colors significantly bluer than the bluest mw counterparts . while some of the faintest objects may hardly be distinguished from bright open clusters , the typical family member appears quite similar to classical globulars ( two typical examples are shown in fig . [ map ] , see * ? ? ? * ; * ? ? ? * ; * ? ? ? * ) . in sec 3.4 we will provide some evidences suggesting that clusters of similar age and total luminosity may be lacking in the milky way . hereafter , we call them `` blue luminous compact clusters '' ( blccs ) the peculiar colors of blccs have been first reported by @xcite , @xcite , @xcite , and this class of objects then received growing attention @xcite , although a systematic study is still lacking . in particular , @xcite noted 14 such blue clusters , not included in the list of open cluster candidates by @xcite , and better consistent with a gc morphology . their absolute luminosities spanned the luminosity range @xmath6 , and their positions in a two - color diagram pointed to a possibly young age . for ten of these objects , @xcite provided supplementary ubvr photometry indicating a global luminosity around @xmath7 l@xmath8 . based on stellar population models , their estimated age appeared to be less than a few @xmath9 yrs , with a typical mass between @xmath10 and @xmath11 m@xmath8 . if confirmed , these values indicate that they are higher than galactic open clusters , but comparable to those of young , rich globulars found in the large magellanic cloud @xcite . @xcite , studying the uv - colors of a sample of 49 gc candidates in m31 , listed 11 objects classified as blue clusters based on their location in the two - color diagram , and suggested that they are probably young . on the same line , @xcite noted that their m31 catalog of gc candidates may be contaminated by several young obiects with @xmath12 and they eventually excluded 55 such objects from their analysis of old m31 clusters . as already stressed long ago @xcite , the integrated spectrum and color of a cluster , especially in the blue , are influenced by the metal abundance and the position of the main sequence turnoff stars ( msto ) ( in turn , by the cluster age ) , by the strength of the horizontal branch ( hb ) , and , to a lesser degree , by the overall luminosity function of its composing stellar population . to disentangle the different effects it is thus very important to obtain the color - magnitude distribution of cluster stellar populations . in this regard , @xcite obtained deep hst photometry of individual stars and c - m diagrams for four of these blccs leading to estimate ages in the range 60 - 160 myr and metallicity from solar to 2/5 solar . this clearly supports the evidence that the exceedingly blue integrated colors of blccs are direct consequence of their remarkably young age . @xcite reached similar conclusions for eight blccs by comparing high - quality low - resolution spectra of a sample of m31 clusters with similar data for mw and magellanic clouds globulars . @xcite reported a global sample of 19 blccs in m31 , including 13 `` young '' objects from the @xcite survey , most of them sharing the kinematical properties of a wider sample of `` red '' clusters belonging to the cold thin disc detected by @xcite . in summary , various observations suggest that m31 may have many more young gcs than the mw , and the latest results on the claimed existence of a thin disc subsystem of gcs in m31 , quite large in number and covering a very wide range in metallicity from @xmath13<-2.3 $ ] up to solar and above @xcite , have opened an important debate and actually gave rise to the present study . in fact , as recently discussed by @xcite , @xcite and ( * ? ? ? * and references therein ) , the detection of several old metal - poor clusters with thin - disc kinematics would imply that _ m31 is likely to have had a disc already in place at the very early stages of the galaxy evolution , and _ ( ii ) _ no substantial merger event can have occurred at later epochs as galaxy disc would have been disrupted or , at least , heated ( but see , however , @xcite ) . this conclusion is at odds with the indications found by @xcite , who reported the detection of a wide intermediate - age ( 6 - 8 gyr ) population of metal - rich stars ( with @xmath13>-0.5 $ ] ) in a minor - axis halo field of m31 observed with hst / acs , interpreted as the result of the merging with an almost equal - mass companion . _ as a consequence , it is of primary interest to verify if and how many clusters do belong to the claimed thin disc and , even more important , how old and metal - poor they are . _ in the following section we carry out a revision and a new selection of the candidate sample of blccs based on different ( and partially complementary ) criteria , relying on the _ bologna revised catalog of m31 gcs _ ( hereafter brc ; @xcite ) . then , in sect . 3 , we discuss the properties of the global sample of clusters and in particular their kinematical properties , metallicity and estimated ages . _ we anticipate that our analysis lead us to conclude that most ( if not all ) of the blccs are younger than @xmath4 gyr , more metal rich than @xmath13\sim -1.0 $ ] and they nicely fit the structural and kinematic characteristics of the thin - disc subsystem recently detected by @xcite . _ this issue is further discussed in sec . 4 , where we also summarize the main results of our analysis . to carry out a systematic analysis of the blcc population of the andromeda galaxy , our brc data have been complemented with the kinematic and spectral indices information from @xcite . this choice allowed us to preserve full homogeneity in the comparison of line indices , and take advantage at the same time of a larger sample of clusters and a better accuracy of radial velocity measurements . it has to be noted , however , that our main conclusions remain essentially unchanged if the @xcite dataset would have been used instead . throughout the paper , for m31 we will assume a distance modulus @xmath14 mag @xcite , a systemic radial velocity @xmath15 kms@xmath16 and a galaxy center , to which refer the xy coordinate system , both from @xcite . a single value for the interstellar extinction toward the galaxy is adopted for simplicity , with @xmath17 @xcite and a standard reddening law ( e.g. @xcite ) . we verified that our results would remain substantially unaffected when using , in alternative , the @xcite reddening values . all the m31 clusters comprised in our analysis ( i.e. 67 targets collected in table 1 ) belong to class 1 brc entries , that is they are all _ genuine _ m31 members confirmed either spectroscopically or by means of high - resolution imaging . the clusters are selected according to their intrinsic color [ i.e. @xmath18 ) ] and/or to the strength of their @xmath1 spectral index ( @xmath2 ) . the rationale of these selection criteria is explained in detail in the following sections . figure [ bvub ] is a collection of the gc population for local group galaxies in the reddening - corrected u - b vs. b - v color plane . we joined data for the milky way ( * ? ? ? * and latest updates ) , the lmc @xcite , m33 @xcite , and m31 @xcite . for the mw globulars , the adopted values of the color excess @xmath19 are those originally provided by @xcite , while @xmath20 for lmc and 0.07 for m33 @xcite . as a well recognized feature , the large majority of m31 gcs in the figure tend to bunch around @xmath21 and closely tracks the locus of the mw globulars , suggesting a similar age and metallicity distribution . on the other hand , as noted long ago by @xcite , @xcite , and more recently by @xcite , several m31 globulars spread over much bluer colors , compatible with the young and intermediate - age lmc gcs . plot are probably due to poor photometry in one ( likely u ) of the ubv bands and/or to inaccurate reddening correction . see @xcite and @xcite for a detailed discussion of m31 gc photometry . ] on the basis of the b - v distribution of fig . [ bvub ] , a more immediate selection criterion for blccs in m31 is to pick up those objects bluer than the bluest gc in the mw ( i.e. ngc 7492 , with @xmath22 ; @xcite ) . operationally , we therefore defined a color - selected blcc sample ( referred to as sample `` a '' in our following discussion ) , consisting of 41 objects ( out of 330 confirmed gcs in m31 , according to brc ) with @xmath18 . these objects are reported in table 1 ( with label `` a '' in column 12 ) ; of them , 29 targets have radial velocity estimates in the list of @xcite . we are aware , of course , that the indicative threshold in the integrated @xmath23 color might be a too `` prudent '' selection , as for instance it misses cluster g44 ( see fig . [ bvub ] ) , claimed by @xcite to have an age about 100 myr on the basis of the hst c - m diagram . on the other hand , the other three young clusters of @xcite are correctly picked up , and in any case the relevant uncertainty in the reddening correction for most of the m31 clusters prevents any firm selection criteron based on integrated colors alone . therefore , sample `` a '' likely provides a `` clean '' conservative estimate of the real fraction of young objects in the m31 gc population , and certainly secures our statistics from any contamination of intrisecally `` red '' clusters , even taking into account the claimed photometric errors , in the range 0.05 to 0.15 mag , depending on the cluster location and the source of the photometry . besides the obvious difference in color , it is interesting to further investigate how sample `` a '' clusters characterize vs. the remaining fraction of `` red '' m31 gcs . we therefore extended our analysis to different positional , morphological and spectrophotometric parameters , according to the brc data . left panels of fig . [ hisbv ] summarize the most striking differences between the two cluster populations . the four histograms report the cluster distribution in absolute @xmath24 magnitude , in two spectrophotometric indices , namely the @xcite @xmath25 and the lick h@xmath26 index ( according to the original definition of @xcite ) , and finally vs.the w@xmath27 structural parameter measured in an homogeneous way by @xcite and @xcite for almost all the m31 cluster candidates . in the plots , absolute v magnitudes from the brc are the result of a full revision of all the available photometry , including the hst data . magnitude differences with respect to the extended database of @xcite are typically less than @xmath28 mag , depending in general on a different reddening correction , but with no systematic trend of magnitude residuals with gc color . both narrow - band indices considered in fig . [ hisbv ] have been taken from @xcite ; , a transformation to pseudo - equivalent width ( ew ) units ( in ) has been done through the equation : @xmath29 ( see , e.g. * ? ? ? * ) , where @xmath30 is the original index , in magnitude scale . ] due to their unfluxed observations , however , the instrumental output had to be reduced to a standard system . this has been done relying on a set of 41 m31 gcs in common with @xcite for which both @xmath25 and standard lick indices are provided in the appropriate units ( see fig . [ calib ] ) . as for h@xmath26 , which measures the strength of the balmer absorption line in pseudo - equivalent width , we verified that no transformation was necessary for the @xcite original data , that roughly matched the lick standard system within a high , but still convenient , 0.9 rms . in case of the @xmath25 index , measuring the amplitude of the 4000 balmer jump in the integrated spectra of the clusters , the @xcite standard system was reproduced via a linear transformation in the form @xmath31}\ ] ] with a 0.12 mag rms in the individual point scatter . the corresponding plots in fig . [ hisbv ] have been corrected to the standard system accordingly . finally , the bottom histogram of fig . [ hisbv ] reports the distribution of the @xcite w@xmath27 morphological parameter , that nicely relates to the cluster core radius through a fitting king profile ( see @xcite for further discussion ) . a kolmogorov - smirnov test on the data distribution of the left panels of fig . [ hisbv ] indicates , at a confidence level better than 99.99% , that sample `` a '' blccs differ from the remaining fraction of m31 clusters under all the considered aspects . compared to the gc general distribution , blccs are intrinsically fainter , morphologically less concentrated , and with a shallower balmer jump . the most striking feature , however , is the intensity of the h@xmath26 absorption line , that is much stronger in blccs and essentially out of the range covered by ordinary globulars . this is strongly suggestive of young ages , as we will discuss in better detail in the next sections . the claimed h@xmath26 enhancement for blcc candidates opens our analysis to a complementary selection criterion that , relying on this narrow - band spectrophotometric index , basically overcomes any problem dealing with the ( poorly known ) reddening correction for m31 gcs . the peculiar strength of h@xmath26 absorption in the spectra of several m31 gcs is a well known feature , discussed since long ago by @xcite , @xcite , @xcite , @xcite , @xcite , and @xcite , among others . since homogeneous h@xmath26 values are available for most of m31 clusters from the @xcite database , one could self - consistently rely on this parameter for a more effective diagnostic of blccs . similar to the color - selected blcc sample , on the basis of the observed distribution of fig . [ hisbv ] , we could try a new sample selection picking up blccs among those objects with a lick index h@xmath32 ( see fig . 7 and fig . 8 , below ) . according to this different criterion , a larger number of clusters ( 51 in total , with 25 objects in common with the color - selected sample `` a '' ) will be included in our _ bona fide _ blcc dataset . this will be our sample `` b '' , as reported in table 1 and labelled accordingly in column 12 . among the 45 objects in table 1 with complete @xmath23 and h@xmath26 photometry , only one sample `` a '' cluster should not be comprised in the h@xmath26-selected sample `` b '' ( this is b206-d048 , with @xmath33 and h@xmath34 ) , while 19 sample `` b '' clusters are too red in @xmath23 to be included in sample `` a '' ( thus confirming the presence of a possibly important fraction of strongly reddened young clusters in the m31 gc population ) . the right panels of fig . [ hisbv ] complete the comparison between sample `` a '' and `` b '' properties in the different parameter domain . the substantial analogy of the two plots and the sharper characterization of the _ bona fide _ blcc candidates in the h@xmath26-selected sample makes the latter spectroscopic selection much more safe and efficient in identifying the _ same _ kind of clusters selected according to the color criterion alone . from the previous arguments , from now on we will adopt the two samples defined above and analyse sample `` a '' and `` b '' objects in parallel . full information for these 67 clusters is summarized in table 1 . in particular , photometry and positional data in the table are from the brc ( unless otherwise stated ) , [ fe / h ] , radial velocity and spectroscopic indices h@xmath26 and @xmath25 are from @xcite ( the latter ones , converted according to fig . [ calib ] ) , the morphological parameter w@xmath27 is from @xcite and @xcite , while @xmath35 kinematic residuals are from @xcite . as we already stressed before , a color selection like for sample `` a '' better favours a plain observational approach ( 90% of the objects in table 1 has a measured @xmath23 , while for only 78% h@xmath26 is provided from spectroscopy ) , although it is prone to any uncertainty in the reddening correction ; on the other hand , h@xmath26 selection , such as in sample `` b '' , likely provides a more confident and physical blcc selection , but suffers from more difficult observing constraints . _ quite importantly , however , we will show that , disregarding any preferred selection criterion , the same conclusions hold for both samples . _ as a final remark to our sample selection , we should also note that a number of other blcc candidates ( confirmed or not ) have been picked up by various authors based on different observations and procedures . for the sake of completeness , we have carried out an exhaustive search in the m31 literature assembling a coarser set of reportedly `` young '' objects , listing our results in table 2 ( 21 confirmed gcs ) and table 3 ( 69 gc candidates to be confirmed ) , together with a note to the corresponding reference studies . in the present analysis , however , we will restrain only to the _ bona fide _ samples ( i.e. table 1 data ) , postponing to a future paper a complete review of each individual object . after a thorough analysis of the positional and kinematical properties of m31 gcs , based on the @xcite data , @xcite concluded that the gcs in this galaxy belong to two kinematical components : `` ... a thin , rapidly rotating disc , and a higher velocity dispersion component whose properties resemble that of the bulge of m31 . '' according to these authors , membership to one of the two dynamical components is assessed for each cluster on the basis of its residual velocity with respect to a disc model , picking up disc clusters among those with residual @xmath36 ( in normalized units , see @xcite ) . the claimed probability of mis - classification with this procedure turns to be about 30% . in fig . [ kine ] we compare the spatial distribution ( in the x and y projected distances along the major and minor axes , respectively ) and the @xmath37 vs. x distribution of sample `` a '' ( left panels ) and `` b '' ( right panels ) blcc candidates with the remaining fraction of `` ordinary '' m31 globulars . the thin plot of points in the x vs. y panels mainly maps the brightest agb and carbon stars of m31 , from the 2mass point sources catalogue @xcite , and @xmath38 have been selected to better tune on m31 stellar population . ] and gives therefore a neat picture of the outermost structure of galaxy disc ( @xcite ) , allowing us to appreciate in some detail the overplotted gc distribution . it is evident from the figure that ordinary clusters ( upper panels ) are uniformly distributed all over the apparent body of m31 , up to distances of @xmath39 ( @xmath40 kpc ) from the galaxy center , closely tracing the smooth luminosity profile of m31 . in addition , most of the bulge clusters show clear sign of coherent rotation in the @xmath37 vs. @xmath41 plane , as a part of the rotationally supported structure of the galaxy @xcite . however , the large scatter around the overplotted rotation curve ( taken from * ? ? ? * see his fig . 3.10 ) , indicates that pressure support has a significant role in the overall kinematics of the sample ( see , however , @xcite for a deeper analysis of these clusters ) . on the contrary , blccs seem to avoid the inner regions of the galaxy ( see lower panels in fig . [ kine ] ) and appear well projected onto the outer disc with a strong correlation with the underlying spiral substructures . also the velocity pattern quite well traces the velocity curve of the disc . _ our conclusion , therefore , is that both color - selected and h@xmath26-selected blccs belong to the cold thin disc of m31 , in agreement with previous results of @xcite and @xcite , that were limited , however , to a much smaller subset of clusters . _ to further clarify this issue , in the upper panels of fig . [ residuals ] we plotted the @xcite residuals ( taken from table 2 therein ) , singling out our _ bona fide _ blccs . it is quite evident from the plots that blccs stand out naturally as a very low-@xmath35 family . almost the whole blcc population ( independent of the adopted selection ) , in fact , matches the @xcite disc - membership criterion , and only a few such clusters exceed @xmath42 . this feature is summarized also in the lower panels of fig . [ residuals ] , where the shaded histograms give the residual distributions of blccs compared to the whole sample of `` ordinary '' gcs . complementing the @xcite results , we can conclude that @xmath43 _ both color- and h@xmath26-selected blccs can effectively comprise a significant fraction of `` disc clusters '' , being natural tracers of the cold thin disc subsystem of m31 ( @xcite , but see also @xcite ) . _ @xmath43 _ blccs appear to constitute a separate family for what concerns a number of non - kinematical properties like , for instance , the absolute magnitudes , the h@xmath26 and @xmath25 indices ( at least ) , and the structural parameters . _ in a quite common view , a blue integrated color for a cluster is actually an indication of a young age as its integrated color is dominated by the bright blue stars at the msto point . on this basis , even the early studies of the m31 clusters ( see e.g. @xcite ) immediately noticed that most of the blue cluster candidates should be young . the combined information provided by h@xmath26 and the balmer jump , @xmath25 , allow in principle to tackle the problem of dating blccs on finer detail ; both indices , in fact , are contributed by the warmer stellar component in a simple stellar population ( ssp ) , and can selectively probe the msto temperature location @xcite , leading therefore to an indirect estimate of age in a stellar aggregate . note that , for its photometric characteristics , the @xmath25 index recalls , to some extent , the broad - band u - b color , but its narrower wavelength baseline makes the @xcite index much less sensitive to the reddening and therefore a better tracer of the intrinsic properties of a stellar cluster . index is defined in a magnitude scale as @xmath44 , where @xmath45 and @xmath46 are the luminosity densities ( per unit wavelength ) in two 200 wide bands centered at 3900 and 4100 , respectively . this has to be compared with the johnson u and b bands , centered respectively at 3600 and 4400 . according to @xcite compilation , the expected color excess can be computed as @xmath47 and @xmath48 . [ foot ] ] figure [ hbd_teo ] will be our reference plot to summarize the main features of a h@xmath26 vs. @xmath25 diagram , from the theoretical point of view . in the figure we reported the expected locus for main sequence stars of different temperature , up to @xmath49 k and their inferred spectral type , according to the @xcite calibration , as labeled . our calculations rely on the new bluered theoretical library of stellar seds @xcite , from which we extracted the grid of model atmospheres with @xmath50 and 5 , and [ fe / h ] = 2.0 , 1.0 and solar . the vanishing sensitivity of both spectral indices to the contribution of late - type stars makes the h@xmath26 vs. @xmath25 information nearly independent of the exact post - ms details of the c - m diagram , as far as we compare with the gc data . this eases a very immediate and straightforward comparison of the theoretical stellar locus with the integrated indices of blccs , providing a first effective constraint to age , through a msto calibration linking @xmath51 ( see , e.g. * ? ? ? * ) . in a more elaborated scheme relying on ssp models , however , one should also account for the supplementary contribution of blue hb stars , that are expected to play a role at some stage of late evolution of low - mass stars modulating the integrated sed of old stellar populations , typically beyond 10 gyr . a more accurate prediction of the integrated output for ssps of different metallicity and hb morphology is reported in the figure , according to the @xcite population synthesis models . the effect of hb morphology is sketched for the 15 gyr ssp models mainly resulting in a sensible enhancement of the h@xmath26 strength . it is clear , however , that in no cases the late evolutionary scenario could account for very strong h@xmath26-lined clusters ( i.e. h@xmath52 ) , where a prevailing contribution of f - type stars calls for a genuinely blue msto , and therefore for a young age , certainly less than a few gyrs . the distribution of m31 gcs in the h@xmath26 vs. @xmath25 plane is displayed in fig . [ hbd ] . big solid dots are the 21 sample `` a '' blccs from table 1 with available h@xmath26 and @xmath25 indices ( by definition , sample `` b '' comprises all the 40 solid and open dots above the h@xmath53 threshold , as indicated in the figure ) . the mw gc data from @xcite ( solid triangles ) are also reported for comparison . the stellar loci of fig . [ hbd_teo ] ( only those for @xmath54 , and different metallicity , for the sake of clarity ) have also been overplotted , together with the ssp locus for different age ( and a red hb morphology ) , according to @xcite . it is evident from the figure that , while the bulk of ordinary gcs ( including the mw globulars ) has spectral properties consistent with an age of several gyrs , the blccs indices are clearly dominated by a younger stellar component of a - f stars of moderately high metallicity ( mostly solar or slightly sub - solar ) . this places a confident upper limit to blcc age at @xmath4 gyr . note also ( fig . 8) that only 4 on 40 blccs lie on the branch of the theoretical stellar loci typical of ages lower than @xmath55 myr ( @xmath56 ) , while the bulk of the population is consistent with ages larger than @xmath57 myr . as we have been commenting in previous section , a more standard assessment of the blcc distribution with respect to the overall gc population in m31 can be carried out relying on the @xmath59 color . for its nature , this choice would allow a more comfortable match with the theoretical output of population synthesis models , and certainly ease the inclusion of a larger gc database with available data , although observations are obviously more critically plagued by reddening uncertainty ( cf . footnote [ foot ] ) . in fig . [ hbub ] the m31 ( dereddened ) data are compared with the expected ssp evolution according to four different population synthesis codes , namely ( * ? ? ? * and further web updates ) , @xcite , @xcite , and @xcite . again , models agree in the overall classification scheme , with the age of h@xmath26-poor gcs fully consistent with old ( i.e. @xmath60 gyr ) evolutionary scenarios and blccs better matched by young ( @xmath61 gyr ) ssps . selection ( see sec . 2.3 ) , might be noticing an old and very metal - poor composing stellar population . this interesting cluster does not appear in fig . [ hbd ] and [ hbub ] as it lacks @xmath59 and @xmath25 indices ( see table 1 ) . ] the emerging evidence , from the combined analysis of the blcc age range ( fig . [ hbd ] and [ hbub ] ) , coupled with the kinematical information of fig . [ kine ] leads eventually to the important conclusion that _ over 25% of the @xcite sample ( that is at least a 15% of the whole sample of confirmed m31 clusters ) consists of young blccs , located in the outskirts of the thin disc of m31 and sharing the kinematic properties of this galactic component . _ this global evidence is quite new and points to the existence of a significant difference between the m31 and mw globular cluster systems , which may have deep implications on the understanding of formation and evolution of the two parent galaxies . another important issue in the @xcite discussion concerns the claimed similarity of `` disc '' and `` bulge '' gc metallicity that , like in the mw , should be regarded as a striking sign of coevality for the whole gc population in m31 . in fact , if our _ bona fide _ blccs were mostly metal - poor , as apparently deduced from the metallicity values reported by @xcite , one would be left with the embarrassing scenario where the vast majority of the most metal - poor clusters in m31 should be young and likely members of the thin - disc galaxy subsystem . though not impossible in principle , this conclusion deserves however a deeper check . the @xcite metallicity scale comes from the @xcite [ fe / h ] estimates , relying on the empirical calibration of three narrow - band lick indices ( namely , the g - band , mgb and fe5335 ) vs. a coarse set of m31 gcs with available [ fe / h ] determination from the literature . given the prevailing contribution of _ old _ gc in this sample , the @xcite calibration actually introduces a subtly circular bias in the inferred value of [ fe / h ] for blccs due to the well recognized age - metallicity degeneracy @xcite , that makes young metal - rich clusters to resemble old metal - poor ones . this is shown in the three panels of fig . [ dmetal ] , where we plotted the [ fe / h ] difference as inferred from the lick index determinations for each individual cluster . from the figure one can immediately recognize a systematic bias induced by the [ fe / h ] estimates from the g - band . as a matter of fact , while the metallicity scale , as derived from the fe5335 and mgb indices , basically agrees in average ( see middle panel in fig . [ dmetal ] ) , for the blccs ( solid dots in the plots ) the g - band line strength tends to yield a significantly lower ( roughly @xmath62 dex ) value of [ fe / h ] . this systematic trend especially affects blccs , whose low age mimics the effect of metal deficiency . in a ssp , this ch - molecular feature maximizes its sensitivity to the coolest stars at the tip of the asymptotic ( agb ) and red giant branches ( rgb ) @xcite , and is therefore a mixed age / metallicity indicator , through a composite and complex dependence on the agb morphology . conversely , both the mg and fe indices are better sensitive to the bulk of stars in the rgb , via chemical opacity , and more confidently trace ssp metallicity @xcite . as a further argument , we note that even the observed morphology of the @xcite c - m diagrams for the four blccs observed so far with hst seem to indicate a metallicity in the range @xmath63\lesssim 0.0 $ ] . this also basically agrees with our `` first - look '' estimate of [ fe / h ] from fig . [ hbd ] , where the blccs h@xmath26 vs. @xmath25 distribution suggests a moderately enhanced metallicity ( i.e. @xmath13 \gtrsim -1.0 $ ] dex ) . the actual classification of blccs as `` young globular clusters '' or `` massive open clusters '' is somehow a question of semantics . the real point that is worth to investigate is if these clusters have a counterpart in our own galaxy , e.g. if clusters of similar age _ and _ luminosity do exist in the milky way . while the luminosity range spanned by blccs is comprised within that of ordinary globular clusters , the age distribution of present - day mw globulars is obviously not consistent with the young age of blccs . on the other hand , galactic open clusters ( oc ) are comparably young but they appear less luminous on average than blccs . the luminosity functions ( lfs ) of the sample b clusters and of the galactic ocs are compared in the left panel of fig . [ oc ] . the plot shows that galactic ocs with ages similar to blccs are sistematically fainter , the two histograms barely overlaps . the only galactic ocs that reach the luminosity range covered by blccs are younger than 30 myr ( e.g. they are clusters whose luminosity budget is dominated by a few massive stars , much different f.i . from the blccs studied by @xcite ) , while fig . 7 and fig . 8 above indicates that @xmath64 per cent of blccs is likely older than @xmath57 gyr . hence , regardless of the quite high incompleteness that probably still affects the lf of m31 clusters for @xmath65 , the milky way lacks ocs as luminous as blccs in the proper age range ( 200 myr @xmath66 age @xmath66 2 gyr ) . conversely , the right panel of fig . [ oc ] show that the luminosity range spanned by m31 blccs is very similar to that covered by lmc globular clusters of similar age ( data from * ? ? ? the above direct evidences lead to conclude that there is no galactic counterpart to the young massive m31 clusters studied here ; they are much younger than present - day galactic globulars and they are much more luminous than present - day galactic ocs of similar age . on the other hand blcc counterparts are quite common in the lmc . obviously , this conclusion rests on the age estimates of blccs as derived from fig . [ hbd ] , above . the involved uncertaintes leave ( formally ) open the possibility that several blccs have ages @xmath67 myr . if so , they should be interpreted as the counterparts of young open clusters of the milky way . the availability of a deep c - m diagram is probably the only observational test that can eventually establish the real nature of these objects . it remains to be explored how blccs would appear in the future , e.g. , in particular if they will look like classical globulars when they will become comparably old . if we assume blccs to consist of plain ssps , then one should expect their luminosity to fade with time , as far as the composing stellar population becomes older and photometrically dominated by low - mass stars . in particular , for a ssp of roughly solar metallicity and salpeter imf , evolutionary population synthesis models predict a quite tuned luminosity change such as @xmath68 over a wide range of age ( e.g. * ? ? ? * ; * ? ? ? * ) . according to the assumed age of present - day blccs , then one could infer the expected luminosity of these clusters at @xmath69 gyr and more consistently compare with the observed luminosity function of old mw gcs . the results of this illustrative excercise are summarized in fig . [ lumi ] ; it is evident from the figure that , in the more likely case of a current age in the range @xmath70 yrs , blccs would end up at @xmath71 yrs populating the low - luminosity ( and low - mass ) tail of current mw gc distribution . on the contrary , in the more extreme ( and quite unlikely ) case of a current age of only a few @xmath72 yrs we would be left at 10 gyr with extremely faint blccs , certainly out of range of the typical mw gcs . finally , if nowadays blccs are already evolved systems ( i.e. a few gyr or older ) , then at @xmath69 gyr their expected luminosity will not change so much and their distribution would maintain them fully consistent with the bulk of both m31 open clusters and mw gcs . a fair assessment of the present - day age distribution of this kind of clusters is therefore a mandatory step to consistently locate them in the appropriate evolutionary framework . till a few years ago , the m31 cluster system has been commonly viewed as an almost identical ( though much larger ) analog of the mw gc system , apart from small differences in average metallicity ( m31 about 0.2 dex richer than mw ) and controversial peculiarities in various spectral indices ( * ? ? ? * ; * ? ? ? * and references therein ) . recent wide - field imaging ( * ? ? ? * and references therein ) , together with high precision spectroscopy @xcite , have led however to a more detailed recognition and investigation of m31 substructures , including its gc system , leaving space to the idea that the andromeda and mw cluster systems are actually more different than conceived so far . in particular , @xcite have put forward a new scenario suggesting the existence in m31 of quite sizeable population of gcs associated to the thin disc and claimed to be old and metal poor . the natural consequence of such an evidence , if confirmed , is that the thin disc was formed when its oldest metal - poor globulars were in place and that no significant perturbation affected m31 , and in particular its thin disc , since then . an in - depth analysis of the claimed thin - disc gc members , and in particular of the metal - poor ones , may thus greatly help clarifying this important dynamical issue . the present study has tackled the problem from a different point of view , i.e.by building up a revised global sample of m31 gcs with intrinsically bluer color than the mw counterparts and assessing if these _ bona fide _ blccs ( sample `` a '' in our analysis ) would further discriminate with respect to `` normal '' m31 and mw globulars as far as other intrinsic properties ( like e.g. absolute magnitude , metallicity , spectral indices , structural parameters , and age ) and kinematic behaviour are concerned . on the basis of this comparison , a tuned and more physical selection based on the @xmath1 spectral index has been proposed ( sample `` b '' in previous discussion ) . quite interestingly , both color- and @xmath1-selected samples led to fully consistent conclusions of our analysis . it has been possible to convincingly demonstrate here that both samples consist of moderately young ( @xmath61 gyr ) stellar systems , not so metal poor as previously estimated , and basically sharing the thin disc kinematics ( see also * ? ? ? * on this line ) . since ( a ) essentially all the blue clusters in the thin disc subsystem as defined by @xcite are likely young and ( b ) the metallicity values adopted by these authors are sistematically underestimated ( at least for most blccs , see sect . 3.3 ) , serious doubts are casted on the actual presence of metal - poor ( and old ) clusters in the thin disc of m31 as claimed by morrison et al . this argument greatly weakens the possible contrast with the @xcite hypothesis of an equal - mass merging event suffered by m31 6 - 8 gyr ago . given the evidences presented so far and within the framework discussed above , one might naturally ask whether also the `` red '' clusters claimed to be member of the m31 thin disc subsystem actually are old and metal - poor ( as at least some of them seem to appear ) or whether they could display some spread in age and/or metal abundance . the answer to such a question is of paramount importance to settle in finer detail the mechanisms and formation timescale for the m31 thin disc and for the global evolution of the galaxy system as a whole . we would like to warmly thank kathy perrett for providing us with her full database of m31 gcs in electronic form . this research has been partially supported by the italian ministero delluniversit e della ricerca ( miur ) , through the cofin grant p.2002028935 - 001 , assigned to the project _ `` distance and stellar populations in the galaxies of the local group''_. bertone , e. , rodriguez - merino , l.h . , chavez , m. , buzzoni , a. 2003 , in _ galaxy evolution : theory and observations _ , eds . v. avila - reese , c. firmani , c. frenk , & c. allen , rev . astrophys . conf . series , 17 , 91 lccrcrrcrrrrll b008-g060 & 16.56 & 1.10 & 0.50 & 0.41@xmath75 0.38 & 0.397 & 3.50 & 4.96 & 319 & 0.70 & 15.34 & 19.95 & b & + b028-g088 & 16.86 & 0.88 & 0.05 & 1.87@xmath75 0.29 & 0.404 & 3.81 & 4.80 & 434 & 0.72 & 23.65 & 2.67 & b & 10 + b040-g102 & 17.38 & 0.29 & 0.01 & 0.98@xmath75 0.48 & 0.137 & 7.41 & 5.32 & 463 & 0.40 & 35.48 & 11.77 & a , b & 1,2,6,7 + b043-g106 & 16.96 & 0.28 & 0.14 & 2.42@xmath75 0.51 & 0.144 & 5.53 & 5.13 & 414 & 0.42 & 33.68 & 11.21 & a , b & 1,2,3,4,5,6,7,8 + b047-g111 & 17.51 & 0.72 & 0.09 & 1.62@xmath75 0.41 & & 3.53 & 5.57 & 291 & 1.02 & 13.80 & 24.58 & b & + b049-g112 & 17.56 & 0.52 & 0.18 & 2.14@xmath75 0.55 & & 9.31 & 5.83 & 481 & 0.39 & 27.56 & 7.27 & a , b & 7 + b057-g118 & 17.64 & 0.69 & & 2.12@xmath75 0.32 & 0.289 & 5.56 & 4.98 & 437 & 0.27 & 25.00 & 7.02 & b & + b066-g128 & 17.42 & 0.36 & 0.27 & 2.10@xmath75 0.35 & 0.061 & 4.67 & 4.95 & 389 & 0.70 & 29.61 & 13.02 & a , b & 3,5,6,7 + b069-g132 & 18.16 & 0.44 & 0.02 & 1.35@xmath75 0.43 & 0.338 & 7.17 & 5.72 & 295 & 0.27 & 3.44 & 11.91 & a , b & 7,10 + b074-g135 & 16.65 & 0.75 & 0.14 & 1.88@xmath75 0.06 & 0.346 & 3.92 & 5.15 & 435 & 5.61 & 17.38 & 22.06 & b & + b081-g142 & 16.80 & 0.54 & 0.26 & 1.74@xmath75 0.40 & 0.352 & 7.98 & 4.99 & 430 & 0.47 & 25.32 & 12.23 & a , b & 7 + b083-g146 & 17.09 & 0.76 & 0.06 & 1.18@xmath75 0.44 & 0.047 & 3.75 & 5.75 & 367 & 4.15 & 19.90 & 22.04 & b & + b091-g151 & 17.56 & 0.41 & 0.02 & 1.80@xmath75 0.61 & 0.269 & 7.30 & 5.02 & 290 & 0.16 & 2.08 & 7.01 & a , b & 4,7,10 + b114-g175 & 17.28 & 0.42 & 0.33 & & & & 4.90 & & & 3.87 & 0.58 & a & 10 + b160-g214 & 18.02 & 0.55 & 0.05 & 1.17@xmath75 1.25&&&4.78 & 354 & 0.46 & 7.98 & 13.45 & a & 10 + b170-g221 & 17.39 & 0.98 & 0.53 & 0.54@xmath75 0.24 & & 4.52 & 4.63 & 295 & 1.54 & 15.59 & 21.43 & b & + b210-m11 & 17.57 & 0.52 & 0.02 & 1.90@xmath75 0.32 & 0.162 & 6.83 & 4.98 & 265 & 0.11 & 7.69 & 12.70 & a , b & 7,10,12 + b216-g267 & 17.25 & 0.20 & 0.02 & 1.87@xmath75 0.39 & 0.177 & 5.66 & 5.72 & 84 & 0.10 & 26.91 & 0.92 & a , b & 2,3,5,7,8,10,12 + b222-g277 & 17.43 & 0.68 & 0.47 & 0.93@xmath75 0.95 & 0.349 & 8.47 & 6.91 & 311 & 1.07 & 10.14 & 16.17 & b & 10,11 + b223-g278 & 17.81 & 0.15 & 0.14 & 1.13@xmath75 0.51 & 0.297 & 4.44 & 5.65 & 101 & 0.01 & 26.39 & 3.81 & a , b & 7,10,12 + b237-g299 & 17.10 & 0.77 & 0.16 & 2.09@xmath75 0.28 & 0.167 & 7.43 & 5.08 & 86 & 4.54 & 21.81 & 17.48 & b & + b281-g288 & 17.67 & 0.84 & 0.50 & 0.87@xmath75 0.52 & 0.364 & 5.56 & 6.20 & 203 & 0.76 & 16.85 & 15.09 & b & + b295-g014 & 16.75 & 0.71 & 0.27 & 1.71@xmath75 0.15 & & 4.77 & 4.92 & 423 & & 85.95 & 19.69 & b & 8 + b303-g026 & 18.22 & 0.24 & 0.46 & 2.09@xmath75 0.41 & 0.252 & 5.78 & 5.46 & 464 & 1.84 & 65.51 & 5.53 & a , b & 7 + b307-g030 & 17.32 & 0.87 & 1.01 & 0.41@xmath75 0.36 & & 5.76 & 6.35 & 407 & 3.02 & 57.96 & 4.58 & b & + b314-g037 & 17.63 & 0.59 & 0.34 & 1.61@xmath75 0.32 & 0.402 & 5.19 & 6.48 & 485 & 0.51 & 69.94 & 10.77 & b & 3,8,11 + b315-g038 & 16.47 & 0.07 & 0.02 & 2.35@xmath75 0.54 & 0.116 & 4.77 & 5.18 & 559 & 0.18 & 55.65 & 0.83 & a , b & 1,2,3,5,7,8,9,12 + b318-g042 & 17.02 & 0.17 & 0.42 & & & & 5.31 & & & 52.16 & 1.09 & a & 1,2,3,4,5,7,8 + b319-g044 & 17.61 & 0.72 & 0.64 & 2.27@xmath75 0.47 & 0.113 & 5.37 & 5.49 & 535 & 0.21 & 52.03 & 1.54 & b & 1,2,3,5,7,9 + b321-g046 & 17.67 & 0.22 & 0.22 & 2.39@xmath75 0.41 & 0.229 & 6.29 & 5.52 & 518 & 0.05 & 55.54 & 7.17 & a , b & 3,7,8,11,12 + b322-g049 & 17.75 & 0.06 & 0.28 & & & & 5.14 & & & 46.31 & 0.57 & a & 6,7,8,11,12 + b327-g053 & 16.58 & 0.32 & 0.35 & 2.33@xmath75 0.49 & 0.160 & 4.09 & 4.87 & 528 & 0.25 & 47.70 & 3.23 & a , b & 3,4,5,6,7,8,11,12 + b331-g057 & 18.19 & 0.25 & & & & & 6.31 & & & 4.76 & 36.35 & a & + b342-g094 & 17.73 & 0.30 & 0.64 & 1.62@xmath75 0.02 & 0.003 & 7.06 & 6.52 & 479 & 0.66 & 40.46 & 12.04 & a , b & 1,2,6,7,9 + b354-g186 & 17.81 & 0.13 & 0.69 & & & & 5.45 & & & 35.36 & 26.68 & a & + b355-g193 & 17.76 & 0.53 & 0.02 & 1.62@xmath75 0.43 & 0.240 & 4.39 & 4.24 & 114 & 5.25 & 34.00 & 24.37 & a , b & + b358-g219 & 15.22 & 0.49 & 0.19 & & & & 5.06 & & & 64.79 & 58.32 & a & + b367-g292 & 18.45 & 0.32 & 0.17 & 2.32@xmath75 0.53 & 0.097 & 6.21 & 5.64 & 152 & 0.58 & 53.03 & 12.32 & a , b & 7 + b368-g293 & 17.92 & 0.26 & 0.36 & & & & 5.75 & & & 41.80 & 3.36 & a & 1,2,3,5,7,9 + b374-g306 & 18.31 & 0.44 & 0.33 & 1.90@xmath75 0.67 & 0.281 & 4.07 & 5.27 & 96 & 0.88 & 41.08 & 10.68 & a , b & 7 + b376-g309 & 18.06 & 0.45 & 0.23 & & & & 5.59 & & & 42.12 & 10.79 & a & 3,5,7,8 + b380-g313 & 17.01 & 0.47 & 0.33 & 2.31@xmath75 0.45 & 0.187 & 6.52 & 6.67 & 13 & 1.14 & 58.47 & 2.07 & a , b & 7,8,11,12 + b431-g027 & 17.73 & 0.49 & 0.29 & & & & 5.59 & & & 59.11 & 9.36 & a & 7 + b443-d034 & 18.20 & 0.80 & 0.52 & 2.37@xmath75 0.46 & & 6.72 & 5.40 & 532 & 0.08 & 50.47 & 4.58 & b & + b448-d035 & 17.49 & 0.61 & 0.01 & 2.16@xmath75 0.19 & & 6.70 & 6.54 & 552 & 0.16 & 43.17 & 2.77 & b & 7 + b451-d037 & 18.66 & 0.19 & & 2.13@xmath75 0.43 & & 3.50 & 4.40 & 514 & 0.20 & 33.01 & 2.57 & a , b & 7 + b453-d042 & 17.30 & 0.87 & 0.16 & 2.09@xmath75 0.53 & 0.234 & 4.12 & 5.27 & 446 & 0.47 & 23.69 & 5.79 & b & + b458-d049 & 17.84 & 0.49 & 0.91 & 1.18@xmath75 0.67 & 0.085 & 6.19 & 5.98 & 521 & 0.83 & 26.50 & 6.22 & a , b & 7 + b475-v128 & 17.56 & 0.31 & 0.10 & 2.00@xmath75 0.14 & & 5.96 & 7.24 & 120 & 1.02 & 45.00 & 3.92 & a , b & 7 + b480-v127 & 17.91 & 0.65 & 0.39 & 1.86@xmath75 0.66 & 0.188 & 5.19 & 4.89 & 135 & 0.50 & 44.30 & 8.38 & b & + b483-d085 & 18.46 & 0.27 & 0.08 & 2.96@xmath75 0.35 & & 5.58 & 5.87 & 53 & 0.09 & 58.16 & 0.58 & a , b & 7 + b484-g310 & 18.10 & 0.52 & 0.58 & 1.95@xmath75 0.59 & 0.110 & 5.70 & 5.45 & 104 & 0.06 & 46.62 & 8.51 & a , b & 7,12 + b486-g316 & 17.52 & 0.35 & 0.93 & & & & 5.02 & & & 9.45 & 41.39 & a & 7 + b189d - g047 & & & & 1.19@xmath75 0.29 & 0.064 & 4.24 & & 584&0.62 & 45.82 & 0.24 & b & + vdb0 & 15.28 & 0.23 & 0.37 & & & & & & & 47.41 & 4.31 & a & 3,4,5,6 + nb21-au5 & 17.86 & 0.31 & & & & & & & & 0.88 & 0.85 & a & + nb67-au13 & 16.14 & 0.48 & 0.03 & 1.43@xmath75 0.13 & & & & 113 & 1.66 & 1.69 & 3.75 & a & 7 + nb83 & 16.68 & 0.56 & 0.03 & 1.26@xmath75 0.16 & & & & 150 & 1.30 & 4.24 & 0.89 & a & + b006d - d036 & 18.00 & & & 2.16@xmath75 0.32 & 0.026 & 5.04 & & 522 & 0.51 & 36.31 & 2.25 & b & + b012d - d039 & 18.40 & & & 1.22@xmath75 0.41 & 0.182 & 7.09 & & 478 & 0.10 & 26.72 & 6.07 & b & + b015d - d041 & 17.80 & & & 1.14@xmath75 0.30 & & 7.32 & & 445 & 0.41 & 19.24 & 9.37 & b & + b111d - d065 & 18.10 & & & 1.80@xmath75 0.36 & 0.082 & 5.55 & & 130 & 0.81 & 27.38 & 2.27 & b & + b195d & 15.19 & 0.22 & 0.30 & 1.64@xmath75 0.19 & 0.166 & 4.29 & & 552 & 0.45 & 47.19 & 4.17 & a , b & + b206d - d048 & 19.06 & 0.12 & & 2.01@xmath75 0.99 & & 2.53 & & 490 & 0.07 & 27.97 & 6.40 & a & + b257d - d073 & 17.40 & & & 1.99@xmath75 0.19 & 0.501 & 5.49 & & 114 & 1.25 & 46.00 & 3.87 & b & + dao47 & & & & 1.13@xmath75 0.57 & 0.229 & 4.03 & & 490 & 0.35 & 33.09 & 7.67 & b & + v031 & 17.43 & 0.68 & 0.39 & 1.59@xmath75 0.06 & 0.497 & 5.84 & & 433 & 0.02 & 19.01 & 7.27 & b & 12 + [ tab1 ] lccrcrrcrrrrl b015-v204 & 17.79 & 1.41 & & 0.35@xmath75 0.96 & & 0.54 & 5.56 & 460 & 0.04 & 26.56 & 7.91 & 10 + b030-g091 & 17.39 & 1.93 & 0.71 & 0.39@xmath75 0.36 & & 1.62 & 4.69 & 380 & 1.62 & 24.83 & 1.24 & 10 + b090 & 18.80 & & & 1.39@xmath75 0.80 & & 3.28 & 4.62 & 428 & 0.43 & 13.12 & 4.60 & 10 + b101-g164 & 16.87 & 0.81 & 0.38 & & & & 4.49 & & & 8.01 & 2.55 & 10 + b102 & 16.58 & 0.62 & 0.12 & 1.57@xmath75 0.10 & 0.224 & 2.74 & 3.98 & 236 & 0.17 & 12.67 & 13.29 & 7 + b117-g176 & 16.34 & 0.65 & 0.45 & 1.33@xmath75 0.45 & 0.266 & 2.70 & 4.76 & 531 & 2.68 & 16.19 & 10.13 & 8 + b146 & 16.95 & 1.49 & 0.54 & & & & 4.86 & & & 1.51 & 3.23 & 10 + b154-g208 & 16.82 & 1.32 & 0.55 & & & & 4.95 & & & 3.17 & 4.21 & 10 + b164-v253 & 17.94 & 1.04 & & 0.09@xmath75 0.40 & 0.559 & 1.48 & 4.74 & 294 & 0.07 & 1.00 & 7.25 & 10 + b197-g247 & 17.63 & 1.08 & 0.19 & 0.43@xmath75 0.36 & & 1.14 & 4.94 & 9 & 1.30 & 18.58 & 1.06 & 10 + b214-g265 & 17.65 & 0.61 & 0.28 & 1.00@xmath75 0.61 & & 3.24 & 4.90 & 258 & 1.26 & 17.19 & 5.53 & 7 + b232-g286 & 15.67 & 0.72 & 0.10 & 1.83@xmath75 0.14 & 0.242 & 3.13 & 4.73 & 179 & 2.52 & 12.52 & 17.88 & 12 + b292-g010 & 16.99 & 0.90 & 0.02 & & & & 5.30 & & & 58.32 & 47.44 & 11 + b311-g033 & 15.44 & 0.96 & 0.14 & 1.96@xmath75 0.07 & 0.120 & 2.72 & 4.53 & 463 & 2.15 & 57.58 & 1.24 & 12 + b324-g051@xmath76 & 16.91 & 0.66 & 0.79 & & & & 4.87 & & & 3.10 & 36.44 & 8,11,12 + b328-g054 & 17.86 & 0.89 & 0.48 & & & & 5.44 & & & 3.20 & 35.57 & 7,10 + b347-g154 & 16.50 & 0.73 & 0.15 & & & & 5.16 & & & 27.83 & 26.66 & 8,12 + b423 & 17.72 & 0.60 & 0.10 & & & & 6.33 & & & 47.66 & 31.80 & 7 + b468 & 17.79 & 0.70 & & & & & 7.52 & & & 66.43 & 58.30 & 7 + nb16 & 17.55 & 0.66 & & 1.36@xmath75 0.12 & & & & 115 & 1.35 & 1.97 & 4.21 & 8 + b150d & 17.55 & 0.61 & 0.04 & & & & & & & 31.77 & 57.17 & 7 + [ tab2 ] lcrrrrl b060-g121 & 16.75 & 0.71 & 0.08 & 14.10 & 0.38 & 7 + b070-g133 & 17.07 & 0.54 & 0.19 & 10.83 & 0.56 & + b089 & 18.18 & 0.10 & 0.32 & 16.05 & 17.93 & 7 + b100-g163 & 17.91 & 0.88 & 0.15 & 22.53 & 13.75 & 7 + b108-g167 & 17.47 & 0.89 & 1.01 & 7.30 & 2.49 & 7,10 + b145 & 18.10 & 0.32 & 1.48 & 0.95 & 4.82 & 6,7 + b150-g203 & 16.80 & 1.10 & 0.39 & 5.97 & 0.87 & 10 + b157-g212 & 17.73 & 0.65 & 0.11 & 0.39 & 7.36 & 7 + b173-g224 & 18.27 & 0.02 & & 10.21 & 2.61 & + b192-g242 & 18.28 & 0.20 & 0.32 & 23.73 & 4.21 & 7,10 + b195 & 18.57 & 0.40 & 0.97 & 3.39 & 17.95 & 7 + b323 & 17.59 & 0.47 & 0.17 & 51.25 & 4.57 & + b330-g056 & 17.72 & 0.95 & 0.09 & 5.33 & 36.93 & 3 + b362 & 17.61 & 0.65 & 0.08 & 29.90 & 3.14 & 10 + b371-g303 & 17.54 & 0.48 & 0.39 & 40.53 & 7.04 & 7 + b414 & 17.98 & 0.50 & 0.08 & 4.58 & 93.72 & + b442-d033 & 17.94 & 0.39 & 0.16 & 47.36 & 1.87 & 7 + b452-g069 & 17.78 & 0.38 & 0.05 & 45.89 & 7.66 & 7 + b460 & 18.35 & 0.54 & 0.23 & 85.52 & 54.13 & + b469-g220 & 17.58 & 0.53 & 0.04 & 46.76 & 28.06 & 3,7 + b477-d075 & 18.46 & 0.32 & & 35.11 & 6.78 & + b508 & 17.12 & 0.46 & 0.27 & 54.77 & 57.54 & + b190d - g048 & 18.17 & 0.24 & 0.29 & 46.03 & 0.26 & + g085-v015 & 17.39 & 0.23 & 0.05 & 43.73 & 11.86 & + b028d - g100 & 18.04 & 0.40 & 0.21 & 26.71 & 4.77 & + g137 & 17.81 & 0.02 & 1.16 & 5.75 & 12.76 & + g270 & 17.30 & 0.46 & 0.77 & 23.70 & 2.39 & + h126 & 16.76 & 0.42 & 0.03 & 43.75 & 13.32 & + nb39-au6 & 17.94 & 0.21 & 0.40 & 0.18 & 0.79 & + nb42 & 18.49 & 0.52 & & 1.49 & 0.56 & + nb47-au3 & 18.75 & 0.10 & & 1.44 & 3.59 & + b065d - nb69 & 16.83 & 0.37 & & 1.41 & 2.52 & + nb79 & 18.27 & 0.53 & & 2.87 & 2.01 & + nb107 & 18.65 & 0.31 & & 2.28 & 1.62 & + b134d & 18.19 & 0.46 & 0.19 & 74.79 & 46.10 & + b137d & 18.51 & 0.54 & 0.05 & 59.80 & 45.32 & + b139d & 18.59 & 0.44 & & 116.18 & 2.35 & + b147d & 17.96 & 0.49 & 0.44 & 47.16 & 49.24 & + b158d & 16.50 & 0.33 & 0.02 & 103.80 & 5.69 & + b162d & 17.85 & 0.44 & 0.52 & 91.16 & 0.17 & + b171d & 17.80 & 0.43 & 0.08 & 57.63 & 14.65 & + b173d & 17.45 & 0.55 & 0.43 & 54.91 & 15.16 & + b196d & 18.79 & 0.21 & 0.94 & 54.20 & 10.79 & + b216d & 18.21 & 0.30 & 0.01 & 28.36 & 21.50 & + b218d & 18.19 & 0.44 & 0.02 & 41.70 & 28.85 & + b220d & 16.94 & 0.53 & 0.06 & 65.70 & 55.73 & + b225d & 18.15 & 0.50 & 0.19 & 55.82 & 50.27 & + b227d & 16.91 & 0.01 & 0.03 & 44.15 & 41.65 & + b246d & 18.05 & 0.09 & 0.77 & 49.50 & 15.14 & + b253d & 17.68 & 0.53 & 0.03 & 16.92 & 42.36 & + b261d & 17.60 & 0.46 & 0.04 & 60.74 & 11.30 & + b270d & 17.50 & 0.40 & 0.34 & 10.14 & 36.21 & + b272d & 16.34 & 0.51 & 0.37 & 24.27 & 64.72 & + b286d & 17.76 & 0.47 & 0.91 & 7.48 & 59.92 & + b293d & 17.91 & 0.54 & 0.12 & 29.58 & 81.90 & + b303d & 18.27 & 0.31 & 0.91 & 43.52 & 97.03 & + b312d & 18.35 & 0.53 & & 24.72 & 88.18 & + b320d & 17.78 & 0.45 & 0.56 & 72.08 & 13.85 & + b322d & 18.26 & 0.17 & 0.29 & 36.49&101.57 & + b324d & 18.74 & 0.55 & 0.03 & 58.96 & 27.38 & + da046@xmath76 & 18.75 & 0.53 & & 47.09 & 16.89 & 7 + da052 & 18.42 & 0.14 & 0.50 & 24.11 & 10.15 & 7 + da069@xmath76 & 17.48 & 0.26 & & 42.11 & 6.66 & 7 + v014 & 17.39 & 0.35 & 0.04 & 43.86 & 12.20 & 1 + v034 & 17.48 & 0.13 & 1.15 & 17.30 & 4.24 & + v133 & 18.36 & 0.06 & 1.26 & 51.54 & 5.58 & + v270 & 18.07 & 0.27 & 0.67 & 12.26 & 9.93 & + sh06 & 16.54 & 0.28 & 0.71 & 27.99 & 27.16 & + bh02 & 18.50 & 0.40 & & 49.25 & 1.15 & + [ tab3 ]

classical spins on a two dimensional triangular lattice with antiferromagnetic interactions are reconsidered . we find that the kosterlitz - thouless transition associated to the symmetry appears at a temperature 0.0020(2 ) below the ising transition at 0.5122(1 ) associated to the symmetry . the ising transition has critical exponents different from the standard ones . using extensive monte carlo simulations for equilibrium and dynamical properties we show that the lack of universality observed in previous studies is due to finite size corrections not taken account . likewise the kosterlitz - thouless transition has a critical exponent larger than the corresponding standard value . also the helicity jump at the critical temperature is smaller than in the ferromagnetic case in disagreement with theoretical predictions . we try using the concept of an `` quasi universality class '' to reconcile the standard critical behavior observable at higher temperatures with the different quasi universal one close to the critical region . pacs number(s ) : 75.10.hk , 05.70.fh , 64.60.cn , 75.10.-b

cond-mat0001122

the improvement in micro - fabrication has increased greatly the experimental studies of the josephson - junction arrays of weakly coupled superconducting islands.@xcite phase transitions in these arrays are similar to those in two dimensional @xmath0 spin systems and the application of a magnetic field introduces additional frustration effects . this explains partly the revival of interest in studying @xmath5 frustrated @xmath0 spin systems . the theoretical problem is connected to the presence of two symmetries and their coupling . indeed the continuous @xmath0 symmetry leads to a kosterlitz - thouless transition driven by the unbinding of vortex - antivortex pairs , while the frustration introduce an additional ising symmetry . also for helium-3 film an ising symmetry due to the p - wave order parameter exists besides the continuous symmetry of a phase like for helium-4 films.@xcite contrary to the three dimensional case , the critical region in films could be accessible to experimental observations . the model we are going to analyze has also a link to the physics of early universe,@xcite where more complicated couplings between symmetries are considered , extensive research has been done on two dimensional frustrated spin systems like the fully frustrated @xmath0 model @xcite or related zig - zag models,@xcite the triangular model we are discussing,@xcite the @xmath6 model,@xcite the @xmath0-ising model @xcite and the villain model.@xcite also other systems have the same symmetries : the 19 vertex model,@xcite the @xmath7 quantum spins,@xcite the coulomb gas representation,@xcite the @xmath8 model.@xcite or the rsos model@xcite therefore it is of great interest to compare the results for these different models in order to find out whether a universal critical behavior exists . a priori only two possibilities exist for the critical behavior . first , if the two phase transitions connected to the two symmetries are at the same temperature , then a new universal behavior should result . secondly if the transitions are at two different temperatures one could expect that the transitions are of the standard ising and kosterlitz - thouless type , especially if the transition temperatures are further apart . the presence of topological defects or vortices could change the reasoning . indeed numerous studies show transitions at different critical temperatures , however with exponents for the ising transition which vary from model to model and therefore can not belong to the standard nor to one universality class . the reason could be the existence of a line of fixed points generating the various exponents found in numerical works.@xcite a second proposition assumes only one new fixed point with the variation of critical exponents due to improper finite sizes corrections.@xcite thirdly olson suggested that a large screening length around the critical temperature of the kosterlitz - thouless ( kt ) transition prevents to see the true standard ising behavior.@xcite in this article we will show that several studies which indicate a non - universality of the exponents are due to finite size corrections . we are therefore led to favor the existence of only one fixed point . moreover to reconcile the result of studies done in the finite size scaling region very close to the critical temperature with those done at slightly higher temperatures we will discuss introduce the concept of an `` quasi universality class '' and give a physical interpretation as function of the size of vortex . the behavior of the kosterlitz - thouless ( kt ) transition have been less studied numerically . especially the critical temperature is difficult to find even for simpler ferromagnet @xmath0 system . we propose a new way using the binder parameter to overcome this difficulty and we obtain a critical temperature 0.002 below the ising transition . we obtain a helicity jump smaller than the jump of the ferromagnetic system and an exponent @xmath3 that is larger than @xmath9 for the ferromagnet , contrary to the theoretical predictions.@xcite also for the kt transition there is a discrepancy between simulations at or near the critical temperature and at high temperatures . indeed in this last region the exponent @xmath10 . we introduce a `` quasi universality class '' to explain this crossover . a combination of the metropolis algorithm and over - relaxation algorithm reduce the cpu time by an order of magnitude than the metropolis alone . with longer simulations , up to 10 times compared to previous studies , we gain two order better statistics . for the determination of critical exponents we have used the finite size scaling method . however , since the possible presence of a screening length could prevent to see the `` true '' behavior,@xcite we have also utilized the properties of the system in the short time critical dynamic @xcite which allows us to verify our results . both methods agree . the outline of the article will be the following . section ii is devoted to the presentation of the model . the ising and the kosterlitz - thouless transitions are studied in section iii and iv respectively , discussion and conclusion are disclosed in the last section . to avoid repetition with a previous article@xcite we will refer often to it . we study the @xmath0 spins on triangular lattices with antiferromagnetic interactions . the hamiltonian is given by : @xmath11 where @xmath12 is a two component classical vector of length unit , @xmath13 is the antiferromagnetic coupling constant ( @xmath14 ) , @xmath15 varies between 0 and @xmath16 and @xmath17 are the next nearest - neighbors . the competition between the interactions gives the famous `` @xmath18 '' structure where the spins are not collinear ( fig . [ fig.gs ] ) and where the frustration is divided amongst all links . due to this structure , the simulated lattice sizes must be a multiple of 3 . we simulate the sizes @xmath1912 , 18 , 24 , 36 , 48 , 60 , 81 , 105 , 123 , 150 . two ground states exist , not related by a global rotation , and therefore , in addition of the symmetry @xmath1 from the continuous aspect of the spins , an ising symmetry is present . to compute the order parameter for the @xmath1 symmetry we divide the lattice in three sublattices ( @xmath20 ) with only parallel spins in the ground state . after having calculated the magnetization of each sublattice @xmath21 we sum them to obtain @xmath22 : @xmath23 @xmath24 is the total number of the lattice sites . the order parameter @xmath25 of the ising symmetry is the sum of the chiralities @xmath26 of each cell ( fig . [ fig.gs ] ) . @xmath27 } \ , \\ \kappa & = & { 1 \over n ' } \,\big| \ , \sum_{i } { \bf \kappa}_i \ , \big| \ , \end{aligned}\ ] ] where the summation is over all cells and @xmath28 is their number . the chirality @xmath26 of one triangle is parallel to the @xmath29-axis and equal to @xmath30 in the ground state only . we note that we could take as a definition the sum of the signs of chiralities . the result should be similar because the length of the chiralities @xmath26 is not relevant for the critical behavior . to compute the critical properties we have to define for each temperature the following quantities : @xmath31}^2\rangle\end{aligned}\ ] ] @xmath32 is the energy , @xmath33 is the magnetic susceptibility per site , @xmath34 are cumulants used to obtain the critical exponent @xmath35 , @xmath36 are the fourth order cumulants , @xmath37 where @xmath38 is the coordinate of the site @xmath39 following one axe , @xmath40 is the helicity@xcite corresponding to the increment of the free energy for a long wavelength twist of the spin system,@xcite @xmath41 means the thermal average . this chapter is devoted to the ising symmetry . the first part is related to the properties in equilibrium , i.e. we calculate the various quantities after a time @xmath42 to thermalize the system much greater than the correlation time @xmath43 . the average is done on a time @xmath44 which is also much greater than @xmath43 . in a second part we will use the short time critical dynamic recently introduced , i.e. the reaction of the system to a quench at the critical temperature from an initial state . no thermalization is used ( @xmath45 ) . as explained in our previous article,@xcite we use a combination of @xmath46 metropolis steps and @xmath47 over - relaxation steps.@xcite the over - relaxation algorithm is `` microcanonical '' in the sense that the energy does not change under a step . this algorithm reduces considerably the autocorrelation time . we have thus two parameters ( @xmath46 and @xmath47 ) to fit in order to minimize the cpu time @xmath48 . our implementation for the over - relaxation algorithm is six times quicker than the metropolis algorithm : @xmath49 in fig . [ fig.nmet.nor ] we have plotted the autocorrelation time @xmath43 for the chirality @xmath25 , at the critical temperature @xmath50 calculated below , multiplied by @xmath46 , as function of @xmath51 in a log - log plot for a lattice size @xmath52 . @xmath43 is calculated with the method explained in the appendix of our previous article . the data are well described by : @xmath53 with @xmath54 and @xmath55 . using ( [ eqn.tau.met ] ) it is not difficult to show that the minimum of @xmath48 ( [ eqn.f.cpu ] ) occurs for : @xmath56 which is 7.95 for @xmath57 . we have determined @xmath58 for several lattice sizes @xmath59 . if we divide the result by @xmath59 , we obtain the ratio @xmath60 : @xmath61 , @xmath62 , @xmath63 , @xmath64 , @xmath65 . this ratio is nearly constant whatever @xmath66 and @xmath59 are . this is in accordance with a conjecture of adler@xcite stating that this ratio is proportional to the correlation length , i.e. the size in the finite size region where we have done the simulations . we have chosen @xmath67 and @xmath68 for the simulations . we show in fig . [ fig.tau.l ] @xmath48 for the chirality as function of the size of the lattice @xmath59 in a log - log plot for various algorithms at the critical temperature @xmath50 . a similar behavior is obtained for @xmath69 , i.e. the order parameter associated to the @xmath1 symmetry . @xmath70 is the metropolis algorithm alone , @xmath71 is used in combination with one step of an over - relaxation algorithm , @xmath72 in combination with @xmath73 steps of an over - relaxation algorithm , while @xmath74 correspond to the use of the wolff cluster algorithm.@xcite the slope of @xmath70 gives us the dynamical exponent @xmath75 we will discuss the value of this exponent at the end of this section . if we compare @xmath48 for @xmath70 and @xmath72 we observe that the gain is about 10 , which means that for the same time of simulation we obtain ten times better statistics ( i.e. `` independent '' data ) . we note that wolff s algorithm is less effective than the metropolis algorithm . this is understandable because this algorithm uses only one link at each time to construct the cluster and we know that three links must be taken into account ( at least one cell ) , therefore the algorithm can not generate the `` good '' cluster . indeed each cluster has about @xmath76 of the sites of the lattice , which is too high . moreover even if we would be able to construct a good cluster , there is no guarantee that the method to flip the spins inside would be efficient.@xcite we follow the procedure explained in the appendix of our previous article.@xcite we use in this work the histogram method@xcite which allows us from a simulation done at @xmath77 to obtain thermodynamic quantities at @xmath78 close to @xmath77 . however to reduce the systematic errors we do not save histograms for @xmath79 , @xmath80 as function of energy but save the data , that is the energy @xmath32 , the magnetization @xmath22 and the chirality @xmath25 . to avoid the use of a large space on the hard disk the data is saved only every @xmath81 sweeps ( see table [ table.tau ] ) . this method slightly increases our statistical errors ( approximative of 20% ) . however the systematic errors decrease and we have a better control of the total errors . since the data of a monte - carlo simulation are not independent , the calculation of errors must be done carefully . for simple quantities like the magnetization , the calculation is done with the standard formula of statistic but with the number of `` independent '' data @xmath82 equal to @xmath83.@xcite @xmath84 is the number of monte carlo steps . however for more complicated quantities we have to consider the correlations between the components of the formula and use , for example , the jackknife procedure.@xcite formula ( a8-a14 ) of our previous article@xcite can be used but we have to change @xmath84 to @xmath85 and @xmath43 to @xmath86 . however the error for the helicity is not given . the helicity can be written : @xmath87 where @xmath88 and @xmath89 are given by ( [ eqn.heli ] ) . applying the same method as in refs.@xcite , it is not difficult to obtain : @xmath90 } \ . \end{aligned}\ ] ] the simulations have been done using sizes between 12 to 150 . in the table [ table.tau ] we gave some details of the simulations where @xmath42 is the number of monte carlo steps ( i.e. one metropolis step followed by @xmath91 over - relaxation steps ) to thermalize the system ; @xmath44 is the number of monte carlo steps to average . the third column gives the autocorellation time , the fourth the time between two consecutive measures @xmath81 . the last column is the number of `` independent '' data . we report the last line in table [ table.revue.i ] , it should be compared to previous studies . our statistic is two orders greater than previous studies for similar sizes . this allows us to obtain better precisions for the quantities , typically one or two orders smaller , and therefore the critical exponents are more reliable . in particular we will see that the finite size corrections are important and this explains the variation of exponents found in previous studies ( see discussion at the end of this section ) . our first task is to find the critical temperature @xmath92 . the most effective way is to use binder s cumulant ( [ formuleuk ] ) in the finite size scaling ( fss ) region . we record the variation of @xmath93 with @xmath78 for various system sizes in fig . [ figcm ] and then locate @xmath92 at the intersection of these curves@xcite since the ratio of @xmath93 for two different lattice sizes @xmath59 and @xmath94 should be 1 at @xmath92 . due to the presence of residual corrections to finite size scaling , one has actually to extrapolate the results taking the limit ( ln@xmath58)@xmath95 in the upper part of fig . [ figtc2.all ] . we observe a strong correction for the small sizes . however for the biggest sizes the fit seems good enough and we can extrapolate @xmath92 as @xmath96 we note from the figure that it is a lower bound . the estimate for the universal quantity @xmath97 at the critical temperature is @xmath98 this value is far away from the two dimensional ising value @xmath99@xcite which is a strong indication that the universality class associated to the chirality order parameter is not of ising type . we note that it is not compatible with the value of the @xmath6 model 0.6269(7)@xcite which , in the standard formulation , means that the two systems belong to two different universality classes . however since this transition is a coupling between two it is not certain that this quantity stays universal . at @xmath100 the critical exponents can be determined by log log fits . we obtain @xmath101 from @xmath102 and @xmath103 ( fig . [ figv ] ) , @xmath104 from @xmath105 and @xmath106 ( fig . [ figx.all ] ) , and @xmath107 from @xmath25 ( see fig . [ figm ] ) . we observe in these figures a strong correction to a direct power law . it is worth noticing however that @xmath108 shows smaller corrections . using only the three ( four for @xmath106 ) largest terms ( @xmath19105 , 123 , 150 ) we obtain : @xmath109 the uncertainty of @xmath92 is included in the estimation of the errors . the large values of our errors are due to the use of only few sizes for our fits . the values obtained for four consecutive sizes are interesting . for the exponent @xmath35 we obtain : @xmath110 , @xmath111 , @xmath112 , @xmath113 , @xmath114 , @xmath115 , @xmath116 . these values cover a large range of the data obtained by previous studies ( see table [ table.revue.i ] ) and we strongly suspect that the lack of universality ( at least in the critical exponents ) is related to the corrections previously ignored . in the conclusion of this section we will show which important physical informations we are able to obtain from the sign of these corrections . we have tried to introduce a correction to calculate the exponents , for example for @xmath117 , we obtain @xmath118 and values for critical exponents fully compatible with ( [ formule.nu.k]-[formule.betasnu.k ] ) . the values given in ( [ formule.nu.k]-[formule.betasnu.k ] ) use the properties of the free energy at the critical temperature . but an error on @xmath92 leading to an error on the exponents , it is therefore advisable to find them without the help of @xmath92 . this can be done using the whole finite size scaling region with the method given in.@xcite it consists to plot , for example , the susceptibility @xmath119 as function of @xmath93 , choosing the exponents so that the curves collapse . this fit is more reliable than the fit at the critical temperature since as it does not depend only on the results at @xmath92 but on a large region of temperature . however the errors are a little bit larger . we show in fig . [ x2.1]-[x2.3 ] the results for three choices of @xmath104 1.82 , 1.78 and 1.74 . obviously the second is the best and we obtain @xmath120 compatible with the result of ( [ formule.gammasnu.k ] ) . in fig . [ fig.eta ] we have plotted the exponent @xmath121 calculated with a direct fit of the susceptibility ( like in fig . [ figx.all ] ) but changing the temperature . in the same diagram we have plotted the result ( @xmath122 ) found by the fit of @xmath105 as function of @xmath93 . this gives an estimate of the critical temperature ( see figure ) compatible with @xmath123 . the results are summarized in table [ table3 ] . in the next section we will try to compute the properties of this transition using the short time critical behavior . this method does not suffer from the same problems as the finite size scaling method and allows us to check our results . finally we will compare our results with those of previous studies . in this section we want to study our system using the dynamical behavior near the critical temperature . this particular method consists of quenching the system from zero temperature to the critical temperature and studying the short time dynamics , i.e. before reaching the equilibrium properties . it seems strange that the system shows critical behavior although it has not reached equilibrium . indeed for a long time this fact has not been utilized in numerical simulations . however it has been demonstrated that between a time @xmath124 and a longer time @xmath125 a region exists where the calculation of the critical properties of the transition is possible @xcite ( see the review of zheng@xcite ) . in this region the system has lost the non - universal informations ( @xmath126 ) but the correlation length @xmath127 is much smaller than the correlation length @xmath128 of the system in the equilibrium phase . since we are at @xmath129 the correlation length @xmath128 is infinite but in our simulation it is upper - bounded by the size @xmath59 of the lattice and the condition written as @xmath130 . after @xmath125 this condition is not satisfied and the system shows a crossover to equilibrium properties . in the region @xmath131 the system shows very simple power law even for the kosterlitz - thouless transition.@xcite the quantities to compute are the same as in formula ( [ formule.m]-[formuleuk ] ) but the average @xmath132 is now taken on different realizations , i.e. starting from the ground state with different random numbers for the metropolis procedure . to determine the range of time and the size @xmath59 that we use in the simulations we have to find @xmath125 where the system begins to show a crossover to the equilibrium properties . in fig . [ fig.a4 ] we show @xmath133 as function of the time @xmath134 for different sizes @xmath59 at the critical temperature @xmath135 in a log - log plot . we observe that for small systems @xmath136 . it grows for larger systems and since the sizes @xmath137 and @xmath138 give similar results for @xmath133 in fig . [ fig.a4 ] @xmath125 is bigger than @xmath139 . therefore we can use the entire region to calculate the critical exponents for the greatest size . we have studied systems with size @xmath138 for a simulation time @xmath134 up to 10,000 and we have averaged over 6000 realizations . this reflects more than one order statistics better than previous studies on the fully frustrated @xmath0 model ( ff@xmath0).@xcite for the algorithm we use only the metropolis algorithm ( model a in the classification of ref . @xcite ) . here we do not use the over - relaxation algorithm . the errors are calculated dividing the data in three sets of 2000 . we want now to verify the critical temperature @xmath140 found in the previous section . for this purpose we have simulated our system for three different temperatures at and around this value . we know that at the critical temperature the magnetization shows a power law behavior , a straight line in a log - log plot . for @xmath141 the system shows important corrections as observed in fig . [ fig.a2 ] . with our knowledge of the critical temperature found above , we are able to calculate very precise exponents . in the short time critical dynamic the quantities ( [ formule.m]-[formuleuk ] ) have at @xmath129 the behavior : @xmath142 where @xmath143 is the dynamical exponent and @xmath144 the dimension of the space , i.e. @xmath145 . we first compute @xmath146 using ( [ form.u.t ] ) . in fig . [ fig.u.t ] we have plotted @xmath93 as function of @xmath134 . we see that from @xmath147 the curve shows a linear behavior . using a fit from @xmath148 to @xmath149 ( shown by the two arrows ) we are able to obtain @xmath150 this value is consistent with the value found in equilibrium properties ( @xmath151 ) , but we are more confident in the result of dynamical properties which are less subject to systematic errors ( see appendix of ref . it is also consistent with the value of @xmath152 found studying the equilibrium properties of the fully frustrated @xmath0 model.@xcite nevertheless it is in contradiction with the value @xmath153 found studying the dynamical properties of the last model.@xcite however a slight change of the critical temperature could have a strong influence on @xmath146 and the calculation have been done with more than one order less statistics than our work . to obtain the other exponents we could use the formula ( [ form.m.t]-[form.v1.t ] ) , however we prefer to obtain directly @xmath101 and @xmath154 plotting @xmath93 and @xmath105 as function of @xmath102 and @xmath121 plotting @xmath105 as function of @xmath93 . this has been done in fig . [ fig.u.x.v1 ] and fig . [ fig.x.u ] . fits have been done for @xmath155 represented by the arrows . we have also calculated @xmath156 plotting @xmath157 as function of @xmath102 ( not shown ) . we obtain : @xmath158 our results are summarized in table [ table3 ] . the exponents calculated by the two methods ( equilibrium and dynamical ) agree very well . however we would prefer the results from short time critical dynamics since they are less sensitive to the finite sizes corrections . before discussing the interpretation of our results , we will compare them to those found previously on various models . in table [ table.revue.i ] we review the numerical studies related to our work . we think that the errors quoted in this table are mostly too optimistic estimates . we first compare our work with previous simulations using the same methods as this work ( monte carlo finite size scaling - mc fss ) . there are two important columns : the maximum size @xmath159 used ( fourth column ) and the number of independent data @xmath82 ( seventh column ) . indeed we have seen that strong finite size corrections exist in our system and we can not see them if the largest size is too small . when comparing the calculations with a similar largest size ( @xmath160),@xcite we observe that our statistics ( @xmath82 ) is between 100 to 1000 times better . this allows us to see the corrections not seen before and to explain the lack of universality observed in previous studies ( see subsection `` equilibrium properties '' ) . the results , for the triangular lattice , in favor of the ising ferromagnetic universality class ( @xmath161)@xcite are not reliable due to too small statistics ( @xmath162 ) . for the other methods none of the authors have reported the strong corrections which should exist . nevertheless almost all results are compatible with ours ( @xmath163 ) and are in favor of a single universality class . only results on the @xmath0-ising model seem to show different universality classes but simulation sizes are so small that results are not conclusive . our results are thus strongly in favor of the picture of@xcite which suggests the phase diagram drawn in fig . [ fig.i - xy ] . @xmath88 and @xmath164 are free parameters . similar phase diagrams appear for the @xmath6 model,@xcite the zig - zag model@xcite and the rsos model@xcite . a variation in free parameters changes only the initial point of the renormalization group flow on the line @xmath165 but all trajectories converge to the same fixed point @xmath166 , i.e. a single universality class . the finite size corrections will be more or less important following the initial point . to verify this interpretation it would be useful to test several initial points of the line @xmath167 of the @xmath0-ising model with similar sizes as in this work ( @xmath168 ) to see if the systems belong to the same universality class and , maybe more interesting , to observe the corrections to scaling . the picture should be very similar to the potts model with disorder where this crossover , and therefore the corrections , have been observed.@xcite one other numerical study also interesting should be the calculation of the properties of the 19-vertex model by the density - matrix renormalization group ( dmrg ) . the model has already been studied by monte carlo transfer matrix @xcite but only for small sizes ( @xmath169 ) . on the contrary the dmrg allows to treat @xmath59 chains of infinite sizes . we note that the method has been proved to be very efficient for the 19-vertex for the ferromagnetic system ( i.e. with different internal parameter).@xcite we now discuss the results of the monte carlo ( mc ) simulation in the high temperature ( ht ) phase . in this region we have to keep the correlation length @xmath170 much smaller than the size @xmath59 of the lattice : @xmath171 . studies on the ff@xmath0 model have been done by nicolaides@xcite and by jose and ramirez.@xcite they found two very different results . however with a close look on their simulations we have found that the results of nicolaides were not reliable and his errors more than strongly underestimated . therefore the ht mc seems in favor of a new universality class but with an exponent @xmath172 larger than found in the finite size scaling mc ( @xmath173 ) . on the other hand olson@xcite has done a ht mc on the villain model , i.e. where the spinwave are decoupled from the vortex , and has observed a very interesting behavior . he proposed that a screening length @xmath174 exists , due to the kosterlitz - thouless transition and , that in addition to the condition @xmath175 , the sizes have to satisfy the condition @xmath176 . in this case the system shows a standard ising behavior @xmath161 while it is in the crossover to this behavior if the condition @xmath177 is not respected . indeed he found some indication that the exponents @xmath178 when the two conditions are satisfied . however since the numerical results are not extremely precise , further simulations are needed to prove this interpretation . let s assume that this last interpretation is correct . it does not prove that there is no new fixed points contrary to the claim of olson . indeed we have shown in this work and in @xcite that the corrections lower the result for @xmath35 ( from 0.90 to 0.80 ) and therefore that our system is not in the crossover to the ferromagnetic ising behavior @xmath161 . we then conclude that a new `` stable '' fixed point exists for a range of size @xmath59 . we call this new fixed point an `` quasi fixed point '' by similarity with the 3d case where an `` almost second order '' exists for a certain range of size @xmath59 before showing the `` true '' first order transition.@xcite this can be understood if we admit that the kosterlitz - thouless transition coupled the spins at large distance by the intermediate of @xmath179 . therefore it has tendency to bring the critical behavior of the ising symmetry to the mean - field solution ( @xmath180 ) exact for infinite interactions . since @xmath179 is finite above the critical temperature the thermodynamic limit of this `` quasi fixed point '' is not stable . to be exhaustive we mention the study of the transition @xmath0-potts(@xmath181).@xcite we have shown that the transition is of first order whenever @xmath182 . in this case the transition will stay of first order even in the thermodynamic limit , since the correlation length is finite . therefore the new fixed point ( or his absence ) is stable . this fact is rather against the interpretation of olson but it is possible that the stability of the new fixed point changes between @xmath183 ( ising ) and @xmath184 . on the contrary it would say that for the @xmath0-ising model , the two behaviors in the finite size scaling region ( @xmath185 ) and far away of the critical temperature ( @xmath161 ) are different and stable in the thermodynamic limit . we will present in this section our results for the kosterlitz - thouless ( kt ) transition@xcite associated to the @xmath1 symmetry . this transition is characterized by the unbounding of the vortex - antivortex pair at the critical temperature . at low temperature there are only some pairs of vortex - antivortex while at high temperature alone vortex ( or antivortex ) can exist . this picture is considered as correct in the ferromagnetic and frustrated case . however the two cases could show some differences for the jump of the helicity at the critical temperature @xmath186 and for the exponent @xmath187 . the helicity @xmath40 is the answer of the system to a twist in one direction . it has been shown that at @xmath186 this quantity jump from 0 to an universal quantity@xcite for all ferromagnetic systems . moreover the behavior of @xmath40 as function of the lattice size has been predicted.@xcite the use of @xmath40 has been proved to be very useful for the ferromagnetic case and in particular to determine the critical temperature.@xcite the situation is not so clear for frustrated systems . it has been suggested that the jump at @xmath186 could be `` non universal '' i.e. different from the ferromagnetic case@xcite and no scaling with the system size has been proved . therefore rather than using the helicity to obtain informations on the transition , we will use a new method for the kt transition that we have introduced in.@xcite it consists in using binder s cumulant to study this transition . it was proved in these articles that , contrary to the common belief , the binder cumulant for ferromagnetic @xmath0 systems crosses for different sizes , allowing thus an estimate of the critical temperature and moreover an estimate of the exponent @xmath187 without the precise knowledge of the critical temperature . the ferromagnetic @xmath187 has been proved equal to @xmath188 @xcite with logarithm corrections@xcite while it has been predicted to be smaller in frustrated cases.@xcite in this section we will first present our results for the equilibrium properties and second verify them using the short time dynamical properties . we proceed in a similar way as for the ising order parameter , replacing @xmath25 by @xmath22 . we record the variation of @xmath190 ( [ formuleum ] ) with the temperature for various system sizes in fig . [ figcm.kt ] . we want to underline the differences between our result in the frustrated case and in the ferromagnetic case : the crossing region is one order of magnitude less than for the standard @xmath0 model ( compare with fig . 1 of ref . we then locate the intersection of these curves and plot the results in the lower part of fig . [ figtc2.all ] . let us first consider a power law behavior at @xmath191 for this system . we then have to consider a linear fit for ( ln@xmath58)@xmath192 . we observe corrections for the smallest sizes @xmath1912 , 18 and 24 but the others seem to converge to the temperature @xmath193 we note from the figure that it is an upper bound for the critical temperature , i.e. it can not reach @xmath194 and therefore we obtain two transitions . secondly we consider the behavior to be exponential as in the standard @xmath0 model . in this case fig . 2 of ref . @xcite shows that a linear fit could be wrong and that a `` crossover '' to a different critical temperature could be observed for bigger @xmath58 , i.e. greater sizes . however contrary to the ferromagnetic @xmath0 model , the region of crossing is very small and the different linear fits tend only to one critical temperature . we observe the same behavior for the @xmath6 model.@xcite we conclude that the linear fit works well enough and we will show below strong arguments in favor of the temperature ( [ eqn.tc.m ] ) . with the help of the critical temperature we have found an estimate of @xmath190 at @xmath186 fitting the value with a law @xmath195 : @xmath196 the exponent @xmath187 is obtained by a log - log fit of @xmath197 as function of @xmath59 shown in fig . [ figx.all ] . we obtain @xmath198 the fit has been done using only the four largest sizes ( @xmath19 81 , 105 , 123 and 150 ) , disregarding the smallest sizes which show small corrections . we note that for small sizes @xmath189 is smaller , for example if we use only the sizes from 12 to 60 we obtain @xmath199 . therefore if we take into account the corrections , the exponent moves away from the standard value 0.25 for the ferromagnetic systems . this shows that we are not in a crossover to this last behavior . our value is in agreement with those found for the @xmath6 model.@xcite our results are summarized in table [ table3 ] . from a theoretical point of view the @xmath200 transition has an exponential behavior , i.e. a correlation length of the form @xmath201 $ ] , however a power law behavior like ( @xmath202 ) can not be excluded _ numerically_. in the latter case the critical exponent @xmath35 can be calculated with the cumulant @xmath203 ( [ formulev2.m ] ) . we have obtained @xmath204 and the results show important corrections . this value being in contradiction with the value @xmath205 found in the @xmath6 model , we can exclude a power law behavior . as for the ising order parameter , the calculation of the exponents has been done at the critical temperature but an error on @xmath186 leads to errors on the exponents , it is then interesting to find them without the help of @xmath206 . this can be done using the same method as described before . we have shown in ref . @xcite that this method is accurate enough in order to obtain @xmath187 whatever the type of the behavior is ( power law or exponential ) . in fig . [ x2.1.kt]-[x2.3.kt ] we show our results for three values of @xmath189 , 0.34 , 0.375 , 0.41 . obviously the second value is the best and obtain : @xmath207 which is compatible with ( [ formule.eta.m ] ) . in fig . [ fig.eta ] we have plotted the exponent @xmath189 calculated with a direct fit of the susceptibility ( like in fig . [ figx.all ] ) but changing the temperature . in the same diagram we have plotted the result ( @xmath208 ) found by the fit of @xmath197 as function of @xmath190 . this gives an estimate of the critical temperature ( see figure ) compatible with @xmath186 . we now want to demonstrate that the two transitions can not appear at the same temperature . first we have shown in fig . [ figtc2.all ] that @xmath194 is a lower bound for the critical temperature while @xmath209 is a upper bound . moreover if the ising transition appears at @xmath210 , the exponent @xmath211 must be equal to 0.115 ( @xmath212 ) by a direct log - log fit at this temperature ( see fig . [ fig.eta ] ) . in fig . [ x2.4 ] we have plotted @xmath105 as function of @xmath93 for this value . the curves do not collapse . now if the kosterlitz - thouless transition appears at @xmath140 then @xmath213 ( see fig . [ fig.eta ] ) and the curves @xmath197 as function of @xmath190 do not collapse in fig . [ x2.5.kt ] . therefore , even if the critical temperatures are different by only 0.4% , we are able to conclude that the kt transition appears at lower temperature than the ising transition . moreover we want to stress the fact that the phase diagram of the ff@xmath0@xcite is not in contradiction with our picture . in this work the authors have shown that with varying a parameter ( @xmath214 ) it is possible to decouple the two symmetries with the ising transition appearing at lower temperature than the kt transition . however in this case the ising transition is due to the contraction to 0 ( or @xmath215 ) of the turn angle between the spins and not to the mixed of different chirality signs.@xcite therefore two physical properties exist near @xmath216 which lead to a complicated phase diagram with a cross of the critical lines . we present know our results for the helicity @xmath40 ( [ eqn.heli ] ) . in fig . [ fig.heli ] we have plotted @xmath40 as function of the temperature . @xmath217 is the temperature of simulation , @xmath186 is our estimate for the critical temperature from above . @xmath218 is the universal jump for a ferromagnetic system . @xmath219 is the best fit of @xmath40 with the scaling form as function of the system size @xmath59 valid in ferromagnetic systems@xcite : @xmath220 where @xmath221 is the jump at @xmath129 and @xmath222 a free parameter . we note that our results are very similar to those of lee and lee.@xcite since it is difficult to obtain , with the histogram method , informations for the biggest sizes too far away from the temperature of simulation ( @xmath223 ) we take their result for @xmath219 : @xmath224 this value does not agree with @xmath209 found above . if we admit @xmath219 as critical temperature the exponent @xmath189 must be equal to 0.22 ( see fig . [ fig.eta ] ) and the curves @xmath197 as function of @xmath190 do not collapse in fig . [ x2.4.kt ] . therefore we can doubt the validity of ( [ eqn.heli.l ] ) for the frustrated case . more interesting is the comparison of @xmath40 at @xmath210 with the ferromagnetic jump @xmath225 . in fig . [ fig.heli.tc ] we have plotted these two quantities as functions of the size @xmath59 . our results suggest a jump at the critical temperature smaller than the ferromagnetic jump in contradiction of the suggestion of.@xcite this surprising result could induce some doubts about our previous results . in order to verify them , we will present now the dynamical properties of the kt transition . this analysis is very similar to those done for the ising symmetry . zheng and co - workers@xcite have proved that a system quenched from a zero temperature to the critical temperature @xmath129 shows that a kt transition has a similar picture as the ising transition , i.e. a power law behavior as function of the time @xmath134 : @xmath226 if the system is not quenched to the critical temperature corrections are present and the behavior ceases to be linear in a log - log plot . in fig . [ fig.a3 ] we have plotted @xmath79 as function of @xmath134 for different temperatures . we observe a linear behavior for @xmath210 and clearly a deviation from this behavior for the other temperatures and particular for @xmath227 . this is a strong indication that @xmath186 is the good choice of the critical temperature . we have calculated the exponents @xmath189 and @xmath228 at @xmath229 from ( [ form.m.t.kt]-[form.u.t.kt ] ) by similar methods used for the ising symmetry . since the average has been done only on one sample of 500 configurations we are not able to calculate the errors . we obtain : @xmath230 the value of @xmath189 is in agreement with those found in equilibrium ( see table [ table3 ] ) . in conclusion we found that , numerically , the kosterlitz - thouless in the finite size scaling ( fss ) region has a complete different behavior than for ferromagnetic systems . we found an exponent @xmath231 larger and a jump of the helicity smaller in frustrated case . our exponent @xmath189 is in agreement with our results for the @xmath6 model @xmath232 ( see table [ table3 ] ) with a similar method as this work , and with the result found in the ff@xmath0 , @xmath233 , using a transfer matrix method.@xcite however they are in disagreement with the values found in the high temperature ( ht ) region @xmath234.@xcite although new and surprising , we believe that our results are reliable and present hereafter the numerical arguments in favor of them : * fig . [ figtc2.all ] is very suggestive for the value of the critical temperature and shows that the behavior of the kosterlitz - thouless transition in frustrated systems is very different from the ferromagnetic case.@xcite * the behavior of @xmath197 as function of @xmath190 ( fig . [ x2.1.kt]-[x2.3.kt ] ) which give @xmath189 . * the universality of our results for @xmath189 for different models ( see table [ table3 ] ) . * the dynamical properties shown in fig . [ fig.a3 ] are an another indication that our choice for critical temperature is correct . moreover the exponent @xmath189 is in good agreement with those found in equilibrium properties . since the two calculations ( in the ht and fss regions ) seem correct we suggest an hypothesis similar to those given for the ising symmetry , i.e. different behavior in the fss and ht regions . we give hereafter a physical interpretation of this behavior . at high temperature the radius of the vortex are very small and they are slightly connected together while the correlation length for the ising symmetry is bigger than the radius of the vortex . the situation is different in the ffs region . in this case the radius of the vortex are large enough and the properties should depend on the domain walls due to the ising symmetry with the ising correlation length @xmath235 smaller or of the same size as the radius of the vortex . this argument is exactly the opposite of those given to explain the difference of behavior for the ising transition : the `` true '' ferromagnetic ising behavior appears only when @xmath235 is much larger than a screening length @xmath236 , i.e. the @xmath179 . this difference of argumentation reflects the differences between the two symmetries : the kosterlitz - thouless transition is a transition driven by topological defects ( vortex - antivortex ) , i.e. by `` local '' behavior , contrary to the ising transition . we call our picture `` quasi universality class '' in a similar way as the ising symmetry . the two interpretations diverge in the sense that in the thermodynamic limit ( infinite size ) the new universality class for the ising symmetry could be not stable ( however see ref . @xcite ) while it is for the kosterlitz - thouless transition . however the two behaviors are unstable in the high temperature region , i.e. sufficiently far away from the transitions . this article is devoted to the study of the triangular frustrated @xmath0 spin system in two dimensions by extensive monte carlo simulations . this system is characterized by two symmetries : a @xmath1 symmetry due to the continuous nature of the @xmath0 spins and an ising symmetry due to the degeneracy of the ground state . our study is restricted to the region where the system sizes that we simulate are smaller than the values of the correlation lengths of the @xmath1 symmetry like those the ising symmetry should have in an infinite system . we have shown that the kosterlitz - thouless transition connected to the @xmath1 symmetry appears at lower temperature than the ising symmetry . our very large statistics ( two order larger than previous studies ) allow us to take into account the corrections to the scaling laws . with the knowledge of corrections we have shown that the ising transition belongs to a new universality class , i.e. the renormalization group flow tends toward a new `` stable '' fixed point . these corrections explain also the lack of universality found by several authors . knowing that the transition has a different behavior in the high temperature phase we have introduced the idea of an `` quasi fixed point '' or `` quasi universality class '' . in this interpretation this new fixed point could be unstable in the thermodynamic limit but the true behavior can not be reached in our limited accessible sizes in numerical simulations . accepting this interpretation , the situation is similar to those who appear in frustrated systems in three dimensions.@xcite we note that our interpretation goes further than the interpretation of olson@xcite who predicts just an impossibility to obtain the true ising behavior in the finite size scaling region . we have also studied the kosterlitz - thouless transition associated to the @xmath1 symmetry . we have shown that this transition belongs to a new universality class characterized by an universal exponent @xmath237 and a helicity jump smaller than those in the ferromagnetic case in disagreement with theoretical predictions . we have introduced the concept of an `` quasi universality class '' for this transition to explain the different behaviors observed between the finite size scaling and the high temperature regions . in this interpretation the new behavior is stable in the thermodynamic limit near the critical temperature but not when the temperature is much larger than the critical temperature . we have given a physical interpretation of this new concept . to bring new informations and verify our predictions for these systems we have three possible ways . the first way is to improve the theoretical approaches of these systems . in particular it should be extremely interesting to have a theoretical result for exponents of the new fixed point for the ising symmetry . however since it is difficult to describe the coupling between the two symmetries , the precise calculation of exponents seems , for the moment , an impossible task . another possibility is the experimental approach in the josephson - junction array@xcite or in films of @xmath238he.@xcite however one of the problems of experiments is to have access to the interesting quantities . for example it is usually difficult to measure the chirality in these systems because it is not related directly to measurable quantities . the easiest way to obtain reliable information seem to be numerical simulations . indeed the use of powerful computers and fast algorithms allows to obtain precise results out of reach some years ago . we have to study different models to observe the universality class in the finite size scaling region but also in the high temperature region to verify the picture predicted in this work . this work was supported by the alexander von humboldt foundation . we are grateful to professors k. d. schotte and i. peschel for discussions and to l. beierlein for critically reading the manuscript . + @xmath2 ( eq)&0.5122(1)&0.6320(20)&0.815(20)&1.450(42)&0.227(9)@xmath239&0.0896(71)&2.30(4 ) + @xmath2 ( dyn)&&&0.818(9)&1.445(20)&0.235(7)@xmath239&0.0967(13)&2.39(5 ) + @xmath240 ( eq ) & 0.56465(8)&0.6269(7)&0.795(20)&1.391(43)&0.250(10)@xmath239&0.101(10)&2.29(4 ) + @xmath1 ( eq)&0.5102(1)&0.6497(12 ) & & & 0.365(10 ) & & + @xmath1 ( dyn ) & & & & & 0.36&&2.10 + @xmath241 ( eq)&0.56271(5)&0.638(5 ) & & & 0.345(5 ) & & c.j . lobb , physica b * 152 * , 1 ( 1988 ) . ling , h.j . lezec , j.s . tsai , j. fujita , h. numata , y. nakamura , y. ochiai , chao tang , p.m. chaikin , and s. bhattacharya , phys . * 76 * , 2989 ( 1996 ) . halsey , j. phys . c * 18 * , 2437 ( 1985 ) . korshunov , j. stat . phys . * 43 * , 17 ( 1986 ) . v. kotsubo , k.d . hahn , and j.p . parpia , phys . lett . * 58 * , 804 ( 1987 ) . j. xu and c. crooker , phys . lett . * 65 * , 3005 ( 1990 ) . t. vachaspati , phys . lett . * 76 * , 188 ( 1996 ) ; contemp . phys . * 39 * , 225 ( 1998 ) . j. villain , j. phys . c * 10 * , 1717 ( 1977 ) . s. teitel and c. jayaprakash , phys . . lett . * 51 * , 1999 ( 1983 ) ; phys . b * 27 * , 598 ( 1983 ) b. berge , h.t . diep , a. ghazali , and p. lallemand , phys . b * 34 * , 3177 ( 1986 ) nicolaides , j. phys . a * 24 * , l231 ( 1991 ) . jose and g. ramirez - santiago , phys . lett . * 77 * , 4849 ( 1996 ) ; * 68 * , 1224 ( 1992 ) ; phys . rev . b * 49 * , 9567 ( 1994 ) . thjissen and h.j.f . knops , phys . b * 42 * , 2438 ( 1990 ) . e. granato , m.p . nightingale , phys . b * 48 * , 7438 ( 1993 ) . s. lee and k.c . lee , phys . b * 49 * , 15184 ( 1994 ) . luo , l. schlke , and b. zheng , phys . lett . * 81 * , 180 ( 1998 ) . j. lee , j.m . kosterlitz , and e. granato , phys . b * 43 * , 11531 ( 1991 ) . m. benakly and e. granato , phys . b * 55 * , 8361 ( 1997 ) . boubcheur and h.t . diep , phys . b * 58 * , 5163 ( 1998 ) . s. miyashita and h. shiba , j. phys . jpn * 53 * , 1145 ( 1984 ) . s. lee and k.c . lee , phys . b * 57 * , 8472 ( 1998 ) . l. caprioti , r. vaia , a. cuccoli , and v. togneti , phys . b * 58 * , 273 ( 1998 ) . j. f. fernandez , m. puma , and r. f. angulo , phys . rev . b * 44 * , 10057 ( 1991 ) . d. loison and p. simon , accepted by phys . b , accessible at http://www.physik.fu-berlin.de/@xmath242loison/articles/reference18.html m.p . nightingale , e. granato , and j.m . kosterlitz , phys . b * 52 * , 7402 ( 1995 ) . j. lee , e. granato , and j.m . kosterlitz , phys . lett . * 66 * , 2090 ( 1991 ) ; phys . b * 44 * , 4819 ( 1991 ) . p. olson , phys . lett . * 75 * , 2758 ( 1995 ) ; * 77 * , 4850 ( 1996 ) ; phys . rev . b * 55 * , 3585 ( 1997 ) . knops , b. nienhuis , h.j.f . knops , and j.w . blte , phys . rev . b * 50 * , 1061 ( 1994 ) . e. granato , phys . b * 45 * , 2557 ( 1992 ) . m. yosefin and e. domany , phys . rev . * b 32 * , 1778 ( 1985 ) . minnhagen , phys . lett * 54 * , 2351 ( 1985 ) ; phys . b * 32 * , 3088 ( 1985 ) grest , phys . b * 39 * , 9267 ( 1989 ) . lee , phys . b * 49 * , 3317 ( 1994 ) . choi and d. stoud , phys . b * 32 * , 5773 ( 1985 ) . g.s . jeon , s.y . park , and m.y . choi , phys . b * 55 * , 14088 ( 1997 ) . s. lee , k.c . lee , and j.m . kosterlitz cond - mat/9612242 ( unpublished ) . fisher , m.n . barber , and d. jasnow , phys . rev . a * 8 * , 1111 ( 1973 ) . t. ohta and d. jasnow , phys . b * 20 * , 139 ( 1979 ) . m. creutz , phys . d * 36 * , 515 ( 1987 ) adler , phys . rev . * d 23 * , 2901 ( 1981 ) ; j. apostolakis , c.f . baillie , and g.c . fox , phys . d * 43 * , 2687 ( 1991 ) . u. wolff , phys . lett . * 62 * , 361 ( 1989 ) ; nucl . b * 322 * , 759 ( 1989 ) . s. caracciolo , a. pelissetto , and a.d . sokal , nucl . b ( proc . suppl . ) * 20 * , 55 ( 1991 ) ; * 20 * , 68 ( 1991 ) . a. m. ferrenberg and r. h. swendsen , phys . * 61 * , 2635 ( 1988 ) ; * 63 * , 1195 ( 1989 ) . h. mller and k. binder , j. stat . * 8 * , 1 ( 1973 ) . b. efron , _ the jackknife , the bootstrap and other resampling plans _ ( siam , philadelphia , pa , 1982 ) . k. binder , z. phys . b * 43 * , 119 ( 1981 ) , phys . rev * 47 * , 693 ( 1981 ) . g. kamienarz and h.w.j . blte , j. phys . a 26 ( 1993 ) 201 . d. loison , physica a * 271 * , 157 ( 1999 ) . z. phys . b * 73 * , 539 ( 1989 ) . huse , phys . b * 40 * , 304 ( 1989 ) . k. humayun and a.j . bray , j. phys . a * 24 * , 1915 ( 1991 ) . b. zhang , int . phys . * 12*. 1419 ( 1998 ) . luo and b. zheng , mod . b * 11 * , 615 ( 1997 ) . p.c . hohenberg and b. halperin , rev . 49 * , 435 ( 1977 ) . s. groe pawig and k. pinn , int . c 9 , 727 ( 1998 ) g. zumbach , phys . lett . * 71 * , 2421 ( 1993 ) ; nucl . b * 413 * , 771 ( 1994 ) . d. loison and k.d . schotte , eur . j. b ( 1999 ) ( in press ) , accessible at http://www.physik.fu-berlin.de/@xmath242loison/articles/reference10.html . d. loison , submitted to phys . b , accessible at http://www.physik.fu-berlin.de/@xmath242loison/articles/reference19.html . m. picco , cond - mat/9802029 ( unpublished ) . y. honda and t. horiguchi , phys . e * 56 * , 3920 ( 1997 ) ; j. mag . mag . mat . * 177 - 181 * , 155 ( 1998 ) . berezinskii , soviet physics jetp * 32 * , 493 ( 1971 ) ; j.m . kosterlitz and d.j . thouless , j. phys . c * 6 * , 1181 ( 1973 ) . kosterlitz , j. phys . c * 7 * , 1046 ( 1974 ) ; d.r . nelson and j.m . kosterlitz , phys . lett . * 39 * , 1201 ( 1977 ) . h. weber and p. minnhagen , phys . b * 37 * , 5986 ( 1988 ) . p. olson , phys . b * 52 * , 4527 ( 1995 ) . d. loison , j. phys . : condens . matter , * 11 * , l401 ( 1999 ) . d. loison , phys . a ( in press ) , accessible at http://www.physik.fu-berlin.de/@xmath242loison/articles/reference13.html . amit , y.y . goldschmidt , and g. grinstein , j. phys . a * 13 * , 585 ( 1980 ) ; l.p . kadanoff and a.b . zisook , nucl . b * 180 * , 61 ( 1981 ) ; w. janke , phys . b * 55 * , 3580 ( 1997 ) . table [ table.tau ] : number of monte carlo steps to thermalize @xmath42 and to average @xmath44 as function of the lattice size . @xmath243 are calculated with shorter mc runs . we collect the data every @xmath81 . the last column gives the number of independent measurements . table [ table.revue.i ] : results for the two dimensional frustrated @xmath0 systems with a breakdown of symmetry @xmath244 . critical exponents are associated to the ising symmetry . i= ising ; ff@xmath0= fully frustrated @xmath0 ; cg= coulomb gas ; ta= triangular antiferromagnetic ; mc= monte carlo ; ht= high temperatures ; fss= finite size scaling ; micro = microcanonic tm= transfer matrix ; dyn= dynamic ; rsrg= real space renormalization group ; rsos= restricted solid on solid . the indices @xmath245 for the zig - zag or the @xmath0-ising model refer to different choice of internal free parameter . @xmath246 ; table [ table3 ] : summary of our results for the ising symmetry ( @xmath2 ) and the @xmath0 symmetry @xmath1 . the results for the @xmath6 model is from ref . ( eq)=equilibrium properties ; ( dyn)=short time dynamical properties . fig . [ fig.tau.l ] : cpu time ( proportional to the autocorrelation time @xmath43 ) for the standard metropolis algorithm ( @xmath70-square ) , in combination with one over - relaxation step ( @xmath71-diamond ) , in combination with @xmath249 over - relaxation steps ( @xmath72-circle ) , and for the wolff s cluster algorithm ( @xmath250-triangle ) . [ figtc2.all ] : crossing @xmath78 plotted vs inverse logarithm of the scale factor @xmath252 . the upper part of the figure corresponds to @xmath93 while the lower part to @xmath190 . we obtain @xmath194 and @xmath253 with a linear fit ( see text for comments ) . [ figv ] : values of @xmath102 and @xmath103 as function of @xmath59 in log log scale at @xmath123 . the value of the slopes gives @xmath254 . strong corrections for small sizes are visible . only the three largest sizes are used for the fits . the estimated statistical errors are smaller than the symbols . fig . [ figx.all ] : values of @xmath105 and @xmath106 as function of @xmath59 in log log scale at @xmath123 and @xmath197 at @xmath186 . the slopes give @xmath255 . strong corrections for the small sizes for @xmath105 are visible . only the three largest sizes are used for the fit for @xmath105 and the four largest for the fits for @xmath106 and @xmath197 . the estimated statistical errors are smaller than the symbols . fig . [ figm ] : values of @xmath157 as function of @xmath59 in log log scale at @xmath123 . the slope gives @xmath107 . strong corrections for the small sizes are visible . only the three largest sizes are used for the fits . the estimated statistical errors are smaller than the symbol . fig . [ fig.eta ] : @xmath211 and @xmath189 as function of critical temperature chosen . @xmath122 and @xmath208 are the estimate from a direct fit of @xmath33 as function of @xmath36 . this give estimates of the critical temperatures in agreement with ( [ eqn.tc.k ] ) and ( [ eqn.tc.m ] ) . black circles show the result using dynamical properties . [ fig.a2 ] : @xmath157 as function of the time @xmath134 for various system sizes at the critical temperature @xmath140 . the curves for the biggest sizes @xmath137 and @xmath138 collapse within the errors . fig . [ fig.u.x.v1 ] : binder cumulant @xmath93 and susceptibility @xmath105 as function of @xmath102 at @xmath140 . the slopes give @xmath101 and @xmath154 respectively . we use only data between @xmath148 and @xmath149 ( shown by the arrows ) for the fit . [ fig.i - xy ] : phase diagram of the ising-@xmath0 model . solid and dotted lines indicate continuous and first - order transitions respectively . the filled black circles show the possible fixed points . @xmath88 and @xmath164 are free parameter . similar phase diagram appears for the @xmath6 model@xcite , the zig - zag model@xcite and the rsos model@xcite . [ figcm.kt ] : binder s parameter @xmath190 for the @xmath1 order parameter function of the temperature for various sizes @xmath59 . the arrow shows the temperature of simulation @xmath223 . the scales is similar to those of fig . [ figcm ] fig . [ fig.heli ] : helicity @xmath40 for the @xmath1 symmetry as function of the temperature for various sizes @xmath59 . the arrows show the temperature of simulation @xmath223 , the critical temperatures for the @xmath1 symmetry @xmath186 and the ising symmetry @xmath123 , @xmath219 showing the best fit with the scaling form ( [ eqn.heli.l ] ) . the dashed line @xmath218 represents the universal ferromagnetic jump .

we introduce and investigate a class of complex semi - infinite banded toeplitz matrices satisfying the condition that the spectra of their principal submatrices accumulate onto a real interval when the size of the submatrix grows to . we prove that a banded toeplitz matrix belongs to this class if and only if its symbol has real values on a jordan curve located in . surprisingly , it turns out that , if such a jordan curve is present , the spectra of all the submatrices have to be real . the latter claim is also proven for matrices given by a more general symbol . further , the limiting eigenvalue distribution of a real banded toeplitz matrix is related to the solution of a determinate hamburger moment problem . we use this to derive a formula for the limiting measure using a parametrization of the jordan curve . we also describe a jacobi operator , whose spectral measure coincides with the limiting measure . we show that this jacobi operator is a compact perturbation of a tridiagonal toeplitz matrix . our main results are illustrated by several concrete examples ; some of them allow an explicit analytic treatment , while some are only treated numerically .

1702.00741

with a laurent series @xmath2 with complex coefficients @xmath3 , one can associate a semi - infinite toeplitz matrix @xmath4 whose elements are given by @xmath5 in the theory of toeplitz matrices , @xmath6 is referred to as the symbol of @xmath4 . according to what properties of @xmath4 are of interest , the class of symbols is usually further restricted . for the spectral analysis of @xmath4 , the special role is played by the wiener algebra which consists of symbols @xmath6 defined on the unit circle @xmath7 whose laurent series is absolutely convergent for @xmath8 . in this case , @xmath4 is a bounded operator on the banach space @xmath9 , for any @xmath10 , and its spectral properties are well - known , see @xcite . rather than actual spectrum of @xmath4 , this paper focuses on an asymptotic spectrum of @xmath4 , meaning the set of all limit points of eigenvalues of matrices @xmath11 , as @xmath12 , where @xmath11 denotes the @xmath13 principal submatrix of @xmath4 . an intimately related subject concerns the asymptotic eigenvalue distribution of toeplitz matrices for which the most complete results were obtained if the symbol of the matrix is a laurent polynomial . more precisely , one considers symbols of the form @xmath14 for which the toeplitz matrix @xmath15 is banded and is not lower or upper triangular . this subject has an impressive history going a century back to the famous work of szeg on the asymptotic behavior of the determinant of @xmath16 for @xmath12 . the most essential progress was achieved by schmidt and spitzer @xcite who proved that the eigenvalue - counting measures @xmath17 where @xmath18 are the eigenvalues of @xmath16 ( repeated according to their algebraic multiplicity ) and @xmath19 the dirac measure supported at @xmath20 , converge weakly to a _ limiting measure _ @xmath21 , as @xmath12 . recall that @xmath22 converges weakly to @xmath21 , as @xmath23 , if and only if @xmath24 for every bounded and continuous function @xmath25 on @xmath26 . the measure @xmath21 is supported on the _ limiting set _ @xmath27 the limit which appears above exists , see @xcite or ( * ? ? ? moreover , schmidt and spitzer derived another very useful description of @xmath28 which reads @xmath29 where @xmath30 are ( not necessarily distinct ) roots of the polynomial @xmath31 arranged in the order of increase of their absolute values , i.e. , @xmath32 ( if there is a chain of equalities above , then the labeling of roots within this chain is arbitrary . ) the latter description allows to deduce the fundamental analytical and topological properties of @xmath28 . first , the set @xmath28 equals the union of a finite number of pairwise disjoint open analytic arcs and a finite number of the so - called exceptional points . a point @xmath33 is called _ exceptional _ , if either @xmath34 is a branch point , i.e. , if the polynomial @xmath31 has a multiple root , or if there is no open neighborhood @xmath35 of @xmath34 such that @xmath36 is an analytic arc starting and terminating on the boundary @xmath37 . second , @xmath28 is a compact and connected set , see @xcite . let us also mention that @xmath21 is absolutely continuous with respect to the arclength measure in @xmath26 and its density has been derived by hirschmann @xcite . as a starting point , we focus on a characterization of laurent polynomial symbols @xmath38 for which @xmath39 . in this case , we say that @xmath38 _ belongs to the class _ @xmath40 ; see definition [ def : class_r ] . besides all the symbols @xmath38 for which @xmath15 is self - adjoint , the class @xmath40 contains also symbols of certain non - self - adjoint matrices . surprisingly , it turns out that , for @xmath41 , the spectra of all submatrices @xmath16 are real ; see theorem [ thm : summary ] below . describing the class @xmath40 can be viewed as a problem of characterization of the subclass of banded toeplitz matrices with asymptotically real spectrum . let us point out that , in this context , the consideration of the set of the limit spectral points , i.e. @xmath28 , is a more relevant and interesting problem than the study of the spectrum of @xmath15 realized as an operator acting on @xmath9 , for some @xmath10 . indeed , note that @xmath15 is self - adjoint if and only if @xmath42 where @xmath7 is the unit circle . on the other hand , @xmath43 coincides with the essential spectrum of @xmath15 , see ( * ? ? ? * cor . 1.10 ) . thus , if @xmath15 is non - self - adjoint , then there is no chance for @xmath44 to be purely real . from a broader perspective , results of the present paper contribute to the study of the spectral properties of non - self - adjoint operators ; a rapidly developing area which has recently attracted attention of many mathematicians and physicists , see , e.g. @xcite . particularly noteworthy problem consists in finding classes of non - self - adjoint operators that have purely real spectrum . this is usually a mathematically challenging problem ( see , for example , the proof of reality of the spectrum of the imaginary cubic oscillator in @xcite ) which , in addition , may be of physical relevance . at the moment , the non - self - adjoint operators whose spectrum is known to be real mainly comprise either very specific operators @xcite or operators which are in a certain sense close to being self - adjoint @xcite . in particular , there are almost no criteria for famous non - self - adjoint families ( such as toeplitz , jacobi , hankel , schrdinger , etc . ) guaranteeing the reality of their spectra . from this point of view and to the best of our knowledge , the present article provides the first relevant results of such a flavor for the class of banded toeplitz matrices . our main result is the following characterization of the class @xmath40 . [ thm : summary ] let @xmath38 be a complex laurent polynomial as in . the following statements are equivalent : 1 . @xmath39 ; 2 . the set @xmath45 contains ( an image of ) a jordan curve ; 3 . for all @xmath46 , @xmath47 . it is a very peculiar feature of banded toeplitz matrices that the asymptotic reality of the eigenvalues ( claim ( i ) ) forces all eigenvalues of all submatrices to be real ( claim ( iii ) ) . hence , if , for instance , the @xmath48 matrix @xmath49 has a non - real eigenvalue , there is no chance for the limiting set @xmath28 to be real . the implication ( iii)@xmath50(i ) is clearly trivial . the implication ( i)@xmath50(ii ) is proven in theorem [ thm : main_neces ] , while ( ii)@xmath50(iii ) is established in theorem [ thm : gamma_impl_real_spec ] for even more general class of symbols . these results are worked out within the first subsection of section [ sec : main ] . the second subsection of section [ sec : main ] primarily studies the density of the limiting measure @xmath21 for real laurent polynomial symbols @xmath41 in more detail . namely , we provide an explicit description of the density of @xmath21 in terms of the symbol @xmath38 and the jordan curve present in @xmath45 provided that the curve admits a polar parametrization . second , we show that the limiting measure is a solution of a particular hamburger moment problem whose moment sequence is determined by the symbol @xmath41 . the positive definiteness of the moment hankel matrix then provides a necessary ( algebraic ) condition for @xmath38 to belong to @xmath40 . finally , we prove an integral formula for the determinant of the moment hankel matrix and discuss the possible sufficiency of the condition of positive definiteness of the hankel moment matrices for @xmath38 to belong to @xmath40 . since , for @xmath41 , the support of @xmath21 is real and bounded , there is a unique self - adjoint bounded jacobi operator @xmath51 whose spectral measure coincides with the limiting measure . some properties of @xmath51 and the corresponding family of orthogonal polynomials are investigated in section [ sec : jac_op ] . mainly , we prove that @xmath51 is a compact perturbation of a jacobi operator with constant diagonal and off - diagonal sequences . also the weyl @xmath52-function of @xmath51 is expressed in terms of @xmath38 . section [ sec : ex_numer ] provides several concrete examples and numerical computations illustrating the results of sections [ sec : main ] and [ sec : jac_op ] . for general @xmath53-diagonal and 4-diagonal toeplitz matrices @xmath15 , explicit conditions in terms of @xmath38 guaranteeing that @xmath41 are deduced . further , for the @xmath53-diagonal and slightly specialized 4-diagonal toeplitz matrices , the densities of the limiting measures as well as the associated jacobi matrices are obtained fully explicitly . moreover , we show that the corresponding orthogonal polynomials are related to certain well - known families of orthogonal polynomials , which involve chebyshev polynomials of the second kind and the associated jacobi polynomials . their properties such as the three - term recurrence , the orthogonality relation , and an explicit representation are given . yet for another interesting example generalizing the previous two , we are able to obtain some partial results . these examples are presented in subsections [ subsec : ex1 ] , [ subsec : ex2 ] , and [ subsec : ex3 ] . finally , the last part of the paper contains various numerical illustrations and plots of the densities of the limiting measures and the distributions of eigenvalues in the situations whose complexity does not allow us to treat them explicitly . [ def : class_r ] laurent polynomial @xmath38 of the form is said to _ belong to the class @xmath40 _ , denoted by @xmath41 , if and only if @xmath39 , where @xmath28 is given by . since @xmath28 is a compact connected set , the inclusion @xmath39 actually implies that @xmath28 is a closed finite interval . we start with the proof of the implication ( i)@xmath50(ii ) from theorem [ thm : summary ] . for this purpose , we take a closer look at the structure of the set @xmath45 where the symbol is real - valued . let us stress that @xmath54 . recall also that a _ curve _ is a continuous mapping from a closed interval to a topological space and a _ jordan curve _ is a curve which is , in addition , simple and closed . occasionally , we will slightly abuse the term _ curve _ sometimes meaning the mapping and sometimes the image of such a mapping . clearly , @xmath55 if and only if @xmath56 and the latter condition can be turned into a polynomial equation @xmath57 where @xmath58 $ ] , @xmath59 , and @xmath60 . consequently , @xmath45 is a finite union of pairwise disjoint open analytic arcs , i.e. , images of analytic mappings from open intervals to @xmath26 , and a finite number of branching points which are the critical points of @xmath38 . it might be convenient to add the points @xmath61 and @xmath0 to @xmath45 and introduce the set @xmath62 endowed with the topology induced from the riemann sphere @xmath63 . observe that the set @xmath64 coincides with the so - called _ net of the rational function _ @xmath38 , cf . @xcite . recall that @xmath64 is a closed subset of @xmath63 . in addition , @xmath64 contains neither isolated points nor curves with end - points ( i.e. , @xmath65 such that there is no open neighborhood @xmath35 of @xmath66 such that @xmath67 is an analytic arc starting and terminating on @xmath37 ) , as one verifies by using general principles ( the mean - value property for harmonic functions and the open mapping theorem ) . in total , @xmath64 is a finite union of jordan curves in @xmath63 of which at most one is entirely located in @xmath1 . indeed , assuming the opposite of the latter claim , one can always find a region ( an open connected set ) in @xmath1 such that @xmath38 has real values on its boundary . recall that , if @xmath25 is a function analytic in a region @xmath68 , continuous up to the boundary @xmath69 , and real on the boundary , then @xmath25 is a ( real ) constant on @xmath70 , as one deduces by using the maximum modulus principle and the open mapping theorem . thus one concludes that under this assumption there exists a non - empty region on which @xmath38 is a constant function , a contradiction . note that , if the jordan curve in @xmath45 is present , it contains @xmath61 in its interior . notice also that , since @xmath71 , as @xmath72 , @xmath64 looks locally at @xmath61 as a star graph with @xmath73 edges and the angle between two consecutive edges is @xmath74 . similarly , since @xmath75 , as @xmath76 , @xmath64 looks locally at @xmath0 as a star graph with @xmath77 edges with equal angles of magnitude @xmath78 . as an illustration of the two situations which may occur for the net of @xmath38 , see figure [ fig : net_illust_yes_no ] . with the above information about the structure of the net @xmath64 , the verification of the following lemma is immediate . [ lem : path0inf_jordan ] the set @xmath45 contains a jordan curve if and only if every path in @xmath63 connecting @xmath61 and @xmath0 has a non - empty intersection with @xmath45 . now , we are ready to show that the existence of a jordan curve in @xmath45 is necessary for @xmath79 , i.e. , the implication ( i)@xmath50(ii ) from theorem [ thm : summary ] . [ thm : main_neces ] if @xmath79 , then @xmath45 contains a jordan curve . set @xmath80 note that @xmath38 maps a neighborhood of @xmath61 and a neighborhood of @xmath0 onto a neighborhood of @xmath0 and that for @xmath34 large in the modulus , the preimages @xmath81 are close to @xmath61 , while @xmath82 are close to @xmath0 . hence there exist neighborhoods of @xmath61 and @xmath0 with empty intersections with @xmath83 . next , let us show that any path in @xmath63 connecting @xmath61 and @xmath0 has a non - empty intersection with @xmath83 . let @xmath84\to s^{2}$ ] denote such a path . first , note that , for any @xmath85 , the function @xmath86 is continuous on @xmath1 and that the value @xmath87 appears at least once in the @xmath88-tuple @xmath89 for all @xmath90 . further , since @xmath91 , the value @xmath87 , for @xmath92 in a right neighborhood of @xmath61 , appears among the first @xmath93 values @xmath94 and does not appear in the remaining @xmath95 values @xmath96 . similarly , since @xmath97 , the value @xmath87 , for @xmath92 in a left neighborhood of @xmath98 , appears in @xmath96 and is not among @xmath94 . since @xmath99 is a continuous map there must exist @xmath100 such that @xmath101 . hence @xmath102 . finally , recalling , one has @xmath103 , hence @xmath104 . consequently , we obtain that any path in @xmath63 joining @xmath61 and @xmath0 has a non - empty intersection with @xmath45 and theorem [ thm : main_neces ] follows from lemma [ lem : path0inf_jordan ] . the second part of this subsection is devoted to the proof of the implication ( ii)@xmath50(iii ) of theorem [ thm : summary ] . recall that every jordan curve in @xmath26 is homeomorphic to the unit circle @xmath105 and therefore it gives rise to a homeomorphic mapping @xmath106 . in addition , the unit circle is naturally parametrized by the polar angle @xmath107 , where @xmath108 $ ] . in the following , we always choose the parametrization of a jordan curve @xmath109 , with @xmath108 $ ] , which can be viewed as the composition of the above two mappings . note that the mapping @xmath110 reflects the points in @xmath1 with respect to @xmath7 . for a given jordan curve @xmath99 with @xmath61 in its interior , we define a new jordan curve @xmath111 by reflecting @xmath99 with respect to @xmath7 , i.e. , @xmath112 . in other words , if @xmath99 is parametrized as @xmath113 with a positive function @xmath114 and real function @xmath115 , then @xmath116 we also introduce the notation @xmath117}|\gamma(t)| \ ; \mbox { and } \ ; r_{\gamma}:=\max_{t\in[-\pi,\pi]}|\gamma(t)|.\ ] ] in the following lemma , @xmath118 stands for the standard inner product on @xmath119 ; further , to any vector @xmath120 , we associate the polynomial @xmath121 . [ lem : quadr_form_gamma ] let @xmath99 be a jordan curve with @xmath61 in its interior and @xmath6 be a function given by the laurent series @xmath122 which is absolutely convergent in the annulus @xmath123 . then , for all @xmath124 and @xmath46 , one has @xmath125 the proof proceeds by a straightforward computation . we have @xmath126 where we have used that @xmath127 for any @xmath128 . [ cor : quadr_form_unit_circ ] for any element @xmath6 of the wiener algebra , @xmath124 , and @xmath46 , one has @xmath129 choose @xmath130 , @xmath108 $ ] , in lemma [ lem : quadr_form_gamma ] . [ thm : gamma_impl_real_spec ] let @xmath99 be a jordan curve with @xmath61 in its interior and @xmath6 be a function given by the laurent series @xmath122 which is absolutely convergent in the annulus @xmath131 . suppose further that @xmath132 , for all @xmath108 $ ] . then @xmath133 in particular , if the laurent series of the symbol @xmath6 from theorem [ thm : gamma_impl_real_spec ] converges absolutely for all @xmath134 , then @xmath135 , for all @xmath46 , provided that @xmath136 contains a jordan curve ( which has to contain @xmath61 in its interior ) . this applies for laurent polynomial symbols especially , which yields the implication ( ii)@xmath50(iii ) of theorem [ thm : summary ] . let @xmath137 and @xmath138 be the normalized eigenvector corresponding to @xmath21 . then , by using corollary [ cor : quadr_form_unit_circ ] , one obtains @xmath139 on the other hand , by applying lemma [ lem : quadr_form_gamma ] twice , one gets @xmath140 where the last equality holds since @xmath132 , for all @xmath108 $ ] , by the assumption . finally , by using together with and applying lemma [ lem : quadr_form_gamma ] again , one arrives at the equality @xmath141 hence @xmath142 . [ rem : hessenberg_real_coef ] note that the entries of the toeplitz matrix considered in theorem [ thm : gamma_impl_real_spec ] are allowed to be complex . clearly , there exists toeplitz matrices satisfying assumptions of theorem [ thm : gamma_impl_real_spec ] with non - real entries , for instance , any self - adjoint toeplitz matrix with non - real entries whose symbol belongs to the wiener algebra . on the other hand , if a toeplitz matrix is at the same time an upper or lower hessenberg matrix , then its entries have to be real . more precisely , if the symbol has the form @xmath143 and the assumptions of theorem [ thm : gamma_impl_real_spec ] are fulfilled , then @xmath144 for all @xmath145 . indeed , it is easy to verify that @xmath146 satisfies the recurrence @xmath147 since , by theorem [ thm : gamma_impl_real_spec ] , all eigenvalues of @xmath11 are real , @xmath148 for any @xmath46 . hence one can use that @xmath149 and the above recurrence to prove by induction that @xmath144 for all @xmath145 . for the analysis of this subsection , we restrict ourself with real banded toeplitz matrices @xmath15 . more precisely , we consider symbols @xmath150 our first goal is to provide a more detailed description of the limiting measure @xmath21 for the symbols in terms of the jordan curve @xmath99 present in @xmath45 . for the sake of simplicity , we focus on the situation when the jordan curve admits the polar parametrization : @xmath151 , \label{eq : gamma_polar_par}\ ] ] where @xmath152 for all @xmath108 $ ] . we should mention that so far we did not observe any example of @xmath41 where the jordan curve in @xmath45 would intersect a radial ray in more than one point ; so the parametrization would be impossible for such a curve . in particular , all the examples presented in section [ sec : ex_numer ] where @xmath99 is known explicitly satisfy . incidentally , the problem of description of the limiting measure for @xmath41 is closely related to the classical hamburger moment problem . standard references on this subject are @xcite . solution _ of the hamburger moment problem with a moment sequence @xmath153 , @xmath154 , is a probability measure @xmath21 supported in @xmath155 with all moments finite and equal to @xmath156 , i.e. , @xmath157 by the well - known hamburger theorem , the hamburger moment problem with a moment sequence @xmath158 , @xmath154 , has a solution if and only if the hankel matrix @xmath159 is positive definite for all @xmath46 . the solution may be unique ( the determinate case ) or they can be infinitely many ( the indeterminate case ) . however , if there is a solution whose support is compact , then it is unique , see , for example , ( * ? ? ? 1.5 ) . for a later purpose , let us also recall the formula @xmath160 which is usually attributed to g. szeg . in fact , szeg proved for any symbol @xmath38 belonging to the wiener class , imposing however , an additional assumption which is equivalent to the self - adjointness of @xmath15 . later on , m. kac showed that remains valid for any @xmath38 from the wiener class without the additional assumption , see @xcite . [ lem : b_in_r_sol_hmp ] let @xmath41 have the form . then the hamburger moment problem with the moment sequence given by @xmath161 has a solution which is unique and coincides with the limiting measure @xmath21 . first , note that @xmath162 , for all @xmath163 , since the coefficients of @xmath38 are assumed to be real . moreover , since @xmath41 , @xmath164 and it is a compact set . hence , by the above discussion , it suffices to verify that @xmath21 satisfies . to this end , recall that @xmath21 is the weak limit of the measures @xmath22 given by . then , by using the limit formula , one gets @xmath165 for all @xmath163 . [ cor : b_in_r_impl_pd_hankel ] if @xmath41 is of the form , then the hankel matrix with entries given by is positive definite for all @xmath46 . note that , if the coefficients of @xmath41 are real , then @xmath166 and @xmath45 is symmetric with respect to the real line . the jordan curve @xmath99 present in @xmath45 intersects the real line at exactly two points ( negative and positive ) which are the critical points of @xmath38 . [ thm : main_suff_generalized ] suppose that @xmath41 is as in and the jordan curve @xmath99 contained in @xmath45 admits the polar parametrization . further , let @xmath167 be the number of critical points of @xmath38 in @xmath168 and @xmath169 are such that @xmath170 for all @xmath171 . then @xmath172 restricted to @xmath173 is strictly monotone for all @xmath174 , and the limiting measure @xmath175 , where @xmath176 is an absolutely continuous measure supported on @xmath177:=b(\gamma([\phi_{i-1},\phi_{i}]))$ ] whose density is given by @xmath178 for all @xmath179 and all @xmath174 . in the @xmath180 sign is used when @xmath172 increases on @xmath181 , and the @xmath182 sign is used otherwise . notice that @xmath183.$ ] further , observe that , in particular , theorem [ thm : main_suff_generalized ] provides a description of the limiting measure for all self - adjoint real banded toeplitz matrices since , in this case , the jordan curve is just the unit circle . an illustration of this situation is given in example 5 in subsection [ subsec : var_numer ] . first , note that , for any @xmath184 , the first derivative of the smooth function @xmath185 does not change sign on @xmath173 because @xmath186 if and only if @xmath187 for some @xmath171 . hence @xmath172 restricted to @xmath173 is either strictly increasing or strictly decreasing and the same holds for the the inverse @xmath188 . consequently , by formula , a positive measure @xmath176 supported on @xmath177 $ ] is well - defined for all @xmath184 . let us denote @xmath189 . in the second part of the proof , we verify that the measure @xmath190 which is supported on @xmath191 $ ] has the same moments as the limiting measure @xmath21 . then @xmath192 by lemma [ lem : b_in_r_sol_hmp ] . let @xmath163 be fixed . by deforming the positively oriented unit circle into the jordan curve @xmath99 , one gets the equality @xmath193 next , by using the symmetry @xmath194 , for @xmath134 , parametrization , and the fact that @xmath172 is real - valued , one arrives at the expression @xmath195 by splitting the above integral according to the positions of the critical points of @xmath38 on the arc of @xmath99 in the upper half - plane , one further obtains @xmath196 finally , since @xmath172 restricted to @xmath173 is monotone , we can change the variable @xmath197 in each of the above integrals getting @xmath198 which concludes the proof . generically , there is no critical point of @xmath38 located on the curve @xmath99 in the upper half - plane , i.e. , @xmath199 in theorem [ thm : main_suff_generalized ] . hence the only critical points located on the curve @xmath99 are the two intersection points of @xmath99 and the real line . in this case , the statement of theorem [ thm : main_suff_generalized ] gets a simpler form . although it is just a particular case of theorem [ thm : main_suff_generalized ] , we formulate this simpler statement separately below . [ thm : main_suff ] suppose that @xmath41 is as in and the jordan curve @xmath99 contained in @xmath45 admits the polar parametrization . moreover , let @xmath200 for all @xmath201 . then @xmath172 restricted to @xmath202 is either strictly increasing or decreasing ; the limiting measure @xmath21 is supported on the interval @xmath203:=b(\gamma([0,\pi]))$ ] and its density satisfies @xmath204 for @xmath205 , where the @xmath180 sign is used when @xmath172 increases on @xmath202 , and the @xmath182 sign is used otherwise . since @xmath172 is a one - to - one mapping from @xmath206 $ ] onto @xmath203 $ ] under the assumptions of theorem [ thm : main_suff ] , we can use it for a reparametrization of the arc of @xmath99 in the upper half - plane . namely , we denote by @xmath207\to\{z\in{{\mathbb{c}}}\mid { \mathop\mathrm{im}\nolimits}z\geq0\}$ ] the new parametrization of the arc of @xmath99 to distinguish the notation . recalling , one observes that @xmath208.\ ] ] then the limiting measure is determined by the change of the argument of the jordan curve contained in @xmath45 provided that the parametrization @xmath209 is used . more concretely , if we denote the distribution function of @xmath21 by @xmath210 , then formula can be rewritten as @xmath211 yet another equivalent formulation reads @xmath212 for @xmath213 $ ] . let us once more go back to the hamburger moment problem with the moment sequence given by . we derive an integral formula for @xmath214 . let @xmath38 be as in and @xmath215 be the hankel matrix with the entries given by . then , for all @xmath46 , one has @xmath216 consequently , if @xmath38 has the form and @xmath41 , then @xmath217 let @xmath46 be fixed . first , by using the definition of the determinant , we get @xmath218 where @xmath219 , @xmath220 is the diagonal matrix with the entries @xmath221 and @xmath222 is the vandermonde matrix with the entries @xmath223 by applying the well - known formula for the determinant of the vandermonde matrix , one arrives at the equation @xmath224\ ! \left[\prod_{1\leq i < j\leq n}\left(b\left(e^{{{\rm i}}t_{j}}\right)-b\left(e^{{{\rm i}}t_{i}}\right)\right)\right]\ ! { { { \rm d}}}t_{1}{{{\rm d}}}t_{2}\dots{{{\rm d}}}t_{n}. \label{eq : det_h_n_first_in_proof}\ ] ] second , we apply a symmetrization trick to the identity . let @xmath225 be a permutation of the set @xmath226 . note that the second term in the square brackets in , i.e. , the vandermonde determinant , is antisymmetric as a function of variables @xmath227 . thus , if we change variables in so that @xmath228 , we obtain @xmath229\ ! \left[\prod_{1\leq i < j\leq n}\left(b\left(e^{{{\rm i}}s_{j}}\right)-b\left(e^{{{\rm i}}s_{i}}\right)\right)\right]\ ! { { { \rm d}}}s_{1}\dots{{{\rm d}}}s_{n}.\ ] ] now , it suffices to divide both sides of the above equality by @xmath230 , sum up over all permutations @xmath225 , and recognize one more vandermonde determinant on the right - hand side . this yields . the second statement of the theorem follows immediately from corollary [ cor : b_in_r_impl_pd_hankel ] and the already proven formula . note that @xmath231\!,\ ] ] i.e. , @xmath232 is a laurent polynomial in the indeterminates @xmath233 . the condition tells us that the constant term of @xmath232 has to be positive for all @xmath234 . consequently , the inequalities yield a necessary condition for the symbol @xmath38 of the form to belong to the class @xmath40 . in principle , these conditions can be formulated as an infinite number of inequalities in terms of the coefficients of @xmath38 ( though in a very complicated form ) . we finish this subsection with a discussion on a possible converse of the implication from corollary [ cor : b_in_r_impl_pd_hankel ] . it is not clear now , whether , for the symbol @xmath38 of the form , the positive definiteness of all the hankel matrices @xmath215 is a sufficient condition for @xmath38 to belong to @xmath40 . if @xmath235 for all @xmath46 , then the hamburger moment problem with the moment sequence has a solution , say @xmath190 , which is unique . indeed , the uniqueness follows from the fact that the moment sequence @xmath156 does not grow too rapidly as @xmath236 , see ( * ? ? ? 1.5 ) . to see this , one has to realize that the spectral radius of @xmath16 is majorized by the ( spectral ) norm of @xmath16 . this norm can be estimated from above as @xmath237 now , taking into account , one observes that the moment sequence @xmath156 grows at most geometrically since @xmath238 where @xmath239 stands for the eigenvalues of @xmath16 counted repeatedly according to their algebraic multiplicity . similarly as in the proof of lemma [ lem : b_in_r_sol_hmp ] , one verifies that the @xmath52-th moment of the limiting measure @xmath21 equals @xmath156 . hence , assuming @xmath235 for all @xmath46 , both measures @xmath21 and @xmath190 have the same moments . this implies that the cauchy transforms @xmath240 and @xmath241 of measures @xmath21 and @xmath190 , respectively , coincide on a neighborhood of @xmath0 since @xmath242 it can be shown that the equality @xmath243 hold for all @xmath244 , @xmath245 , with @xmath246 as in . this , however , does not imply @xmath192 . the measures @xmath21 and @xmath190 would coincide , if @xmath243 for almost every @xmath244 ( with respect to the lebesgue measure in @xmath26 ) . this can be obtained by imposing an additional assumption on @xmath28 requiring its complement to be connected . indeed , since the cauchy transform of a measure is analytic outside the support of the measure , the equality @xmath243 can be extended to all @xmath247 by analyticity provided that @xmath248 is connected . clearly , both sets @xmath28 and @xmath249 have zero lebesgue measure and therefore @xmath192 . it follows from the above discussion that , if there exists @xmath38 of the form such that @xmath235 , for all @xmath46 , and @xmath250 , then the set @xmath248 has to be disconnected . let us stress that it is by no means clear for which symbols @xmath38 the set @xmath248 is connected . to our best knowledge , the only exception corresponds to symbols @xmath38 which are trinomials for which the set @xmath248 is known to be connected as pointed out in @xcite . on the other hand , the relatively simple examples of @xmath38 where @xmath28 separates the plane @xmath26 are known , see ( * ? ? ? one might believe that , if @xmath16 is a lower ( or upper ) hessenberg matrix , i.e , @xmath251 ( or @xmath252 ) in , then @xmath248 is connected . it seems that it is not the case neither . it is not our intention to prove it analytically ; we provide only a numerical evidence given by figure [ fig : separ ] . in lemma [ lem : b_in_r_sol_hmp ] , we have observed that , for @xmath41 of the form , the limiting measure is a solution of the hamburger moment problem . it is well known that the hamburger moment problem is closely related with jacobi operators and orthogonal polynomials @xcite . the aim of this section is to investigate spectral properties of the jacobi operator whose spectral measure coincides with the limiting measure @xmath21 provided that @xmath41 . for the general theory of jacobi operators , we refer the reader to @xcite . recall that @xmath253 is the hilbert space of square summable sequences indexed by @xmath254 endowed with the standard inner product denoted here by @xmath118 , and @xmath255 stands for the canonical basis of @xmath253 . [ thm : jac_op_rel_b_in_r ] for every @xmath41 of the form , there exists a bounded self - adjoint jacobi operator @xmath51 acting on @xmath253 such that its spectral measure @xmath256 satisfies @xmath257 where @xmath21 is the limiting measure of @xmath15 ; in particular , @xmath258 . moreover , for @xmath259 , the weyl @xmath52-function @xmath260 of @xmath51 satisfies the equation @xmath261 where @xmath262 is an arbitrary number such that @xmath263 . by the general theory , for every determinate hamburger moment problem with a moment sequence @xmath264 , @xmath154 , there exists a self - adjoint jacobi operator @xmath265 determined uniquely by its diagonal and off - diagonal sequence : @xmath266 and @xmath267 , @xmath46 , see ( * ? ? ? * chp . 4 ) . the sequences @xmath268 , @xmath269 are expressible in terms of the moment sequence @xmath264 by the formulas ( * ? ? ? * eq . ( 2.118 ) ) @xmath270 and @xmath271 where @xmath272 is obtained from @xmath273 by deleting its @xmath274th row and its @xmath275st column . the projection - valued spectral measure @xmath276 of the self - adjoint operator @xmath265 determines the measure @xmath277 which coincides with the solution of the determinate hamburger moment problem with the moment sequence @xmath264 . moreover , @xmath278 and therefore @xmath279 is bounded if and only if @xmath265 is a bounded operator . these facts together with lemma [ lem : b_in_r_sol_hmp ] and the equality @xmath280 yield all the claims of the statement except for the formula . let us now derive the expression for the weyl @xmath52-function of @xmath51 . by expressing the resolvent operator of @xmath51 in terms of its spectral measure , one gets @xmath281 in other words , the weyl @xmath52-function of @xmath51 is nothing else but the cauchy transform of the limiting measure @xmath21 . let @xmath282 be fixed . take an arbitrary @xmath283 such that @xmath284 . in the course of the proof of ( * ? ? ? * lem . 11.11 ) , it was shown that @xmath285 where @xmath52 is an integer . differentiating the above equation with respect to @xmath34 , one obtains @xmath286 on the other hand , it is also well - known that @xmath287 see , e.g. , ( * ? ? ? 4.2 ) . by using the above formula together with the identity @xmath288 which , in its turn , is obtained by the logarithmic differentiation of the equation @xmath289 expressing the relation between the roots and the leading and constant coefficients of the polynomial @xmath290 , we obtain @xmath291 combining and , one arrives at the equality @xmath292 which , together with , yields . recall that the density of @xmath21 can be recovered from @xmath293 by using the stieltjes perron inversion formula , see , for example , ( * ? ? ? * chp . 2 and append . namely , one has @xmath294 for all @xmath295 which are not exceptional points . let us now apply the close relation between the spectral properties of jacobi operators and the orthogonal polynomials . theorem [ thm : jac_op_rel_b_in_r ] tells us that for any @xmath41 of the form , there exists a family of orthogonal polynomials @xmath296 , determined by the three - term recurrence @xmath297 with the initial conditions @xmath298 and @xmath299 , where coefficients @xmath300 and @xmath301 are given by and ( @xmath302 is arbitrary ) . this family satisfies the orthogonality relation @xmath303 where @xmath114 stands for the density of the limiting measure @xmath21 supported on @xmath304 $ ] . for @xmath305 , one has to set @xmath306 in . particular examples of these families of polynomials are examined in section [ sec : ex_numer ] . the map sending @xmath41 to the corresponding jacobi parameters @xmath300 and @xmath301 is interesting ; however , it is unlikely that in a concrete situation , one can use general formulas and to obtain the diagonal and off - diagonal sequences of @xmath51 explicitly . consequently , it is desirable at least to know the asymptotic behavior of @xmath307 and @xmath308 , as @xmath12 . [ thm : compact ] let @xmath41 of the form and @xmath304 $ ] . then @xmath51 is a compact perturbation of the jacobi matrix with the constant diagonal sequence @xmath309 and the constant off - diagonal sequence @xmath310 , i.e. , @xmath311 the statement is a consequence of rakhmanov s theorem ( * ? ? ? 4 ) ; see also thm . 1 in loc . cit . referring to a weaker older result due to p. nevai @xcite which is still sufficient for our purposes . the latter theorem implies that , if 1 . @xmath265 is a bounded self - adjoint jacobi operator such that @xmath312 $ ] , 2 . @xmath313 for almost every @xmath314 $ ] , where @xmath114 denotes the density of the lebesgue absolutely continuous component of the measure @xmath315 @xmath276 being the spectral measure of @xmath265 , then @xmath265 is a compact perturbation of the jacobi operator with constant diagonal sequence @xmath316 and constant off - diagonal sequence @xmath317 . in case of @xmath51 , theorem [ thm : jac_op_rel_b_in_r ] implies that @xmath318 $ ] . in fact , since @xmath319 does not contain isolated points , the discrete part of the spectrum of the self - adjoint operator @xmath51 is empty and hence @xmath320 $ ] . taking into account , it suffices to show that the density of the limiting measure @xmath21 is positive almost everywhere on @xmath321 . the latter fact is true since the density is positive on every analytic arc of @xmath28 , i.e. , everywhere except possibly at the exceptional points which are finitely many , see ( * ? ? ? 4c ) . using the terminology of orthogonal polynomials , the statement of theorem [ thm : compact ] says that the family of orthogonal polynomials @xmath296 determined by belongs to the blumenthal nevai class @xmath322 , see @xcite . first , we take a look at the simplest nontrivial situation when @xmath15 is a tridiagonal toeplitz matrix . since the entry on the main diagonal only causes a shift of the spectral parameter and the matrix @xmath15 itself can be scaled by a nonzero constant we may assume , without loss of generality , that @xmath15 belongs to the one parameter family of tridiagonal toeplitz matrices with the symbol @xmath323 where @xmath324 . first , we decide for what parameter @xmath6 , the symbol @xmath41 . by theorem [ thm : summary ] , @xmath325 is a necessary condition from which one easily deduces that @xmath326 , if @xmath41 . next , for @xmath326 , the symbol @xmath38 is real - valued on the circle @xmath327 , @xmath108 $ ] . thus , by using theorem [ thm : summary ] once more , we conclude that @xmath41 if and only if @xmath326 . one way to deduce the limiting measure @xmath21 , for @xmath326 , is to solve the corresponding hamburger moment problem with the moment sequence . a straightforward use of the binomial formula provides us with the constant term of the laurent polynomial @xmath328 , for @xmath329 , giving the identity @xmath330 recall that the wallis integral formula reads @xmath331 changing the variable in the above integral by @xmath332 , one obtains @xmath333 or , equivalently , @xmath334 where @xmath335 hence , taking also into account that @xmath114 is an even function , we have @xmath336 with @xmath156 given by . consequently , the measure @xmath21 supported on @xmath337 $ ] with the above density @xmath114 is the unique solution of the hamburger moment problem with the moment sequence . by lemma [ lem : b_in_r_sol_hmp ] , this measure is the desired limiting measure @xmath21 . for the distribution function of @xmath21 , one obtains @xmath338.\ ] ] this measure is well - known as the arcsine measure supported on the interval @xmath337 $ ] . in particular , it is the so - called equilibrium measure of @xmath337 $ ] , see e.g. ( * ? ? ? an alternative way to deduce @xmath21 is based on formula . to obtain a suitable expression for the weyl @xmath52-function , one can use the generating function for the moments which reads as @xmath339 this identity together with the von neumann series expansion of the resolvent operator yields @xmath340 for @xmath341 $ ] . by substituting the latter expression into and evaluating the limit , one rediscovers . further , let us examine the structure of the jacobi matrix @xmath51 and the corresponding family of orthogonal polynomials . first , since the density is an even function on the interval symmetric with respect to @xmath61 , the diagonal sequence @xmath301 vanishes , as it follows , for example , from ( * ? ? ? * thm . 4.2 ( c ) ) . to compute the off - diagonal sequence @xmath300 , we need to evaluate @xmath214 where @xmath215 is the hankel matrix with the entries given by moments . the evaluation of @xmath214 is treated in lemma [ lem : det_h_n_tridiag ] in the appendix . by using lemma [ lem : det_h_n_tridiag ] together with , one immediately gets @xmath342 and @xmath343 for @xmath344 . thus , the self - adjoint jacobi operator associated with @xmath38 has the matrix representation @xmath345 where @xmath326 . the corresponding family of orthogonal polynomials @xmath296 generated by the recurrence ( and initial conditions given therein ) satisfies the orthogonality relation @xmath346 which one verifies by using together with the formula and lemma [ lem : det_h_n_tridiag ] . polynomials @xmath296 do not belong to any family listed in the hypergeometric askey scheme @xcite . however , they can be written as the following linear combination of chebyshev polynomials of the second kind , @xmath347 for @xmath46 ( here @xmath348 ) . the above equation and the hypergeometric representation of chebyshev polynomials , see ( * ? ? ? * eq . ( 9.8.36 ) ) , can be used to obtain the explicit formula @xmath349 where @xmath350 denotes the greatest integer less or equal to a real number @xmath351 . let us examine the case of symbols with @xmath251 and @xmath352 . without loss of generality , we can set @xmath353 and @xmath354 which yields the symbol @xmath38 of the form @xmath355 where @xmath356 and @xmath357 . first , we discuss for which parameters @xmath358 and @xmath359 we have @xmath41 . by remark [ rem : hessenberg_real_coef ] , if @xmath41 , @xmath358 and @xmath359 have to be real . further , according to theorem [ thm : summary ] , if @xmath41 , then @xmath360 . the characteristic polynomial of @xmath361 reads @xmath362 by inspection of the discriminant of the above cubic polynomial with real coefficients , one concludes that its roots are real if and only if @xmath363 . next , we show that , for @xmath364 and @xmath363 , @xmath45 contains a jordan curve . note that , if @xmath363 , the equation @xmath365 have all roots non - zero real and they can not degenerate to a triple root . consequently , all critical points of @xmath38 are real . at the same time , these points are the intersection points of the net @xmath64 . taking into account that only one curve in @xmath64 ( the real line ) passes through @xmath61 , because @xmath366 , as @xmath72 , and two curves ( the extended real line and one more ) pass through @xmath0 , because @xmath367 , as @xmath76 , one concludes that there has to be another arc in @xmath45 passing through a real critical point of @xmath38 . this arc necessarily closes into a jordan curve located in @xmath45 . altogether , theorem [ thm : summary ] implies that @xmath41 if and only if @xmath364 and @xmath363 . next , we will derive the limiting measure @xmath21 explicitly in a special case when the symbol takes the form @xmath368 where @xmath369 . this corresponds to @xmath370 , @xmath371 in , but we additionally add a real constant to @xmath38 . clearly , if we add a real constant to @xmath38 , the net @xmath64 does not change and @xmath28 is just shifted by the constant . hence , we may use the previous discussion to conclude that @xmath38 , given by , belongs to the class @xmath40 for all @xmath369 . note that the critical points of @xmath38 are @xmath372 and @xmath373 , hence the measure @xmath21 is supported on the closed interval with the endpoints @xmath374 and @xmath375 . by using the binomial formula , one verifies that the constant term of @xmath328 equals @xmath376 the generating function for the moments @xmath156 can be expressed in terms of the gauss hypergeometric series as @xmath377 for the sake of simplicity of the forthcoming formulas and without loss of generality , we set @xmath378 . we can apply the identity @xmath379 valid for @xmath380 and @xmath381 , see ( * ? ? ? * eq . 7.3.3.4 , p. 486 ) , to the right - hand side of with @xmath382 and deduce the formula for the weyl @xmath52-function of the jacobi operator @xmath51 which reads @xmath383 for @xmath384 $ ] . by evaluating the limit in , one obtains @xmath385 this density appeared earlier in connection with faber polynomials @xcite ; see also @xcite . next , we examine the operator @xmath51 and the corresponding family of orthogonal polynomials in detail . first , we derive formulas for the diagonal sequence @xmath301 and off - diagonal sequence @xmath300 of @xmath51 . in @xcite , the hankel determinant @xmath386 has been evaluated with @xmath387 , see ( * ? ? ? * eq . ( 1 ) ) . the slightly more general identity with the additional parameter @xmath6 can be justified by using the same argument as in the first paragraph of the proof of lemma [ lem : det_h_n_tridiag ] . moreover , the first equation from ( * ? ? ? * eq . ( 25 ) ) yields @xmath388 ( in the notation used in ( * ? ? ? * eq . ( 25 ) ) , @xmath389 coincides with @xmath390 and @xmath391 coincides with @xmath392 ) . by substituting the identities and in the general formulas and , one obtains @xmath393 and @xmath394 the corresponding family of monic orthogonal polynomials generated by the recurrence is orthogonal with respect to the density for @xmath378 . taking into account and the identity , one gets the orthogonality relation @xmath395 the polynomial sequence @xmath296 does not coincide with any family listed in the askey scheme @xcite either . nevertheless , @xmath396 can be expressed as a linear combination of the associated jacobi polynomials introduced and studied by j. wimp in @xcite . following notation from @xcite , the associated jacobi polynomials @xmath397 constitute a three - parameter family of orthogonal polynomials generated by the same recurrence as the jacobi polynomials @xmath398 , but every occurrence of @xmath274 in the coefficients of the recurrence relation defining the jacobi polynomials is replaced by @xmath399 , see ( * ? ? ? * eq . ( 12 ) ) . set @xmath400 and @xmath401 , where @xmath402 is the pochhammer symbol . then , putting again @xmath378 , one has @xmath403 where @xmath404 , @xmath405 , and @xmath406 . verification of is done in a completely routine way by showing that both sides satisfy the same recurrence relation with the same initial conditions . finally , since wimp derived the explicit formula for associated jacobi polynomials in ( * ? ? ? * theorem 1 ) , one can make use of this result together with to compute the explicit expression for @xmath407 . however , the computation is somewhat lengthy and the resulting formula is rather cumbersome ; we omit the details and state only the final result for the record . for @xmath145 , one has @xmath408 where @xmath409 and @xmath410 both cases treated in the previous subsections , where the limiting measure was derived fully explicitly , can be thought of as special cases of the more general symbol @xmath411 ( up to a shift by a constant term ) , where @xmath412 and @xmath369 . by using the binomial formula , one computes @xmath413 first , we prove that @xmath41 for all @xmath414 and @xmath369 by showing that @xmath45 contains a jordan curve . without loss of generality , we may again put @xmath387 ( otherwise one can take the @xmath415 multiple of the curve @xmath99 given below ) . define @xmath416 , \label{eq : def_gamma_spec_exam}\ ] ] where @xmath417 . the value @xmath418 is determined by the corresponding limit . figure [ fig : jordan_spec_ex ] shows the jordan @xmath99 for three special choices of @xmath93 and @xmath95 . expressing the sine function in terms of complex exponentials , one easily verifies that @xmath419 , \label{eq : b_comp_gamma_spec_exam}\ ] ] where the value at @xmath420 is the respective limit . thus , @xmath38 restricted to the image of @xmath99 is real - valued and theorem [ thm : summary ] implies @xmath41 . the parametrization has the polar form . moreover , @xmath38 has no critical point on the arc of @xmath99 in the upper half - plane since @xmath421 as one readily verifies . taking into account that @xmath422 we see that @xmath172 is strictly decreasing on @xmath202 . consequently , by theorem [ thm : main_suff ] , the limiting mesure is supported on the interval @xmath423 \label{eq : supp_mu_spec_exam}\ ] ] and , according to , the distribution function of @xmath21 satisfies @xmath424 it seems that , for general parameters @xmath93 and @xmath95 , the density of @xmath21 can not be expressed explicitly because the function in can not be inverted ; see figure [ fig : density_spec_ex ] for numerical plots of the density of @xmath21 . let us point out that , besides the cases @xmath251 and either @xmath252 or @xmath352 , one can also derive explicitly the density for the self - adjoint 5-diagonal case , i.e. @xmath425 ( @xmath387 ) . the resulting formula reads @xmath426 since the weyl @xmath52-function can be expressed as @xmath427.\ ] ] the above formula follows from the generating function of @xmath264 which , for @xmath428 , reads @xmath429 where we make use of the identity with @xmath430 . derivation of a closed formula for the sequences @xmath300 and @xmath301 determining the jacobi operator @xmath51 for general values of @xmath93 and @xmath95 , seems to be out of our reach . even a closed expression for the determinant of the hankel matrix @xmath215 determined by the sequences @xmath431 corresponding to the special cases with @xmath387 , @xmath251 and @xmath432 or @xmath433 and @xmath352 , respectively , is unknown , to the best of our knowledge . in fact , numerical experiments with the sequence of these determinants show a presence of huge prime factors which might indicate that no closed formula , similar to , exists for the hankel determinants . nevertheless , theorem [ thm : compact ] and equation yield the limit formulas : @xmath434 for all @xmath414 . most of the results of this paper are devoted to banded toeplitz matrices but theorem [ thm : gamma_impl_real_spec ] is applicable to matrices with more general symbols . a combination of the previous example with the symbol and theorem [ thm : gamma_impl_real_spec ] provides us with more examples of possibly non - self - adjoint toeplitz matrices given by a more general symbol . for instance , one can proceed as follows . if @xmath25 is a function analytic on @xmath1 which maps @xmath435 to @xmath155 and @xmath38 is as in , then the symbol @xmath436 satisfies the assumptions of theorem [ thm : gamma_impl_real_spec ] with the jordan curve given by . consequently , @xmath437 for all @xmath46 . to be even more concrete , the possible choices of @xmath25 comprise , e.g. , @xmath438 , producing many examples of non - self - adjoint and non - banded toeplitz matrices whose principal submatrices have purely real spectrum . to illustrate a computational applicability of theorems [ thm : main_suff_generalized ] and [ thm : main_suff ] , we add two more complicated examples treated numerically using wolfram mathematica . in particular , to emphasize the connection between the existence of a jordan curve in @xmath45 and the reality of the limiting set of eigenvalues of the corresponding toeplitz matrices , we provide some numerical plots on this account . we plot the numerically obtained density by applying theorem [ thm : main_suff ] to the symbol @xmath439 whose net is shown in figure [ fig : net_illust_yes_no ] ( left ) . although we can only check the validity of the assumptions of theorem [ thm : main_suff ] numerically , looking at figure [ fig : net_illust_yes_no ] it is reasonable to believe the jordan curve present in @xmath45 can be parametrized by polar coordinates . further , a numerical computation indicates that no critical point of @xmath38 lies on the jordan curve in the upper half - plane . the real critical points of @xmath38 closest to the origin are @xmath440 and @xmath441 and the corresponding critical values are @xmath442 and @xmath443 . so the limiting measure @xmath21 should be supported approximately on the interval @xmath444 $ ] which is in agreement with the numerical results obtained by an implementation of the algorithm for the computation of @xmath28 suggested in @xcite . the density of @xmath21 is plotted in figure [ fig : density_ex4 ] . for an illustration of theorem [ thm : main_suff_generalized ] which allows to have non - real critical points of @xmath38 on the arc of jordan curve in @xmath45 , we consider the symbol @xmath445 the corresponding toeplitz matrix @xmath15 is self - adjoint and hence the jordan curve in @xmath45 is the unit circle . the function @xmath38 has two critical points with positive imaginary part on the unit circle , namely @xmath446 and the corresponding values are @xmath447 see figure [ fig : net_ex5 ] . theorem [ thm : main_suff_generalized ] tells us that each arc of the unit circle between two critical points gives rise to a measure . labeling the measures in agreement with theorem [ thm : main_suff_generalized ] ( starting at @xmath98 and traversing the arcs of the unit circle in the counterclockwise direction ) we get measures @xmath448 , @xmath449 , and @xmath450 , which , since @xmath451 and @xmath452 , are supported approximately on the intervals @xmath453 $ ] , @xmath454 $ ] , and @xmath455 $ ] , respectively . the limiting measure then equals @xmath456 . the illustration of the corresponding densities are given in figures [ fig : densities_ex5 ] and [ fig : density_ex5 ] . the graph of the density of @xmath21 suggests that the eigenvalues of @xmath16 cluster with higher density to the left of the point @xmath457 which has also been observed numerically , see figure [ fig : eigen_plot_ex5 ] . our final plots are devoted to an illustration of the connection between the presence of a jordan curve in @xmath45 and the reality of @xmath28 . we go back to the example 4 once again and introduce an additional parameter @xmath458 at @xmath459 getting the symbol @xmath460 in figure [ fig : break_ex ] , it is shown that , as the parameter @xmath358 increases from @xmath461 to @xmath462 , the jordan curve in @xmath45 gets `` destroyed '' by an incoming curve from the net of @xmath38 . at the same time , @xmath28 changes from a real interval to a set with non - real values . the set @xmath28 is plotted by using the algorithm from @xcite . the authors wish to express their gratitude to professor a. b. j. kuijlaars for pointing out an error in the first version of the manuscript and to professor m. duits for his interest in the topic . the following determinant formula presented in lemma [ lem : det_h_n_tridiag ] was used in subsection [ subsec : ex1 ] . it can be deduced from the well - known identity for the hankel transform of the sequence of central binomial coefficients . although the derivation is elementary , it is not completely straightforward and therefore we provide it with its proof for the convenience of a reader . first note that , if @xmath215 and @xmath464 are two hankel matrices with @xmath465 and @xmath466 such that @xmath467 , for some @xmath356 , then @xmath468 where @xmath469 . therefore it suffices to verify the statement for @xmath387 . next we make use of the direct sum decomposition @xmath475 where @xmath476 since @xmath477 , for all @xmath329 , both subspaces @xmath478 and @xmath479 are @xmath215-invariant , i.e. , @xmath480 and @xmath481 . the matrix @xmath215 decomposes accordingly as @xmath482 where @xmath483 thus , by using formulas in , one obtains @xmath484 evaluation of a special hankel determinant of binomial coefficients . in _ fifth colloquium on mathematics and computer science _ , discrete math . , ai . assoc . discrete math . , nancy , 2008 , pp . 251267 . chebyshev quadrature for measures with a strong singularity . in _ proceedings of the international conference on orthogonality , moment problems and continued fractions ( delft , 1994 ) _ ( 1995 ) , vol . 65 , pp . 207214 .

the homunculus reflection nebula around carinae provides a rare opportunity to observe the spectrum of a star from more than one direction . in the case of car , the nebula s geometry is known well enough to infer how line profiles vary with latitude . we present stis spectra of several positions in the homunculus , showing directly that car has an aspherical stellar wind . p cygni absorption in balmer lines depends on latitude , with relatively high velocities and strong absorption near the polar axis . stronger absorption at high latitudes is surprising , and it suggests _ higher mass flux toward the poles _ , perhaps resulting from radiative driving with equatorial gravity darkening on a rotating star . reflected profiles of he i lines are more puzzling , offering clues to the wind geometry and ionization structure . during car s high - excitation state in march 2000 , the wind was fast and dense at the poles , with higher ionization at low latitudes . older stis data obtained since 1998 reveal that this global stellar - wind geometry changes during car s 5.5 year cycle , and may suggest that this star s _ spectroscopic events are shell ejections_. whether or not a companion star triggers these outbursts remains ambiguous . the most dramatic changes in the wind occur at low latitudes , while the dense polar wind remains relatively undisturbed during an event . the apparent stability of the polar wind also supports the inferred bipolar geometry . the wind geometry and its variability have critical implications for understanding the 5.5 year cycle and long - term variability , but do not provide a clear alternative to the binary hypothesis for generating car s x - rays .

astro-ph0301394

at visual wavelengths , the dusty bipolar `` homunculus '' ejecta structure around @xmath0 car is primarily a hollow reflection nebula ( see thackeray 1961 ; visvanathan 1967 ; meaburn 1987 ; hillier & allen 1992 ; hamann et al . 1994 ; etc . ) . therefore , by observing localized parts of the homunculus , _ we can indirectly see the star s spectrum from a range of directions in space _ potentially allowing us to reconstruct the shape of @xmath0 car s wind , which is expected to lack spherical symmetry . independent of the large - scale bipolar morphology , hints of non - sphericity have been noted by , e.g. , viotti et al . ( 1989 ) , hillier & allen ( 1992 ) , davidson et al . ( 1995 ) , davidson & humphreys ( 1997 ) , and zethson et al . indeed , hillier & allen ( 1992 ) found a difference between direct and reflected he i emission that originates in the stellar wind , and early space telescope imaging spectrograph ( _ hst_/stis ) observations showed that the reflected h@xmath1 profile depends on location within the homunculus . departures from spherical symmetry are obviously critical for theories of winds and instabilities in the most massive stars ( see lamers & cassinelli 1999 , and papers in wolf , stahl , & fullerton 1998 ) . asphericity in @xmath0 car s wind may also be critical for understanding its 5.5 year cycle ( damineli 1996 ) , its long term variability , and excitation of its ejecta . in addition to its classic status as a very massive unstable star , @xmath0 car is uniquely favorable for this problem because the dusty homunculus has a fairly definite three - dimensional structure ; each projected location on the nebula provides reflected light from a viewing direction whose stellar latitude is approximately known ( davidson et al . 2001 ; smith 2002 ) . we have obtained long - slit stis spectra optimized for studies of the reflected latitude - dependent effects . we find that some features do in fact depend on viewing direction . wind velocities are highest around the pole , as one might expect if stellar rotation reduces the effective gravity and escape velocity at the equator . dominant outflow velocities in the stellar wind are 400 to 600 km s@xmath2 , but some faster material ( @xmath31000 km s@xmath2 ) had been seen ( viotti et al . 1989 ; damineli et al . 1998 ) , and our data now show that these high velocities are polar . we also find a much less anticipated phenomenon : broad p cyg absorption components of h@xmath1 and h@xmath4 , usually weak in our direct view of @xmath0 car , _ are quite deep at higher ( not lower ) latitudes_. traditionally , a slow equatorial wind is supposed to be denser than the associated fast polar wind and the mass - loss rate may increase with decreasing @xmath5 ( friend & abbott 1986 ) ; some examples are b[e ] stars ( zickgraf et al . 1986 ) , wind - compressed disks or zones ( bjorkman & cassinelli 1993 ; ignace , cassinelli , & bjorkman 1996 ; owocki , cranmer , & blondin 1994 ) , and bistable winds ( lamers & pauldrach 1991 ) . our observations of @xmath0 car , however , suggest that its wind is densest near the poles despite higher velocities there . other interpretations are conceivable , but require detailed models of line transfer . if @xmath0 car s wind densities are highest near the poles as one might guess from the deeper hydrogen p cyg absorption , then this tends to support recent comments by owocki and others concerning the importance of non - radial forces and gravity darkening in line - driven winds of massive stars ( owocki , cranmer , & gayley 1996 ; owocki & gayley 1997 ; owocki , gayley , & cranmer 1998 ; maeder & meynet 2000 ; meader & desjacues 2001 ) . specifically , these studies predict that if a luminous star is rotating fast enough for gravity darkening to be important ( i.e. , at @xmath670% of the `` critical '' rotation speed ) , then stronger radiative flux at the hot poles will tend to drive a higher mass flux . glatzel ( 1998 ) and maeder ( 1999 ) have discussed modifications to the `` critical '' rotation rate near the eddington luminosity . we describe new stis observations in 2 below , hydrogen line profiles in 3 , and helium profiles in 4 . in 5 we discuss various likely implications of the observed wind geometry . then in 6 we examine older stis spectra and discuss temporal variability of the wind structure . finally , in 7 we mention consequences that a variable , bipolar wind may have for interpretations of @xmath0 car s 5.5-year spectroscopic cycle . these observations are part of an intensive _ hst_/stis project to study @xmath0 car and the homunculus during their current 5.5-year spectroscopic cycle ( davidson et al . 1999 , 2000 , 2001 ; gull et al.1999 ; gull & ishibashi 2001 ) . early observations in march 1998 and february 1999 used only a narrow 0.1@xmath7-wide stis slit with short exposures , inadequate for outer parts of the homunculus ( although we did obtain a few deeper exposures at wavelengths near h@xmath1 that were offset from the star ; see below ) . moreover , difficulties in scheduling _ hst _ prevented the slit from being aligned closely with the homunculus bipolar axis during those earlier observations . therefore , we planned a special set of stis observations with a wider 52@xmath80@xmath92 aperture and longer exposures , and with a slit orientation closer to the homunculus axis at position angle 138@xmath109@xmath11318@xmath109 , as shown in figure 1 . here we discuss observations sampled with stis gratings g430 m and g750 m at standard tilts that gave central wavelengths of 4961 , 6768 , and 7283 . achieved on 2000 march 13 , these are the same data used in a paper on the homunculus shape ( davidson et al.2001 ) . in order to reduce scattered light we occulted the star with a 0.65@xmath7 bar across the slit , but other stis observations gave a direct view of the star a week later ( see below ) . figure 2 shows examples of long - slit stis spectra for a few bright reflected emission lines in @xmath0 car s stellar wind ( reflected continuum has been subtracted ) . to study positional variation of line profiles in the homunculus , we made several spectral extractions at positions along the slit listed in table 1 . positive offsets are nw of the star , and negative offsets are se . the spectrum at each offset position has a net redshift , due to reflection by expanding dust . we corrected for this effect by aligning the blue side of the h@xmath1 emission line profile at 10 times the continuum flux ( however , a different choice would not alter the results substantially because the profiles match well ) . the net shift applied at each position is given as @xmath12 in table 1 . we used these values of @xmath12 to correct the apparent velocities of other emission lines discussed later . each resulting line profile samples the spectrum of the central star seen from different positions in the reflection nebula , and each corresponds to a different latitude as depicted in figure 3 . this technique works because the polar lobes are mostly hollow , as indicated by limb - brightening in thermal - infrared images ( e.g. , smith et al . 1998 , 2002 ) . the case of @xmath0 car is unique , because the geometry of the bright , hollow reflection nebula is known sufficiently well that we can correlate position in the nebula with stellar latitude ( assuming that the polar axis of the star and nebula are aligned in this case the assumption is justified by the results , see 3 ) . davidson et al.(2001 ) estimated the shape and orientation of the homunculus from the same dataset used here , and we use their `` compromise '' shape ( shown in figure 3 ) to derive the latitude for each extraction position as listed in table 1 and depicted in figure 3 . since the homunculus may not be fully axisymmetric , our estimated latitudes have uncertainties of roughly @xmath135@xmath14 . the axis for the southeast homunculus lobe is tilted at @xmath15 relative to our line of sight ( see davidson et al . 2001 ; smith 2002 ) . the 2000 march 13 observations used a 0@xmath965 bar to occult the central star , but other stis spectra obtained on 2000 march 20 sampled the star with a 0@xmath91-wide slit . a 0@xmath915 segment was extracted from the 2000 march 20 data for all emission lines discussed here . we also examined the spectrum of one position in the equatorial ejecta located @xmath35@xmath7 ne of the star , constructed from several adjacent slit positions used to map the homunculus at wavelengths near h@xmath1 on 2000 march 21 . the extracted spectrum covers an effective area of 1@xmath7@xmath161@xmath7 , represented by the box labeled `` ne 5 '' in figure 1 . this spectrum is discussed in 5.4 ( and see table 1 ) . temporal variations should not introduce significant errors since the datasets were taken only 1 week apart , while differences in light travel times for reflected light from the se polar lobe are only a few weeks or less . thus , the observed variation in line profiles _ along the slit _ can not be due merely to temporal variations in the spectrum . in 6 we investigate temporal variability of reflected profiles as seen in stis spectra obtained with the g750 m grating ( central wavelength = 6768 ) on 1998 march 19 and 1999 february 21 . these are long - slit spectra similar to those obtained on 2000 march 13 , but were obtained using a 0@xmath91-wide slit located 0@xmath94 nw of the star at position angle 152@xmath101@xmath11332@xmath101 , as shown by the dotted line in figure 1 . transformation between position along the slit and corresponding stellar latitude is less straightforward than for the march 2000 data . for the 1998 and 1999 data , we estimated latitudes for each offset position using contours superimposed on images of the homunculus ( davidson et al . uncertainty in latitude for these extractions is about @xmath135@xmath14 . spectra of the star for comparison with reflected spectra were obtained simultaneously , and processed in the same manner as stellar spectra discussed above . in context with variability during @xmath0 car s 5.5 year spectroscopic cycle ( damineli 1996 ) , stis observations in march 1998 , february 1999 , and march 2000 occur at phases of 0.04 , 0.21 , and 0.4 , respectively ( see 5.1.2 and 7 ) . in 3 we discuss a small portion of the ultraviolet stis fuv / mama spectrum of @xmath0 carinae , obtained in echelle mode with the e140 m grating on 2000 march 23 , using a 0@xmath92 @xmath16 0@xmath92 aperture centered on the central star . these data are used here to help understand aspects of the optical spectra ; a detailed analysis of the uv spectra will be presented in a future publication . figure 2 shows that reflected line profiles vary with position along the slit . the primary change occurs in strength and velocity range of hydrogen p cygni absorption , which is strongest 5@xmath7 to 6@xmath7 se of the star . p cygni absorption is weak in the directly - observed star and in reflected light from the nw lobe . h@xmath4 and h@xmath1 profiles agree well ( figures 4 and 5 ) ; their scattering wings and emission peaks vary little across the se lobe , but absorption components depend on position . absorption traces a more direct and narrow line of sight between reflecting dust in the polar lobes and the star ( this is particularly true for balmer absorption , which occurs at large radii compared to the continuum ) , whereas emission comes from a range of latitudes . since each position along the slit translates to a line of sight down onto a different latitude on the central star , p cygni absorption components give direct evidence that the density and/or ionization structure of @xmath0 car s current stellar wind is non - spherical . the possible cause of this aspherical structure is addressed in 5 . to quantify the latitudinal dependence of velocity , we measured `` terminal '' speed @xmath17 and minimum - flux speed @xmath18 for each tracing in figure 4 . figure 6 shows how we defined these two velocities observationally , and resulting measurements are listed in table 1 . with data from table 1 , figure 7 indicates how observed expansion velocities vary with latitude . changes in @xmath17 are greater than the uncertainty , which could be up to 100 km s@xmath2 for high latitudes . for our line of sight , @xmath17 measured in balmer lines is within @xmath350 km s@xmath2 of the terminal velocity indicated by uv wind lines , like si ii @xmath191527 , also observed in march 2000 using the stis fuv / mama detector ( figure 8) . several aspects of hydrogen emission - line profiles in figures 4 and 5 , and trends in figure 7 deserve comment : \1 . _ higher velocities near the pole . _ blueshifted absorption is seen at speeds up to @xmath31100 km s@xmath2 , which is much faster than values usually quoted for @xmath0 car s stellar wind . the fastest material between 5@xmath7 and 6@xmath7 se of the star nearly coincides with the polar axis of the homunculus . in figure 7 , the measured quantities @xmath17 and @xmath18 both follow a similar trend of smoothly increasing speed with latitude , and there appears to be a rather extreme increase in @xmath17 close to the polar axis . \2 . _ deeper absorption near the pole . _ perhaps the most significant and surprising result of this investigation is that the deepest p cygni absorption in balmer lines is seen at the polar axis , with absorption weakening _ progressively _ toward lower latitudes ( see figure 5 ) . at latitudes below about 40@xmath14 , the normal balmer p cygni absorption is absent in the wind . in dense winds like @xmath0 car s , the strength of p cygni absorption in balmer lines is sensitive . it depends on the population of the @xmath20 state and is not necessarily a direct tracer of column density . weak balmer absorption at low to mid latitudes could be due to higher ionization in that part of @xmath0 car s wind , lower density , or a subtle combination of both effects . _ symmetry about the equatorial plane and polar axis . _ measured values of @xmath17 and @xmath18 plotted together in figure 7 follow the same trend regardless of which polar lobe or which side of the se polar axis they correspond to . also , h@xmath1 emission and absorption profiles on either side of the pole are identical . this indicates that the wind outflow pattern is _ axisymmetric_. electron scattering wings and line peaks seen at various positions in the se polar lobe all have the same profile shape ( excluding blueshifted absorption ) , but these reflected profiles all differ from our direct view of the central star . the difference is considerable almost a factor of 2 in the peak intensity ( figure 5 has a logarithmic scale ) . our direct view of the central star shows stronger emission at velocities between about @xmath21250 and @xmath22500 km s@xmath2 in both h@xmath4 and h@xmath1 , but broad scattering wings are consistent with profiles seen in reflection . apparently , when we look at the star we see extra line emission ( or extra continuum opacity ) that is not seen by reflected lines of sight ; the opposite case would have been easier to explain . a similar but lesser effect appeared in the _ hst_/fos data on h@xmath4 ( davidson et al . 1995 ; see also hillier et al . 2001 ) . later ( in 6 and figure 16 ) we show that this characteristic of the star s emission persists at roughly the same level throughout @xmath0 car s spectroscopic variability cycle . models of the central star s spectrum ( as seen in march 1998 ) by hillier et al . ( 2001 ) showed a related problem : the model adequately reproduced the electron scattering wings , but underestimated peak emission of balmer lines . the model also overestimated the strength of p cygni absorption . thus , the atmosphere model used by hillier et al . approximates the reflected spectrum from the polar lobes in march 2000 more closely than it does the central star s spectrum , although the observed outflow velocities at high latitudes exceed the 500 km s@xmath2 terminal velocity adopted in their model . this problem may be related to wind geometry , since hillier et al . assumed spherical symmetry . _ fe ii profiles . _ figures 9 and 10 show observed emission - line profiles for fe ii @xmath194925 , extracted from positions across the homunculus and the central star in the same manner as balmer line profiles shown in figures 4 and 5 . fe ii @xmath194925 is a bright permitted line in @xmath0 car s wind , and shows a broad profile similar to many other fe ii lines . fe ii and h@xmath1 emission show a similar latitudinal dependence . namely , the velocity and to some extent the strength of p cygni absorption increase with latitude . also , fe ii emission from the central star is stronger than at reflected positions . thus , a solution to this problem for balmer line emission should also account for the fe ii discrepancy . more importantly , the similar behavior of these two lines suggests that they arise in the same outer zone of @xmath0 car s wind , as expected because the ionization of fe ii is coupled to h by charge exchange in dense stellar winds . figure 2 shows the long - slit spectrum across the major axis of the homunculus for he i @xmath197067 and @xmath196680 . tracings for both he i @xmath196680 and @xmath197067 at several reflected positions across the nebula and for the central star are shown in figures 9 and 10 . reflected line profiles in the se lobe seem to have three velocity components in figure 2 , caused by a combination of a narrow central component from circumstellar gas , plus a broad plateau or double - peaked line arising in the wind . contamination by reflected narrow he i @xmath197067 emission from circumstellar ejecta like the weigelt blobs ( see davidson et al.1995 , 1997 ) needs to be examined before the wind lines can be interpreted . at 2@xmath7 to 4@xmath7 se of the star , narrow emission is conspicuous but the broad component and the continuum emission fade . this suggests that obscuring dust lies along this particular line of sight to the star , projecting _ a shadow _ onto the se polar lobe . this same position corresponds to a relatively dark region in color _ hst _ images of the homunculus ( see figure 1 and morse et al . intervening material must be close to the star ( within @xmath33500 au ) and compact ( @xmath23200 au ) , because it obscures the star along this line of sight but does not block the weigelt blobs . the dust in question may be in condensations analogous to the weigelt blobs themselves . dense dust knots seen at thermal - infrared wavelengths ( smith et al . 2002 ) are good candidates for such blobs that are casting shadows . perhaps some other dark spots in the homunculus are also shadows cast by inner debris , rather than holes or dense patches . changes in broad stellar - wind profiles are complicated , and may be affected by a hypothetical companion star . however , some clues to the overall wind structure can be gathered . broad he i @xmath197067 emission in the se lobe ( see figure 9 ) has a nearly symmetric profile . this broad profile , combined with the central narrow component , creates the triple - peaked appearance of he i @xmath197067 in the se lobe in figure 2 . the red side of he i lines disappears in the reflected spectra from the nw polar lobe , in the spectrum of the star , and in reflected spectra in the se polar lobe at positions near the star ( see figures 9 and 10 ) . the blue side of the broad he i line profiles is much stronger in the nw lobe , but roughly the same for the se lobe and central star ( figure 10 ) . in other words , low latitudes ( below @xmath350@xmath14 ) seen from the nw lobe have asymmetric profiles with extra blueshifted emission , whereas high latitudes in the se lobe show symmetric profiles and extra redshifted emission compared to the central star . it is more difficult to ascertain the level of axisymmetry in he i emission than for balmer lines , because he i lines do not show such drastic changes near the pole ; he i emission does appear to be symmetric about the equatorial plane . the central star and reflected emission from the nw lobe ( low latitudes ) also show p cygni absorption at 400 km s@xmath2 in he i lines , but the reflected spectra in the se lobe ( high latitudes ) do not . balmer lines show the opposite trend ; namely , stronger p cygni absorption at high latitudes . important he i lines at 10830 and 20581 also show p cygni absorption ( hamann et al . 1994 ; damineli et al . 1998 ; smith 2001 , 2002 ) ; it would be interesting to see how these vary with latitude . continued monitoring of direct and reflected he i line profiles during @xmath0 car s 5.5 year cycle with stis may provide a powerful diagnostic of the stellar wind s complex ionization structure . latitudinal variations of h and he i lines described above show that the speed , density , and ionization in @xmath0 car s wind are non - spherical . do these indicate an inherently axisymmetric wind , or is the structure influenced by a hypothetical hot companion star ? one can imagine suitable parameters for a binary model , but in several ways described below the observations more closely match theoretical predictions for a dense line - driven wind from a rotating star ( owocki et al . 1996 , 1998 ; owocki 2002 ) . thus , we favor rotation of the primary star as the dominant factor shaping @xmath0 car s wind . figure 7 shows maximum outflow velocities increasing rapidly toward the poles . if the terminal speed @xmath17 is proportional to the escape speed as usually occurs in stellar - wind theory , and if the surface rotation rate @xmath24 does not perceptibly depend on stellar latitude @xmath4 , then one might expect a simple latitude dependence : @xmath25 where @xmath26 and @xmath27 at the equator . higher polar velocities are generally expected from any rotating star with lower effective gravity at the equator . an example of this trend is shown in figure 7 for @xmath28 , and obviously does not match the observed latitude dependence exactly . but this failure is not very surprising since the assumption ignores possible complications ; line - driven wind theory may be an oversimplification for @xmath0 car , opacity may depend on latitude , the flow is not necessarily radial , etc . however , the ratio of the polar to equatorial velocities does seem marginally consistent with @xmath29 . the rapid rise in @xmath17 toward the pole is somewhat problematic , especially since no high - speed redshifted emission is seen there . fast material may be concentrated in a narrow cone near the polar axis , so it may not contribute significantly to the h@xmath1-emitting volume in the wind . this must be addressed with quantitative modeling . a normal binary model offers no obvious explanation for high polar wind velocities , especially since one expects the orbital plane to be roughly aligned with the equator of the star and homunculus . high velocities seen in reflected light from the se polar lobe give the first direct indication that the polar axis of the homunculus is aligned with the rotation axis of the central star . the alignment of these axes has important consequences for the formation of the bipolar lobes and equatorial ejecta around @xmath0 car . namely , it means that axial symmetry and ejection physics during the great eruption may be directly linked to the star s rotation ( see 5.5 ) . high velocities seen in reflection also indicate fast - moving material somewhere inside the homunculus , but the absorption probably occurs relatively close to the star and not in the middle of the polar lobes or outside them ( see 6.3 ) . values for @xmath17 at high latitudes are almost twice the expansion velocities for gas and dust in the homunculus ( see figure 7 ) . therefore , we should expect shock - excited emission inside the polar lobes as the stellar wind overtakes slower material from the great eruption ( smith 2001 , 2002 ; smith & davidson 2001 ) . the highest speeds we observe in @xmath0 car s wind and homunculus are insufficient for @xmath0 car s hard x - rays , which presumably require shock velocities of roughly 3000 km s@xmath2 ( corcoran et al . 1995 ; ishibashi et al.1999 ; pittard & corcoran 2002 ) . the weakness or absence of p cygni absorption in balmer lines at low latitudes and the strange latitudinal behavior of he i lines make a colliding - wind binary model worth considering . a hot companion star s wind might sculpt part of the dense primary wind ( see , e.g. , pittard & corcoran 2002 ) and ionize its equatorial zones , but a binary model does not give an obvious explanation for the fast polar wind . if high speeds are induced by a companion star , we might expect them primarily at low latitudes instead . however , one argument against a companion star dominating the wind structure is that the non - spherical wind appears to be _ axisymmetric _ , as noted in 3.1 . in figure 7 , velocities from both hemispheres and both sides of the polar axis follow the same smooth latitudinal dependence , and p cygni absorption in hydrogen lines has the same depth on either side of the pole . the evidence for axisymmetry is not as clear in he i emission profiles , but they are consistent with an axisymmetric wind . if @xmath0 carinae has a companion star that causes colliding - wind x - ray emission , it must have a very eccentric orbit ( see contribution in morse et al . the stis observations described here were obtained in march 2000 , when a companion would be nearly at apastron . the shock would be between the two winds , while the far side of the primary star s wind should be mostly unaffected a seriously asymmetric arrangement . therefore the basic stellar - wind geometry most likely results from the star s rotation . a hypothetical hot companion may help to explain the he i emission and other effects , but this has not been demonstrated and requires theoretical work . to understand the wind structure , we first need to recall how density and ionization govern the behavior of p cygni absorption in h@xmath1 and h@xmath4 . balmer absorption in dense stellar winds is sensitive to changes in wind ionization structure , but that ionization structure in turn depends sensitively on density . najarro , hillier , & stahl ( 1997 ) present a detailed analysis of the behavior of h and he i lines in the spectrum of p cygni . much of their analysis may be applicable to @xmath0 car s wind ( see also hillier et al . they find that in dense winds of hot stars , the appearance of balmer lines falls into one of two regimes , depending on h recombination . dense winds that remain fully ionized show pure h emission profiles . if hydrogen recombines , however , the population of the @xmath30=1 state increases dramatically , and so the lyman series and lyman continuum become optically thick . because of high optical depth in ly@xmath1 , the @xmath30=2 state acts as a metastable level and gives rise to strong balmer absorption . essentially , najarro et al . find that strong balmer absorption requires sufficient density to cause h recombination in the outer wind . this is evidently the case for @xmath0 car s polar wind ; much weaker balmer absorption toward low latitudes then suggests that h remains mostly ionized there . najarro et al.s calculations show that even a small reduction in the mass - loss rate can prevent h recombination ; hillier et al . ( 2001 ) comment that for @xmath0 car , dramatic changes in p cygni absorption are seen if the mass - loss rate in their model is lowered by factors of 2 to 4 . thus , we can explain increasing strength of balmer absorption toward the pole if @xmath0 car s wind has _ a latitudinal density gradient , with higher densities toward the pole_. evidently , the wind begins to reach a critical level for recombination near latitudes of 40@xmath14 . this is analogous to ( but opposite ) the `` bistability '' model for b[e ] supergiants ( lamers & pauldrach 1991 ; pauldrach & puls 1990 ) . figure 11 shows a hypothetical picture of latitudinal structure in @xmath0 car s stellar wind during its normal high - excitation state . the inner circle represents the stellar radius , which is somewhat ambiguous because the wind is opaque ( see hillier et al . 2001 , davidson et al . figure 11 depicts two different zones in @xmath0 car s outer wind : 1 ) a high - density , low - ionization zone where h recombines and produces balmer absorption ( shaded ) , and 2 ) a lower density , high - ionization zone ( hatched ) producing the he i emission , or at least relatively transparent so that more of the underlying he i zone is seen , and where h remains ionized . in reality , of course , demarcation between these zones is not so sharp . more theoretical work is needed to understand the observed line profiles ( especially he i ) , but this picture qualitatively satisfies the main observational requirements of preceding sections : _ polar view : _ seen from the pole , the spectrum shows deep , high - velocity p cygni absorption in balmer and fe ii lines through dense regions of the recombined wind . looking through relatively transparent extended he i emission zones on either side of the star , an observer sees a symmetric he i profile , with blue- and red - shifted velocities of similar intensity out to about @xmath13400 km s@xmath2 . p cyg absorption is weak or absent in he i lines toward the poles , since there is no continuum source behind these extended he i - emitting zones . _ low latitudes : _ the low density and high - ionization at low latitudes inhibit hydrogen p cyg absorption in balmer lines and favor he i emission . an observer will see stronger blueshifted he i lines from deeper in the wind than other latitudes , since redshifted he i emission is occulted by the star . _ intermediate latitudes ( our line of sight ) : _ we see weaker hydrogen p cygni absorption than a polar observer because we see more continuum from deeper in the hotter parts of the wind ; for a polar observer , all the continuum and electron scattering is behind the absorbing region of the wind . we also see mostly blueshifted he i emission , but not as strong as for viewpoints at lower latitudes because the he i is spread over a wider velocity range , and because our line - of - sight is nearly along the transition between the two zones . the recombined polar region of the wind does not emit as much he i and is responsible for the missing ( or weaker ) blueshifted peak in the star s he i line profile as compared to the nw lobe . the above results suggest that @xmath0 car s wind has prolate mass flux , with higher velocities , higher densities , and lower ionization toward the pole . this is surprising , because reduced effective gravity at the equator of a rotating star might lead one to expect the opposite ; for example , rapidly rotating b[e ] supergiants are thought to have low - ionization equatorial density enhancements , as noted earlier . however , the prolate mass flux we observe in @xmath0 car during its normal high - excitation state is not without theoretical precedent . considering only centrifugal forces in an expanding and rotating wind , one might expect material to be deflected toward the equator ( e.g. , bjorkmann & cassinelli 1993 ) . however , in a line driven wind other factors are important ; radiation - hydrodynamical simulations that include non - radial forces with cak ( castor , abbott , & klein 1975 ) formalism for line - driven winds show that equatorial density enhancements can be inhibited ( owocki et al . 1996 , 1998 ) . furthermore , with gravity darkening , oblateness , and other effects , owocki et al . predict enhanced mass flux toward the polar directions of a rotating star . analytic scaling laws from cak theory ( their equation 46 ) can be modified to give an approximate relation ( see owocki et al . 1996 ) for the dependence of local mass - loss rate on flux @xmath31 and gravity @xmath5 , given by @xmath32 where @xmath1 is the force multiplier power - law index that is usually adopted in cak theory . however , the prologue to von zeipel s theorem predicts that @xmath33 for a rotating star . if this relation applies , if rotation is responsible for @xmath0 car s asymmetry , and if line - driven wind theory is a fair approximation , then equation ( 2 ) implies that local mass flux @xmath34 should be proportional to @xmath5 , which depends on latitude . since @xmath35 tends to be more or less proportional to @xmath36 , we expect , very crudely , a wind density dependence @xmath37 . thus , the polar wind density can easily be higher than that near the equator . figure 7 suggests that the ratio might be as large as 2 , which , as noted in 5.2 , may allow deep p cygni absorption features at the pole but not at low latitudes . even if the above argument does not strictly apply , some modifications to the theory for luminous stars tend to produce strong polar mass flux ( see , e.g. , maeder 1999 ; maeder & desjaques 2001 ) . here we have ignored obvious possibilities of latitudinal opacity variations and of acceleration by the radiation continuum ; a quantitative investigation is merited but will be difficult . the essential requirement of our proposed interpretation of @xmath0 car s wind asymmetry that higher densities toward the pole lead to recombination at large distances in the wind ( as discussed by najarro et al . 1997 ) seems theoretically plausible if @xmath0 car is rotating rapidly enough to cause equatorial gravity darkening ( see owocki et al . 1996 , 1998 ; owocki & gayley 1997 ) . and h@xmath4 absorption features represent the nlte population of the @xmath38 level , which depends on various subtle factors . some of these , such as three - dimensional velocity gradients , are not treated adequately in any published wind models . ] gravity darkening on the star would not necessarily contradict the higher ionization observed at low / mid latitudes in the wind , because that latitude - dependent ionization is regulated by the local density instead of an anisotropic uv source ( 5.2 ) . for instance , latitudes at @xmath325@xmath14 from the equator should still `` see '' the hot pole of the star . material much closer to the equator will have a more difficult time getting uv radiation from the hot poles , leading one to expect lower ionization there ; this may indeed be the case , as discussed in the next section . while higher polar velocities are a generic property of winds from rotating stars , higher polar mass flux is a distinctive property of line - driven winds with gravity darkening . thus , our observations appear to support the hypothesis that @xmath0 car s present - day stellar wind is in fact radiatively driven . this may not hold true during the great eruption . several authors have discussed evidence for strange kinematics in @xmath0 car s equatorial ejecta ( zethson et al . 1999 ; smith & gehrz 1998 ; morse et al . 2001 ; davidson et al . 2001 ; davidson & humphreys 1997 ) . not all equatorial structures are coeval with the polar lobes ejected during the great eruption . the homunculus ragged _ equatorial _ structure distinguishes it from other bipolar nebulae . equatorial outflow in the present - day wind may be important for understanding this larger - scale morphology . figure 4 includes the equatorial h@xmath1 profile observed @xmath35@xmath7 ne of the central star ( see figure 1 ) , and table 1 lists values of @xmath17 and @xmath18 for this feature . these outflow velocities are higher than the expected trend at the equator , and deep p cyg absorption there contradicts the observation that blueshifted absorption weakens with decreasing latitude . thus , the reflected spectrum from ne 5 suggests that these trends reverse at the equator perhaps indicating the presence of a dense equatorial disk . ne 5 might be peculiar since it is projected along a line from the star to the nn `` jet '' ( see morse et al . 1998 ; meaburn et al . 1993 ) . however , other stis data obtained in november 1998 ( not shown ) , with the slit crossing other extended parts of the equatorial skirt , also show obvious p cygni absorption at the equator . binary models that invoke a strong wind from a companion star ( e.g. , pittard & corcoran 2002 ) would seem to impede the survival of such a thin equatorial disk . perhaps some material escaping from mid - latitudes on the star is deflected toward the equator in which case the same process may cause both depletion of the outer wind density at mid - latitudes and increased density near the equatorial plane . various theoretical ideas including centrifugal wind compression and magnetic fields may produce an outflowing equatorial disk in this way ( bjorkman & cassinelli 1993 ; washimi & shibata 1993 ; balick & matt 2001 ; ud - doula & owocki 2002 ; lee , saio , & osaki 1991 ) . alternatively , low ionization in the `` disk '' may be related to the star s gravity darkening ( 5.3 ) . the bipolar geometry depicted in figure 11 naturally leads one to wonder if the present - day stellar wind s bipolarity on small scales ( within @xmath3100 au ) is related to the shape and structure of the homunculus nebula on much larger scales . conditions in @xmath0 car s wind during the great eruption were very different than today , but let s assume for the moment that the geometry of the mass loss was similar then . suppose that latitudinal dependence on density in @xmath0 car s wind 160 years ago resembled figure 11 , but that during the eruption the mass - loss rate increased by a factor of 100 for 20 years turned on and off like a faucet . in that case , most mass would have been ejected toward the poles during that time , forming two thick polar caps . their radial thickness would be roughly @xmath39 . this simple estimate matches observations ; the homunculus polar caps are indeed roughly 1@xmath7 thick at radii around 8@xmath7 from the star ( smith et al . 1998 , 2002 ; davidson et al . 2001 ) . this outburst scenario differs from previous hydrodynamic models for the formation of the homunculus that invoke wind interactions to explain the observed bipolarity ( e.g. , frank et al . 1995 , 1998 ; dwarkadas & balick 1998 ; langer et al . we should not expect post - eruption hydrodynamic effects to dominate the structure of the homunculus , since material ejected during the eruption was a factor of 100 more dense but had comparable velocities to the pre- and post - eruption winds . ram pressure in the post - eruption wind is insufficient to dominate the shape of the nebula or cause significant acceleration . proper motions of the homunculus ejecta ( morse et al.2001 ; smith & gehrz 1998 ; currie et al . 1996 ) indicate linear expansion . a bipolar eruption of a rotating star seems the best explanation for the basic homunculus morphology ( see owocki & gayley 1997 ; maeder & desjacques 2001 ; dwarkadas & owocki 2002 ) . small - scale structure in the homunculus lobes may also be related to details in the eruptive wind . davidson et al . ( 2001 ) noted that the smallest observed feature size , about 0.1@xmath7 ( see figure 1 and morse et al . 1998 ) , corresponds to the relevant speed of sound multiplied by the time elapsed since the great eruption . the observed characteristic scale size is larger , but does not contradict this basic idea : that the clumpy ( `` granular '' ) small - scale structure in each bipolar lobe represents features in the eruptive wind that were frozen into the freely expanding ejecta . given the quantitative details mentioned above , the corrugated structures probably did _ not _ result from an instability in the flow long after the eruption ( kelvin - helmholtz , rayleigh - taylor - like , or their kin ) . the structure , which pictorially resembles solar granulation , may have resulted from wind instabilities or from effects near the stellar surface ( the sonic point ) . according to shaviv ( 2000 ) , porous structure should develop in a super - eddington wind like that of @xmath0 car during 18381858 . this porosity may be imprinted in clumps or holes seen in the polar lobes , and may prove more useful than the large - scale lobe shapes for understanding the great eruption itself . the proposed wind structure with high ionization at low latitudes and dense optically - thick low - ionization zones near the poles helps explain several peculiarities : \1 . high - ionization [ ne iii ] and [ fe iii ] emission regions exist a few hundred au from the star even though the dense wind is expected to absorb the star s ionizing photons ( e.g. , see model in hillier et al . this is one motivation for the binary hypothesis , wherein a second star can provide ionizing radiation . however , the non - spherical wind that we propose may allow uv photons to escape from the primary star at low latitudes . this occurs despite gravity darkening at the stellar surface ; the wind is opaque near the poles but semi - transparent at low latitudes . the `` weigelt blobs , '' bright emission - line sources northwest of the star , are indeed thought to be located near the equatorial plane ( davidson et al . 1995 , 1997 ) . the brightness of the weigelt blobs is also a problem , since they seem to reflect more of the star s light than they intercept , for their presumed size ( see davidson & humpherys 1986 ) . dust is continually forming in @xmath0 car s wind , evidenced by the unresolved hot dust in the core of the homunculus seen in infrared images ( e.g. , smith et al . 1998 , 2002 ) . if dust forms primarily in dense polar regions of the wind , it might help obscure the star along our line of sight more than toward the weigelt blobs . the simultaneous existence of separate high- and low - excitation regions in the wind helps explain the range of spectral types from b2 to b8 ia seen in the far - uv spectrum of @xmath0 car by ebbets et al.(1997 ) . \4 . the spherical wind model used by hillier et al . ( 2001 ) overestimates the depth of the p cygni absorption components in balmer and lines . figures 4 and 10 show that both species show deeper absorption in reflected spectra from the se lobe than in our direct line of sight to the star . this supports the hypothesis ( point number 2 above ) that dense polar zones of @xmath0 car s wind dominate the optical spectrum , and the spherical model used by hillier et al.more closely approximates polar wind conditions . in addition to the weigelt blobs , extended _ equatorial _ ejecta around @xmath0 car show peculiar excitation that might be less mysterious if @xmath0 car has a stronger uv flux at low latitudes . equatorial regions would also see enhanced ly@xmath1 emission escaping from the polar wind , relevant to some proposed excitation mechanisms ( johansson & hamann 1993 ; hamann et al . notable extended equatorial features are the `` purple haze '' in _ hst _ images ( morse et al . 1998 ) , gas in the inner torus seen in infrared and optical images ( smith et al . 1998 , 2002 ) , the bizarre lines near @xmath32507 ( see davidson et al . 1995 , 1997 ; johansson et al . 1998 ; hamann et al . 1999 ; johansson & letokhov 2001 ) , and other and metal lines in the equatorial skirt with strange velocities and excitation ( zethson et al . extended equatorial gas in the circumstellar torus also shows peculiar _ variability _ ( duncan et al . 1995 ; smith et al . 2000 ; smith & gehrz 2000 ) , which may be related to spectroscopic changes discussed below . so far we have discussed the wind during its normal high - excitation state as observed in march 2000 , but @xmath0 car also has occasional `` spectroscopic events '' in which the spectrum temporarily changes for reasons that are not yet understood . originally these events were interpreted as single - star shell ejections ( zanella , wolf , & stahl 1984 ; bidelman et al . 1993 ) , but a companion star seems likely , given the 5.5-year periodicity associated with the events ( damineli 1996 ; whitelock et al . 1994 ; corcoran et al . 1995 ; duncan et al . 1995 ; damineli et al . 1997 , 1998 ; davidson 1999 ) . limited hst / stis data were obtained during the most recent spectroscopic event in 199798 and at a few later times . davidson et al . ( 1999 ) described some results for the star viewed directly , and remarked that the spectroscopic changes resembled a shell ejection ( see also gull et al . 1999 ; davidson 1999 , 2001 ) . we can apply the same latitude - dependent reflection method described above to the march 1998 and february 1999 stis data , where h@xmath1 and he i @xmath196680 were sampled with long exposure times on a particular slit orientation and placement shown in figure 1 . unfortunately this slit position was not the same as for the march 2000 observations discussed above , but it crossed the homunculus in a similar way . here we describe significant changes in the reflected spectra ; a report on the directly - viewed star is in preparation ( davidson et al . ) . figure 12 shows grayscale representations of long - slit spectra for he i @xmath196680 in march 1998 and february 1999 , and figure 13 demonstrates variability of he i line profiles with tracings of the central star , as well as reflected spectra from the nw and se lobes . equivalent widths measured for these tracings are plotted in figure 14 . at positions in the nw polar lobe , he i @xmath196680 is absent in march 1998 and extremely weak in february 1999 . positions in the nw lobe during march 1998 reflect the spectrum of the star 1 to 3 months earlier ( soon after the event ) , and show that this feature nearly vanished , at least at these low latitudes . stis spectra of the central star during the event in late december 1997 show that the he i @xmath196680 flux remained detectable along our direct line of sight ( figure 13 ) . in the se polar lobe the equivalent width in figure 14 is nearly constant . also , in march 1998 the he i emission line in the se lobe is stronger than it is in the nw polar lobe almost a year later in february 1999 . this could not be the case if changes in the he i line were spherically symmetric ( the time delay for positions in the se lobe is only a few weeks ) . in 4 we noted that the se lobe in march 2000 showed a broad , roughly symmetric stellar - wind he i profile , plus reflected narrow emission from slow - moving inner ejecta such as the weigelt blobs . the narrow line disappeared during 1998 and early 1999 ; compare figure 12 to figure 2 . meanwhile , as shown in figure 13 , the polar view of the stellar wind component ( `` se lobe '' ) changed in a way quite unlike that observed directly on the star . the latitude dependence of he i variability probably indicates that _ the wind s asphericity was changing _ during and after the 199798 event . figure 14 shows that broad he i emission faded mainly at low latitudes , while the polar wind seemed relatively undisturbed . h@xmath1 emission seen directly in the spectrum of the central star has remained roughly constant in brightness relative to the continuum and in width , but its p cyg absorption changed during 199899 ( see davidson et al . 1999 ; davidson 2001 ) . changes in the latitudinal dependence of reflected p cygni absorption are striking ( figure 12 ) ; tracings of the h@xmath1 line are shown in figures 15 and 16 , and corresponding velocities are listed in table 2 . in our direct view of the star , deep p cyg absorption around @xmath40 km s@xmath2 occurred in early 1998 during the event but virtually disappeared before february 1999 . meanwhile , however , strong p cyg absorption persisted at high latitudes , and its extreme velocity grew from @xmath41 km s@xmath2 in 1998 to @xmath42 km s@xmath2 in 2000 ( se lobe , figures 15 & 16 ) . deep p cyg absorption seen at low latitudes in march 1998 disappeared two years later . thus , viewing the spectra seen in early 1998 rather than 2000 , we realize that _ the strong latitude dependence described in 5 above did not apply during the 199798 spectroscopic event . _ at that time the depth of the p cyg absorption was less latitude - dependent , while the extreme observed speeds that we have denoted @xmath43 temporarily decreased at high latitudes but increased at low latitudes ( figure 17 ) . the wind was evidently disrupted through a wide latitude range , most likely by a `` shell ejection '' that increased the density especially at low latitudes . this may help to explain why hillier et al . ( 2001 ) could reproduce the early-1998 spectrum fairly well with a spherical wind model , but found later spectra more problematic . in two ways the temporal variability appears very consistent with the hypothetical bipolar wind geometry sketched in figure 11 ( 5 ) . first , a moderate shell ejection can easily decrease the ionization in the low - latitude wind , but should have less effect on the polar wind which is normally denser and less ionized . qualitatively , at least , the observed behavior matched this expectation . second , the 199899 temporary disappearance of the highest observable velocities ( @xmath31000 km s@xmath2 , figure 5 ) suggests that the fast material occurs in the polar wind _ per se _ within 0.5@xmath7 of the star , and not at larger radii in the homunculus lobes . our reason for saying this is that the recombination time @xmath44 is too long for rapid response , i.e. , longer than a year , at radii more than a few hundred au from the star . finally , we note that if the march 2000 data midway through the 5.5-year spectroscopic cycle represent the `` normal '' state , then figures 16 and 17 indicate that the post - event recovery extended over more than a year . this is not what one would expect in a simple binary model , since the hypothetical companion star should have moved a considerable distance , @xmath310 au , from the primary in less than 6 months after periastron passage ( see , e.g. , davidson 1999 ) . the results discussed above suggest that the 199798 spectroscopic event was essentially a disturbance of the primary wind , and not , for instance , an eclipse or occultation . most likely the term `` shell ejection '' is valid to some extent , as advocated by zanella et al.(1984 ) , davidson ( 1999 ) , and davidson et al . none of the observations directly requires a companion star ; a single - star model remains conceivable , but a binary model is not precluded either . here , largely independent of whether the system is a binary , we describe a qualitative scenario for a spectroscopic event with the observed latitude dependences , and we list a variety of ways in which it appears consistent with observations . we assume that stronger hydrogen p cygni absorption implies higher local mass density as discussed in 5 ; other interpretations may exist but would be more complicated . detailed calculations are obviously needed but will be quite difficult , since a proper model must be two- or three - dimensional , with continuum- and line - driven gas dynamics , rotation , realistic nlte level populations and radiative transfer , etc . figure 18 shows the likely arrangement of zones in @xmath0 car s wind at three sample times : before , during , and soon after a spectroscopic event . this is merely a schematic diagram , and boundaries between zones are not necessarily sharp . figure 18@xmath45 usually applies : ionizing uv can escape through the lower - density low - latitude regions as conjectured in 5 . then , during an event ( figure 18@xmath46 ) , the low - latitude wind becomes denser , mostly non - ionized , and opaque to ionizing photons . a density - increase factor of the order of 2 may suffice and the extra mass ejected might be as little as @xmath47 @xmath48 . meanwhile , a relatively weaker high - latitude increase has less effect on the emergent spectrum ; lessened values of @xmath17 and @xmath18 shown in figures 16 and 17 may indicate that hydrogen recombines closer to the star , or that the polar wind slows as its mass increases . a few months later ( figure 18@xmath49 ) , the ejected `` shell '' gradually becomes transparent . altogether this scheme , where changing density at low latitudes acts as a sort of _ valve _ regulating the escape of uv photons , can account for nearly all observed characteristics of each spectroscopic event : \1 . narrow [ ne iii ] , [ fe iii ] , he i , and related emission lines temporarily fade the phenomenon that led zanella et al . ( 1984 ) to propose a uv - quashing `` shell ejection . '' we now know that these features originate in slow - moving _ low - latitude _ ejecta close to the star ( davidson et al . 1995 , 1997 ; smith et al . 2000 ) . broad ( i.e. , wind ) components of some low - excitation lines such as fe ii become stronger ( damineli et al . this is qualitatively unsurprising if the volume and mass of low - excitation material increase as the wind becomes quasi - spherical ( figure 18@xmath46 ) . \5 . in the time period between spectroscopic events , colliding - wind x - ray binary models for @xmath0 car ( ishibashi et al . 1999 ; corcoran et al . 2001 ; pittard & corcoran 2002 ) seem to require a primary mass - loss rate of only a few times 10@xmath50 @xmath48 yr@xmath2 , a factor of three to ten less than the familiar value based on other considerations ( hillier et al . 2001 ; davidson & humphreys 1997 ) . however , if the primary wind is not spherical in its `` normal '' state , then the x - ray parameters refer mainly to local densities around the shock interface , presumably within the low - latitude , lower density part of the primary wind sketched in figure 18@xmath45 . corcoran et al.(2001 ) also invoke a sudden density increase to explain the collapse of the x - ray flux during an event , which seems consistent with the shell - ejection scenario depicted in figure 18 . the radio - wavelength behavior of @xmath0 car includes varying free - free emission near the equatorial plane ( duncan et al . the source is compact during an event , and later expands irregularly as ionizing photons penetrate to larger distances from the star . near - infrared images , also representing free - free continuum emission mostly at low latitudes , show changes morphologically similar to the radio emission ( smith & gehrz 2000 ) . the 7 mm flux , produced by free - free emission in outer parts of the wind , fades during a spectroscopic event ( abraham & damineli 1999 ) . this can be explained if some normally ionized ( low - latitude ) portions of the wind recombine during the event . ( millimeter - wavelength continuum mostly originates within 100 au of the star , while centimeter wavelengths mentioned above in point 6 represent ejecta farther out . ) \8 . the variable geometry and emitting volume of the wind may cause serious line profile variations like those that are observed . as noted by mcgregor et al . ( 1999 ) and davidson et al . ( 2000 ) , this makes it difficult or impossible to correlate radial velocities with possible orbital motion . there is almost no obvious large - scale photoionization in the high - latitude zones around the star ; the [ ne iii ] and similar features mentioned in point 1 above appear to be fairly close to the equatorial plane based on their doppler velocities , and so is the centimeter - wavelength emission mentioned in point 6 . since there is no shortage of gas at various radii within the homunculus lobes ( see ishibashi et al . 2001 ) , we conclude that ionizing uv photons are scarce at high latitudes . this may be an objection to the most likely type of binary model , wherein a hot companion star produces ionizing radiation . since the primary wind does not seem dense enough to block ionizing photons at the hypothetical apoastron distance ( @xmath3 30 au ) , we should expect to see bright photoionized material in near - polar locations around the star , contrary to observations . the single - star geometry sketched in figure 18@xmath45 , on the other hand , can emit ionizing radiation toward low latitudes only , consistent with the observed large - scale ionization . in most binary schemes for @xmath0 car s 5.5-year cycle ( e.g. , damineli et al . 1997 , 2000 ; and various authors in morse et al . 1999 ) , a hot companion star in an eccentric orbit is the main perpetrator . it provides a fast wind needed for the x - rays , and far greater in the energy budget it supplies ionizing uv radiation to excite emission lines in surrounding ejecta . emergent uv and x - rays are modulated by the secondary star s varying radial location in the dense primary wind . in the latitude - dependent shell ejection scenario that we propose , however , much ( perhaps nearly all ) of the uv radiation comes from the primary rather than the secondary star . a companion star may still be needed for the hard x - rays from @xmath0 car ( corcoran et al.1998 , 2001 ; ishibashi et al . 1999 ; ishibashi 2001 ; pittard & corcoran 2002 ; etc . ) , but at wavelengths other than x - rays i.e. , for 99.99% of the energy flow the main role of a hypothetical secondary star is to regulate the periodicity . it may supply the most extreme uv photons , but this is unproven . most likely a spectroscopic event occurs near periastron passage when the secondary star tidally disturbs the surface layers of an already unstable primary . this is feasible , at a separation more than 3 times the primary s radius , only because radiation pressure and ( probably ) rotation magnify the tidal effect . in 5.3 we noted how @xmath0 car s wind geometry qualitatively matches theoretical predictions by owocki et al . ( 1996 ) for a rotating star with gravity darkening . in that formulation , gravity darkening followed @xmath51 ( von zeipel 1924 ) , and simple scaling arguments would lead to @xmath52 . this is probably an oversimplification for @xmath0 car s optically - thick wind , but the agreement between the observed wind structure and the associated predictions from theory suggest that it is not too far - fetched . therefore , the pole - to - equator velocity ratio of @xmath32 ( figure 7 ) during the normal high - excitation state implies , very crudely , that @xmath53 . thus , if @xmath0 car s polar temperature is close to the often - adopted value of 30,000 k ( e.g. , davidson & humphreys 1997 ; hillier et al . 2001 ) , then the corresponding temperature at low latitudes could be dangerously close to 21,000 k , where an important opacity increase toward lower temperatures occurs ( see pauldrach & puls 1990 ; lamers & pauldrach 1991 ) . intuition then guides us to ask : what happens if the rotation rate is slightly increased ? such a change should increase the amount of gravity darkening , decrease @xmath54 , and raise opacity at the equator i.e. rotation may trigger an equatorial shell ejection . the rotation in question may be innate or it may have been forced by a companion star , but either way the bipolar homunculus strongly suggests that rotation is indeed significant in @xmath0 car . thus we conjecture that _ systematic variations of the surface rotation rate play a role in @xmath0 car s 5.5-year cycle . _ the reasoning , mentioned earlier by davidson ( 1999 ) , goes like this : \3 . afterward , the star very quickly readjusts to a new state of quasi - hydrostatic equilibrium . the dense inner region containing most of the mass is scarcely affected , but the outermost surviving layers must expand appreciably outward and therefore now have slower rotation rates . \4 . at this point the star s dense center rotates faster than the less massive outer layers . therefore angular momentum will diffuse outward , gradually accelerating the surface rotation . according to simple order - of - magnitude turbulent diffusion estimates , a spin - up timescale of a few years seems plausible following the ejection of less than 10@xmath55 @xmath48 thus we conjecture that the surface rotation rate gradually increases during the 5.5-year period , with possible effects on the low - latitude temperature ( gravity darkening ) , on the wind , etc . rotation is always far slower than the naive `` break - up '' limit , but toward the end of the cycle it may be fast enough to have a destabilizing effect as noted earlier . internal thermal readjustment can meanwhile play a related role . most likely this development affects a binary model , and tidal friction near periastron may play a long - term role . however , one can also imagine a _ single - star _ cycle based on the above reasoning . in such a model , 5.5 years is the time required for spin - up to permit another mass ejection event . there is no obvious physical reason why such a periodicity can not be as regular as the observed cycle . other single - star alternatives for short - term periodicity exist as well ( e.g. , stothers 2000 ; guzik et al . 1999 ; davidson 1999 ) . these could act independently , or they could be assisted or regularized by interactions with a companion . the wind geometry , shaped by rotation and/or a companion star , may also affect long - term spectroscopic variations . feast et al . ( 2001 ) found in old spectrograms that @xmath0 car was continuously in a low - excitation ( `` event '' ) state for many years before 1920 . most likely the wind was then quasi - spherical as in figure 18@xmath46 . the large amount of mass ejected in @xmath0 car s 19th century eruptions ( see humphreys et al . 1999 ) represented most of the star s prior radius , so we _ expect _ the surface of @xmath0 car to have been rotating very slowly a century ago . without rotation , there was no reason for the wind to have the modern non - spherical shape shown in figure 18@xmath45 ; in particular there would not have been a lower - density zone near the equatorial plane which today allows uv photons to escape . therefore , no matter whether the hard uv comes from the primary or from a secondary star , such radiation could have been absorbed within the dense primary wind at low latitudes hence no high - excitation emission lines then . later , we must suppose , the surface rotation rate gradually increased for the reason noted above , until it became sufficient to establish the current geometry and high - excitation spectrum sometime around 1940 when a sudden brightening occurred ( de vaucouleurs & eggen 1952 ; thackeray 1953 ; gaviola 1953 ) . this coincidence suggests that the presence of spectroscopic events is linked to changes in the primary star , which dominates the observed optical flux . although @xmath0 car is an extreme case , rotation - induced wind structure and angular momentum diffusion may play important roles for mass ejections from other luminous blue variables . this is not a new revelation , but our latitude - dependent study of @xmath0 car suggests the possibilities quite vividly . the authors benefitted from interesting and informative conversations with stan owocki , joe cassinelli , and roberta humphreys . an anonymous referee gave several suggestions that improved the presentation of the paper . is grateful for the support of a nasa / gsrp fellowship from goddard space flight center . additional support was provided by nasa through grant number go-8327 ( p.i . : davidson ) from the space telescope science institute , which is operated by the association for research in astronomy , inc . , under nasa contract nas 5 - 26555 , and through stis gto resources and funding ( p.i . : gull ) . gull , t.r . , ishibashi , k. , davidson , k. , and the cycle 7 stis go team , 1999 , in asp conf . ser . 179 , eta carinae at the millenium , ed . j.a . morse , r.m . humphreys , & a. damineli ( san francisco : asp ) , 144 + 7.3 & 0.44 & 785 & 53 & 560 & 435 & nw lobe + + 5.4 & 0.57 & 500 & 45 & 475 & 410 & nw lobe + + 3.3 & 1.00 & 350 & 35 & & 210 & fan , no p cyg abs . + + 2.6 & 0.38 & 220 & 28 & & 150 & fan , no p cyg abs . + star & 0.15 & 0 & 49 & 540 & 410 & central star + 1.0 & 0.51 & 70 & 56 & 640 & 430 & se lobe + 1.5 & 0.33 & 105 & 61 & 630 & 450 & se lobe + 2.5 & 0.63 & 60 & 66 & 700 & 460 & se lobe , shadow + 3.0 & 0.38 & 100 & 69 & 760 & 480 & se lobe + 3.8 & 0.38 & 125 & 73 & 780 & 480 & se lobe + 4.3 & 0.38 & 125 & 77 & 875 & 480 & se lobe + 5.2 & 0.38 & 165 & 83 & 1000 & 550 & se lobe + 5.8 & 0.25 & 185 & 86 & 1150 & 590 & se lobe , polar axis + 7.0 & 0.51 & 270 & 82 & 1050 & 540 & se lobe + 8.0 & 0.38 & 350 & 74 & 870 & 460 & se lobe + ne 5 & 1.00 & 195 & 0 & 750 & 440 & equatorial ejecta + position & width & @xmath12 & latitude & @xmath17(1998 ) & @xmath18(1998 ) & @xmath17(1999 ) & @xmath18(1999 ) & comments + ( arcsec ) & ( arcsec ) & km s@xmath2 & ( deg ) & km s@xmath2 & km s@xmath2 & km s@xmath2 & km s@xmath2 & + + 7.4 & 0.65 & 930 & 56 & 660 & 520 & & & nw lobe + + 6.0 & 1.84 & 710 & 49 & 650 & 500 & 480 & 360 & nw lobe + + 4.6 & 0.62 & 450 & 43 & & & & 300 & nw lobe + + 3.2 & 0.95 & 225 & 36 & & 170 & & 170 & fan , no p cyg abs . + + 2.0 & 0.85 & 60 & 31 & 750 & 580 & & & fan , no p cyg abs . + star & 0.15 & 0 & 49 & 680 & 520 & 510 & 460 & central star + 1.1 & 0.29 & 70 & 55 & 740 & 590 & & & se lobe + 1.6 & 0.47 & 75 & 60 & 850 & 540 & 580 & 460 & se lobe + 2.3 & 0.55 & 75 & 64 & 780 & 550 & & 500 & se lobe + 2.9 & 0.70 & 90 & 67 & 780 & 540 & & 500 & se lobe + 3.8 & 0.57 & 100 & 71 & 610 & 480 & 580 & 460 & se lobe + 4.4 & 0.37 & 145 & 73 & 640 & 490 & 640 & 490 & se lobe + 5.0 & 0.49 & 165 & 76 & 650 & 500 & 680 & 500 & se lobe + 5.7 & 0.52 & 192 & 79 & 640 & 500 & 690 & 500 & se lobe + 6.9 & 0.52 & 270 & 78 & 620 & 480 & 650 & 480 & se lobe + 7.5 & 0.37 & 345 & 75 & 590 & 480 & 580 & 450 & se lobe + 7.9 & 0.39 & 400 & 72 & 600 & 480 & 580 & 460 & se lobe +